Original Research

Recent studies have emphasized the importance of tight metabolic control in combination with state-of-the-art management of other risk factors to prevent macrovascular and microvascular complications in patients with type 2 diabetes mellitus.^1-5 Guidelines for diabetes care recommend systematic monitoring of patients’ health status, including metabolic control, cardiovascular risk factors, and desired outcome of care.^6-8

The formulation of clinical guidelines alone, however, is insufficient to improve actual care.^9,10 Strategies to reinforce the guidelines in daily practice include monitoring the patient’s clinical condition over a given period of time, feedback to the clinician about the outcome, audit of clinical performance, academic detailing by peers, and evidence-based guidelines.^10-12 Monitoring and feedback with systematic follow-up of relevant treatment targets enhanced a proactive approach to patients,¹³ which is a key factor for successful diabetes care.¹⁴ As large numbers of patients with type 2 diabetes are treated in family practice, it is important that target-specific monitoring fit into the overall primary care function of family practice and that it answer the needs, demands, and expectations of patients.

Since 1985, the Nijmegen University Department of Family Practice has been developing a computer-assisted practice network, the Nijmegen Academic Research Network CMR/NMP, to study chronic diseases.^15,16 The objectives of this network are to support care for patients with chronic diseases and to create an optimal setting for clinical research under family practice conditions. This paper analyzes the outcome of diabetes care in the CMR/NMP 7 years after the introduction of an audit-enhanced monitoring system (AEMS).¹⁷

The aims were to assess (1) the outcome of care compared with external guideline criteria and the results of clinical trials, and (2) the relationship of outcome to process of care measures and to patient-related and practice-related factors.

Methods

Study population

Data were collected at the 10 family practices in the CMR/NMP, with 25 family physicians and a patient list of approximately 46,000 in 1999.¹⁶ All patients meeting World Health Organization criteria for the diagnosis of type 2 diabetes mellitus and under treatment by a family physician in 1993 and 1999 were included in the AEMS.^15,18 Patients who were treated with insulin within 1 year of diagnosis and who continued to take it were considered to have type 1 diabetes mellitus. All other patients were regarded as type 2, regardless of current treatment. For this study we included all type 2 diabetes patients under treatment by their family physician in 1993 and 1999. Patients who had died or who had moved to another area or been admitted to a residential nursing home before the end of the year were excluded, as were those who had been newly diagnosed during the year.

Audit-enhanced monitoring system

Since 1989, data have been collected on all type 2 diabetes patients at the time of diagnosis and during all regular (quarterly) diabetes-related outpatient visits. In 1992, a structured annual review, based on guidelines from the Dutch College of Family Physicians,¹⁹ was added. Starting in 1992, monitoring has consisted of the assessment of (1) compliance with 3 monthly control visits and an annual review visit; (2) glycemic control (ie, fasting blood glucose and Hb A_1c); (3) diabetes-related complications (ie, retinopathy, creatinine clearance, and foot problems); (4) cardiovascular risk factors (ie, smoking behavior, blood pressure, and lipid profile); (5) cardiovascular morbidity (ie, myocardial infarction, angina pectoris, heart failure, peripheral vascular disease, transient ischemic attack, or cerebrovascular accident). In addition, all reasons for dropping out, including cause of death, were recorded. Morbidity and causes of death were defined as in the International Classification of Health Problems in Primary Care.

To facilitate data collection, a computerized Research Registration System (RRS) was developed. The system was integrated into a standard Dutch electronic record system for family practice (Promedico, Euroned). The RRS generates templates for recording data at the quarterly or annual diabetes control visits into the patient’s electronic record. Templates guide the delivery of care and a reminder system is integrated into the RRS. Office assistants contact patients who do not come in for visits at regular intervals, both those (< 1%) who usually do not come in and those who are supposed to but fail to do so.

Family physicians sent the RRS data files to the University Department of Family Practice, where they were processed into a feedback report on process of care and outcome of care measures on 3 levels: (1) total study population; (2) practice population; (3) individual patient. Process and outcome measures were compared with external criteria based on guidelines from the Dutch College of Family Medicine and with average performance at the other practices. Feedback items were selected in consultation with the participating physicians. In this way, feedback corresponded with daily practice needs. During the project, the feedback was gradually extended from process to outcome measures. The feedback was standard to all practices.

Feedback was discussed at University Department of Family Medicine meetings, which maintained uniform registration and safeguarded the progress of the project. The feedback was also sent to every practice and participating GP. This report contained practice-level as well as physician-level data. The Figure demonstrates one way in which data are presented at the meetings and shows the percentage of patients who attended their annual diabetes control visit in the year studied.

Targets for care

Targets for care consisted of 2 elements: process and outcome measures. The key marker for process of care was compliance to the annual diabetes control visit. Key markers for desired outcome of care were (1) Hb A_1c < 8.5%,¹⁹ (2) blood pressure less than 160/90 mm Hg (revised to 150/85 mm Hg in 1999),^8,19 and (3) lipids in accordance with Dutch guidelines for general practice⁸: (a) cholesterol < 5 mmol (192 mg/dL) for patients with cardiovascular morbidity; (b) cholesterol/HDL ratio < 5.0 in smokers without cardiovascular morbidity; and (c) cholesterol/HDL < 6.0 in nonsmokers without cardiovascular morbidity. These guidelines for lipid-lowering therapy are based on sex, a life expectancy of at least 5 years, smoking status, presence of cardiovascular morbidity, total cholesterol levels, high-density lipoprotein (HDL) cholesterol levels, and triglyceride levels. If even 1 of these variables is absent, the potential value of lipid lowering cannot be determined.⁸

Analysis

Cross-sectional analysis was performed on the outcome of diabetes care in patients with type 2 diabetes who were treated by their family physicians in 1993 and 1999. The comparison was based on all patients who had been treated for the full calendar year in 1993 and 1999; therefore, it was based on a dynamic population. Process and outcome measures are compared using the chi-squared, unpaired t, or Mann–Whitney test, as appropriate. Results are expressed as means plus or minus standard deviations or as proportions. Multilevel analysis was performed to assess factors that contributed to the variance in compliance with the annual review and the desired glycemic level (Hb A_1c < 8.5%).

Results

In 1993, 540 type 2 diabetes patients (prevalence 1.3%) were included in the AEMS. Of these, 51 had been newly diagnosed (incidence 1.2/1000); 37 had been treated by a specialist (7%); and 20 did not participate (4%). Excluding the 108 patients in the latter 3 categories left a total of 432 patients for analysis. In 1999, 851 patients were included (prevalence 1.9%). Of these, 138 had been newly diagnosed (incidence 3.0/1000); 88 had been treated by a specialist (10%); and 31 did not participate (4%). Excluding the 257 patients in those 3 categories left 594 for analysis. Table 1 shows the baseline characteristics of patients included in the analysis.

Annual review was attended by 73% of patients in 1993 and 84% of patients in 1999 (Table 2). Increased compliance was achieved at all the practices, although differences between practices remained in 1999 (Figure). Univariate analysis showed that compliance with the annual review in 1999 was related to the practice (P = .001) but not to patient factors such as sex, age, duration of diabetes, therapy regimen, or cardiovascular morbidity, even after adjusting for blood glucose levels. Patients who did not attend their annual diabetes control visit had statistically significantly higher fasting blood glucose levels than patients who did comply (8.9 mmol/L [160 mg/dL] vs 8.2 mmol [147 mg/dL], P = .03). In 1993, 59% of patients had visited an ophthalmologist in the previous 2 years versus 80% in 1999.

In 1993, Hb A_1c was measured in 51% of patients with a mean of 8.2%. In 1999, compliance in measurement of Hb A_1c improved to 82%, with a mean Hb A_1c level of 7.1% (P = .0001, Table 3). The percentages of patients with an Hb A_1c level of more than 8.5% decreased from 41% to 13% (P = .001). These outcomes were associated with changes in treatment (P = .001): a decrease in patients treated with diet only (22% in 1993 vs 13% in 1999) and with oral hypoglycemic monotherapy (45% in 1993 vs 37% in 1999); an increase in patients treated with combination therapy using 2 or more oral hypoglycemic agents (22% in 1993 vs 31% in 1999); and an increase in insulin therapy (11% in 1993 vs 19% in 1999). Univariate analysis showed that poor glycemic control (Hb A_1c > 8.5%) in 1999 was related to the therapy regimen (P = .001) but not to sex, age, duration of diabetes, cardiovascular morbidity, or practice. The glycemic control in patients treated with combination therapy or insulin was poorer than in patients treated with diet only or oral hypoglycemic monotherapy, probably reflecting the fact that patients with less severe disease are managed with single agents and diet.

Compliance with measurement of blood pressure improved from 72% to 83% during the study period (Table 3). However, the percentage of patients with a systolic blood pressure below 150 mm Hg or a diastolic blood pressure below 85 mm Hg did not change between 1993 and 1999 whether patients were hypertensive or not. In hypertensive patients with type 2 diabetes, the mean diastolic blood pressure decreased from 88 mm Hg to 85 mm Hg (P = .004), but mean systolic blood pressure did not change.

The mean cholesterol level was lower in 1999 than in 1993 (6.2 vs 5.4 mmol/L; 238 mg/dL vs 207 mg/dL, P = .0001), as was the mean triglyceride level (2.54 mmol/L vs 2.07 mmol/L; 221 mg/dL vs 180 mg/dL, P = .0003). In both years, data regarding which patients could be considered for lipid-lowering therapy were available for 63% and 82%, respectively. In 1993, a far higher proportion of patients had failed to reach lipid target levels than was the case in 1999 (48% vs 28%, respectively, P = .001).

Multilevel analysis showed that paying an annual diabetes control visit (a process outcome) was related to the practice (intraclass correlation coefficient [ICC] = 0.29) but not to patient factors. Reaching the glycemic target level of Hb A_1c < 8.5%, however, was not related to practice factors (ICC = 0.003).

TABLE 1
Chacteristics of type 2 diabetes patients under family physician care in 1993 and 1999

TABLE 2
Process of care for type 2 diabetes patients under family physician care, 1993 vs 1999

TABLE 3
Outcomes of care for type 2 diabetes patients under family physician care, 1993 vs 1999

FIGURE
Percentage of patients with annual review (target = 75%) in 1999 (n=594)

Discussion

During 7 years of structured audit-enhanced monitoring of patients with type 2 diabetes in an academic family practice research network, the intermediate measures of diabetes care improved. In particular, the mean Hb A1c of 7.1% can be seen as a measure of good quality of care. The number of patients treated according to Dutch family practice guidelines (a process of care outcome) also increased.⁸

While our data were collected during normal daily care (effectiveness), the findings come close to the outcome of care under ideal trial conditions (efficacy).²¹ In the UK Prospective Diabetes Study (UKPDS), the median Hb A_1c level for all newly diagnosed patients in the group with intensive blood glucose control over 10 years reached a comparable level of 7.0%.¹ Thus, the outcome of our study approaches that achieved under trial conditions. When we analyzed patients without outcome data as poorly controlled (worst-case scenario), Hb A_1c was less than 8.5% in 28%.

The trend of improvement in glycemic control could have been a result of improved overall diabetes care in the Netherlands during the study period. Data about the outcome of diabetes care in the family medicine setting in the Netherlands during the study period are scarce and, when available, are derived from other research networks. In these networks a mean Hb A_1c of 7.0% to 7.6% was reached.²² Yet indicators from other studies suggest that our results were far better than outcome from usual care. Recently published data on such outcomes in family medicine in the Netherlands showed that Hb A_1c, blood pressure, and lipids were measured in less than 30% of patients.^23,24 Outcomes from usual care as reported in research studies appear to be strongly biased by selection and probably cannot serve as a valid reference value.

The disappointing effect on the percentage of patients who reached the target blood pressure could have resulted from evaluating the data prematurely. When the study began, the primary objective was to improve glycemic control. Shortly after the publication of the Scandinavian Simvastatin Survival Study (4S)³ and the UKPDS,¹² the guidelines of the Dutch College of Family Physicians were changed⁸ and more attention was paid to blood pressure and lipid control. This new approach was discussed with the participating family physicians. Consequently, the target for blood pressure was revised from 160/90 mm Hg to 150/85 mm Hg and lipid-lowering therapy was tailored to each patient’s cardiovascular risk profile. The 1999 outcome with respect to blood pressure and lipid control was measured only 1 year after these changes had been announced. Nevertheless, mean diastolic blood pressure in hypertensive patients and total cholesterol and triglyceride levels decreased significantly, and more patients reached target levels for lipids in 1999 than in 1993.

Our outcome was reached through enhanced compliance to guidelines. Therefore, the outcome in 1999 was based on a larger percentage of available patients. Because the AEMS studied a dynamic group of patients, the study groups in 1993 and 1999 were not identical. Theoretically, improvement in outcome could have been reached by including more easily manageable patients. However, no patient factors such as sex, age, duration of diabetes, treatment modality, or cardiovascular morbidity were related to compliance with annual review. The higher fasting blood glucose levels in patients who were noncompliant with annual review probably reflected under-treatment rather than more severe illness status. Therefore, we are confident that the findings reflect improved overall diabetes care.

The data on process measures in this study compare favorably with those of multipractice audits of diabetes care in the United Kingdom.^25-29 The high prevalence rate of 2.0% (exclusively patients with type 2 diabetes) supports the validity of our data.^25-27 Among our patients, 96% had been seen at least once during the previous year. In the large studies by Khunti and Bennett, only 85% had been seen during the previous year.^25,26 The mean annual compliance rate of nearly 85% with Hb A_1c and blood pressure measurements in our study was high. In particular, compliance with lipid control^25,26,28 and funduscopy^26,28 was better in our study. In 2 longitudinal studies that used an organized care system in which feedback was provided to the participating family physicians, compliance rates in process measures of up to 75% were reported.^30,31

The outcomes of this study were achieved in an academic family practice research network, with specific facilities for the proactive supervision of patients with chronic diseases. These results cannot and should not be generalized to “routine” family practice. Monitoring and feedback in routine family practice are in themselves insufficient to improve the quality of care.¹⁰ Care assessment should preferably take a more comprehensive approach in which evidence-based goals for care are formulated, care is improved to reach those goals, and care is measured to see whether those goals have been achieved.¹¹ Our academic network provides this comprehensive approach.

The electronic Research Registration System played an important role in the audit-enhanced monitoring. In the pilot phase of the project, paper records were used. Although using paper records had clear disadvantages, one could expect to achieve similar results using such records in combination with a central electronic data bank.

Conclusions

Outcomes of diabetes care in our family research setting were comparable with those reported in randomized controlled trials. Therefore, it is possible for the management of diabetes in family practice to be efficacious. This finding should encourage more efforts by physicians in family practice to bridge the gap between efficacy and effectiveness.

Important differences remained in achieved process measures between the academic family practices. While the outcome of diabetes care in the network was favorable, the outcome of treatment was unsatisfactory in a substantial number of patients. Further implementation strategies must be developed. The differences in achieved process measures were probably unrelated to socioeconomic differences between the practice populations, since the practice pairs 1/3 and 7/10 served comparable communities and had different levels of compliance.

Our study demonstrated that a high quality of diabetes care in family practice can be achieved. Audit-enhanced monitoring, which will provide the greatest benefit to the most patients with type 2 diabetes mellitus, should be implemented as part of a quality improvement system.

ACKNOWLEDGMENTS

The authors wish to thank the family physicians and practice nurses for their continuing support and data collection.

Recent studies have emphasized the importance of tight metabolic control in combination with state-of-the-art management of other risk factors to prevent macrovascular and microvascular complications in patients with type 2 diabetes mellitus.^1-5 Guidelines for diabetes care recommend systematic monitoring of patients’ health status, including metabolic control, cardiovascular risk factors, and desired outcome of care.^6-8

The formulation of clinical guidelines alone, however, is insufficient to improve actual care.^9,10 Strategies to reinforce the guidelines in daily practice include monitoring the patient’s clinical condition over a given period of time, feedback to the clinician about the outcome, audit of clinical performance, academic detailing by peers, and evidence-based guidelines.^10-12 Monitoring and feedback with systematic follow-up of relevant treatment targets enhanced a proactive approach to patients,¹³ which is a key factor for successful diabetes care.¹⁴ As large numbers of patients with type 2 diabetes are treated in family practice, it is important that target-specific monitoring fit into the overall primary care function of family practice and that it answer the needs, demands, and expectations of patients.

Since 1985, the Nijmegen University Department of Family Practice has been developing a computer-assisted practice network, the Nijmegen Academic Research Network CMR/NMP, to study chronic diseases.^15,16 The objectives of this network are to support care for patients with chronic diseases and to create an optimal setting for clinical research under family practice conditions. This paper analyzes the outcome of diabetes care in the CMR/NMP 7 years after the introduction of an audit-enhanced monitoring system (AEMS).¹⁷

The aims were to assess (1) the outcome of care compared with external guideline criteria and the results of clinical trials, and (2) the relationship of outcome to process of care measures and to patient-related and practice-related factors.

Methods

Study population

Data were collected at the 10 family practices in the CMR/NMP, with 25 family physicians and a patient list of approximately 46,000 in 1999.¹⁶ All patients meeting World Health Organization criteria for the diagnosis of type 2 diabetes mellitus and under treatment by a family physician in 1993 and 1999 were included in the AEMS.^15,18 Patients who were treated with insulin within 1 year of diagnosis and who continued to take it were considered to have type 1 diabetes mellitus. All other patients were regarded as type 2, regardless of current treatment. For this study we included all type 2 diabetes patients under treatment by their family physician in 1993 and 1999. Patients who had died or who had moved to another area or been admitted to a residential nursing home before the end of the year were excluded, as were those who had been newly diagnosed during the year.

Audit-enhanced monitoring system

Since 1989, data have been collected on all type 2 diabetes patients at the time of diagnosis and during all regular (quarterly) diabetes-related outpatient visits. In 1992, a structured annual review, based on guidelines from the Dutch College of Family Physicians,¹⁹ was added. Starting in 1992, monitoring has consisted of the assessment of (1) compliance with 3 monthly control visits and an annual review visit; (2) glycemic control (ie, fasting blood glucose and Hb A_1c); (3) diabetes-related complications (ie, retinopathy, creatinine clearance, and foot problems); (4) cardiovascular risk factors (ie, smoking behavior, blood pressure, and lipid profile); (5) cardiovascular morbidity (ie, myocardial infarction, angina pectoris, heart failure, peripheral vascular disease, transient ischemic attack, or cerebrovascular accident). In addition, all reasons for dropping out, including cause of death, were recorded. Morbidity and causes of death were defined as in the International Classification of Health Problems in Primary Care.

To facilitate data collection, a computerized Research Registration System (RRS) was developed. The system was integrated into a standard Dutch electronic record system for family practice (Promedico, Euroned). The RRS generates templates for recording data at the quarterly or annual diabetes control visits into the patient’s electronic record. Templates guide the delivery of care and a reminder system is integrated into the RRS. Office assistants contact patients who do not come in for visits at regular intervals, both those (< 1%) who usually do not come in and those who are supposed to but fail to do so.

Family physicians sent the RRS data files to the University Department of Family Practice, where they were processed into a feedback report on process of care and outcome of care measures on 3 levels: (1) total study population; (2) practice population; (3) individual patient. Process and outcome measures were compared with external criteria based on guidelines from the Dutch College of Family Medicine and with average performance at the other practices. Feedback items were selected in consultation with the participating physicians. In this way, feedback corresponded with daily practice needs. During the project, the feedback was gradually extended from process to outcome measures. The feedback was standard to all practices.

Feedback was discussed at University Department of Family Medicine meetings, which maintained uniform registration and safeguarded the progress of the project. The feedback was also sent to every practice and participating GP. This report contained practice-level as well as physician-level data. The Figure demonstrates one way in which data are presented at the meetings and shows the percentage of patients who attended their annual diabetes control visit in the year studied.

Targets for care

Targets for care consisted of 2 elements: process and outcome measures. The key marker for process of care was compliance to the annual diabetes control visit. Key markers for desired outcome of care were (1) Hb A_1c < 8.5%,¹⁹ (2) blood pressure less than 160/90 mm Hg (revised to 150/85 mm Hg in 1999),^8,19 and (3) lipids in accordance with Dutch guidelines for general practice⁸: (a) cholesterol < 5 mmol (192 mg/dL) for patients with cardiovascular morbidity; (b) cholesterol/HDL ratio < 5.0 in smokers without cardiovascular morbidity; and (c) cholesterol/HDL < 6.0 in nonsmokers without cardiovascular morbidity. These guidelines for lipid-lowering therapy are based on sex, a life expectancy of at least 5 years, smoking status, presence of cardiovascular morbidity, total cholesterol levels, high-density lipoprotein (HDL) cholesterol levels, and triglyceride levels. If even 1 of these variables is absent, the potential value of lipid lowering cannot be determined.⁸

Analysis

Cross-sectional analysis was performed on the outcome of diabetes care in patients with type 2 diabetes who were treated by their family physicians in 1993 and 1999. The comparison was based on all patients who had been treated for the full calendar year in 1993 and 1999; therefore, it was based on a dynamic population. Process and outcome measures are compared using the chi-squared, unpaired t, or Mann–Whitney test, as appropriate. Results are expressed as means plus or minus standard deviations or as proportions. Multilevel analysis was performed to assess factors that contributed to the variance in compliance with the annual review and the desired glycemic level (Hb A_1c < 8.5%).

Results

In 1993, 540 type 2 diabetes patients (prevalence 1.3%) were included in the AEMS. Of these, 51 had been newly diagnosed (incidence 1.2/1000); 37 had been treated by a specialist (7%); and 20 did not participate (4%). Excluding the 108 patients in the latter 3 categories left a total of 432 patients for analysis. In 1999, 851 patients were included (prevalence 1.9%). Of these, 138 had been newly diagnosed (incidence 3.0/1000); 88 had been treated by a specialist (10%); and 31 did not participate (4%). Excluding the 257 patients in those 3 categories left 594 for analysis. Table 1 shows the baseline characteristics of patients included in the analysis.

Annual review was attended by 73% of patients in 1993 and 84% of patients in 1999 (Table 2). Increased compliance was achieved at all the practices, although differences between practices remained in 1999 (Figure). Univariate analysis showed that compliance with the annual review in 1999 was related to the practice (P = .001) but not to patient factors such as sex, age, duration of diabetes, therapy regimen, or cardiovascular morbidity, even after adjusting for blood glucose levels. Patients who did not attend their annual diabetes control visit had statistically significantly higher fasting blood glucose levels than patients who did comply (8.9 mmol/L [160 mg/dL] vs 8.2 mmol [147 mg/dL], P = .03). In 1993, 59% of patients had visited an ophthalmologist in the previous 2 years versus 80% in 1999.

In 1993, Hb A_1c was measured in 51% of patients with a mean of 8.2%. In 1999, compliance in measurement of Hb A_1c improved to 82%, with a mean Hb A_1c level of 7.1% (P = .0001, Table 3). The percentages of patients with an Hb A_1c level of more than 8.5% decreased from 41% to 13% (P = .001). These outcomes were associated with changes in treatment (P = .001): a decrease in patients treated with diet only (22% in 1993 vs 13% in 1999) and with oral hypoglycemic monotherapy (45% in 1993 vs 37% in 1999); an increase in patients treated with combination therapy using 2 or more oral hypoglycemic agents (22% in 1993 vs 31% in 1999); and an increase in insulin therapy (11% in 1993 vs 19% in 1999). Univariate analysis showed that poor glycemic control (Hb A_1c > 8.5%) in 1999 was related to the therapy regimen (P = .001) but not to sex, age, duration of diabetes, cardiovascular morbidity, or practice. The glycemic control in patients treated with combination therapy or insulin was poorer than in patients treated with diet only or oral hypoglycemic monotherapy, probably reflecting the fact that patients with less severe disease are managed with single agents and diet.

Compliance with measurement of blood pressure improved from 72% to 83% during the study period (Table 3). However, the percentage of patients with a systolic blood pressure below 150 mm Hg or a diastolic blood pressure below 85 mm Hg did not change between 1993 and 1999 whether patients were hypertensive or not. In hypertensive patients with type 2 diabetes, the mean diastolic blood pressure decreased from 88 mm Hg to 85 mm Hg (P = .004), but mean systolic blood pressure did not change.

The mean cholesterol level was lower in 1999 than in 1993 (6.2 vs 5.4 mmol/L; 238 mg/dL vs 207 mg/dL, P = .0001), as was the mean triglyceride level (2.54 mmol/L vs 2.07 mmol/L; 221 mg/dL vs 180 mg/dL, P = .0003). In both years, data regarding which patients could be considered for lipid-lowering therapy were available for 63% and 82%, respectively. In 1993, a far higher proportion of patients had failed to reach lipid target levels than was the case in 1999 (48% vs 28%, respectively, P = .001).

Multilevel analysis showed that paying an annual diabetes control visit (a process outcome) was related to the practice (intraclass correlation coefficient [ICC] = 0.29) but not to patient factors. Reaching the glycemic target level of Hb A_1c < 8.5%, however, was not related to practice factors (ICC = 0.003).

TABLE 1
Chacteristics of type 2 diabetes patients under family physician care in 1993 and 1999

TABLE 2
Process of care for type 2 diabetes patients under family physician care, 1993 vs 1999

TABLE 3
Outcomes of care for type 2 diabetes patients under family physician care, 1993 vs 1999

FIGURE
Percentage of patients with annual review (target = 75%) in 1999 (n=594)

Discussion

During 7 years of structured audit-enhanced monitoring of patients with type 2 diabetes in an academic family practice research network, the intermediate measures of diabetes care improved. In particular, the mean Hb A1c of 7.1% can be seen as a measure of good quality of care. The number of patients treated according to Dutch family practice guidelines (a process of care outcome) also increased.⁸

While our data were collected during normal daily care (effectiveness), the findings come close to the outcome of care under ideal trial conditions (efficacy).²¹ In the UK Prospective Diabetes Study (UKPDS), the median Hb A_1c level for all newly diagnosed patients in the group with intensive blood glucose control over 10 years reached a comparable level of 7.0%.¹ Thus, the outcome of our study approaches that achieved under trial conditions. When we analyzed patients without outcome data as poorly controlled (worst-case scenario), Hb A_1c was less than 8.5% in 28%.

The trend of improvement in glycemic control could have been a result of improved overall diabetes care in the Netherlands during the study period. Data about the outcome of diabetes care in the family medicine setting in the Netherlands during the study period are scarce and, when available, are derived from other research networks. In these networks a mean Hb A_1c of 7.0% to 7.6% was reached.²² Yet indicators from other studies suggest that our results were far better than outcome from usual care. Recently published data on such outcomes in family medicine in the Netherlands showed that Hb A_1c, blood pressure, and lipids were measured in less than 30% of patients.^23,24 Outcomes from usual care as reported in research studies appear to be strongly biased by selection and probably cannot serve as a valid reference value.

The disappointing effect on the percentage of patients who reached the target blood pressure could have resulted from evaluating the data prematurely. When the study began, the primary objective was to improve glycemic control. Shortly after the publication of the Scandinavian Simvastatin Survival Study (4S)³ and the UKPDS,¹² the guidelines of the Dutch College of Family Physicians were changed⁸ and more attention was paid to blood pressure and lipid control. This new approach was discussed with the participating family physicians. Consequently, the target for blood pressure was revised from 160/90 mm Hg to 150/85 mm Hg and lipid-lowering therapy was tailored to each patient’s cardiovascular risk profile. The 1999 outcome with respect to blood pressure and lipid control was measured only 1 year after these changes had been announced. Nevertheless, mean diastolic blood pressure in hypertensive patients and total cholesterol and triglyceride levels decreased significantly, and more patients reached target levels for lipids in 1999 than in 1993.

Our outcome was reached through enhanced compliance to guidelines. Therefore, the outcome in 1999 was based on a larger percentage of available patients. Because the AEMS studied a dynamic group of patients, the study groups in 1993 and 1999 were not identical. Theoretically, improvement in outcome could have been reached by including more easily manageable patients. However, no patient factors such as sex, age, duration of diabetes, treatment modality, or cardiovascular morbidity were related to compliance with annual review. The higher fasting blood glucose levels in patients who were noncompliant with annual review probably reflected under-treatment rather than more severe illness status. Therefore, we are confident that the findings reflect improved overall diabetes care.

The data on process measures in this study compare favorably with those of multipractice audits of diabetes care in the United Kingdom.^25-29 The high prevalence rate of 2.0% (exclusively patients with type 2 diabetes) supports the validity of our data.^25-27 Among our patients, 96% had been seen at least once during the previous year. In the large studies by Khunti and Bennett, only 85% had been seen during the previous year.^25,26 The mean annual compliance rate of nearly 85% with Hb A_1c and blood pressure measurements in our study was high. In particular, compliance with lipid control^25,26,28 and funduscopy^26,28 was better in our study. In 2 longitudinal studies that used an organized care system in which feedback was provided to the participating family physicians, compliance rates in process measures of up to 75% were reported.^30,31

The outcomes of this study were achieved in an academic family practice research network, with specific facilities for the proactive supervision of patients with chronic diseases. These results cannot and should not be generalized to “routine” family practice. Monitoring and feedback in routine family practice are in themselves insufficient to improve the quality of care.¹⁰ Care assessment should preferably take a more comprehensive approach in which evidence-based goals for care are formulated, care is improved to reach those goals, and care is measured to see whether those goals have been achieved.¹¹ Our academic network provides this comprehensive approach.

The electronic Research Registration System played an important role in the audit-enhanced monitoring. In the pilot phase of the project, paper records were used. Although using paper records had clear disadvantages, one could expect to achieve similar results using such records in combination with a central electronic data bank.

Conclusions

Outcomes of diabetes care in our family research setting were comparable with those reported in randomized controlled trials. Therefore, it is possible for the management of diabetes in family practice to be efficacious. This finding should encourage more efforts by physicians in family practice to bridge the gap between efficacy and effectiveness.

Important differences remained in achieved process measures between the academic family practices. While the outcome of diabetes care in the network was favorable, the outcome of treatment was unsatisfactory in a substantial number of patients. Further implementation strategies must be developed. The differences in achieved process measures were probably unrelated to socioeconomic differences between the practice populations, since the practice pairs 1/3 and 7/10 served comparable communities and had different levels of compliance.

Our study demonstrated that a high quality of diabetes care in family practice can be achieved. Audit-enhanced monitoring, which will provide the greatest benefit to the most patients with type 2 diabetes mellitus, should be implemented as part of a quality improvement system.

ACKNOWLEDGMENTS

The authors wish to thank the family physicians and practice nurses for their continuing support and data collection.

Recent studies have emphasized the importance of tight metabolic control in combination with state-of-the-art management of other risk factors to prevent macrovascular and microvascular complications in patients with type 2 diabetes mellitus.^1-5 Guidelines for diabetes care recommend systematic monitoring of patients’ health status, including metabolic control, cardiovascular risk factors, and desired outcome of care.^6-8

The formulation of clinical guidelines alone, however, is insufficient to improve actual care.^9,10 Strategies to reinforce the guidelines in daily practice include monitoring the patient’s clinical condition over a given period of time, feedback to the clinician about the outcome, audit of clinical performance, academic detailing by peers, and evidence-based guidelines.^10-12 Monitoring and feedback with systematic follow-up of relevant treatment targets enhanced a proactive approach to patients,¹³ which is a key factor for successful diabetes care.¹⁴ As large numbers of patients with type 2 diabetes are treated in family practice, it is important that target-specific monitoring fit into the overall primary care function of family practice and that it answer the needs, demands, and expectations of patients.

Since 1985, the Nijmegen University Department of Family Practice has been developing a computer-assisted practice network, the Nijmegen Academic Research Network CMR/NMP, to study chronic diseases.^15,16 The objectives of this network are to support care for patients with chronic diseases and to create an optimal setting for clinical research under family practice conditions. This paper analyzes the outcome of diabetes care in the CMR/NMP 7 years after the introduction of an audit-enhanced monitoring system (AEMS).¹⁷

The aims were to assess (1) the outcome of care compared with external guideline criteria and the results of clinical trials, and (2) the relationship of outcome to process of care measures and to patient-related and practice-related factors.

Methods

Study population

Data were collected at the 10 family practices in the CMR/NMP, with 25 family physicians and a patient list of approximately 46,000 in 1999.¹⁶ All patients meeting World Health Organization criteria for the diagnosis of type 2 diabetes mellitus and under treatment by a family physician in 1993 and 1999 were included in the AEMS.^15,18 Patients who were treated with insulin within 1 year of diagnosis and who continued to take it were considered to have type 1 diabetes mellitus. All other patients were regarded as type 2, regardless of current treatment. For this study we included all type 2 diabetes patients under treatment by their family physician in 1993 and 1999. Patients who had died or who had moved to another area or been admitted to a residential nursing home before the end of the year were excluded, as were those who had been newly diagnosed during the year.

Audit-enhanced monitoring system

Since 1989, data have been collected on all type 2 diabetes patients at the time of diagnosis and during all regular (quarterly) diabetes-related outpatient visits. In 1992, a structured annual review, based on guidelines from the Dutch College of Family Physicians,¹⁹ was added. Starting in 1992, monitoring has consisted of the assessment of (1) compliance with 3 monthly control visits and an annual review visit; (2) glycemic control (ie, fasting blood glucose and Hb A_1c); (3) diabetes-related complications (ie, retinopathy, creatinine clearance, and foot problems); (4) cardiovascular risk factors (ie, smoking behavior, blood pressure, and lipid profile); (5) cardiovascular morbidity (ie, myocardial infarction, angina pectoris, heart failure, peripheral vascular disease, transient ischemic attack, or cerebrovascular accident). In addition, all reasons for dropping out, including cause of death, were recorded. Morbidity and causes of death were defined as in the International Classification of Health Problems in Primary Care.

To facilitate data collection, a computerized Research Registration System (RRS) was developed. The system was integrated into a standard Dutch electronic record system for family practice (Promedico, Euroned). The RRS generates templates for recording data at the quarterly or annual diabetes control visits into the patient’s electronic record. Templates guide the delivery of care and a reminder system is integrated into the RRS. Office assistants contact patients who do not come in for visits at regular intervals, both those (< 1%) who usually do not come in and those who are supposed to but fail to do so.

Family physicians sent the RRS data files to the University Department of Family Practice, where they were processed into a feedback report on process of care and outcome of care measures on 3 levels: (1) total study population; (2) practice population; (3) individual patient. Process and outcome measures were compared with external criteria based on guidelines from the Dutch College of Family Medicine and with average performance at the other practices. Feedback items were selected in consultation with the participating physicians. In this way, feedback corresponded with daily practice needs. During the project, the feedback was gradually extended from process to outcome measures. The feedback was standard to all practices.

Feedback was discussed at University Department of Family Medicine meetings, which maintained uniform registration and safeguarded the progress of the project. The feedback was also sent to every practice and participating GP. This report contained practice-level as well as physician-level data. The Figure demonstrates one way in which data are presented at the meetings and shows the percentage of patients who attended their annual diabetes control visit in the year studied.

Targets for care

Targets for care consisted of 2 elements: process and outcome measures. The key marker for process of care was compliance to the annual diabetes control visit. Key markers for desired outcome of care were (1) Hb A_1c < 8.5%,¹⁹ (2) blood pressure less than 160/90 mm Hg (revised to 150/85 mm Hg in 1999),^8,19 and (3) lipids in accordance with Dutch guidelines for general practice⁸: (a) cholesterol < 5 mmol (192 mg/dL) for patients with cardiovascular morbidity; (b) cholesterol/HDL ratio < 5.0 in smokers without cardiovascular morbidity; and (c) cholesterol/HDL < 6.0 in nonsmokers without cardiovascular morbidity. These guidelines for lipid-lowering therapy are based on sex, a life expectancy of at least 5 years, smoking status, presence of cardiovascular morbidity, total cholesterol levels, high-density lipoprotein (HDL) cholesterol levels, and triglyceride levels. If even 1 of these variables is absent, the potential value of lipid lowering cannot be determined.⁸

Analysis

Cross-sectional analysis was performed on the outcome of diabetes care in patients with type 2 diabetes who were treated by their family physicians in 1993 and 1999. The comparison was based on all patients who had been treated for the full calendar year in 1993 and 1999; therefore, it was based on a dynamic population. Process and outcome measures are compared using the chi-squared, unpaired t, or Mann–Whitney test, as appropriate. Results are expressed as means plus or minus standard deviations or as proportions. Multilevel analysis was performed to assess factors that contributed to the variance in compliance with the annual review and the desired glycemic level (Hb A_1c < 8.5%).

Results

In 1993, 540 type 2 diabetes patients (prevalence 1.3%) were included in the AEMS. Of these, 51 had been newly diagnosed (incidence 1.2/1000); 37 had been treated by a specialist (7%); and 20 did not participate (4%). Excluding the 108 patients in the latter 3 categories left a total of 432 patients for analysis. In 1999, 851 patients were included (prevalence 1.9%). Of these, 138 had been newly diagnosed (incidence 3.0/1000); 88 had been treated by a specialist (10%); and 31 did not participate (4%). Excluding the 257 patients in those 3 categories left 594 for analysis. Table 1 shows the baseline characteristics of patients included in the analysis.

Annual review was attended by 73% of patients in 1993 and 84% of patients in 1999 (Table 2). Increased compliance was achieved at all the practices, although differences between practices remained in 1999 (Figure). Univariate analysis showed that compliance with the annual review in 1999 was related to the practice (P = .001) but not to patient factors such as sex, age, duration of diabetes, therapy regimen, or cardiovascular morbidity, even after adjusting for blood glucose levels. Patients who did not attend their annual diabetes control visit had statistically significantly higher fasting blood glucose levels than patients who did comply (8.9 mmol/L [160 mg/dL] vs 8.2 mmol [147 mg/dL], P = .03). In 1993, 59% of patients had visited an ophthalmologist in the previous 2 years versus 80% in 1999.

In 1993, Hb A_1c was measured in 51% of patients with a mean of 8.2%. In 1999, compliance in measurement of Hb A_1c improved to 82%, with a mean Hb A_1c level of 7.1% (P = .0001, Table 3). The percentages of patients with an Hb A_1c level of more than 8.5% decreased from 41% to 13% (P = .001). These outcomes were associated with changes in treatment (P = .001): a decrease in patients treated with diet only (22% in 1993 vs 13% in 1999) and with oral hypoglycemic monotherapy (45% in 1993 vs 37% in 1999); an increase in patients treated with combination therapy using 2 or more oral hypoglycemic agents (22% in 1993 vs 31% in 1999); and an increase in insulin therapy (11% in 1993 vs 19% in 1999). Univariate analysis showed that poor glycemic control (Hb A_1c > 8.5%) in 1999 was related to the therapy regimen (P = .001) but not to sex, age, duration of diabetes, cardiovascular morbidity, or practice. The glycemic control in patients treated with combination therapy or insulin was poorer than in patients treated with diet only or oral hypoglycemic monotherapy, probably reflecting the fact that patients with less severe disease are managed with single agents and diet.

Compliance with measurement of blood pressure improved from 72% to 83% during the study period (Table 3). However, the percentage of patients with a systolic blood pressure below 150 mm Hg or a diastolic blood pressure below 85 mm Hg did not change between 1993 and 1999 whether patients were hypertensive or not. In hypertensive patients with type 2 diabetes, the mean diastolic blood pressure decreased from 88 mm Hg to 85 mm Hg (P = .004), but mean systolic blood pressure did not change.

The mean cholesterol level was lower in 1999 than in 1993 (6.2 vs 5.4 mmol/L; 238 mg/dL vs 207 mg/dL, P = .0001), as was the mean triglyceride level (2.54 mmol/L vs 2.07 mmol/L; 221 mg/dL vs 180 mg/dL, P = .0003). In both years, data regarding which patients could be considered for lipid-lowering therapy were available for 63% and 82%, respectively. In 1993, a far higher proportion of patients had failed to reach lipid target levels than was the case in 1999 (48% vs 28%, respectively, P = .001).

Multilevel analysis showed that paying an annual diabetes control visit (a process outcome) was related to the practice (intraclass correlation coefficient [ICC] = 0.29) but not to patient factors. Reaching the glycemic target level of Hb A_1c < 8.5%, however, was not related to practice factors (ICC = 0.003).

TABLE 1
Chacteristics of type 2 diabetes patients under family physician care in 1993 and 1999

TABLE 2
Process of care for type 2 diabetes patients under family physician care, 1993 vs 1999

TABLE 3
Outcomes of care for type 2 diabetes patients under family physician care, 1993 vs 1999

FIGURE
Percentage of patients with annual review (target = 75%) in 1999 (n=594)

Discussion

During 7 years of structured audit-enhanced monitoring of patients with type 2 diabetes in an academic family practice research network, the intermediate measures of diabetes care improved. In particular, the mean Hb A1c of 7.1% can be seen as a measure of good quality of care. The number of patients treated according to Dutch family practice guidelines (a process of care outcome) also increased.⁸

While our data were collected during normal daily care (effectiveness), the findings come close to the outcome of care under ideal trial conditions (efficacy).²¹ In the UK Prospective Diabetes Study (UKPDS), the median Hb A_1c level for all newly diagnosed patients in the group with intensive blood glucose control over 10 years reached a comparable level of 7.0%.¹ Thus, the outcome of our study approaches that achieved under trial conditions. When we analyzed patients without outcome data as poorly controlled (worst-case scenario), Hb A_1c was less than 8.5% in 28%.

The trend of improvement in glycemic control could have been a result of improved overall diabetes care in the Netherlands during the study period. Data about the outcome of diabetes care in the family medicine setting in the Netherlands during the study period are scarce and, when available, are derived from other research networks. In these networks a mean Hb A_1c of 7.0% to 7.6% was reached.²² Yet indicators from other studies suggest that our results were far better than outcome from usual care. Recently published data on such outcomes in family medicine in the Netherlands showed that Hb A_1c, blood pressure, and lipids were measured in less than 30% of patients.^23,24 Outcomes from usual care as reported in research studies appear to be strongly biased by selection and probably cannot serve as a valid reference value.

The disappointing effect on the percentage of patients who reached the target blood pressure could have resulted from evaluating the data prematurely. When the study began, the primary objective was to improve glycemic control. Shortly after the publication of the Scandinavian Simvastatin Survival Study (4S)³ and the UKPDS,¹² the guidelines of the Dutch College of Family Physicians were changed⁸ and more attention was paid to blood pressure and lipid control. This new approach was discussed with the participating family physicians. Consequently, the target for blood pressure was revised from 160/90 mm Hg to 150/85 mm Hg and lipid-lowering therapy was tailored to each patient’s cardiovascular risk profile. The 1999 outcome with respect to blood pressure and lipid control was measured only 1 year after these changes had been announced. Nevertheless, mean diastolic blood pressure in hypertensive patients and total cholesterol and triglyceride levels decreased significantly, and more patients reached target levels for lipids in 1999 than in 1993.

Our outcome was reached through enhanced compliance to guidelines. Therefore, the outcome in 1999 was based on a larger percentage of available patients. Because the AEMS studied a dynamic group of patients, the study groups in 1993 and 1999 were not identical. Theoretically, improvement in outcome could have been reached by including more easily manageable patients. However, no patient factors such as sex, age, duration of diabetes, treatment modality, or cardiovascular morbidity were related to compliance with annual review. The higher fasting blood glucose levels in patients who were noncompliant with annual review probably reflected under-treatment rather than more severe illness status. Therefore, we are confident that the findings reflect improved overall diabetes care.

The data on process measures in this study compare favorably with those of multipractice audits of diabetes care in the United Kingdom.^25-29 The high prevalence rate of 2.0% (exclusively patients with type 2 diabetes) supports the validity of our data.^25-27 Among our patients, 96% had been seen at least once during the previous year. In the large studies by Khunti and Bennett, only 85% had been seen during the previous year.^25,26 The mean annual compliance rate of nearly 85% with Hb A_1c and blood pressure measurements in our study was high. In particular, compliance with lipid control^25,26,28 and funduscopy^26,28 was better in our study. In 2 longitudinal studies that used an organized care system in which feedback was provided to the participating family physicians, compliance rates in process measures of up to 75% were reported.^30,31

The outcomes of this study were achieved in an academic family practice research network, with specific facilities for the proactive supervision of patients with chronic diseases. These results cannot and should not be generalized to “routine” family practice. Monitoring and feedback in routine family practice are in themselves insufficient to improve the quality of care.¹⁰ Care assessment should preferably take a more comprehensive approach in which evidence-based goals for care are formulated, care is improved to reach those goals, and care is measured to see whether those goals have been achieved.¹¹ Our academic network provides this comprehensive approach.

The electronic Research Registration System played an important role in the audit-enhanced monitoring. In the pilot phase of the project, paper records were used. Although using paper records had clear disadvantages, one could expect to achieve similar results using such records in combination with a central electronic data bank.

Conclusions

Outcomes of diabetes care in our family research setting were comparable with those reported in randomized controlled trials. Therefore, it is possible for the management of diabetes in family practice to be efficacious. This finding should encourage more efforts by physicians in family practice to bridge the gap between efficacy and effectiveness.

Important differences remained in achieved process measures between the academic family practices. While the outcome of diabetes care in the network was favorable, the outcome of treatment was unsatisfactory in a substantial number of patients. Further implementation strategies must be developed. The differences in achieved process measures were probably unrelated to socioeconomic differences between the practice populations, since the practice pairs 1/3 and 7/10 served comparable communities and had different levels of compliance.

Our study demonstrated that a high quality of diabetes care in family practice can be achieved. Audit-enhanced monitoring, which will provide the greatest benefit to the most patients with type 2 diabetes mellitus, should be implemented as part of a quality improvement system.

ACKNOWLEDGMENTS

The authors wish to thank the family physicians and practice nurses for their continuing support and data collection.

Unintended pregnancy is a significant public health issue. More than half of all pregnancies are unintended at the time of conception; approximately half of those end as births and half as induced abortions.¹ Forty-eight percent of women have at least one unplanned pregnancy, and 28% of women have at least one unplanned birth during their reproductive lifetime.² Unintended pregnancies and births are associated with numerous harmful behaviors and adverse outcomes.^3,4

Breastfeeding is currently promoted as the preferred method of feeding for infants for at least 1 year because of its multiple immediate and long-term benefits for both mother and child.^5,6 Yet in 1998 only 64% of US mothers were breastfeeding at the time of hospital discharge and 29% at 6 months postpartum, which is well below the Healthy People 2010 goals of 75% and 50%, respectively, for those intervals.⁷

We hypothesized that women with unintended pregnancies are less likely to breastfeed their infants than those with intended ones. We quantified the association between the intendedness of pregnancy at the time of conception and breastfeeding behavior, both the initiation of any breastfeeding and the continuation of nonexclusive breastfeeding for at least 16 weeks, for first singleton births to US mothers. We then explored other factors which might affect breastfeeding practices.

Methods

Study design

This study is a secondary data analysis of the 1995 Cycle 5 of the National Survey of Family Growth (NSFG), a periodic population-based survey conducted by the National Center for Health Statistics and the Centers for Disease Control which focuses on women’s health and pregnancy. A national probability sample of 14,000 civilian noninstitutionalized women aged 15 years to 44 years was selected from among households that responded to the 1993 National Health Interview Survey, with an oversampling of minority women. Personal interviews were conducted between January and October of 1995 with 10,847 of these women. The data were then adjusted for a response rate of 79% and weighted so that findings would reflect the US population as a whole. Full details of the NSFG survey methods are described elsewhere.⁸

The data set contains information on 21,332 pregnancies and 14,958 live births (Figure). After excluding multiple gestations (n=154), subsequent births to the same mother (n=7930), and neonatal adoptions or deaths (n=141), the final sample contained 6733 first singleton live births. To study the initiation of breastfeeding, women who breastfed at all were compared with those who did not. To study duration of breastfeeding, women who breastfed for 16 or more weeks were compared with those who did not. In this second set of analyses, the 1459 women who breastfed for between 0 and 16 weeks and the 33 women who were breastfeeding at the time of the interview and whose children had been born within 16 weeks of that date were excluded.

FIGURE
Study sample

Variable definitions

Pregnancies were categorized as either intended or unintended at conception, using new definitions established by the Institute of Medicine in 1995.³ Pregnancies were considered intended if a woman had stopped using birth control because she wanted to become pregnant. Unintended pregnancies were classified into 1 of 2 categories: (1) mis-timed: wanted pregnancies that occurred sooner than desired, or (2) unwanted: pregnancies that occurred while a woman was using contraception and had not ever wanted to have a(nother) baby. The 170 women who described their pregnancy intention status as “didn’t know” or “didn’t care” were excluded from analyses involving intention status as defined above. There were 2 breastfeeding outcomes in this study: (1) initiation of breastfeeding, including women who reported any breastfeeding at all, and (2) duration of non-exclusive breastfeeding for at least 16 weeks.

Maternal demographics, intrapartum and postpartum behaviors, and birth outcomes were considered as potential confounders. As missing data were imputed in the public use data file, information on each variable was complete, except where noted below. Maternal age was determined at the time of conception. Education was defined as completed years of schooling at the time of the interview. Race was categorized as white, black, Hispanic, or other. Marital status was defined as either married or not married. Socioeconomic status was measured continuously as a percentage of the 1995 poverty level. Information on prenatal care was only available for births in the last 5 years (n=1266). For the 1241 women who reported that they had received prenatal care, the mean weeks of gestation at the time of the first prenatal visit were calculated. Maternity leave was defined as the use of maternity leave, paid or unpaid, for women who were working during the pregnancy (n=3662). For those who took any leave, the mean length of that leave was calculated in weeks. The 33 women for whom data were not available and the 5 women who answered that they took 0 weeks of maternity leave were not included in this calculation. Infant variables were considered categorically: mode of delivery, either vaginal or cesarean; prematurity, birth at less than or equal to 36 weeks; and low birth weight, less than or equal to 5.5 pounds.

Statistical analysis

The statistical significance of descriptive variables was determined using 2-sample t tests and chi-square tests, with women who did not breastfeed at all as the comparison group.

To compare our results with existing literature, we calculated crude odds ratios of not breastfeeding and used chi-square tests to assess the statistical significance of these associations. The reference group was always women whose pregnancies were intended. Variables identified as potential confounders were age, race, marital status, poverty level, education, maternity leave, mode of delivery, prematurity, and low birth weight. Final logistic regression models were adjusted for those variables that changed crude odds ratios by 10% or more: age, race, marital status, poverty level, and education. Adjusted odds ratios (ORs) of not breastfeeding and 95% confidence intervals (CIs) are reported.

Effect modification was assessed by dichotomizing each of the 5 confounding variables in the following manner: teen versus 20 years or older; white versus black versus Hispanic (here, the 193 women in the sample who defined their race as other were not included); married versus unmarried; high school education or higher versus less than a high school education; and below the poverty level versus at or above the poverty level. Each stratified analysis was adjusted for the other 4 factors. Interaction terms were created and P values for heterogeneity were calculated for all logistic models. For race, an additional interaction term was created to compare white with non-white (black and Hispanic combined) women.

These data are contained on the National Survey of Family Growth Cycle 5 1995 CD-ROM, Series 23, No. 3 and were exported using SETS version 1.22a (National Center for Health Statistics, Hyattsville, MD). All analyses were performed with SAS version 6.12 (SAS Institute, Cary, NC). Odds ratios are weighted using sampling weights provided in the data set. SUDAAN version 7.5.3 was used to obtain standard errors (Research Triangle Institute, Research Triangle Park, NC).

Results

In the total sample of 6733 first singleton live births to US mothers, 3267 (48.5%) of women initiated breastfeeding, compared with 3466 (51.5%) who did not. In the entire sample, 1775 (26.4%) continued to breastfeed nonexclusively for at least 16 weeks.

The breastfeeding initiation rate was 55.9%, 37.4%, and 28.0% for women with intended, mis-timed, and unwanted pregnancies, respectively. By 16 weeks, 32.6%, 17.0%, and 15.5% of women, respectively, were still breastfeeding. For all women who breastfed, the mean number of weeks of breastfeeding was 24.4 (standard deviation = 24.9; range = < 1 week to 4.0 years). Only 3.9% of women who breastfed did so for more than 2 years.

Table 1 shows characteristics of mothers and infants by breastfeeding behavior. Women who breastfed, both initially and for at least 16 weeks, were older at conception and had had more years of education than women who did not breastfeed at all. They were more likely to be white and less likely to be black. A similar percentage of women in each group were Hispanic. Percentage of the poverty level, a proxy for socioeconomic status, was higher for those who breastfed at all, but similar for those who continued breastfeeding and those who did not breastfeed. Rates of prenatal care and mean weeks at first prenatal visit were similar in all groups. Among women who were employed during their pregnancies, almost two thirds took maternity leave, regardless of breastfeeding behavior. Mean length of maternity leave was 3.4 weeks longer among women who continued to breastfeed than among women who did not breastfeed at all. The percentage of vaginal deliveries was similar among groups. Both premature and low birth weight infants were more common among women who did not breastfeed.

The association between the intendedness of pregnancy and breastfeeding behavior is reported in Table 2. Crude odds ratios show that women with any type of unintended pregnancy were more likely not to initiate breastfeeding than women whose pregnancies were intended. Some, but not all, of this association can be attributed to confounding by demographic factors. Having an unintended pregnancy was not associated with any significant difference in the initiation of breastfeeding, after adjusting for age, race, marital status, poverty level, and education. While women with mis-timed pregnancies were as likely to initiate breastfeeding as those whose pregnancies were intended (OR = 1.03; 95% CI, 0.88-1.21), women with unwanted pregnancies were more likely not to start breastfeeding (OR = 1.76; 95% CI, 1.26-2.44).

Table 2 also describes the association between pregnancy intention status and the continuation of nonexclusive breastfeeding for at least 16 weeks. In contrast to the initiation of breastfeeding, duration of breastfeeding was affected by the intendedness of pregnancy in every comparison. Adjusted odds ratios show that women with either type of unintended pregnancy were more likely not to continue breastfeeding than those with intended ones (OR = 1.28; 95% CI, 1.06-1.54). As with breastfeeding initiation, this association is being driven by the unwanted pregnancies. Women with unwanted pregnancies were more likely not to continue breastfeeding (OR = 1.69; 95% CI, 1.12-2.55).

Each of these associations was then evaluated for effect modification. As seen in Table 3, only race was an important factor. In the total sample, 56.3% of white women, 55.4% of Hispanic women, and 24.7% of black women breastfed at all; and 41.6%, 41.2%, and 12.6% of white, Hispanic, and black women breastfed for at least 16 weeks. White women with unwanted pregnancies were more likely not to initiate breastfeeding (OR = 2.50; 95% CI, 1.54-4.05) and more likely not to continue breastfeeding (OR = 2.56; 95% CI, 1.34-4.87) than white women with intended pregnancies. These differences in breastfeeding behaviors for unwanted pregnancies were not seen for either Hispanic or black women. For each stratified analysis, a single P value for heterogeneity was calculated to compare white women with non-white women (Hispanic and black women combined). The only significant difference by race was for unwanted pregnancies. The P value for heterogeneity was 0.01 for both initiation and continuation of breastfeeding. Stratified analyses for age, marital status, education, poverty level, and year of birth showed similar odds ratios across strata and nonsignificant P values for heterogeneity in every case (analyses not shown).

TABLE 1
Characteristics of mothers and infants by breastfeeding status (n=6733)

TABLE 2
Unintended pregnancy and breastfeeding behavior in the United States

TABLE 3
Effect of race on unintended pregnancy and breastfeeding behavior

Discussion

In this study of first-time US mothers, women who breastfed were demographically different from those who did not, but had relatively similar maternal behaviors and infant characteristics. After controlling for these demographic differences, having an unwanted pregnancy was associated with a lower likelihood of both initiating breastfeeding and continuing to breastfeed. In addition, race was an important effect modifier for unwanted pregnancies.

The demographic findings of this study are consistent with the current breastfeeding literature: US women who breastfeed tend to be older, white, married, well-educated, and of a higher socioeconomic status than those who do not.⁹ The main findings of this study are also consistent with the only other study that has examined the relationship between unintended pregnancy and breastfeeding behavior.¹⁰ A cross-sectional sample of 27,700 women who gave birth to a live baby were asked prior to postpartum discharge whether they had intended to become pregnant and their plans for breastfeeding. After controlling for education, race, Medicaid status, maternal age younger than 20, and any tobacco use during pregnancy, the authors found that women whose pregnancies were unintended were more likely not to initiate breastfeeding or to breastfeed exclusively. Adjusted odds ratios of not breastfeeding ranged from 1.10 to 1.41, depending on intention status, and all were statistically significant. In contrast to our study, a major limitation of that study was that the measured outcome was intent to breastfeed at hospital discharge, which may have differed greatly from actual breastfeeding behavior.

The interaction seen in our analysis between intention status and race is initially surprising because, in general, white and Hispanic women breastfeed at much higher rates than black women. But if a pregnancy was unwanted, white women were much less likely to breastfeed than either black or Hispanic women. Neither socioeconomic status nor educational level is the explanation, as both of these factors were controlled for in stratified analyses. Perhaps Hispanic and black women are more accepting of unintended pregnancy than white women and these results reflect cultural differences. Further studies which examine other aspects of unintended pregnancy with respect to race will help to further clarify the reasons for this finding.

Strengths and limitations

Our population-based study has several strengths. The data set provides a large national sample with excellent representation of minority women; statistical oversampling and weighting allow these data to reflect the entire national population. Adjustment of data for nonresponse lessens the risk of selection bias. Furthermore, the study sample was restricted to first births to limit the effect of previous birth experiences on postpartum behaviors. Therefore our results are generalizable to all first-time mothers in the United States.

A major limitation of our study is that information was not collected on several factors, such as substance use both during pregnancy and after birth, that might influence the relationship between pregnancy intention status and breastfeeding behavior. The work of Dye and colleagues,¹⁰ discussed above, found that prenatal tobacco, but not alcohol or drug use, was a significant confounder. Information was also not available on health service–related factors that may contribute to breastfeeding success, such as breastfeeding in the delivery room, length of hospital stay, and participation in educational programs.¹¹

Given that data were collected for the NSFG during personal interviews at differing lengths of time after a pregnancy, inaccuracy is possible. Although the survey does not include corroboration from other sources, such as medical records or birth certificates, it is reassuring that, as an example, rates of prenatal care in our study are similar to those of other nationally reported rates for 1995 (98.1% in our study and 98.8% in National Vital Statistics Reports).¹² Potential misclassification with respect to such medical outcomes as prematurity would be nondifferential and only bias odds ratios toward the null. The extended time between conception and measurement of maternal attitudes increases the uncertainty that a mother will accurately recall both her pregnancy intentions at conception and her breastfeeding practices. Women are more likely to recall a pregnancy carried to birth as intended, but this phenomenon would only bias the results if it also applied to breastfeeding practices, which is unlikely.¹³ While breastfeeding practices may not be exactly recalled, there is no obvious reason for differential reporting.

Conclusions

Our study has clinical implications for first-time US mothers. A recent national goal of the Institute of Medicine is that all pregnancies be planned.³ One of the many benefits of decreasing unintended pregnancy may be to increase breastfeeding rates closer to the Healthy People 2010 goals. In addition, a new hypothesis is suggested by the results of this study: Clinicians should promote breastfeeding differently for women with intended and unintended pregnancies. Future research will evaluate the importance of incorporating pregnancy intention status into patient-centered counseling. In the interim, women with unwanted pregnancies, especially white women, should be targeted for counseling, as they could benefit from breastfeeding, not just for medical reasons but for psychological and economic ones as well.

Acknowledgments

We would like to thank Larry Culpepper, MD, MPH, for his guidance.

Unintended pregnancy is a significant public health issue. More than half of all pregnancies are unintended at the time of conception; approximately half of those end as births and half as induced abortions.¹ Forty-eight percent of women have at least one unplanned pregnancy, and 28% of women have at least one unplanned birth during their reproductive lifetime.² Unintended pregnancies and births are associated with numerous harmful behaviors and adverse outcomes.^3,4

Breastfeeding is currently promoted as the preferred method of feeding for infants for at least 1 year because of its multiple immediate and long-term benefits for both mother and child.^5,6 Yet in 1998 only 64% of US mothers were breastfeeding at the time of hospital discharge and 29% at 6 months postpartum, which is well below the Healthy People 2010 goals of 75% and 50%, respectively, for those intervals.⁷

We hypothesized that women with unintended pregnancies are less likely to breastfeed their infants than those with intended ones. We quantified the association between the intendedness of pregnancy at the time of conception and breastfeeding behavior, both the initiation of any breastfeeding and the continuation of nonexclusive breastfeeding for at least 16 weeks, for first singleton births to US mothers. We then explored other factors which might affect breastfeeding practices.

Methods

Study design

This study is a secondary data analysis of the 1995 Cycle 5 of the National Survey of Family Growth (NSFG), a periodic population-based survey conducted by the National Center for Health Statistics and the Centers for Disease Control which focuses on women’s health and pregnancy. A national probability sample of 14,000 civilian noninstitutionalized women aged 15 years to 44 years was selected from among households that responded to the 1993 National Health Interview Survey, with an oversampling of minority women. Personal interviews were conducted between January and October of 1995 with 10,847 of these women. The data were then adjusted for a response rate of 79% and weighted so that findings would reflect the US population as a whole. Full details of the NSFG survey methods are described elsewhere.⁸

The data set contains information on 21,332 pregnancies and 14,958 live births (Figure). After excluding multiple gestations (n=154), subsequent births to the same mother (n=7930), and neonatal adoptions or deaths (n=141), the final sample contained 6733 first singleton live births. To study the initiation of breastfeeding, women who breastfed at all were compared with those who did not. To study duration of breastfeeding, women who breastfed for 16 or more weeks were compared with those who did not. In this second set of analyses, the 1459 women who breastfed for between 0 and 16 weeks and the 33 women who were breastfeeding at the time of the interview and whose children had been born within 16 weeks of that date were excluded.

FIGURE
Study sample

Variable definitions

Pregnancies were categorized as either intended or unintended at conception, using new definitions established by the Institute of Medicine in 1995.³ Pregnancies were considered intended if a woman had stopped using birth control because she wanted to become pregnant. Unintended pregnancies were classified into 1 of 2 categories: (1) mis-timed: wanted pregnancies that occurred sooner than desired, or (2) unwanted: pregnancies that occurred while a woman was using contraception and had not ever wanted to have a(nother) baby. The 170 women who described their pregnancy intention status as “didn’t know” or “didn’t care” were excluded from analyses involving intention status as defined above. There were 2 breastfeeding outcomes in this study: (1) initiation of breastfeeding, including women who reported any breastfeeding at all, and (2) duration of non-exclusive breastfeeding for at least 16 weeks.

Maternal demographics, intrapartum and postpartum behaviors, and birth outcomes were considered as potential confounders. As missing data were imputed in the public use data file, information on each variable was complete, except where noted below. Maternal age was determined at the time of conception. Education was defined as completed years of schooling at the time of the interview. Race was categorized as white, black, Hispanic, or other. Marital status was defined as either married or not married. Socioeconomic status was measured continuously as a percentage of the 1995 poverty level. Information on prenatal care was only available for births in the last 5 years (n=1266). For the 1241 women who reported that they had received prenatal care, the mean weeks of gestation at the time of the first prenatal visit were calculated. Maternity leave was defined as the use of maternity leave, paid or unpaid, for women who were working during the pregnancy (n=3662). For those who took any leave, the mean length of that leave was calculated in weeks. The 33 women for whom data were not available and the 5 women who answered that they took 0 weeks of maternity leave were not included in this calculation. Infant variables were considered categorically: mode of delivery, either vaginal or cesarean; prematurity, birth at less than or equal to 36 weeks; and low birth weight, less than or equal to 5.5 pounds.

Statistical analysis

The statistical significance of descriptive variables was determined using 2-sample t tests and chi-square tests, with women who did not breastfeed at all as the comparison group.

To compare our results with existing literature, we calculated crude odds ratios of not breastfeeding and used chi-square tests to assess the statistical significance of these associations. The reference group was always women whose pregnancies were intended. Variables identified as potential confounders were age, race, marital status, poverty level, education, maternity leave, mode of delivery, prematurity, and low birth weight. Final logistic regression models were adjusted for those variables that changed crude odds ratios by 10% or more: age, race, marital status, poverty level, and education. Adjusted odds ratios (ORs) of not breastfeeding and 95% confidence intervals (CIs) are reported.

Effect modification was assessed by dichotomizing each of the 5 confounding variables in the following manner: teen versus 20 years or older; white versus black versus Hispanic (here, the 193 women in the sample who defined their race as other were not included); married versus unmarried; high school education or higher versus less than a high school education; and below the poverty level versus at or above the poverty level. Each stratified analysis was adjusted for the other 4 factors. Interaction terms were created and P values for heterogeneity were calculated for all logistic models. For race, an additional interaction term was created to compare white with non-white (black and Hispanic combined) women.

These data are contained on the National Survey of Family Growth Cycle 5 1995 CD-ROM, Series 23, No. 3 and were exported using SETS version 1.22a (National Center for Health Statistics, Hyattsville, MD). All analyses were performed with SAS version 6.12 (SAS Institute, Cary, NC). Odds ratios are weighted using sampling weights provided in the data set. SUDAAN version 7.5.3 was used to obtain standard errors (Research Triangle Institute, Research Triangle Park, NC).

Results

In the total sample of 6733 first singleton live births to US mothers, 3267 (48.5%) of women initiated breastfeeding, compared with 3466 (51.5%) who did not. In the entire sample, 1775 (26.4%) continued to breastfeed nonexclusively for at least 16 weeks.

The breastfeeding initiation rate was 55.9%, 37.4%, and 28.0% for women with intended, mis-timed, and unwanted pregnancies, respectively. By 16 weeks, 32.6%, 17.0%, and 15.5% of women, respectively, were still breastfeeding. For all women who breastfed, the mean number of weeks of breastfeeding was 24.4 (standard deviation = 24.9; range = < 1 week to 4.0 years). Only 3.9% of women who breastfed did so for more than 2 years.

Table 1 shows characteristics of mothers and infants by breastfeeding behavior. Women who breastfed, both initially and for at least 16 weeks, were older at conception and had had more years of education than women who did not breastfeed at all. They were more likely to be white and less likely to be black. A similar percentage of women in each group were Hispanic. Percentage of the poverty level, a proxy for socioeconomic status, was higher for those who breastfed at all, but similar for those who continued breastfeeding and those who did not breastfeed. Rates of prenatal care and mean weeks at first prenatal visit were similar in all groups. Among women who were employed during their pregnancies, almost two thirds took maternity leave, regardless of breastfeeding behavior. Mean length of maternity leave was 3.4 weeks longer among women who continued to breastfeed than among women who did not breastfeed at all. The percentage of vaginal deliveries was similar among groups. Both premature and low birth weight infants were more common among women who did not breastfeed.

The association between the intendedness of pregnancy and breastfeeding behavior is reported in Table 2. Crude odds ratios show that women with any type of unintended pregnancy were more likely not to initiate breastfeeding than women whose pregnancies were intended. Some, but not all, of this association can be attributed to confounding by demographic factors. Having an unintended pregnancy was not associated with any significant difference in the initiation of breastfeeding, after adjusting for age, race, marital status, poverty level, and education. While women with mis-timed pregnancies were as likely to initiate breastfeeding as those whose pregnancies were intended (OR = 1.03; 95% CI, 0.88-1.21), women with unwanted pregnancies were more likely not to start breastfeeding (OR = 1.76; 95% CI, 1.26-2.44).

Table 2 also describes the association between pregnancy intention status and the continuation of nonexclusive breastfeeding for at least 16 weeks. In contrast to the initiation of breastfeeding, duration of breastfeeding was affected by the intendedness of pregnancy in every comparison. Adjusted odds ratios show that women with either type of unintended pregnancy were more likely not to continue breastfeeding than those with intended ones (OR = 1.28; 95% CI, 1.06-1.54). As with breastfeeding initiation, this association is being driven by the unwanted pregnancies. Women with unwanted pregnancies were more likely not to continue breastfeeding (OR = 1.69; 95% CI, 1.12-2.55).

Each of these associations was then evaluated for effect modification. As seen in Table 3, only race was an important factor. In the total sample, 56.3% of white women, 55.4% of Hispanic women, and 24.7% of black women breastfed at all; and 41.6%, 41.2%, and 12.6% of white, Hispanic, and black women breastfed for at least 16 weeks. White women with unwanted pregnancies were more likely not to initiate breastfeeding (OR = 2.50; 95% CI, 1.54-4.05) and more likely not to continue breastfeeding (OR = 2.56; 95% CI, 1.34-4.87) than white women with intended pregnancies. These differences in breastfeeding behaviors for unwanted pregnancies were not seen for either Hispanic or black women. For each stratified analysis, a single P value for heterogeneity was calculated to compare white women with non-white women (Hispanic and black women combined). The only significant difference by race was for unwanted pregnancies. The P value for heterogeneity was 0.01 for both initiation and continuation of breastfeeding. Stratified analyses for age, marital status, education, poverty level, and year of birth showed similar odds ratios across strata and nonsignificant P values for heterogeneity in every case (analyses not shown).

TABLE 1
Characteristics of mothers and infants by breastfeeding status (n=6733)

TABLE 2
Unintended pregnancy and breastfeeding behavior in the United States

TABLE 3
Effect of race on unintended pregnancy and breastfeeding behavior

Discussion

In this study of first-time US mothers, women who breastfed were demographically different from those who did not, but had relatively similar maternal behaviors and infant characteristics. After controlling for these demographic differences, having an unwanted pregnancy was associated with a lower likelihood of both initiating breastfeeding and continuing to breastfeed. In addition, race was an important effect modifier for unwanted pregnancies.

The demographic findings of this study are consistent with the current breastfeeding literature: US women who breastfeed tend to be older, white, married, well-educated, and of a higher socioeconomic status than those who do not.⁹ The main findings of this study are also consistent with the only other study that has examined the relationship between unintended pregnancy and breastfeeding behavior.¹⁰ A cross-sectional sample of 27,700 women who gave birth to a live baby were asked prior to postpartum discharge whether they had intended to become pregnant and their plans for breastfeeding. After controlling for education, race, Medicaid status, maternal age younger than 20, and any tobacco use during pregnancy, the authors found that women whose pregnancies were unintended were more likely not to initiate breastfeeding or to breastfeed exclusively. Adjusted odds ratios of not breastfeeding ranged from 1.10 to 1.41, depending on intention status, and all were statistically significant. In contrast to our study, a major limitation of that study was that the measured outcome was intent to breastfeed at hospital discharge, which may have differed greatly from actual breastfeeding behavior.

The interaction seen in our analysis between intention status and race is initially surprising because, in general, white and Hispanic women breastfeed at much higher rates than black women. But if a pregnancy was unwanted, white women were much less likely to breastfeed than either black or Hispanic women. Neither socioeconomic status nor educational level is the explanation, as both of these factors were controlled for in stratified analyses. Perhaps Hispanic and black women are more accepting of unintended pregnancy than white women and these results reflect cultural differences. Further studies which examine other aspects of unintended pregnancy with respect to race will help to further clarify the reasons for this finding.

Strengths and limitations

Our population-based study has several strengths. The data set provides a large national sample with excellent representation of minority women; statistical oversampling and weighting allow these data to reflect the entire national population. Adjustment of data for nonresponse lessens the risk of selection bias. Furthermore, the study sample was restricted to first births to limit the effect of previous birth experiences on postpartum behaviors. Therefore our results are generalizable to all first-time mothers in the United States.

A major limitation of our study is that information was not collected on several factors, such as substance use both during pregnancy and after birth, that might influence the relationship between pregnancy intention status and breastfeeding behavior. The work of Dye and colleagues,¹⁰ discussed above, found that prenatal tobacco, but not alcohol or drug use, was a significant confounder. Information was also not available on health service–related factors that may contribute to breastfeeding success, such as breastfeeding in the delivery room, length of hospital stay, and participation in educational programs.¹¹

Given that data were collected for the NSFG during personal interviews at differing lengths of time after a pregnancy, inaccuracy is possible. Although the survey does not include corroboration from other sources, such as medical records or birth certificates, it is reassuring that, as an example, rates of prenatal care in our study are similar to those of other nationally reported rates for 1995 (98.1% in our study and 98.8% in National Vital Statistics Reports).¹² Potential misclassification with respect to such medical outcomes as prematurity would be nondifferential and only bias odds ratios toward the null. The extended time between conception and measurement of maternal attitudes increases the uncertainty that a mother will accurately recall both her pregnancy intentions at conception and her breastfeeding practices. Women are more likely to recall a pregnancy carried to birth as intended, but this phenomenon would only bias the results if it also applied to breastfeeding practices, which is unlikely.¹³ While breastfeeding practices may not be exactly recalled, there is no obvious reason for differential reporting.

Conclusions

Our study has clinical implications for first-time US mothers. A recent national goal of the Institute of Medicine is that all pregnancies be planned.³ One of the many benefits of decreasing unintended pregnancy may be to increase breastfeeding rates closer to the Healthy People 2010 goals. In addition, a new hypothesis is suggested by the results of this study: Clinicians should promote breastfeeding differently for women with intended and unintended pregnancies. Future research will evaluate the importance of incorporating pregnancy intention status into patient-centered counseling. In the interim, women with unwanted pregnancies, especially white women, should be targeted for counseling, as they could benefit from breastfeeding, not just for medical reasons but for psychological and economic ones as well.

Acknowledgments

We would like to thank Larry Culpepper, MD, MPH, for his guidance.

Unintended pregnancy is a significant public health issue. More than half of all pregnancies are unintended at the time of conception; approximately half of those end as births and half as induced abortions.¹ Forty-eight percent of women have at least one unplanned pregnancy, and 28% of women have at least one unplanned birth during their reproductive lifetime.² Unintended pregnancies and births are associated with numerous harmful behaviors and adverse outcomes.^3,4

Breastfeeding is currently promoted as the preferred method of feeding for infants for at least 1 year because of its multiple immediate and long-term benefits for both mother and child.^5,6 Yet in 1998 only 64% of US mothers were breastfeeding at the time of hospital discharge and 29% at 6 months postpartum, which is well below the Healthy People 2010 goals of 75% and 50%, respectively, for those intervals.⁷

We hypothesized that women with unintended pregnancies are less likely to breastfeed their infants than those with intended ones. We quantified the association between the intendedness of pregnancy at the time of conception and breastfeeding behavior, both the initiation of any breastfeeding and the continuation of nonexclusive breastfeeding for at least 16 weeks, for first singleton births to US mothers. We then explored other factors which might affect breastfeeding practices.

Methods

Study design

This study is a secondary data analysis of the 1995 Cycle 5 of the National Survey of Family Growth (NSFG), a periodic population-based survey conducted by the National Center for Health Statistics and the Centers for Disease Control which focuses on women’s health and pregnancy. A national probability sample of 14,000 civilian noninstitutionalized women aged 15 years to 44 years was selected from among households that responded to the 1993 National Health Interview Survey, with an oversampling of minority women. Personal interviews were conducted between January and October of 1995 with 10,847 of these women. The data were then adjusted for a response rate of 79% and weighted so that findings would reflect the US population as a whole. Full details of the NSFG survey methods are described elsewhere.⁸

The data set contains information on 21,332 pregnancies and 14,958 live births (Figure). After excluding multiple gestations (n=154), subsequent births to the same mother (n=7930), and neonatal adoptions or deaths (n=141), the final sample contained 6733 first singleton live births. To study the initiation of breastfeeding, women who breastfed at all were compared with those who did not. To study duration of breastfeeding, women who breastfed for 16 or more weeks were compared with those who did not. In this second set of analyses, the 1459 women who breastfed for between 0 and 16 weeks and the 33 women who were breastfeeding at the time of the interview and whose children had been born within 16 weeks of that date were excluded.

FIGURE
Study sample

Variable definitions

Pregnancies were categorized as either intended or unintended at conception, using new definitions established by the Institute of Medicine in 1995.³ Pregnancies were considered intended if a woman had stopped using birth control because she wanted to become pregnant. Unintended pregnancies were classified into 1 of 2 categories: (1) mis-timed: wanted pregnancies that occurred sooner than desired, or (2) unwanted: pregnancies that occurred while a woman was using contraception and had not ever wanted to have a(nother) baby. The 170 women who described their pregnancy intention status as “didn’t know” or “didn’t care” were excluded from analyses involving intention status as defined above. There were 2 breastfeeding outcomes in this study: (1) initiation of breastfeeding, including women who reported any breastfeeding at all, and (2) duration of non-exclusive breastfeeding for at least 16 weeks.

Maternal demographics, intrapartum and postpartum behaviors, and birth outcomes were considered as potential confounders. As missing data were imputed in the public use data file, information on each variable was complete, except where noted below. Maternal age was determined at the time of conception. Education was defined as completed years of schooling at the time of the interview. Race was categorized as white, black, Hispanic, or other. Marital status was defined as either married or not married. Socioeconomic status was measured continuously as a percentage of the 1995 poverty level. Information on prenatal care was only available for births in the last 5 years (n=1266). For the 1241 women who reported that they had received prenatal care, the mean weeks of gestation at the time of the first prenatal visit were calculated. Maternity leave was defined as the use of maternity leave, paid or unpaid, for women who were working during the pregnancy (n=3662). For those who took any leave, the mean length of that leave was calculated in weeks. The 33 women for whom data were not available and the 5 women who answered that they took 0 weeks of maternity leave were not included in this calculation. Infant variables were considered categorically: mode of delivery, either vaginal or cesarean; prematurity, birth at less than or equal to 36 weeks; and low birth weight, less than or equal to 5.5 pounds.

Statistical analysis

The statistical significance of descriptive variables was determined using 2-sample t tests and chi-square tests, with women who did not breastfeed at all as the comparison group.

To compare our results with existing literature, we calculated crude odds ratios of not breastfeeding and used chi-square tests to assess the statistical significance of these associations. The reference group was always women whose pregnancies were intended. Variables identified as potential confounders were age, race, marital status, poverty level, education, maternity leave, mode of delivery, prematurity, and low birth weight. Final logistic regression models were adjusted for those variables that changed crude odds ratios by 10% or more: age, race, marital status, poverty level, and education. Adjusted odds ratios (ORs) of not breastfeeding and 95% confidence intervals (CIs) are reported.

Effect modification was assessed by dichotomizing each of the 5 confounding variables in the following manner: teen versus 20 years or older; white versus black versus Hispanic (here, the 193 women in the sample who defined their race as other were not included); married versus unmarried; high school education or higher versus less than a high school education; and below the poverty level versus at or above the poverty level. Each stratified analysis was adjusted for the other 4 factors. Interaction terms were created and P values for heterogeneity were calculated for all logistic models. For race, an additional interaction term was created to compare white with non-white (black and Hispanic combined) women.

These data are contained on the National Survey of Family Growth Cycle 5 1995 CD-ROM, Series 23, No. 3 and were exported using SETS version 1.22a (National Center for Health Statistics, Hyattsville, MD). All analyses were performed with SAS version 6.12 (SAS Institute, Cary, NC). Odds ratios are weighted using sampling weights provided in the data set. SUDAAN version 7.5.3 was used to obtain standard errors (Research Triangle Institute, Research Triangle Park, NC).

Results

In the total sample of 6733 first singleton live births to US mothers, 3267 (48.5%) of women initiated breastfeeding, compared with 3466 (51.5%) who did not. In the entire sample, 1775 (26.4%) continued to breastfeed nonexclusively for at least 16 weeks.

The breastfeeding initiation rate was 55.9%, 37.4%, and 28.0% for women with intended, mis-timed, and unwanted pregnancies, respectively. By 16 weeks, 32.6%, 17.0%, and 15.5% of women, respectively, were still breastfeeding. For all women who breastfed, the mean number of weeks of breastfeeding was 24.4 (standard deviation = 24.9; range = < 1 week to 4.0 years). Only 3.9% of women who breastfed did so for more than 2 years.

Table 1 shows characteristics of mothers and infants by breastfeeding behavior. Women who breastfed, both initially and for at least 16 weeks, were older at conception and had had more years of education than women who did not breastfeed at all. They were more likely to be white and less likely to be black. A similar percentage of women in each group were Hispanic. Percentage of the poverty level, a proxy for socioeconomic status, was higher for those who breastfed at all, but similar for those who continued breastfeeding and those who did not breastfeed. Rates of prenatal care and mean weeks at first prenatal visit were similar in all groups. Among women who were employed during their pregnancies, almost two thirds took maternity leave, regardless of breastfeeding behavior. Mean length of maternity leave was 3.4 weeks longer among women who continued to breastfeed than among women who did not breastfeed at all. The percentage of vaginal deliveries was similar among groups. Both premature and low birth weight infants were more common among women who did not breastfeed.

The association between the intendedness of pregnancy and breastfeeding behavior is reported in Table 2. Crude odds ratios show that women with any type of unintended pregnancy were more likely not to initiate breastfeeding than women whose pregnancies were intended. Some, but not all, of this association can be attributed to confounding by demographic factors. Having an unintended pregnancy was not associated with any significant difference in the initiation of breastfeeding, after adjusting for age, race, marital status, poverty level, and education. While women with mis-timed pregnancies were as likely to initiate breastfeeding as those whose pregnancies were intended (OR = 1.03; 95% CI, 0.88-1.21), women with unwanted pregnancies were more likely not to start breastfeeding (OR = 1.76; 95% CI, 1.26-2.44).

Table 2 also describes the association between pregnancy intention status and the continuation of nonexclusive breastfeeding for at least 16 weeks. In contrast to the initiation of breastfeeding, duration of breastfeeding was affected by the intendedness of pregnancy in every comparison. Adjusted odds ratios show that women with either type of unintended pregnancy were more likely not to continue breastfeeding than those with intended ones (OR = 1.28; 95% CI, 1.06-1.54). As with breastfeeding initiation, this association is being driven by the unwanted pregnancies. Women with unwanted pregnancies were more likely not to continue breastfeeding (OR = 1.69; 95% CI, 1.12-2.55).

Each of these associations was then evaluated for effect modification. As seen in Table 3, only race was an important factor. In the total sample, 56.3% of white women, 55.4% of Hispanic women, and 24.7% of black women breastfed at all; and 41.6%, 41.2%, and 12.6% of white, Hispanic, and black women breastfed for at least 16 weeks. White women with unwanted pregnancies were more likely not to initiate breastfeeding (OR = 2.50; 95% CI, 1.54-4.05) and more likely not to continue breastfeeding (OR = 2.56; 95% CI, 1.34-4.87) than white women with intended pregnancies. These differences in breastfeeding behaviors for unwanted pregnancies were not seen for either Hispanic or black women. For each stratified analysis, a single P value for heterogeneity was calculated to compare white women with non-white women (Hispanic and black women combined). The only significant difference by race was for unwanted pregnancies. The P value for heterogeneity was 0.01 for both initiation and continuation of breastfeeding. Stratified analyses for age, marital status, education, poverty level, and year of birth showed similar odds ratios across strata and nonsignificant P values for heterogeneity in every case (analyses not shown).

TABLE 1
Characteristics of mothers and infants by breastfeeding status (n=6733)

TABLE 2
Unintended pregnancy and breastfeeding behavior in the United States

TABLE 3
Effect of race on unintended pregnancy and breastfeeding behavior

Discussion

In this study of first-time US mothers, women who breastfed were demographically different from those who did not, but had relatively similar maternal behaviors and infant characteristics. After controlling for these demographic differences, having an unwanted pregnancy was associated with a lower likelihood of both initiating breastfeeding and continuing to breastfeed. In addition, race was an important effect modifier for unwanted pregnancies.

The demographic findings of this study are consistent with the current breastfeeding literature: US women who breastfeed tend to be older, white, married, well-educated, and of a higher socioeconomic status than those who do not.⁹ The main findings of this study are also consistent with the only other study that has examined the relationship between unintended pregnancy and breastfeeding behavior.¹⁰ A cross-sectional sample of 27,700 women who gave birth to a live baby were asked prior to postpartum discharge whether they had intended to become pregnant and their plans for breastfeeding. After controlling for education, race, Medicaid status, maternal age younger than 20, and any tobacco use during pregnancy, the authors found that women whose pregnancies were unintended were more likely not to initiate breastfeeding or to breastfeed exclusively. Adjusted odds ratios of not breastfeeding ranged from 1.10 to 1.41, depending on intention status, and all were statistically significant. In contrast to our study, a major limitation of that study was that the measured outcome was intent to breastfeed at hospital discharge, which may have differed greatly from actual breastfeeding behavior.

The interaction seen in our analysis between intention status and race is initially surprising because, in general, white and Hispanic women breastfeed at much higher rates than black women. But if a pregnancy was unwanted, white women were much less likely to breastfeed than either black or Hispanic women. Neither socioeconomic status nor educational level is the explanation, as both of these factors were controlled for in stratified analyses. Perhaps Hispanic and black women are more accepting of unintended pregnancy than white women and these results reflect cultural differences. Further studies which examine other aspects of unintended pregnancy with respect to race will help to further clarify the reasons for this finding.

Strengths and limitations

Our population-based study has several strengths. The data set provides a large national sample with excellent representation of minority women; statistical oversampling and weighting allow these data to reflect the entire national population. Adjustment of data for nonresponse lessens the risk of selection bias. Furthermore, the study sample was restricted to first births to limit the effect of previous birth experiences on postpartum behaviors. Therefore our results are generalizable to all first-time mothers in the United States.

A major limitation of our study is that information was not collected on several factors, such as substance use both during pregnancy and after birth, that might influence the relationship between pregnancy intention status and breastfeeding behavior. The work of Dye and colleagues,¹⁰ discussed above, found that prenatal tobacco, but not alcohol or drug use, was a significant confounder. Information was also not available on health service–related factors that may contribute to breastfeeding success, such as breastfeeding in the delivery room, length of hospital stay, and participation in educational programs.¹¹

Given that data were collected for the NSFG during personal interviews at differing lengths of time after a pregnancy, inaccuracy is possible. Although the survey does not include corroboration from other sources, such as medical records or birth certificates, it is reassuring that, as an example, rates of prenatal care in our study are similar to those of other nationally reported rates for 1995 (98.1% in our study and 98.8% in National Vital Statistics Reports).¹² Potential misclassification with respect to such medical outcomes as prematurity would be nondifferential and only bias odds ratios toward the null. The extended time between conception and measurement of maternal attitudes increases the uncertainty that a mother will accurately recall both her pregnancy intentions at conception and her breastfeeding practices. Women are more likely to recall a pregnancy carried to birth as intended, but this phenomenon would only bias the results if it also applied to breastfeeding practices, which is unlikely.¹³ While breastfeeding practices may not be exactly recalled, there is no obvious reason for differential reporting.

Conclusions

Our study has clinical implications for first-time US mothers. A recent national goal of the Institute of Medicine is that all pregnancies be planned.³ One of the many benefits of decreasing unintended pregnancy may be to increase breastfeeding rates closer to the Healthy People 2010 goals. In addition, a new hypothesis is suggested by the results of this study: Clinicians should promote breastfeeding differently for women with intended and unintended pregnancies. Future research will evaluate the importance of incorporating pregnancy intention status into patient-centered counseling. In the interim, women with unwanted pregnancies, especially white women, should be targeted for counseling, as they could benefit from breastfeeding, not just for medical reasons but for psychological and economic ones as well.

Acknowledgments

We would like to thank Larry Culpepper, MD, MPH, for his guidance.

Neonatal jaundice is the most common condition for which newborns are tested, treated, and often rehospitalized, resulting in millions of dollars of annual expenditure.^1-3 Two types of jaundice are associated with breastfeeding⁴: early jaundice, or breastfeeding jaundice, caused by the infant receiving insufficient breast milk^5,6; and breast milk jaundice, which develops later in a thriving breastfed infant and is thought to be caused by a substance in the breast milk.^7,8

Diagnosis and treatment of jaundice can begin within the first few days after birth, while the breastfeeding relationship is being established. Despite the ongoing debate on the appropriate protocol for jaundice management^9-11 and a wide variance in physician practice,^12,13 little research has examined the effect of the jaundice experience on the newborn’s mother. The few studies to directly examine the influence of jaundice management on breastfeeding show that protocols such as maternal-infant separation for phototherapy or temporarily suspending breastfeeding are associated with a shorter duration of breastfeeding.^14-16

What remains unclear is how jaundice management affects breastfeeding. Our study adds to existing knowledge by exploring the process by which the maternal experience with a jaundiced newborn affects the mother and her breastfeeding decisions. Qualitative methods, guided by grounded theory, were used because of the paucity of information on this topic and the study’s focus on process.^17-19

Methods

Settings

Two distinct sites in Chicago were chosen to increase the heterogeneity of experiences: a community hospital serving a mixed-income and ethnically diverse population, and an urban teaching hospital serving primarily low-income Latino and African American patients. Breastfeeding initiation rates, tracked by the University of Illinois at Chicago breastfeeding task force, were 70% and 40%, respectively. Institutional Review Board approval was obtained from both sites.

Sample

Two purposeful sampling strategies were employed.²⁰ Criterion sampling was used to recruit mothers, identified through medical record abstraction of all jaundiced infants, who met the following criteria: Spanish or English speaking; exclusively or partially breastfeeding at postpartum discharge; and mother of an otherwise healthy term newborn who had a serum bilirubin level of =10 mg/dL within the first month of life and received care through a study site in 1 or more of the following settings: newborn nursery, outpatient clinic, hospital ward, or home. Maximum variation sampling, which seeks heterogeneity within the sample to permit examination of common themes, was applied to achieve variation in ethnicity, language, age, parity, and jaundice treatment. Eligible mothers were invited by phone to participate in an interview. Sampling continued until data from new interviews confirmed earlier data, signifying that theoretical saturation was achieved.¹⁸

Data collection

Using the literature on hyperbilirubinemia and breastfeeding, an interview guideline was developed addressing the topics in Table 1.²¹ Three female ethnographers (including authors S.K.W. and P.R.H.) conducted in-depth, semi-structured interviews in women’s homes. The interviews were approximately 60 minutes in length given in either Spanish or English. Women were encouraged to lead the conversation, with ethnographers using prompts to guide the discussion toward any topics not addressed and probes to elicit detailed descriptions of the women’s experiences. Audiotaped interviews were transcribed verbatim, and edited by the ethnographer to ensure accuracy and include field notes. Spanish-language interviews were translated into English. Participants received no financial incentives.

Analysis

Interviews were carefully read by all investigators for themes, and codes were developed to represent these themes.^18,22,23 A code book defining each code and listing inclusion and exclusion criteria was developed, and one investigator (S.K.W.) applied codes to the interviews using ATLAS/ti qualitative data analysis software (Scientific Software Development, Berlin, Germany). Intracoder and intercoder (with P.R.H.) agreement were determined to assure consistency of code definitions. Codes with a low level of agreement were redefined and reapplied. Coded text was retrieved and emerging themes analyzed in relation to other themes and variables. Focus was placed on comparing and contrasting women’s experiences to elicit what in the maternal experience with neonatal jaundice influenced infant feeding decisions. We also focused on women’s understanding of the information they received and the relationship between jaundice and breastfeeding.

Results

Of 69 eligible mothers, 11 declined to participate, and 13 could not be reached or scheduled for an interview. Forty-five mothers were interviewed between October 1997 and April 1998 at 2.5 to 14.5 weeks postpartum (mean = 6 weeks). Investigators attempted to hold 2 focus groups with unsuccessful show rates. Individual interviews were conducted with the 2 women who attended these sessions and analyzed with the other interviews. The 24 nonparticipants had similar demographic characteristics to women in the study.

Participants represented a range of sociodemographic and jaundice management characteristics Table 2. Mothers were predominately Latinas of Mexican descent, with a mean age of 27 years (range = 16-38 years). Women born outside the United States had lived in the US from 1 to 25 years (mean = 7 years). More than three quarters of the women lived with the father of the baby. Peak bilirubin levels of all infants ranged from 10.3 to 23.5 mg/dL; 4 infants had peak levels of >20 mg/dL and 7 had peak levels of <12 mg/dL. Thirty-nine infants experienced jaundice within the first 6 days of life, with the majority having nonhemolytic jaundice. Eight infants had breast milk jaundice with peak bilirubin levels occurring between 1 and 2 weeks of age. More than half of the multiparous women had experienced jaundice with a previous infant (n = 14) and three fourths had breastfed a previous child (n = 19).

Though each woman’s experience was unique, a pattern emerged from the women’s discussions that described a process by which their experiences affected the breastfeeding relationship. This process centered on mothers’ interactions with medical professionals during jaundice management and their internalization of the experience.

Jaundice management

Half the women described how their experiences with neonatal jaundice had directly influenced their breastfeeding decisions, positively or negatively, primarily discussing this impact in terms of the breastfeeding instructions they received. Table 3 illustrates the clear pattern seen between a maternal report of breastfeeding orders received from medical professionals and a woman’s feeding status at 2 weeks postpartum, directly after the jaundice experience. Breastfeeding orders were categorized as: continue, conflicting, supplement, suspend, and none. Regardless of parity, women’s interactions with medical professionals related to breastfeeding orders and the level of encouragement they received had the strongest influence on whether women continued to breastfeed.

Mothers exclusively breastfeeding after their experience discussed the encouragement they received from medical staff. Mothers told to continue to breastfeed felt encouraged to breastfeed frequently to help the jaundice go away. All continued to breastfeed for at least 3 weeks, none quit because of their infant’s jaundice. Mothers who returned to exclusive breastfeeding after being told to temporarily suspend breastfeeding or to supplement with formula described being encouraged not to quit breastfeeding and were reassured that their milk was good.

“Right away I wanted to stop breastfeeding, especially if it is me causing him to get that. And they were like, ‘No, no. We’re not telling you to stop. It’s good that you are breastfeeding him.’”

Women exclusively formula feeding because of their experience with jaundice shared 2 separate reasons for not resuming. The first related to not wanting to “take anymore chances” with their infant receiving insufficient milk.

“At the time she was in the hospital they told me to stop breastfeeding her. They wanted to formula feed her. They just said that they think she wasn’t getting enough. They said since they can’t measure how much she drinks that they don’t know how much she is drinking. So I decided, well, I’ll just continue formula feeding her.”

The second related to physical difficulty in reestablishing lactation.

“I breastfed my other three children. … That’s why I tried more to see if he’d latch on, but he didn’t. … Since the beginning, I had the idea that I was only going to breastfeed him, but no.”

Women who continued to supplement with formula because of their experience expressed fear that jaundice would return if they quit supplementation.

“I am still on formula now [7 weeks after experience]. The doctor said he wanted to wait until it is 3 weeks after he is released to wean down. ... I think it caused some damage for me because I am still frightened to really let go of the formula. I may be wrong and maybe it can’t come back at this stage, but I think that something could go wrong and I am still giving formula to make sure that he is getting enough.”

Although mothers whose infants did not receive phototherapy were more likely to be told to continue breastfeeding or to be given no feeding orders than mothers whose infants received phototherapy, there was not a clear pattern between feeding method at 2 weeks postpartum and form of jaundice treatment. Although a few mothers expressed concern about breastfeeding during phototherapy because of having to remove their infant from the light, no mother quit breastfeeding or began supplementing specifically because of treatment. However, many mothers discussed the strong emotional impact that blood work, phototherapy, and the mother-child separation had on them.²⁴

The majority of mothers had prior exposure to neonatal jaundice, approximately one third through personal experience with previous children. Although a few mothers had previous experience with jaundiced infants undergoing phototherapy, only one had been told to stop breastfeeding a previous infant because of jaundice. Even though this mother was not told to stop breastfeeding her current infant, she supplemented with formula because she felt her milk was “no good.”

Maternal internalization of experience

Mothers who received breastfeeding orders to suspend or supplement, or who were given conflicting orders, repeatedly expressed confusion or discontent with these instructions. They commented on the conflict between the medical professional’s advice and their own understanding that breastfeeding was healthier for babies, and on not receiving sufficient explanation to justify changing their feeding method.

“She [doctor] just told me to stop breastfeeding because…. Actually she didn’t tell me why. Which got me confused. I remember I was thinking why did she tell me to stop breastfeeding if breast milk is better for the baby?”

Mothers also shared concerns about nipple confusion affecting their ability to exclusively breastfeed, and feeling that decisions were out of their control.

Lack of understanding about causes of jaundice and feelings of guilt over their role in the etiology were common among all women. More than one third of women expressed guilt that they had caused jaundice either during pregnancy or while breastfeeding.

“I was afraid I did something wrong... that my milk wasn’t coming in right… that I wasn’t feeding her enough or I wasn’t feeding her the right things. Or that my milk was broken ... that I wasn’t making enough or it was wrong somehow. Like it wasn’t meeting her needs.”

Several mothers also expressed belief that their infant may not have gotten jaundice if they had not started breastfeeding.

“I wonder if she would’ve started with the formula, if we would’ve started to supplement with formula from the very beginning, I wonder if it would’ve happened.”

More than half the women discussed breastfeeding as a cause of jaundice. Those women had experienced early neonatal jaundice. More than one third had a previously jaundiced infant and approximately half had previously breastfed. Those who received breastfeeding orders to suspend or supplement, or who were given conflicting orders, were much more likely to discuss breastfeeding as a cause of jaundice than were those who received no breastfeeding orders. Mothers told to continue breastfeeding were least likely to discuss breastfeeding as a cause. No pattern was evident among women by language spoken, parity, bilirubin level, jaundice treatment, or feeding at the interview.

Women’s perceptions of what it is about breastfeeding that causes jaundice related to either quantity or quality.

The causes related to the quantity of breast milk included:

• Insufficient feeding. The infant is not receiving enough breast milk because he or she is not eating enough or the mother is not feeding infant enough.
• Insufficient milk. The mother is not producing enough breast milk.

The causes related to the quality of breast milk included:

• Milk composition. The breast milk caused jaundice, often because it was not good.
• Something in milk. There is something passed through the milk to cause jaundice, such as medicine, hormones, or emotions.

Women who gave explanations for jaundice closer to a biomedical understanding of jaundice, such as insufficient fluid intake, received this information primarily from medical professionals and occasionally from family and friends. Many issues related to feelings of guilt came from their own thoughts, though they were, at times, triggered by something said by a medical professional or family member. Mothers interpreted the information, or lack of information, they received in their attempt to explain why their infant had jaundice.

“The nurse said, the baby won’t be having your breast. All the time the baby will be in the hospital he will only be given formula. Since they never explained why he was that color or anything, I thought maybe my milk was no good. That’s why they told me not to give it to him.”

Discussion

Interactions with medical professionals emerged as the most important factor mediating the impact of the maternal experience with neonatal jaundice on breastfeeding. Encouragement or lack of encouragement from health care professionals played a large role in whether women continued to breastfeed after their experience, which agrees with other findings on the influence, both positive and negative, of interactions with medical professionals on breastfeeding.^25-28 Maternal understanding of and reaction to information received during jaundice management and their subsequent internalization of their experience also played a key role in mothers’ infant feeding decisions. In contrast with previous studies,^14,29 no consistent pattern was seen between breastfeeding continuation and whether infants received blood work only or phototherapy. This difference may be due to our study’s inclusion of multiple settings for phototherapy and differences in the study populations. In addition, earlier studies did not include information on medical professional’s breastfeeding orders.

Limitations

Although the use of qualitative methods allowed in-depth inquiry from the mother’s perspective, it necessitated a small sample limiting generalizability. Generalizability was further limited because the women in the sample were predominately Latina, though study findings were consistent across ethnicities. Data was only collected from the mothers’ point of view, which likely differed from medical professionals’ perceptions. Limiting the sample to mothers who initiated breastfeeding may have excluded mothers who decided not to breastfeed because of a previous jaundice experience. In addition, women whose infants did not develop jaundice because of adequate early breastfeeding support were not interviewed. Careful structuring of the interview guideline and use of experienced ethnographers minimized potential threats to validity through interviewer bias. Regular team meetings to discuss data collection and analysis increased reliability.

Additional research is needed to gain further understanding of mothers’ emotional responses when faced with neonatal conditions like jaundice. While maternal guilt has been acknowledged as a potential problem arising from treatment for neonatal jaundice,^4,7,8 no research has focused on the impact of this guilt on breastfeeding. How do responses like guilt influence perceptions of themselves as breastfeeding mothers and their breastfeeding decisions? The possibility that neonatal jaundice and its management may deprive future children of the opportunity to breastfeed should be examined.

Conclusions

Neonatal jaundice affects many newborns and their families. Besides the monetary cost of treatment, our study results indicate that treatment for jaundice is not completely benign; there are health and emotional costs. Medical professionals must weigh the perceived benefits of treatment decisions and feeding orders against the potential costs to the emotional well being of mothers and newly established breastfeeding relationships. To minimize guilt and enhance maternal understanding about this common condition, professionals need to be aware not only of what information is given to mothers, but how mothers receive and interpret this information. Identifying neonatal jaundice with terms such as breastfeeding jaundice and breast milk jaundice may cause maternal concerns that jaundice is a result of their decision to breastfeed. Medical professionals must provide consistent information and ensure that mothers more fully understand the causes of jaundice and how it relates to breastfeeding, as well as breastfeeding instructions during the experience. The neonatal jaundice experience provides an opportunity for medical professionals to encourage breastfeeding mothers and provide specific guidance on how to maintain a successful breastfeeding relationship.

Acknowledgments

Our study was supported by a grant from the Campus Research Board at the University of Illinois. We would like to thank the women who participated in this study and made time to share their stories with us. We would also like to thank Isabel Martinez, MPH, for her assistance with scheduling and data collection, and Nadine Peacock, PhD, Arden Handler, DrPH, and Rebecca Lipton, PhD, for their comments during the development and analysis of this project.

Neonatal jaundice is the most common condition for which newborns are tested, treated, and often rehospitalized, resulting in millions of dollars of annual expenditure.^1-3 Two types of jaundice are associated with breastfeeding⁴: early jaundice, or breastfeeding jaundice, caused by the infant receiving insufficient breast milk^5,6; and breast milk jaundice, which develops later in a thriving breastfed infant and is thought to be caused by a substance in the breast milk.^7,8

Diagnosis and treatment of jaundice can begin within the first few days after birth, while the breastfeeding relationship is being established. Despite the ongoing debate on the appropriate protocol for jaundice management^9-11 and a wide variance in physician practice,^12,13 little research has examined the effect of the jaundice experience on the newborn’s mother. The few studies to directly examine the influence of jaundice management on breastfeeding show that protocols such as maternal-infant separation for phototherapy or temporarily suspending breastfeeding are associated with a shorter duration of breastfeeding.^14-16

What remains unclear is how jaundice management affects breastfeeding. Our study adds to existing knowledge by exploring the process by which the maternal experience with a jaundiced newborn affects the mother and her breastfeeding decisions. Qualitative methods, guided by grounded theory, were used because of the paucity of information on this topic and the study’s focus on process.^17-19

Methods

Settings

Two distinct sites in Chicago were chosen to increase the heterogeneity of experiences: a community hospital serving a mixed-income and ethnically diverse population, and an urban teaching hospital serving primarily low-income Latino and African American patients. Breastfeeding initiation rates, tracked by the University of Illinois at Chicago breastfeeding task force, were 70% and 40%, respectively. Institutional Review Board approval was obtained from both sites.

Sample

Two purposeful sampling strategies were employed.²⁰ Criterion sampling was used to recruit mothers, identified through medical record abstraction of all jaundiced infants, who met the following criteria: Spanish or English speaking; exclusively or partially breastfeeding at postpartum discharge; and mother of an otherwise healthy term newborn who had a serum bilirubin level of =10 mg/dL within the first month of life and received care through a study site in 1 or more of the following settings: newborn nursery, outpatient clinic, hospital ward, or home. Maximum variation sampling, which seeks heterogeneity within the sample to permit examination of common themes, was applied to achieve variation in ethnicity, language, age, parity, and jaundice treatment. Eligible mothers were invited by phone to participate in an interview. Sampling continued until data from new interviews confirmed earlier data, signifying that theoretical saturation was achieved.¹⁸

Data collection

Using the literature on hyperbilirubinemia and breastfeeding, an interview guideline was developed addressing the topics in Table 1.²¹ Three female ethnographers (including authors S.K.W. and P.R.H.) conducted in-depth, semi-structured interviews in women’s homes. The interviews were approximately 60 minutes in length given in either Spanish or English. Women were encouraged to lead the conversation, with ethnographers using prompts to guide the discussion toward any topics not addressed and probes to elicit detailed descriptions of the women’s experiences. Audiotaped interviews were transcribed verbatim, and edited by the ethnographer to ensure accuracy and include field notes. Spanish-language interviews were translated into English. Participants received no financial incentives.

Analysis

Interviews were carefully read by all investigators for themes, and codes were developed to represent these themes.^18,22,23 A code book defining each code and listing inclusion and exclusion criteria was developed, and one investigator (S.K.W.) applied codes to the interviews using ATLAS/ti qualitative data analysis software (Scientific Software Development, Berlin, Germany). Intracoder and intercoder (with P.R.H.) agreement were determined to assure consistency of code definitions. Codes with a low level of agreement were redefined and reapplied. Coded text was retrieved and emerging themes analyzed in relation to other themes and variables. Focus was placed on comparing and contrasting women’s experiences to elicit what in the maternal experience with neonatal jaundice influenced infant feeding decisions. We also focused on women’s understanding of the information they received and the relationship between jaundice and breastfeeding.

Results

Of 69 eligible mothers, 11 declined to participate, and 13 could not be reached or scheduled for an interview. Forty-five mothers were interviewed between October 1997 and April 1998 at 2.5 to 14.5 weeks postpartum (mean = 6 weeks). Investigators attempted to hold 2 focus groups with unsuccessful show rates. Individual interviews were conducted with the 2 women who attended these sessions and analyzed with the other interviews. The 24 nonparticipants had similar demographic characteristics to women in the study.

Participants represented a range of sociodemographic and jaundice management characteristics Table 2. Mothers were predominately Latinas of Mexican descent, with a mean age of 27 years (range = 16-38 years). Women born outside the United States had lived in the US from 1 to 25 years (mean = 7 years). More than three quarters of the women lived with the father of the baby. Peak bilirubin levels of all infants ranged from 10.3 to 23.5 mg/dL; 4 infants had peak levels of >20 mg/dL and 7 had peak levels of <12 mg/dL. Thirty-nine infants experienced jaundice within the first 6 days of life, with the majority having nonhemolytic jaundice. Eight infants had breast milk jaundice with peak bilirubin levels occurring between 1 and 2 weeks of age. More than half of the multiparous women had experienced jaundice with a previous infant (n = 14) and three fourths had breastfed a previous child (n = 19).

Though each woman’s experience was unique, a pattern emerged from the women’s discussions that described a process by which their experiences affected the breastfeeding relationship. This process centered on mothers’ interactions with medical professionals during jaundice management and their internalization of the experience.

Jaundice management

Half the women described how their experiences with neonatal jaundice had directly influenced their breastfeeding decisions, positively or negatively, primarily discussing this impact in terms of the breastfeeding instructions they received. Table 3 illustrates the clear pattern seen between a maternal report of breastfeeding orders received from medical professionals and a woman’s feeding status at 2 weeks postpartum, directly after the jaundice experience. Breastfeeding orders were categorized as: continue, conflicting, supplement, suspend, and none. Regardless of parity, women’s interactions with medical professionals related to breastfeeding orders and the level of encouragement they received had the strongest influence on whether women continued to breastfeed.

Mothers exclusively breastfeeding after their experience discussed the encouragement they received from medical staff. Mothers told to continue to breastfeed felt encouraged to breastfeed frequently to help the jaundice go away. All continued to breastfeed for at least 3 weeks, none quit because of their infant’s jaundice. Mothers who returned to exclusive breastfeeding after being told to temporarily suspend breastfeeding or to supplement with formula described being encouraged not to quit breastfeeding and were reassured that their milk was good.

“Right away I wanted to stop breastfeeding, especially if it is me causing him to get that. And they were like, ‘No, no. We’re not telling you to stop. It’s good that you are breastfeeding him.’”

Women exclusively formula feeding because of their experience with jaundice shared 2 separate reasons for not resuming. The first related to not wanting to “take anymore chances” with their infant receiving insufficient milk.

“At the time she was in the hospital they told me to stop breastfeeding her. They wanted to formula feed her. They just said that they think she wasn’t getting enough. They said since they can’t measure how much she drinks that they don’t know how much she is drinking. So I decided, well, I’ll just continue formula feeding her.”

The second related to physical difficulty in reestablishing lactation.

“I breastfed my other three children. … That’s why I tried more to see if he’d latch on, but he didn’t. … Since the beginning, I had the idea that I was only going to breastfeed him, but no.”

Women who continued to supplement with formula because of their experience expressed fear that jaundice would return if they quit supplementation.

“I am still on formula now [7 weeks after experience]. The doctor said he wanted to wait until it is 3 weeks after he is released to wean down. ... I think it caused some damage for me because I am still frightened to really let go of the formula. I may be wrong and maybe it can’t come back at this stage, but I think that something could go wrong and I am still giving formula to make sure that he is getting enough.”

Although mothers whose infants did not receive phototherapy were more likely to be told to continue breastfeeding or to be given no feeding orders than mothers whose infants received phototherapy, there was not a clear pattern between feeding method at 2 weeks postpartum and form of jaundice treatment. Although a few mothers expressed concern about breastfeeding during phototherapy because of having to remove their infant from the light, no mother quit breastfeeding or began supplementing specifically because of treatment. However, many mothers discussed the strong emotional impact that blood work, phototherapy, and the mother-child separation had on them.²⁴

The majority of mothers had prior exposure to neonatal jaundice, approximately one third through personal experience with previous children. Although a few mothers had previous experience with jaundiced infants undergoing phototherapy, only one had been told to stop breastfeeding a previous infant because of jaundice. Even though this mother was not told to stop breastfeeding her current infant, she supplemented with formula because she felt her milk was “no good.”

Maternal internalization of experience

Mothers who received breastfeeding orders to suspend or supplement, or who were given conflicting orders, repeatedly expressed confusion or discontent with these instructions. They commented on the conflict between the medical professional’s advice and their own understanding that breastfeeding was healthier for babies, and on not receiving sufficient explanation to justify changing their feeding method.

“She [doctor] just told me to stop breastfeeding because…. Actually she didn’t tell me why. Which got me confused. I remember I was thinking why did she tell me to stop breastfeeding if breast milk is better for the baby?”

Mothers also shared concerns about nipple confusion affecting their ability to exclusively breastfeed, and feeling that decisions were out of their control.

Lack of understanding about causes of jaundice and feelings of guilt over their role in the etiology were common among all women. More than one third of women expressed guilt that they had caused jaundice either during pregnancy or while breastfeeding.

“I was afraid I did something wrong... that my milk wasn’t coming in right… that I wasn’t feeding her enough or I wasn’t feeding her the right things. Or that my milk was broken ... that I wasn’t making enough or it was wrong somehow. Like it wasn’t meeting her needs.”

Several mothers also expressed belief that their infant may not have gotten jaundice if they had not started breastfeeding.

“I wonder if she would’ve started with the formula, if we would’ve started to supplement with formula from the very beginning, I wonder if it would’ve happened.”

More than half the women discussed breastfeeding as a cause of jaundice. Those women had experienced early neonatal jaundice. More than one third had a previously jaundiced infant and approximately half had previously breastfed. Those who received breastfeeding orders to suspend or supplement, or who were given conflicting orders, were much more likely to discuss breastfeeding as a cause of jaundice than were those who received no breastfeeding orders. Mothers told to continue breastfeeding were least likely to discuss breastfeeding as a cause. No pattern was evident among women by language spoken, parity, bilirubin level, jaundice treatment, or feeding at the interview.

Women’s perceptions of what it is about breastfeeding that causes jaundice related to either quantity or quality.

The causes related to the quantity of breast milk included:

• Insufficient feeding. The infant is not receiving enough breast milk because he or she is not eating enough or the mother is not feeding infant enough.
• Insufficient milk. The mother is not producing enough breast milk.

The causes related to the quality of breast milk included:

• Milk composition. The breast milk caused jaundice, often because it was not good.
• Something in milk. There is something passed through the milk to cause jaundice, such as medicine, hormones, or emotions.

Women who gave explanations for jaundice closer to a biomedical understanding of jaundice, such as insufficient fluid intake, received this information primarily from medical professionals and occasionally from family and friends. Many issues related to feelings of guilt came from their own thoughts, though they were, at times, triggered by something said by a medical professional or family member. Mothers interpreted the information, or lack of information, they received in their attempt to explain why their infant had jaundice.

“The nurse said, the baby won’t be having your breast. All the time the baby will be in the hospital he will only be given formula. Since they never explained why he was that color or anything, I thought maybe my milk was no good. That’s why they told me not to give it to him.”

Discussion

Interactions with medical professionals emerged as the most important factor mediating the impact of the maternal experience with neonatal jaundice on breastfeeding. Encouragement or lack of encouragement from health care professionals played a large role in whether women continued to breastfeed after their experience, which agrees with other findings on the influence, both positive and negative, of interactions with medical professionals on breastfeeding.^25-28 Maternal understanding of and reaction to information received during jaundice management and their subsequent internalization of their experience also played a key role in mothers’ infant feeding decisions. In contrast with previous studies,^14,29 no consistent pattern was seen between breastfeeding continuation and whether infants received blood work only or phototherapy. This difference may be due to our study’s inclusion of multiple settings for phototherapy and differences in the study populations. In addition, earlier studies did not include information on medical professional’s breastfeeding orders.

Limitations

Although the use of qualitative methods allowed in-depth inquiry from the mother’s perspective, it necessitated a small sample limiting generalizability. Generalizability was further limited because the women in the sample were predominately Latina, though study findings were consistent across ethnicities. Data was only collected from the mothers’ point of view, which likely differed from medical professionals’ perceptions. Limiting the sample to mothers who initiated breastfeeding may have excluded mothers who decided not to breastfeed because of a previous jaundice experience. In addition, women whose infants did not develop jaundice because of adequate early breastfeeding support were not interviewed. Careful structuring of the interview guideline and use of experienced ethnographers minimized potential threats to validity through interviewer bias. Regular team meetings to discuss data collection and analysis increased reliability.

Additional research is needed to gain further understanding of mothers’ emotional responses when faced with neonatal conditions like jaundice. While maternal guilt has been acknowledged as a potential problem arising from treatment for neonatal jaundice,^4,7,8 no research has focused on the impact of this guilt on breastfeeding. How do responses like guilt influence perceptions of themselves as breastfeeding mothers and their breastfeeding decisions? The possibility that neonatal jaundice and its management may deprive future children of the opportunity to breastfeed should be examined.

Conclusions

Neonatal jaundice affects many newborns and their families. Besides the monetary cost of treatment, our study results indicate that treatment for jaundice is not completely benign; there are health and emotional costs. Medical professionals must weigh the perceived benefits of treatment decisions and feeding orders against the potential costs to the emotional well being of mothers and newly established breastfeeding relationships. To minimize guilt and enhance maternal understanding about this common condition, professionals need to be aware not only of what information is given to mothers, but how mothers receive and interpret this information. Identifying neonatal jaundice with terms such as breastfeeding jaundice and breast milk jaundice may cause maternal concerns that jaundice is a result of their decision to breastfeed. Medical professionals must provide consistent information and ensure that mothers more fully understand the causes of jaundice and how it relates to breastfeeding, as well as breastfeeding instructions during the experience. The neonatal jaundice experience provides an opportunity for medical professionals to encourage breastfeeding mothers and provide specific guidance on how to maintain a successful breastfeeding relationship.

Acknowledgments

Our study was supported by a grant from the Campus Research Board at the University of Illinois. We would like to thank the women who participated in this study and made time to share their stories with us. We would also like to thank Isabel Martinez, MPH, for her assistance with scheduling and data collection, and Nadine Peacock, PhD, Arden Handler, DrPH, and Rebecca Lipton, PhD, for their comments during the development and analysis of this project.

Neonatal jaundice is the most common condition for which newborns are tested, treated, and often rehospitalized, resulting in millions of dollars of annual expenditure.^1-3 Two types of jaundice are associated with breastfeeding⁴: early jaundice, or breastfeeding jaundice, caused by the infant receiving insufficient breast milk^5,6; and breast milk jaundice, which develops later in a thriving breastfed infant and is thought to be caused by a substance in the breast milk.^7,8

Diagnosis and treatment of jaundice can begin within the first few days after birth, while the breastfeeding relationship is being established. Despite the ongoing debate on the appropriate protocol for jaundice management^9-11 and a wide variance in physician practice,^12,13 little research has examined the effect of the jaundice experience on the newborn’s mother. The few studies to directly examine the influence of jaundice management on breastfeeding show that protocols such as maternal-infant separation for phototherapy or temporarily suspending breastfeeding are associated with a shorter duration of breastfeeding.^14-16

What remains unclear is how jaundice management affects breastfeeding. Our study adds to existing knowledge by exploring the process by which the maternal experience with a jaundiced newborn affects the mother and her breastfeeding decisions. Qualitative methods, guided by grounded theory, were used because of the paucity of information on this topic and the study’s focus on process.^17-19

Methods

Settings

Two distinct sites in Chicago were chosen to increase the heterogeneity of experiences: a community hospital serving a mixed-income and ethnically diverse population, and an urban teaching hospital serving primarily low-income Latino and African American patients. Breastfeeding initiation rates, tracked by the University of Illinois at Chicago breastfeeding task force, were 70% and 40%, respectively. Institutional Review Board approval was obtained from both sites.

Sample

Two purposeful sampling strategies were employed.²⁰ Criterion sampling was used to recruit mothers, identified through medical record abstraction of all jaundiced infants, who met the following criteria: Spanish or English speaking; exclusively or partially breastfeeding at postpartum discharge; and mother of an otherwise healthy term newborn who had a serum bilirubin level of =10 mg/dL within the first month of life and received care through a study site in 1 or more of the following settings: newborn nursery, outpatient clinic, hospital ward, or home. Maximum variation sampling, which seeks heterogeneity within the sample to permit examination of common themes, was applied to achieve variation in ethnicity, language, age, parity, and jaundice treatment. Eligible mothers were invited by phone to participate in an interview. Sampling continued until data from new interviews confirmed earlier data, signifying that theoretical saturation was achieved.¹⁸

Data collection

Using the literature on hyperbilirubinemia and breastfeeding, an interview guideline was developed addressing the topics in Table 1.²¹ Three female ethnographers (including authors S.K.W. and P.R.H.) conducted in-depth, semi-structured interviews in women’s homes. The interviews were approximately 60 minutes in length given in either Spanish or English. Women were encouraged to lead the conversation, with ethnographers using prompts to guide the discussion toward any topics not addressed and probes to elicit detailed descriptions of the women’s experiences. Audiotaped interviews were transcribed verbatim, and edited by the ethnographer to ensure accuracy and include field notes. Spanish-language interviews were translated into English. Participants received no financial incentives.

Analysis

Interviews were carefully read by all investigators for themes, and codes were developed to represent these themes.^18,22,23 A code book defining each code and listing inclusion and exclusion criteria was developed, and one investigator (S.K.W.) applied codes to the interviews using ATLAS/ti qualitative data analysis software (Scientific Software Development, Berlin, Germany). Intracoder and intercoder (with P.R.H.) agreement were determined to assure consistency of code definitions. Codes with a low level of agreement were redefined and reapplied. Coded text was retrieved and emerging themes analyzed in relation to other themes and variables. Focus was placed on comparing and contrasting women’s experiences to elicit what in the maternal experience with neonatal jaundice influenced infant feeding decisions. We also focused on women’s understanding of the information they received and the relationship between jaundice and breastfeeding.

Results

Of 69 eligible mothers, 11 declined to participate, and 13 could not be reached or scheduled for an interview. Forty-five mothers were interviewed between October 1997 and April 1998 at 2.5 to 14.5 weeks postpartum (mean = 6 weeks). Investigators attempted to hold 2 focus groups with unsuccessful show rates. Individual interviews were conducted with the 2 women who attended these sessions and analyzed with the other interviews. The 24 nonparticipants had similar demographic characteristics to women in the study.

Participants represented a range of sociodemographic and jaundice management characteristics Table 2. Mothers were predominately Latinas of Mexican descent, with a mean age of 27 years (range = 16-38 years). Women born outside the United States had lived in the US from 1 to 25 years (mean = 7 years). More than three quarters of the women lived with the father of the baby. Peak bilirubin levels of all infants ranged from 10.3 to 23.5 mg/dL; 4 infants had peak levels of >20 mg/dL and 7 had peak levels of <12 mg/dL. Thirty-nine infants experienced jaundice within the first 6 days of life, with the majority having nonhemolytic jaundice. Eight infants had breast milk jaundice with peak bilirubin levels occurring between 1 and 2 weeks of age. More than half of the multiparous women had experienced jaundice with a previous infant (n = 14) and three fourths had breastfed a previous child (n = 19).

Though each woman’s experience was unique, a pattern emerged from the women’s discussions that described a process by which their experiences affected the breastfeeding relationship. This process centered on mothers’ interactions with medical professionals during jaundice management and their internalization of the experience.

Jaundice management

Half the women described how their experiences with neonatal jaundice had directly influenced their breastfeeding decisions, positively or negatively, primarily discussing this impact in terms of the breastfeeding instructions they received. Table 3 illustrates the clear pattern seen between a maternal report of breastfeeding orders received from medical professionals and a woman’s feeding status at 2 weeks postpartum, directly after the jaundice experience. Breastfeeding orders were categorized as: continue, conflicting, supplement, suspend, and none. Regardless of parity, women’s interactions with medical professionals related to breastfeeding orders and the level of encouragement they received had the strongest influence on whether women continued to breastfeed.

Mothers exclusively breastfeeding after their experience discussed the encouragement they received from medical staff. Mothers told to continue to breastfeed felt encouraged to breastfeed frequently to help the jaundice go away. All continued to breastfeed for at least 3 weeks, none quit because of their infant’s jaundice. Mothers who returned to exclusive breastfeeding after being told to temporarily suspend breastfeeding or to supplement with formula described being encouraged not to quit breastfeeding and were reassured that their milk was good.

“Right away I wanted to stop breastfeeding, especially if it is me causing him to get that. And they were like, ‘No, no. We’re not telling you to stop. It’s good that you are breastfeeding him.’”

Women exclusively formula feeding because of their experience with jaundice shared 2 separate reasons for not resuming. The first related to not wanting to “take anymore chances” with their infant receiving insufficient milk.

“At the time she was in the hospital they told me to stop breastfeeding her. They wanted to formula feed her. They just said that they think she wasn’t getting enough. They said since they can’t measure how much she drinks that they don’t know how much she is drinking. So I decided, well, I’ll just continue formula feeding her.”

The second related to physical difficulty in reestablishing lactation.

“I breastfed my other three children. … That’s why I tried more to see if he’d latch on, but he didn’t. … Since the beginning, I had the idea that I was only going to breastfeed him, but no.”

Women who continued to supplement with formula because of their experience expressed fear that jaundice would return if they quit supplementation.

“I am still on formula now [7 weeks after experience]. The doctor said he wanted to wait until it is 3 weeks after he is released to wean down. ... I think it caused some damage for me because I am still frightened to really let go of the formula. I may be wrong and maybe it can’t come back at this stage, but I think that something could go wrong and I am still giving formula to make sure that he is getting enough.”

Although mothers whose infants did not receive phototherapy were more likely to be told to continue breastfeeding or to be given no feeding orders than mothers whose infants received phototherapy, there was not a clear pattern between feeding method at 2 weeks postpartum and form of jaundice treatment. Although a few mothers expressed concern about breastfeeding during phototherapy because of having to remove their infant from the light, no mother quit breastfeeding or began supplementing specifically because of treatment. However, many mothers discussed the strong emotional impact that blood work, phototherapy, and the mother-child separation had on them.²⁴

The majority of mothers had prior exposure to neonatal jaundice, approximately one third through personal experience with previous children. Although a few mothers had previous experience with jaundiced infants undergoing phototherapy, only one had been told to stop breastfeeding a previous infant because of jaundice. Even though this mother was not told to stop breastfeeding her current infant, she supplemented with formula because she felt her milk was “no good.”

Maternal internalization of experience

Mothers who received breastfeeding orders to suspend or supplement, or who were given conflicting orders, repeatedly expressed confusion or discontent with these instructions. They commented on the conflict between the medical professional’s advice and their own understanding that breastfeeding was healthier for babies, and on not receiving sufficient explanation to justify changing their feeding method.

“She [doctor] just told me to stop breastfeeding because…. Actually she didn’t tell me why. Which got me confused. I remember I was thinking why did she tell me to stop breastfeeding if breast milk is better for the baby?”

Mothers also shared concerns about nipple confusion affecting their ability to exclusively breastfeed, and feeling that decisions were out of their control.

Lack of understanding about causes of jaundice and feelings of guilt over their role in the etiology were common among all women. More than one third of women expressed guilt that they had caused jaundice either during pregnancy or while breastfeeding.

“I was afraid I did something wrong... that my milk wasn’t coming in right… that I wasn’t feeding her enough or I wasn’t feeding her the right things. Or that my milk was broken ... that I wasn’t making enough or it was wrong somehow. Like it wasn’t meeting her needs.”

Several mothers also expressed belief that their infant may not have gotten jaundice if they had not started breastfeeding.

“I wonder if she would’ve started with the formula, if we would’ve started to supplement with formula from the very beginning, I wonder if it would’ve happened.”

More than half the women discussed breastfeeding as a cause of jaundice. Those women had experienced early neonatal jaundice. More than one third had a previously jaundiced infant and approximately half had previously breastfed. Those who received breastfeeding orders to suspend or supplement, or who were given conflicting orders, were much more likely to discuss breastfeeding as a cause of jaundice than were those who received no breastfeeding orders. Mothers told to continue breastfeeding were least likely to discuss breastfeeding as a cause. No pattern was evident among women by language spoken, parity, bilirubin level, jaundice treatment, or feeding at the interview.

Women’s perceptions of what it is about breastfeeding that causes jaundice related to either quantity or quality.

The causes related to the quantity of breast milk included:

• Insufficient feeding. The infant is not receiving enough breast milk because he or she is not eating enough or the mother is not feeding infant enough.
• Insufficient milk. The mother is not producing enough breast milk.

The causes related to the quality of breast milk included:

• Milk composition. The breast milk caused jaundice, often because it was not good.
• Something in milk. There is something passed through the milk to cause jaundice, such as medicine, hormones, or emotions.

Women who gave explanations for jaundice closer to a biomedical understanding of jaundice, such as insufficient fluid intake, received this information primarily from medical professionals and occasionally from family and friends. Many issues related to feelings of guilt came from their own thoughts, though they were, at times, triggered by something said by a medical professional or family member. Mothers interpreted the information, or lack of information, they received in their attempt to explain why their infant had jaundice.

“The nurse said, the baby won’t be having your breast. All the time the baby will be in the hospital he will only be given formula. Since they never explained why he was that color or anything, I thought maybe my milk was no good. That’s why they told me not to give it to him.”

Discussion

Interactions with medical professionals emerged as the most important factor mediating the impact of the maternal experience with neonatal jaundice on breastfeeding. Encouragement or lack of encouragement from health care professionals played a large role in whether women continued to breastfeed after their experience, which agrees with other findings on the influence, both positive and negative, of interactions with medical professionals on breastfeeding.^25-28 Maternal understanding of and reaction to information received during jaundice management and their subsequent internalization of their experience also played a key role in mothers’ infant feeding decisions. In contrast with previous studies,^14,29 no consistent pattern was seen between breastfeeding continuation and whether infants received blood work only or phototherapy. This difference may be due to our study’s inclusion of multiple settings for phototherapy and differences in the study populations. In addition, earlier studies did not include information on medical professional’s breastfeeding orders.

Limitations

Although the use of qualitative methods allowed in-depth inquiry from the mother’s perspective, it necessitated a small sample limiting generalizability. Generalizability was further limited because the women in the sample were predominately Latina, though study findings were consistent across ethnicities. Data was only collected from the mothers’ point of view, which likely differed from medical professionals’ perceptions. Limiting the sample to mothers who initiated breastfeeding may have excluded mothers who decided not to breastfeed because of a previous jaundice experience. In addition, women whose infants did not develop jaundice because of adequate early breastfeeding support were not interviewed. Careful structuring of the interview guideline and use of experienced ethnographers minimized potential threats to validity through interviewer bias. Regular team meetings to discuss data collection and analysis increased reliability.

Additional research is needed to gain further understanding of mothers’ emotional responses when faced with neonatal conditions like jaundice. While maternal guilt has been acknowledged as a potential problem arising from treatment for neonatal jaundice,^4,7,8 no research has focused on the impact of this guilt on breastfeeding. How do responses like guilt influence perceptions of themselves as breastfeeding mothers and their breastfeeding decisions? The possibility that neonatal jaundice and its management may deprive future children of the opportunity to breastfeed should be examined.

Conclusions

Neonatal jaundice affects many newborns and their families. Besides the monetary cost of treatment, our study results indicate that treatment for jaundice is not completely benign; there are health and emotional costs. Medical professionals must weigh the perceived benefits of treatment decisions and feeding orders against the potential costs to the emotional well being of mothers and newly established breastfeeding relationships. To minimize guilt and enhance maternal understanding about this common condition, professionals need to be aware not only of what information is given to mothers, but how mothers receive and interpret this information. Identifying neonatal jaundice with terms such as breastfeeding jaundice and breast milk jaundice may cause maternal concerns that jaundice is a result of their decision to breastfeed. Medical professionals must provide consistent information and ensure that mothers more fully understand the causes of jaundice and how it relates to breastfeeding, as well as breastfeeding instructions during the experience. The neonatal jaundice experience provides an opportunity for medical professionals to encourage breastfeeding mothers and provide specific guidance on how to maintain a successful breastfeeding relationship.

Acknowledgments

Our study was supported by a grant from the Campus Research Board at the University of Illinois. We would like to thank the women who participated in this study and made time to share their stories with us. We would also like to thank Isabel Martinez, MPH, for her assistance with scheduling and data collection, and Nadine Peacock, PhD, Arden Handler, DrPH, and Rebecca Lipton, PhD, for their comments during the development and analysis of this project.

Since the initial use of antihypertensive medications in the 1940s, they have been the traditional approach to treatment essential hypertension. Many of these pharmacologic agents, however, are costly and are associated with substantial adverse effects. As a result, interest has been increasing in alternative methods to prevent and treat hypertension. Clinical trials using dietary interventions for the alleviation of hypertension and observational studies have suggested that a number of foods and specific food components may exert an antihypertensive effect.^1-5 Other research, however, has shown no effect.^6-10 Studies specific to oats or cereal fibers have also provided mixed results. Observational studies have noted a reduction in blood pressure (BP),¹¹ but the few clinical trials conducted to date have shown no effect.¹²

Selected whole grains are known to be good sources of soluble fibers. Previous research trials have demonstrated that these fibers can effectively reduce plasma insulin concentrations and provide other health benefits.^13,14 Additionally, elevated insulin levels have been implicated in the etiology of hypertension.¹⁵ Based on this potential biologic mechanism and the previously inconsistent findings, we conducted a 12-week trial to evaluate the clinical effects of soluble fiber–rich whole oat cereals when added to the diet of hyperinsulinemic patients medicated for essential hypertension.

Methods

Study sample

Participants were recruited from a database of treated hypertensive patients provided by a local health maintenance organization (HealthPartners). Initial letters describing the study were mailed to 8000 potential participants. Of these, 524 people responded to the mailing and agreed to a telephone screen to determine eligibility. Among respondents, 212 passed the initial phone screening and were invited to our research clinic (Hypertension and Cholesterol Research Clinic at the University of Minnesota Medical School) for a BP screening and general physical. For inclusion in the study, average screening BP readings (2 sets of readings within 7 days) taken by our team physician could not exceed 160/100. Table 1 lists exclusion criteria. The study protocol was reviewed and approved by the University of Minnesota Human Subjects Committee of the Institutional Review Board.

Eighty-eight volunteers (45 men and 43 women) aged 33 to 67 years met all inclusion criteria and provided written informed consent. All participants had a history of essential mild or moderate hypertension (BP 120/80 to 160/100 mm Hg), and were treated with no more than 1 antihypertensive medication (excluding -adrenergic receptor blocking agents) and/or 1 diuretic medication for at least 1 month before enrollment. Eighty participants were treated with a single antihypertensive medication; 8 required an antihypertensive drug and a diuretic medication to manage their BP. Individuals taking beta blockers were excluded from the study because they often take medications prescribed for more serious cardiovascular conditions, such as cardiac arrhythmias, and medication reduction would be inappropriate under such circumstances. Participants’ primary physicians were also consulted concerning participation and study-related medication changes.

TABLE 1
EXCLUSION CRITERIA

Study design

This randomized controlled parallel-group trial consisted of 3 four-week phases: a Baseline Feeding phase, a Medication Reduction phase, and a Maintenance phase. Eligible individuals were stratified by baseline systolic blood pressure (SBP) (< 140 mm Hg versus 140 mm Hg) and baseline soluble fiber intake (less than 3 grams/day versus 3 to 6 grams/day). At the start of the baseline phase, participants were randomized to either an oats cereal treatment group (n = 45) or a low-fiber cereal control group (n = 43).

The cereal treatments were isocaloric and administered during all 3 phases of the study. Individuals in the oats group received a daily serving of 60 grams (approximately three fourths cup) Quaker Oatmeal (5.61 grams total dietary fiber, 3.25 grams soluble fiber, and 2.83 grams -glucans) and 77 grams (approximately one and one third cups) Quaker Oat Squares (6.07 grams total dietary fiber, 2.98 grams soluble fiber, and 2.59 grams -glucans). Individuals in the control group consumed 65 grams (0.5 cup) Malt-O-Meal Hot Wheat Cereal (2.32 grams total dietary fiber, 0.6 grams soluble fiber) and 81 grams (2 cups) Kellogg’s Crispix (1.2 grams total dietary fiber, < 0.5 grams soluble fiber).

Cereals were dispensed in unlabeled bulk containers to facilitate physician blinding. Remaining cereal was returned and weighed at each of the weekly or biweekly visits at our research clinic. Additionally, participants kept a daily cereal calendar that was reviewed by members of our research staff and used to help determine cereal compliance.

Changes in antihypertensive medication dose were implemented according to the protocol described in the Figure. Participants were asked to maintain their usual lifestyle, physical activity, dietary pattern, and body weight during the 12 weeks of the study. Individuals were invited to participate in a 6-week follow-up phase after the intervention was completed to monitor the residual BP effect after cereal consumption was discontinued.

FIGURE
MEDICATION REDUCTION PROTOCOL

Outcomes measured

The study physician responsible for BP measurement, blood draws, and general patient examinations (described below) was unaware of the cereal group assignment. BP was measured at the clinic twice a week during the first (baseline feeding) and last (maintenance) phases of the study and weekly during the second (medication reduction) phase. Participants reported at approximately the same time of the day for all appointments. BP readings were obtained 24 hours after the last medication dose or, if the patient was unmedicated, at the same time of day as previous study BP readings and after participants had rested quietly in the seated position for at least 5 minutes in an examination room.

The study physician took all readings on the right arm, using a mercury column sphygmomanometer (Korotkoff phase V for diastolic blood pressure [DBP]). Standard cuff size was used unless upper arm circumference exceeded 31 cm, in which case a large cuff with 15 x 35–cm bladders was chosen. Measurements were repeated 4 times in 2-minute intervals. The mean of the last 3 readings was calculated and used in subsequent analyses. Baseline and final study measurements used in the analyses and reported in this paper represent the averages of the first 2 and last 2 study visits.

Preintervention and postintervention blood samples were collected into standard 6-mL serum separator tubes. Samples were analyzed within 24 hours for general chemistry and plasma lipids (total cholesterol, low-density cholesterol [LDL-C], and high-density cholesterol [HDL-C] as well as triglyceride levels) by an accredited independent laboratory and according to standard chemical methods.¹⁶

A written 42-question side effect questionnaire was administered to participants at the beginning of the baseline phase and at the end of the intervention. Participants reported the frequency with which they experienced side effects associated with increased fiber intake (eg, loose stools, flatulence) and hypertension (eg, headaches, dizziness) using a 5-item scale ranging from “never” to “very frequently” (event occurring once or several times daily). Each item of the scale was assigned a value ranging from 1 to 5. Values were tallied across all 42 questions. A final score was assigned to each participant for both time points. Mean scores by group were used in the analyses.

Participants completed a 3-day food record at baseline and at the end of the 12 weeks of intervention. Food records were examined for thoroughness by a licensed nutritionist and used to determine dietary changes. Nutrient intakes were calculated using the Nutrient Data System software (version 2.92) managed by the Nutrition Coordinating Center at the University of Minnesota School of Public Health.¹⁷

Statistics

The sample size calculation was based on a level of significance set at 0.05 and power at 80% to detect a 15% difference in medication reduction. Differences in medication reduction were determined by using the chi-square test of proportions. For continuous variables, Student’s paired and unpaired t tests were performed to determine differences within and between groups. In terms of medication reduction, logistic regression was used to adjust for potential confounders such as body weight and sodium intake. Multiple regression was employed to adjust blood lipid and glucose levels and BP findings for confounding. Because adjustment did not change the interpretation of the data, unadjusted findings are reported. The analyses of the data from this intent-to-treat population, which were determined to include all randomized patients, were conducted using the Statistical Analysis System (SAS Institute, Cary, N.C.). Results are reported as means ± SD unless noted otherwise. All P values are double sided.

Results

All the original 88 participants enrolled, all completed the 12-week trial, and all participated in the 6-week follow-up phase. Instructions to consume all dispensed cereals every day were followed well. Compliance was high for both groups (94.5% for the oat group and 92.7% for the control group) based on the amount of consumed cereal by weight. Randomization was largely effective; there were no apparent differences in baseline characteristics between each of the treatment groups (Table 2). Participants were primarily white (97%), with a mean age of 48 years (range 33 to 67 years).

BP and BP medication changes are summarized in Table 3. Among subjects in the oats group, 73% experienced a BP medication reduction during the intervention and had maintained that by the end of the study, as compared with only 42% in the control group (P < .05). Moreover, those in the oats group who did not experience a medication reduction had a 7-mm Hg decrease in SBP and a 4-mm Hg reduction in DBP. There was a small, nonsignificant change in SBP and DBP among those who did not experience a medication reduction in the control group. Medication reduction did not differ across classes of antihypertensive medication or our stratification variables of baseline soluble fiber intake or BP. Additionally, during the 6-week follow-up phase, 6 of the 18 (33%) individuals in the control group versus 22 of the 33 (67%) in the treatment group resumed taking medication.

Average BP in the oats group was lowered from 140/88 mm Hg at baseline to 134/85 mm Hg by the end of the first 4 weeks. Only the change in systolic BP was statistically significant (P < .05). Over the same 4-week period, the control group experienced a mean change of BP from 138/86 mm Hg to 136/85 mm Hg, which was not significant.

Baseline and postintervention lipid and glucose levels appear in Table 4. There were no significant modifications in any of the lipid parameters for the individuals in the control group, although there was a downward trend in all lipid measures. In the oats treatment group, mean total cholesterol (TC) concentration decreased by 31.7 mg/dL (15% drop). A similar decrease of 22.3 mg/dL (16% drop) was seen in the oats group’s average LDL-C levels. Blood glucose levels in the oats group also improved significantly (P < .01). The mean differences between post study and prestudy values (± SE) between the 2 groups, calculated for the average changes in TC, LDL-C, and glucose experienced by each of the groups, were -24.2 mg/dL (± 6.1), -16.2 mg/dL (± 4.4), and -15.03 mg/dL (± 4.3), respectively.

The frequency of dietary fiber-related and hypertension-related side effects decreased by 22% in the treatment group (Table 4). This finding was not observed in the control group. No weight changes were observed in either group, indicating that participants adjusted their diet to compensate for the addition of the cereals by substituting cereal for their standard breakfast and consuming them in place of afternoon snacks as determined by the food record inspection. Total daily energy intake (kcal/day) remained virtually unchanged when postintervention food intake was compared with intake at baseline. Participants in both groups did experience significant decreases in total fat and saturated fat intake along with significant increases in fiber (both soluble and insoluble), potassium, and calcium. The increase in total fiber intake was greater in the treatment group (P < .01) than in the control group (P < .05). In addition, the treatment group experienced a significant increase in magnesium not observed in the control group.

TABLE 2
BASELINE CHARACTERISTICS*

TABLE 3
ANTIHYPERTENSIVE MEDICATION AND BLOOD PRESSURE CHANGES BY GROUP

TABLE 4
SECONDARY OUTCOME MEASURES BY GROUP*

Discussion

The results of this trial suggest that an increased consumption of soluble fiber-rich, whole-grain, oat-based cereals can significantly reduce antihypertensive medication need among patients being treated for hypertension. Of the 45 participants in the oats group, 33 experienced at least half medication reduction compared with only 18 of the 43 participants in the control group. Positive BP changes were evident during the first 4 weeks of oat cereal treatment; BP levels rose steadily during the 6-week follow-up phase.

In addition, mean BP readings in the oat group participants who did not experience a medication reduction had improved at study completion compared with baseline. A significant number of participants in the refined cereal control group experienced at least half medication reduction (18/43), a finding that might be attributed to the increase in calcium, potassium, and total dietary fiber intake^8,9,18 as well as to the decreased intake of total and saturated fat.¹⁹ Additionally, during the follow-up phase, only 6 of the 18 (33%) versus 22 of the 33 (67%) in the oats group resumed taking their medication. Therefore, part of the medication reduction effect in the control group may have been the result of a greater percentage of participants who did not need their antihypertensive medication. This issue should be considered in the design of future trials.

As always, regression to the mean and the Hawthorne effect might explain some of the outcomes in this trial. However, it is likely that both increased soluble fiber and micronutrient intake explain the decrease in antihypertensive medication need observed in the treatment group. This study was designed to identify not the hypotensive effects of specific cereal components but the effects of a whole food intervention. Our findings are consistent with those of other whole-food interventions, such as the Dietary Approaches to Stop Hypertension (DASH) trial, tested in hypertensive populations.²⁰ Nonetheless, known diet-related determinants of BP (sodium chloride, alcohol, body weight, and level of physical activity) could not explain the treatment effect because no significant differences in these variables existed between the groups.

The soluble fiber fraction of the oat-based cereal intervention is probably partially responsible for the reduction in antihypertensive medication need observed in this trial. Previous studies that tested either soluble fiber supplements or diets rich in soluble fiber have noted significant reductions in BP.^21-23 Improvement in insulin sensitivity has been proposed as the pathway through which soluble fiber improves BP.²⁴ Insulin sensitivity was not determined in this study, yet the oats treatment group experienced a significant improvement in plasma glucose levels. This finding suggests that insulin sensitivity may have been enhanced. Impaired response to insulin was recently shown to precede endothelial dysfunction and subsequent elevations in BP.²⁵ Moreover, soluble fiber supplements and diets high in soluble fiber have been shown to improve insulin sensitivity.^25-28 Other components of whole grains, such as magnesium or grain flavonoids, may also contribute to the favorable medication reduction observed in the oats group.^29,30

This 12-week whole-food intervention trial was not designed to test either the long-term efficacy of oat-based cereals or the likelihood of long-term adherence to the feeding regimen. Nonetheless, a whole-grain oat-based cereal intervention might be an effective way to manage mild (type I) hypertension. The reduction in BP medication that occurred in the oats group was independent of weight change and sodium chloride and alcohol intake, suggesting that soluble fiber–rich whole grains should be added to the current dietary recommendations for people with elevated BP. Moreover, it is possible that the consumption of a diet high in soluble fiber–rich whole grains may prevent or delay the initiation of hypertension drug therapy in at-risk or borderline hypertensive patients. Based on the results from this study, physicians may be justified in recommending to their hypertensive patients a dietary regimen that includes the daily consumption of whole-grain oats (equaling 6 g of soluble fiber) in conjunction with their usual therapy. Such an intervention may be expected to yield results within 4 weeks.

Conclusions

A diet containing soluble fiber-rich whole grains can significantly reduce antihypertensive medication need and improve BP control among treated hypertensives. Combined with the reductions in blood lipids and plasma glucose, the intake of soluble fiber–rich whole oat cereals appears to be an effective nutritional approach in the reduction of cardiovascular disease risk. Future trials will need to investigate the antihypertensive effects of oats in other populations (eg, different racial groups) and determine whether reductions in BP measurements can be sustained for the long term.

Acknowledgment

The research team recognizes Anne Marie Weber-Main, PhD, for her excellent and tireless editorial contributions to this project.

Since the initial use of antihypertensive medications in the 1940s, they have been the traditional approach to treatment essential hypertension. Many of these pharmacologic agents, however, are costly and are associated with substantial adverse effects. As a result, interest has been increasing in alternative methods to prevent and treat hypertension. Clinical trials using dietary interventions for the alleviation of hypertension and observational studies have suggested that a number of foods and specific food components may exert an antihypertensive effect.^1-5 Other research, however, has shown no effect.^6-10 Studies specific to oats or cereal fibers have also provided mixed results. Observational studies have noted a reduction in blood pressure (BP),¹¹ but the few clinical trials conducted to date have shown no effect.¹²

Selected whole grains are known to be good sources of soluble fibers. Previous research trials have demonstrated that these fibers can effectively reduce plasma insulin concentrations and provide other health benefits.^13,14 Additionally, elevated insulin levels have been implicated in the etiology of hypertension.¹⁵ Based on this potential biologic mechanism and the previously inconsistent findings, we conducted a 12-week trial to evaluate the clinical effects of soluble fiber–rich whole oat cereals when added to the diet of hyperinsulinemic patients medicated for essential hypertension.

Methods

Study sample

Participants were recruited from a database of treated hypertensive patients provided by a local health maintenance organization (HealthPartners). Initial letters describing the study were mailed to 8000 potential participants. Of these, 524 people responded to the mailing and agreed to a telephone screen to determine eligibility. Among respondents, 212 passed the initial phone screening and were invited to our research clinic (Hypertension and Cholesterol Research Clinic at the University of Minnesota Medical School) for a BP screening and general physical. For inclusion in the study, average screening BP readings (2 sets of readings within 7 days) taken by our team physician could not exceed 160/100. Table 1 lists exclusion criteria. The study protocol was reviewed and approved by the University of Minnesota Human Subjects Committee of the Institutional Review Board.

Eighty-eight volunteers (45 men and 43 women) aged 33 to 67 years met all inclusion criteria and provided written informed consent. All participants had a history of essential mild or moderate hypertension (BP 120/80 to 160/100 mm Hg), and were treated with no more than 1 antihypertensive medication (excluding -adrenergic receptor blocking agents) and/or 1 diuretic medication for at least 1 month before enrollment. Eighty participants were treated with a single antihypertensive medication; 8 required an antihypertensive drug and a diuretic medication to manage their BP. Individuals taking beta blockers were excluded from the study because they often take medications prescribed for more serious cardiovascular conditions, such as cardiac arrhythmias, and medication reduction would be inappropriate under such circumstances. Participants’ primary physicians were also consulted concerning participation and study-related medication changes.

TABLE 1
EXCLUSION CRITERIA

Study design

This randomized controlled parallel-group trial consisted of 3 four-week phases: a Baseline Feeding phase, a Medication Reduction phase, and a Maintenance phase. Eligible individuals were stratified by baseline systolic blood pressure (SBP) (< 140 mm Hg versus 140 mm Hg) and baseline soluble fiber intake (less than 3 grams/day versus 3 to 6 grams/day). At the start of the baseline phase, participants were randomized to either an oats cereal treatment group (n = 45) or a low-fiber cereal control group (n = 43).

The cereal treatments were isocaloric and administered during all 3 phases of the study. Individuals in the oats group received a daily serving of 60 grams (approximately three fourths cup) Quaker Oatmeal (5.61 grams total dietary fiber, 3.25 grams soluble fiber, and 2.83 grams -glucans) and 77 grams (approximately one and one third cups) Quaker Oat Squares (6.07 grams total dietary fiber, 2.98 grams soluble fiber, and 2.59 grams -glucans). Individuals in the control group consumed 65 grams (0.5 cup) Malt-O-Meal Hot Wheat Cereal (2.32 grams total dietary fiber, 0.6 grams soluble fiber) and 81 grams (2 cups) Kellogg’s Crispix (1.2 grams total dietary fiber, < 0.5 grams soluble fiber).

Cereals were dispensed in unlabeled bulk containers to facilitate physician blinding. Remaining cereal was returned and weighed at each of the weekly or biweekly visits at our research clinic. Additionally, participants kept a daily cereal calendar that was reviewed by members of our research staff and used to help determine cereal compliance.

Changes in antihypertensive medication dose were implemented according to the protocol described in the Figure. Participants were asked to maintain their usual lifestyle, physical activity, dietary pattern, and body weight during the 12 weeks of the study. Individuals were invited to participate in a 6-week follow-up phase after the intervention was completed to monitor the residual BP effect after cereal consumption was discontinued.

FIGURE
MEDICATION REDUCTION PROTOCOL

Outcomes measured

The study physician responsible for BP measurement, blood draws, and general patient examinations (described below) was unaware of the cereal group assignment. BP was measured at the clinic twice a week during the first (baseline feeding) and last (maintenance) phases of the study and weekly during the second (medication reduction) phase. Participants reported at approximately the same time of the day for all appointments. BP readings were obtained 24 hours after the last medication dose or, if the patient was unmedicated, at the same time of day as previous study BP readings and after participants had rested quietly in the seated position for at least 5 minutes in an examination room.

The study physician took all readings on the right arm, using a mercury column sphygmomanometer (Korotkoff phase V for diastolic blood pressure [DBP]). Standard cuff size was used unless upper arm circumference exceeded 31 cm, in which case a large cuff with 15 x 35–cm bladders was chosen. Measurements were repeated 4 times in 2-minute intervals. The mean of the last 3 readings was calculated and used in subsequent analyses. Baseline and final study measurements used in the analyses and reported in this paper represent the averages of the first 2 and last 2 study visits.

Preintervention and postintervention blood samples were collected into standard 6-mL serum separator tubes. Samples were analyzed within 24 hours for general chemistry and plasma lipids (total cholesterol, low-density cholesterol [LDL-C], and high-density cholesterol [HDL-C] as well as triglyceride levels) by an accredited independent laboratory and according to standard chemical methods.¹⁶

A written 42-question side effect questionnaire was administered to participants at the beginning of the baseline phase and at the end of the intervention. Participants reported the frequency with which they experienced side effects associated with increased fiber intake (eg, loose stools, flatulence) and hypertension (eg, headaches, dizziness) using a 5-item scale ranging from “never” to “very frequently” (event occurring once or several times daily). Each item of the scale was assigned a value ranging from 1 to 5. Values were tallied across all 42 questions. A final score was assigned to each participant for both time points. Mean scores by group were used in the analyses.

Participants completed a 3-day food record at baseline and at the end of the 12 weeks of intervention. Food records were examined for thoroughness by a licensed nutritionist and used to determine dietary changes. Nutrient intakes were calculated using the Nutrient Data System software (version 2.92) managed by the Nutrition Coordinating Center at the University of Minnesota School of Public Health.¹⁷

Statistics

The sample size calculation was based on a level of significance set at 0.05 and power at 80% to detect a 15% difference in medication reduction. Differences in medication reduction were determined by using the chi-square test of proportions. For continuous variables, Student’s paired and unpaired t tests were performed to determine differences within and between groups. In terms of medication reduction, logistic regression was used to adjust for potential confounders such as body weight and sodium intake. Multiple regression was employed to adjust blood lipid and glucose levels and BP findings for confounding. Because adjustment did not change the interpretation of the data, unadjusted findings are reported. The analyses of the data from this intent-to-treat population, which were determined to include all randomized patients, were conducted using the Statistical Analysis System (SAS Institute, Cary, N.C.). Results are reported as means ± SD unless noted otherwise. All P values are double sided.

Results

All the original 88 participants enrolled, all completed the 12-week trial, and all participated in the 6-week follow-up phase. Instructions to consume all dispensed cereals every day were followed well. Compliance was high for both groups (94.5% for the oat group and 92.7% for the control group) based on the amount of consumed cereal by weight. Randomization was largely effective; there were no apparent differences in baseline characteristics between each of the treatment groups (Table 2). Participants were primarily white (97%), with a mean age of 48 years (range 33 to 67 years).

BP and BP medication changes are summarized in Table 3. Among subjects in the oats group, 73% experienced a BP medication reduction during the intervention and had maintained that by the end of the study, as compared with only 42% in the control group (P < .05). Moreover, those in the oats group who did not experience a medication reduction had a 7-mm Hg decrease in SBP and a 4-mm Hg reduction in DBP. There was a small, nonsignificant change in SBP and DBP among those who did not experience a medication reduction in the control group. Medication reduction did not differ across classes of antihypertensive medication or our stratification variables of baseline soluble fiber intake or BP. Additionally, during the 6-week follow-up phase, 6 of the 18 (33%) individuals in the control group versus 22 of the 33 (67%) in the treatment group resumed taking medication.

Average BP in the oats group was lowered from 140/88 mm Hg at baseline to 134/85 mm Hg by the end of the first 4 weeks. Only the change in systolic BP was statistically significant (P < .05). Over the same 4-week period, the control group experienced a mean change of BP from 138/86 mm Hg to 136/85 mm Hg, which was not significant.

Baseline and postintervention lipid and glucose levels appear in Table 4. There were no significant modifications in any of the lipid parameters for the individuals in the control group, although there was a downward trend in all lipid measures. In the oats treatment group, mean total cholesterol (TC) concentration decreased by 31.7 mg/dL (15% drop). A similar decrease of 22.3 mg/dL (16% drop) was seen in the oats group’s average LDL-C levels. Blood glucose levels in the oats group also improved significantly (P < .01). The mean differences between post study and prestudy values (± SE) between the 2 groups, calculated for the average changes in TC, LDL-C, and glucose experienced by each of the groups, were -24.2 mg/dL (± 6.1), -16.2 mg/dL (± 4.4), and -15.03 mg/dL (± 4.3), respectively.

The frequency of dietary fiber-related and hypertension-related side effects decreased by 22% in the treatment group (Table 4). This finding was not observed in the control group. No weight changes were observed in either group, indicating that participants adjusted their diet to compensate for the addition of the cereals by substituting cereal for their standard breakfast and consuming them in place of afternoon snacks as determined by the food record inspection. Total daily energy intake (kcal/day) remained virtually unchanged when postintervention food intake was compared with intake at baseline. Participants in both groups did experience significant decreases in total fat and saturated fat intake along with significant increases in fiber (both soluble and insoluble), potassium, and calcium. The increase in total fiber intake was greater in the treatment group (P < .01) than in the control group (P < .05). In addition, the treatment group experienced a significant increase in magnesium not observed in the control group.

TABLE 2
BASELINE CHARACTERISTICS*

TABLE 3
ANTIHYPERTENSIVE MEDICATION AND BLOOD PRESSURE CHANGES BY GROUP

TABLE 4
SECONDARY OUTCOME MEASURES BY GROUP*

Discussion

The results of this trial suggest that an increased consumption of soluble fiber-rich, whole-grain, oat-based cereals can significantly reduce antihypertensive medication need among patients being treated for hypertension. Of the 45 participants in the oats group, 33 experienced at least half medication reduction compared with only 18 of the 43 participants in the control group. Positive BP changes were evident during the first 4 weeks of oat cereal treatment; BP levels rose steadily during the 6-week follow-up phase.

In addition, mean BP readings in the oat group participants who did not experience a medication reduction had improved at study completion compared with baseline. A significant number of participants in the refined cereal control group experienced at least half medication reduction (18/43), a finding that might be attributed to the increase in calcium, potassium, and total dietary fiber intake^8,9,18 as well as to the decreased intake of total and saturated fat.¹⁹ Additionally, during the follow-up phase, only 6 of the 18 (33%) versus 22 of the 33 (67%) in the oats group resumed taking their medication. Therefore, part of the medication reduction effect in the control group may have been the result of a greater percentage of participants who did not need their antihypertensive medication. This issue should be considered in the design of future trials.

As always, regression to the mean and the Hawthorne effect might explain some of the outcomes in this trial. However, it is likely that both increased soluble fiber and micronutrient intake explain the decrease in antihypertensive medication need observed in the treatment group. This study was designed to identify not the hypotensive effects of specific cereal components but the effects of a whole food intervention. Our findings are consistent with those of other whole-food interventions, such as the Dietary Approaches to Stop Hypertension (DASH) trial, tested in hypertensive populations.²⁰ Nonetheless, known diet-related determinants of BP (sodium chloride, alcohol, body weight, and level of physical activity) could not explain the treatment effect because no significant differences in these variables existed between the groups.

The soluble fiber fraction of the oat-based cereal intervention is probably partially responsible for the reduction in antihypertensive medication need observed in this trial. Previous studies that tested either soluble fiber supplements or diets rich in soluble fiber have noted significant reductions in BP.^21-23 Improvement in insulin sensitivity has been proposed as the pathway through which soluble fiber improves BP.²⁴ Insulin sensitivity was not determined in this study, yet the oats treatment group experienced a significant improvement in plasma glucose levels. This finding suggests that insulin sensitivity may have been enhanced. Impaired response to insulin was recently shown to precede endothelial dysfunction and subsequent elevations in BP.²⁵ Moreover, soluble fiber supplements and diets high in soluble fiber have been shown to improve insulin sensitivity.^25-28 Other components of whole grains, such as magnesium or grain flavonoids, may also contribute to the favorable medication reduction observed in the oats group.^29,30

This 12-week whole-food intervention trial was not designed to test either the long-term efficacy of oat-based cereals or the likelihood of long-term adherence to the feeding regimen. Nonetheless, a whole-grain oat-based cereal intervention might be an effective way to manage mild (type I) hypertension. The reduction in BP medication that occurred in the oats group was independent of weight change and sodium chloride and alcohol intake, suggesting that soluble fiber–rich whole grains should be added to the current dietary recommendations for people with elevated BP. Moreover, it is possible that the consumption of a diet high in soluble fiber–rich whole grains may prevent or delay the initiation of hypertension drug therapy in at-risk or borderline hypertensive patients. Based on the results from this study, physicians may be justified in recommending to their hypertensive patients a dietary regimen that includes the daily consumption of whole-grain oats (equaling 6 g of soluble fiber) in conjunction with their usual therapy. Such an intervention may be expected to yield results within 4 weeks.

Conclusions

A diet containing soluble fiber-rich whole grains can significantly reduce antihypertensive medication need and improve BP control among treated hypertensives. Combined with the reductions in blood lipids and plasma glucose, the intake of soluble fiber–rich whole oat cereals appears to be an effective nutritional approach in the reduction of cardiovascular disease risk. Future trials will need to investigate the antihypertensive effects of oats in other populations (eg, different racial groups) and determine whether reductions in BP measurements can be sustained for the long term.

Acknowledgment

The research team recognizes Anne Marie Weber-Main, PhD, for her excellent and tireless editorial contributions to this project.

Since the initial use of antihypertensive medications in the 1940s, they have been the traditional approach to treatment essential hypertension. Many of these pharmacologic agents, however, are costly and are associated with substantial adverse effects. As a result, interest has been increasing in alternative methods to prevent and treat hypertension. Clinical trials using dietary interventions for the alleviation of hypertension and observational studies have suggested that a number of foods and specific food components may exert an antihypertensive effect.^1-5 Other research, however, has shown no effect.^6-10 Studies specific to oats or cereal fibers have also provided mixed results. Observational studies have noted a reduction in blood pressure (BP),¹¹ but the few clinical trials conducted to date have shown no effect.¹²

Selected whole grains are known to be good sources of soluble fibers. Previous research trials have demonstrated that these fibers can effectively reduce plasma insulin concentrations and provide other health benefits.^13,14 Additionally, elevated insulin levels have been implicated in the etiology of hypertension.¹⁵ Based on this potential biologic mechanism and the previously inconsistent findings, we conducted a 12-week trial to evaluate the clinical effects of soluble fiber–rich whole oat cereals when added to the diet of hyperinsulinemic patients medicated for essential hypertension.

Methods

Study sample

Participants were recruited from a database of treated hypertensive patients provided by a local health maintenance organization (HealthPartners). Initial letters describing the study were mailed to 8000 potential participants. Of these, 524 people responded to the mailing and agreed to a telephone screen to determine eligibility. Among respondents, 212 passed the initial phone screening and were invited to our research clinic (Hypertension and Cholesterol Research Clinic at the University of Minnesota Medical School) for a BP screening and general physical. For inclusion in the study, average screening BP readings (2 sets of readings within 7 days) taken by our team physician could not exceed 160/100. Table 1 lists exclusion criteria. The study protocol was reviewed and approved by the University of Minnesota Human Subjects Committee of the Institutional Review Board.

Eighty-eight volunteers (45 men and 43 women) aged 33 to 67 years met all inclusion criteria and provided written informed consent. All participants had a history of essential mild or moderate hypertension (BP 120/80 to 160/100 mm Hg), and were treated with no more than 1 antihypertensive medication (excluding -adrenergic receptor blocking agents) and/or 1 diuretic medication for at least 1 month before enrollment. Eighty participants were treated with a single antihypertensive medication; 8 required an antihypertensive drug and a diuretic medication to manage their BP. Individuals taking beta blockers were excluded from the study because they often take medications prescribed for more serious cardiovascular conditions, such as cardiac arrhythmias, and medication reduction would be inappropriate under such circumstances. Participants’ primary physicians were also consulted concerning participation and study-related medication changes.

TABLE 1
EXCLUSION CRITERIA

Study design

This randomized controlled parallel-group trial consisted of 3 four-week phases: a Baseline Feeding phase, a Medication Reduction phase, and a Maintenance phase. Eligible individuals were stratified by baseline systolic blood pressure (SBP) (< 140 mm Hg versus 140 mm Hg) and baseline soluble fiber intake (less than 3 grams/day versus 3 to 6 grams/day). At the start of the baseline phase, participants were randomized to either an oats cereal treatment group (n = 45) or a low-fiber cereal control group (n = 43).

The cereal treatments were isocaloric and administered during all 3 phases of the study. Individuals in the oats group received a daily serving of 60 grams (approximately three fourths cup) Quaker Oatmeal (5.61 grams total dietary fiber, 3.25 grams soluble fiber, and 2.83 grams -glucans) and 77 grams (approximately one and one third cups) Quaker Oat Squares (6.07 grams total dietary fiber, 2.98 grams soluble fiber, and 2.59 grams -glucans). Individuals in the control group consumed 65 grams (0.5 cup) Malt-O-Meal Hot Wheat Cereal (2.32 grams total dietary fiber, 0.6 grams soluble fiber) and 81 grams (2 cups) Kellogg’s Crispix (1.2 grams total dietary fiber, < 0.5 grams soluble fiber).

Cereals were dispensed in unlabeled bulk containers to facilitate physician blinding. Remaining cereal was returned and weighed at each of the weekly or biweekly visits at our research clinic. Additionally, participants kept a daily cereal calendar that was reviewed by members of our research staff and used to help determine cereal compliance.

Changes in antihypertensive medication dose were implemented according to the protocol described in the Figure. Participants were asked to maintain their usual lifestyle, physical activity, dietary pattern, and body weight during the 12 weeks of the study. Individuals were invited to participate in a 6-week follow-up phase after the intervention was completed to monitor the residual BP effect after cereal consumption was discontinued.

FIGURE
MEDICATION REDUCTION PROTOCOL

Outcomes measured

The study physician responsible for BP measurement, blood draws, and general patient examinations (described below) was unaware of the cereal group assignment. BP was measured at the clinic twice a week during the first (baseline feeding) and last (maintenance) phases of the study and weekly during the second (medication reduction) phase. Participants reported at approximately the same time of the day for all appointments. BP readings were obtained 24 hours after the last medication dose or, if the patient was unmedicated, at the same time of day as previous study BP readings and after participants had rested quietly in the seated position for at least 5 minutes in an examination room.

The study physician took all readings on the right arm, using a mercury column sphygmomanometer (Korotkoff phase V for diastolic blood pressure [DBP]). Standard cuff size was used unless upper arm circumference exceeded 31 cm, in which case a large cuff with 15 x 35–cm bladders was chosen. Measurements were repeated 4 times in 2-minute intervals. The mean of the last 3 readings was calculated and used in subsequent analyses. Baseline and final study measurements used in the analyses and reported in this paper represent the averages of the first 2 and last 2 study visits.

Preintervention and postintervention blood samples were collected into standard 6-mL serum separator tubes. Samples were analyzed within 24 hours for general chemistry and plasma lipids (total cholesterol, low-density cholesterol [LDL-C], and high-density cholesterol [HDL-C] as well as triglyceride levels) by an accredited independent laboratory and according to standard chemical methods.¹⁶

A written 42-question side effect questionnaire was administered to participants at the beginning of the baseline phase and at the end of the intervention. Participants reported the frequency with which they experienced side effects associated with increased fiber intake (eg, loose stools, flatulence) and hypertension (eg, headaches, dizziness) using a 5-item scale ranging from “never” to “very frequently” (event occurring once or several times daily). Each item of the scale was assigned a value ranging from 1 to 5. Values were tallied across all 42 questions. A final score was assigned to each participant for both time points. Mean scores by group were used in the analyses.

Participants completed a 3-day food record at baseline and at the end of the 12 weeks of intervention. Food records were examined for thoroughness by a licensed nutritionist and used to determine dietary changes. Nutrient intakes were calculated using the Nutrient Data System software (version 2.92) managed by the Nutrition Coordinating Center at the University of Minnesota School of Public Health.¹⁷

Statistics

The sample size calculation was based on a level of significance set at 0.05 and power at 80% to detect a 15% difference in medication reduction. Differences in medication reduction were determined by using the chi-square test of proportions. For continuous variables, Student’s paired and unpaired t tests were performed to determine differences within and between groups. In terms of medication reduction, logistic regression was used to adjust for potential confounders such as body weight and sodium intake. Multiple regression was employed to adjust blood lipid and glucose levels and BP findings for confounding. Because adjustment did not change the interpretation of the data, unadjusted findings are reported. The analyses of the data from this intent-to-treat population, which were determined to include all randomized patients, were conducted using the Statistical Analysis System (SAS Institute, Cary, N.C.). Results are reported as means ± SD unless noted otherwise. All P values are double sided.

Results

All the original 88 participants enrolled, all completed the 12-week trial, and all participated in the 6-week follow-up phase. Instructions to consume all dispensed cereals every day were followed well. Compliance was high for both groups (94.5% for the oat group and 92.7% for the control group) based on the amount of consumed cereal by weight. Randomization was largely effective; there were no apparent differences in baseline characteristics between each of the treatment groups (Table 2). Participants were primarily white (97%), with a mean age of 48 years (range 33 to 67 years).

BP and BP medication changes are summarized in Table 3. Among subjects in the oats group, 73% experienced a BP medication reduction during the intervention and had maintained that by the end of the study, as compared with only 42% in the control group (P < .05). Moreover, those in the oats group who did not experience a medication reduction had a 7-mm Hg decrease in SBP and a 4-mm Hg reduction in DBP. There was a small, nonsignificant change in SBP and DBP among those who did not experience a medication reduction in the control group. Medication reduction did not differ across classes of antihypertensive medication or our stratification variables of baseline soluble fiber intake or BP. Additionally, during the 6-week follow-up phase, 6 of the 18 (33%) individuals in the control group versus 22 of the 33 (67%) in the treatment group resumed taking medication.

Average BP in the oats group was lowered from 140/88 mm Hg at baseline to 134/85 mm Hg by the end of the first 4 weeks. Only the change in systolic BP was statistically significant (P < .05). Over the same 4-week period, the control group experienced a mean change of BP from 138/86 mm Hg to 136/85 mm Hg, which was not significant.

Baseline and postintervention lipid and glucose levels appear in Table 4. There were no significant modifications in any of the lipid parameters for the individuals in the control group, although there was a downward trend in all lipid measures. In the oats treatment group, mean total cholesterol (TC) concentration decreased by 31.7 mg/dL (15% drop). A similar decrease of 22.3 mg/dL (16% drop) was seen in the oats group’s average LDL-C levels. Blood glucose levels in the oats group also improved significantly (P < .01). The mean differences between post study and prestudy values (± SE) between the 2 groups, calculated for the average changes in TC, LDL-C, and glucose experienced by each of the groups, were -24.2 mg/dL (± 6.1), -16.2 mg/dL (± 4.4), and -15.03 mg/dL (± 4.3), respectively.

The frequency of dietary fiber-related and hypertension-related side effects decreased by 22% in the treatment group (Table 4). This finding was not observed in the control group. No weight changes were observed in either group, indicating that participants adjusted their diet to compensate for the addition of the cereals by substituting cereal for their standard breakfast and consuming them in place of afternoon snacks as determined by the food record inspection. Total daily energy intake (kcal/day) remained virtually unchanged when postintervention food intake was compared with intake at baseline. Participants in both groups did experience significant decreases in total fat and saturated fat intake along with significant increases in fiber (both soluble and insoluble), potassium, and calcium. The increase in total fiber intake was greater in the treatment group (P < .01) than in the control group (P < .05). In addition, the treatment group experienced a significant increase in magnesium not observed in the control group.

TABLE 2
BASELINE CHARACTERISTICS*

TABLE 3
ANTIHYPERTENSIVE MEDICATION AND BLOOD PRESSURE CHANGES BY GROUP

TABLE 4
SECONDARY OUTCOME MEASURES BY GROUP*

Discussion

The results of this trial suggest that an increased consumption of soluble fiber-rich, whole-grain, oat-based cereals can significantly reduce antihypertensive medication need among patients being treated for hypertension. Of the 45 participants in the oats group, 33 experienced at least half medication reduction compared with only 18 of the 43 participants in the control group. Positive BP changes were evident during the first 4 weeks of oat cereal treatment; BP levels rose steadily during the 6-week follow-up phase.

In addition, mean BP readings in the oat group participants who did not experience a medication reduction had improved at study completion compared with baseline. A significant number of participants in the refined cereal control group experienced at least half medication reduction (18/43), a finding that might be attributed to the increase in calcium, potassium, and total dietary fiber intake^8,9,18 as well as to the decreased intake of total and saturated fat.¹⁹ Additionally, during the follow-up phase, only 6 of the 18 (33%) versus 22 of the 33 (67%) in the oats group resumed taking their medication. Therefore, part of the medication reduction effect in the control group may have been the result of a greater percentage of participants who did not need their antihypertensive medication. This issue should be considered in the design of future trials.

As always, regression to the mean and the Hawthorne effect might explain some of the outcomes in this trial. However, it is likely that both increased soluble fiber and micronutrient intake explain the decrease in antihypertensive medication need observed in the treatment group. This study was designed to identify not the hypotensive effects of specific cereal components but the effects of a whole food intervention. Our findings are consistent with those of other whole-food interventions, such as the Dietary Approaches to Stop Hypertension (DASH) trial, tested in hypertensive populations.²⁰ Nonetheless, known diet-related determinants of BP (sodium chloride, alcohol, body weight, and level of physical activity) could not explain the treatment effect because no significant differences in these variables existed between the groups.

The soluble fiber fraction of the oat-based cereal intervention is probably partially responsible for the reduction in antihypertensive medication need observed in this trial. Previous studies that tested either soluble fiber supplements or diets rich in soluble fiber have noted significant reductions in BP.^21-23 Improvement in insulin sensitivity has been proposed as the pathway through which soluble fiber improves BP.²⁴ Insulin sensitivity was not determined in this study, yet the oats treatment group experienced a significant improvement in plasma glucose levels. This finding suggests that insulin sensitivity may have been enhanced. Impaired response to insulin was recently shown to precede endothelial dysfunction and subsequent elevations in BP.²⁵ Moreover, soluble fiber supplements and diets high in soluble fiber have been shown to improve insulin sensitivity.^25-28 Other components of whole grains, such as magnesium or grain flavonoids, may also contribute to the favorable medication reduction observed in the oats group.^29,30

This 12-week whole-food intervention trial was not designed to test either the long-term efficacy of oat-based cereals or the likelihood of long-term adherence to the feeding regimen. Nonetheless, a whole-grain oat-based cereal intervention might be an effective way to manage mild (type I) hypertension. The reduction in BP medication that occurred in the oats group was independent of weight change and sodium chloride and alcohol intake, suggesting that soluble fiber–rich whole grains should be added to the current dietary recommendations for people with elevated BP. Moreover, it is possible that the consumption of a diet high in soluble fiber–rich whole grains may prevent or delay the initiation of hypertension drug therapy in at-risk or borderline hypertensive patients. Based on the results from this study, physicians may be justified in recommending to their hypertensive patients a dietary regimen that includes the daily consumption of whole-grain oats (equaling 6 g of soluble fiber) in conjunction with their usual therapy. Such an intervention may be expected to yield results within 4 weeks.

Conclusions

A diet containing soluble fiber-rich whole grains can significantly reduce antihypertensive medication need and improve BP control among treated hypertensives. Combined with the reductions in blood lipids and plasma glucose, the intake of soluble fiber–rich whole oat cereals appears to be an effective nutritional approach in the reduction of cardiovascular disease risk. Future trials will need to investigate the antihypertensive effects of oats in other populations (eg, different racial groups) and determine whether reductions in BP measurements can be sustained for the long term.

Acknowledgment

The research team recognizes Anne Marie Weber-Main, PhD, for her excellent and tireless editorial contributions to this project.

In the past decade, bacterial resistance to commonly used antibiotics has risen dramatically.^1,2 While a number of factors have contributed to this problem, overuse of antibiotics by physicians has been implicated.^3-6 An association has been demonstrated between the volume of antibiotic prescriptions and bacterial resistance at both a national^4,5 and a local level.³ Where prescribing by physicians has been reduced, rates of antibiotic resistance have subsequently been observed to decline.^5,6 As a result, physicians have been urged to reduce their use of antibiotics.^7,8 Respiratory infections are the most common reason for the prescribing of antibiotics.⁹ Upper respiratory tract infections (URTIs) and pharyngitis account for 19% to 28% of all antibiotic prescriptions written by family physicians.^9-11

While the use of antibiotics for URTI with sore throat is frequently debated,^12-14 experts continue to recommend such treatment for group A streptococcus (GAS) infections to prevent rheumatic fever.^15,16 However, only 10% to 20% of patients with a sore throat who visit a family physician have a GAS infection,^17-19 whereas antibiotics are prescribed for 50% of URTIs¹⁰ and 90% of cases of tonsillitis.²⁰ Uncertainty as to whether or not a bacterial infection is present and clinical error in estimating the likelihood of a GAS infection are associated with the unnecessary prescription of antibiotics.^21,22 To address clinical uncertainty, a number of prediction rules and clinical scores have been proposed.^23-30 However, physicians taught simply to generate more accurate estimates of the likelihood of a strep infection in this manner do not necessarily lower their use of antibiotics.³¹

We have previously shown that linking score estimates for the likelihood of a GAS infection to explicit management recommendations to take a throat swab or prescribe an antibiotic has the potential to lower antibiotic use significantly.^19,32 In an observational study involving 621 children and adults, this approach would have reduced unnecessary antibiotic prescriptions by 63%.³² We also found a trend toward reduced antibiotic use when physicians were provided with an explicit reminder about the score approach.³³ As a result, we hypothesized that this might also help physicians to learn to adopt the sore throat score approach. Reminders have been found to improve the delivery of preventive health services.^34,35 The objective of this study was to determine whether repeated clinical prompts to community-based family physicians about the score approach could reduce unnecessary antibiotic prescriptions and lower overall antibiotic use for patients with a sore throat.

Methods

In the fall of 1998, a sample of family physicians in the province of Ontario were invited to participate in a trial to reduce antibiotic use in patients with a sore throat. Physicians who had previously participated in practice-based research projects for the College of Family Physicians of Canada and a random sample from the College’s general membership listing were contacted. Those who mailed back a reply card indicating that they wished to participate were randomized to either an intervention or a control group. The study was approved by the University of Toronto Ethics Review Committee.

Both groups of physicians received, by mail, an article describing the clinical score management approach¹⁹; a laminated pocket card summarizing the method; clinical encounter and patient consent forms; and a 1-page survey of practice characteristics. Each physician was asked to enroll 8 patients aged 3 years or older whom they believed to have a new URTI with a sore throat. No attempt was made to further define an eligible presentation to encourage physicians to enroll cases representative of their usual practice. Patients were ineligible if they had taken antibiotics during the previous week, were immunocompromised, or could not understand English. Parents were asked to provide consent for children younger than 16 years of age.

A brief standardized assessment form was completed by the physician for each patient and a throat swab was obtained. The throat swab was submitted to the physician’s local laboratory. A copy of the culture result was forwarded to the study center. Treatment decisions and the management of subsequent culture results were the responsibility of the treating physician.

In the intervention group, physicians were provided with a sticker to apply to the encounter form that listed the score management approach. The sticker contained boxes to be checked by the physicians to calculate the score total and determine appropriate management. Physicians not wishing to use the sticker were prompted on the form to write the score total in a space provided. As a result, physicians in the intervention group received repeated prompts that reminded them to use the score approach each time they completed a clinical encounter form. The control group completed a similar form but without either the sticker or the chart prompts.

The details of the clinical score approach have been previously published.^19,25 Briefly, 4 clinical findings (fever > 38°C, absence of cough, tender anterior cervical adenopathy, tonsillar swelling or exudate) and age < 15 years are each assigned 1 point and totaled. One point is subtracted for age 45 years or more. Explicit recommendations for management are linked to score totals. If the score total is 1 or less, no throat swab or antibiotic is indicated. A throat swab is recommended for a score of 2 or 3 and an antibiotic only if the culture is positive. Either initiating treatment with an antibiotic or taking a throat swab is appropriate for a score of 4 or more.

The main outcome for the study was the prescription of unnecessary antibiotics, defined as a prescription for antibiotic medication given to a patient whose subsequent throat culture was negative for group A streptococcus. The secondary outcome was overall antibiotic use. The sample size was calculated to detect a 30% decrease in unnecessary antibiotic use (2-sided = 0.05, 1- = 0.90), assuming a 40% baseline prescription rate⁹ and a 70% negative culture rate.¹⁷¹⁹ Because groups of patients were treated by the same physician, the sample size was adjusted to take the clustered sampling design into account.³⁶ The intraclass correlation coefficient for prescribing estimated from an earlier study was 0.07.¹⁹ Assuming an average of 5 patients assessed per physician, the sample size was estimated to be 85 physicians and 425 patients in each group.

The clinical characteristics of patients in the intervention and control groups were compared with a chi-square test for categorical variables and a t-test for continuous variables. Associations between prescription rates and the practice and demographic characteristics of the physicians were assessed and adjusted for the clustered sampling with Stata Statistical Software (Release 6, Stata Corp., College Station, Tex.). While clustering improves the efficiency of sampling by requiring participation by fewer physicians, confidence intervals that do not account for the design effect are too narrow. Multiple logistic regression was used to adjust for differences in patient and physician characteristics, taking into account the patient clusters by physician in estimating the effect of the intervention.

Results

One hundred sixty-four physicians agreed to participate and were randomized. Of these, only 97 (59.1%) completed the study and provided patient data (Figure). An equal proportion of physicians in the intervention group (40.2%) and control group (41.5%, P = .87) failed to complete the study. No significant differences were identified between the sex or age of physicians who participated and those who did not participate. Of the participating physicians, 86 (84.3%) returned surveys describing their practice settings.

Patients assessed included 692 children and adults. Of these, 71 (10.3%) were excluded because of a diagnosis of bronchitis (35), sinusitis (16), otitis media (11), or pneumonia (4) or because the patient was less than 3 years old (5). The score approach did not apply to the 4 conditions of exclusion because they involve organisms other than GAS. The remaining 621 patients in the control and intervention groups were similar in demographic and clinical characteristics as well as regarding the prevalence of GAS as documented by throat culture (Table 1). However, a diagnosis of tonsillitis, strep throat, or pharyngitis was more likely to occur in the intervention group (38.6%) than in the control group (28.9%, P = .01). These diagnoses were associated with a higher rate of antibiotic prescribing (54.8%) than were situations in which physicians recorded a URTI or other diagnosis (14.2%, P < .001).

Differences were noted in the characteristics of the treating physicians in each group when considered by patient encounter. Although there were no differences in the age or sex of individual physicians in each group, the participating physicians did not contribute equal numbers of patient encounters. The average number of patients assessed per physician was 3, ranging from a low of 1 patient contributed by some physicians to a high of 8 for others. More patient encounters in the intervention group were contributed by male physicians who had been in practice longer, who worked in smaller communities, and who reported larger practice volumes (Table 1). Physicians from small communities were more likely to diagnose strep throat, tonsillitis, or pharyngitis than were those in larger communities (45.1% vs 28.1%, respectively, P < .001), as were those with higher patient volumes (46.5% vs 30.2%, P = .003).

Certain physician practice characteristics were associated with a patient’s being more likely to receive a prescription for an unnecessary antibiotic (Table 2). For example, physicians were more likely to prescribe unnecessary antibiotics if they saw more than 150 patients per week than if they saw fewer and if they had been in practice for 20 or more years than if they had practiced for a shorter time. In addition, higher overall antibiotic use was associated with higher patient volume and with practicing in a smaller community.

There were no differences between the intervention and control groups in either unnecessary antibiotic prescriptions (20.4% vs 16.1%, respectively, P = .17) or overall antibiotic use (28.1% vs 27.9%, P = .96) (Table 1). However, while the culture reports that were needed to classify prescriptions as unnecessary were available for most (600) patients (96.6%), significantly more culture reports were missing in the control group (5.4%) than in the intervention group (1.2%, P = .007). Antibiotics were prescribed in 59% of the 17 cases with missing culture reports in the control group but for none of the 4 cases with missing culture reports in the intervention group.

Because intervention patients were more likely than controls to have been treated by physicians with higher prescribing characteristics, adjustments were made for the differing physician characteristics and diagnostic practices and for the clustering of patients by physician, using multiple logistic regression (Table 3). After adjustment, the intervention was associated with a nonsignificant reduction in unnecessary antibiotic prescriptions (odds ratio [OR] = 0.76, 95% confidence interval [CI] = 0.42, 1.40) and in overall antibiotic use (OR = 0.57, 95% CI = 0.27, 1.17).

TABLE 1
COMPARISON OF PATIENTS IN CONTROL AND INTERVENTION GROUPS

TABLE 2
ASSOCIATION BETWEEN INDIVIDUAL PHYSICIAN FACTORS* AND ANTIBIOTIC PRESCRIBING, ADJUSTING FOR THE CLUSTERING OF PATIENTS BY PHYSICIAN

TABLE 3
EFFECT OF REPEATED CHART PROMPTS ON PRESCRIBING RATES, ADJUSTING FOR PHYSICIAN FACTORS AND CLUSTERING* OF PATIENTS BY PHYSICIAN (N = 453†)

Figure
FAMILY PHYSICIANS WHO WERE CONTACTED AND WHO COMPLETED THE STUDY

Discussion

The use of repeated chart prompt reminders to family physicians to use a clinical scoring approach in the management of children and adults presenting with URTI and a sore throat did not affect unnecessary antibiotic prescriptions or overall antibiotic use. Problems encountered in conducting this community-based trial may have contributed to the negative result.

Sixty-seven (41%) physicians agreed to be randomized but failed to complete the study. These losses after randomization and the differing sizes of the patient clusters per physician led to differences in the characteristics of the treating physician between the 2 groups. Characteristics associated with higher antibiotic prescribing rates were more common in the intervention group. As a result, despite the randomized design, the 2 patient groups were not initially similar in terms of their likelihood to receive a prescription for an antibiotic. To compensate for these differences, we controlled for the different physician characteristics in the analysis. However, the large number of physician dropouts also resulted in a failure to achieve the planned sample size. As a result, the study had insufficient power to detect the effect size that had been hypothesized.

We had planned the sample size to detect a 30% decrease in unnecessary antibiotic use. The adjusted analysis produced a point estimate of a 23% decrease in unnecessary antibiotic use and a 43% decrease in overall antibiotic use. These point estimates are the same whether or not the clustering is taken into effect; however, the more appropriate clustered analysis increases the estimate for the sample variance, resulting in wider confidence intervals. Examination of the lower 95% confidence interval reveals that the study lacked sufficient power to rule out as much as a 58% reduction in unnecessary antibiotic use. Therefore, while the study failed to find a statistically significant effect from the intervention, it also did not have the power to rule out a clinically important reduction in unnecessary antibiotic use.

We gave information about the clinical scoring approach to physicians in the control group. Doing so may have reduced the study’s ability to detect an effect of the intervention. We did not include a group that had been not exposed to information because we believed that mailed information was the equivalent of “standard” care in terms of changing physician behavior. Mailed information is a common method of informing physicians about new clinical information but has a limited ability to influence clinical behavior.³⁴ However, the rate of antibiotic prescribing in the control group was indeed somewhat lower than is generally reported in the literature.⁹ This finding may be compatible with volunteer bias or the Hawthorne effect. More likely, perhaps, asking the control group to complete encounter forms for multiple patients may have inadvertently reminded them about the score. As a result, the control group may have been contaminated from repeated clinical prompts.

Some problems encountered in this study have been noted by other investigators conducting community-based research in primary care.³⁷ The difficulty of retaining community-based physicians resulted in significant losses after randomization. This situation occurred even though qualifying to be randomized required physicians to mail back a reply card indicating that they wished to participate, suggesting that they were motivated to some degree.³⁷ In addition, they received a modest cash honorarium. Some physicians returned the package stating that circumstances had changed and they would be unable to participate. Many who initially agreed to participate failed to reply despite 3 mailed reminders. The level of dropouts did not become apparent until late in the study. In retrospect, it might have been advisable to phone physicians directly soon after randomization in order to detect problems early. Other physicians could then have been randomly selected from the general membership listing to replace those who had dropped out.

This study found that repeated reminders to physicians to use a clinical score in the management of their patients with a sore throat did not reduce unnecessary antibiotic use. The problems encountered in this community-based intervention trial may have contributed to the negative result. Studies of prescribing behavior may need to stratify physicians before randomization by characteristics, such as patient volume and experience, that are related to prescribing behavior. Including a group that received no information is probably necessary to allow the greatest chance of detecting an effect. Particular attention and resources need to be available to ensure the retention, and replacement if needed, of community-based family physicians participating in research studies.

Acknowledgments

This study was supported by a grant from the Medical Research Council of Canada, Grant No. MA-15088. Dr McIsaac’s work is supported by the Mt. Sinai Hospital and the Family Healthcare Research Unit of the Department of Family and Community Medicine, University of Toronto, Toronto, Canada. This study was conducted in conjunction with the National Research System of the College of Family Physicians of Canada. The cooperation of the Ontario Association of Medical Laboratories is gratefully acknowledged.

In the past decade, bacterial resistance to commonly used antibiotics has risen dramatically.^1,2 While a number of factors have contributed to this problem, overuse of antibiotics by physicians has been implicated.^3-6 An association has been demonstrated between the volume of antibiotic prescriptions and bacterial resistance at both a national^4,5 and a local level.³ Where prescribing by physicians has been reduced, rates of antibiotic resistance have subsequently been observed to decline.^5,6 As a result, physicians have been urged to reduce their use of antibiotics.^7,8 Respiratory infections are the most common reason for the prescribing of antibiotics.⁹ Upper respiratory tract infections (URTIs) and pharyngitis account for 19% to 28% of all antibiotic prescriptions written by family physicians.^9-11

While the use of antibiotics for URTI with sore throat is frequently debated,^12-14 experts continue to recommend such treatment for group A streptococcus (GAS) infections to prevent rheumatic fever.^15,16 However, only 10% to 20% of patients with a sore throat who visit a family physician have a GAS infection,^17-19 whereas antibiotics are prescribed for 50% of URTIs¹⁰ and 90% of cases of tonsillitis.²⁰ Uncertainty as to whether or not a bacterial infection is present and clinical error in estimating the likelihood of a GAS infection are associated with the unnecessary prescription of antibiotics.^21,22 To address clinical uncertainty, a number of prediction rules and clinical scores have been proposed.^23-30 However, physicians taught simply to generate more accurate estimates of the likelihood of a strep infection in this manner do not necessarily lower their use of antibiotics.³¹

We have previously shown that linking score estimates for the likelihood of a GAS infection to explicit management recommendations to take a throat swab or prescribe an antibiotic has the potential to lower antibiotic use significantly.^19,32 In an observational study involving 621 children and adults, this approach would have reduced unnecessary antibiotic prescriptions by 63%.³² We also found a trend toward reduced antibiotic use when physicians were provided with an explicit reminder about the score approach.³³ As a result, we hypothesized that this might also help physicians to learn to adopt the sore throat score approach. Reminders have been found to improve the delivery of preventive health services.^34,35 The objective of this study was to determine whether repeated clinical prompts to community-based family physicians about the score approach could reduce unnecessary antibiotic prescriptions and lower overall antibiotic use for patients with a sore throat.

Methods

In the fall of 1998, a sample of family physicians in the province of Ontario were invited to participate in a trial to reduce antibiotic use in patients with a sore throat. Physicians who had previously participated in practice-based research projects for the College of Family Physicians of Canada and a random sample from the College’s general membership listing were contacted. Those who mailed back a reply card indicating that they wished to participate were randomized to either an intervention or a control group. The study was approved by the University of Toronto Ethics Review Committee.

Both groups of physicians received, by mail, an article describing the clinical score management approach¹⁹; a laminated pocket card summarizing the method; clinical encounter and patient consent forms; and a 1-page survey of practice characteristics. Each physician was asked to enroll 8 patients aged 3 years or older whom they believed to have a new URTI with a sore throat. No attempt was made to further define an eligible presentation to encourage physicians to enroll cases representative of their usual practice. Patients were ineligible if they had taken antibiotics during the previous week, were immunocompromised, or could not understand English. Parents were asked to provide consent for children younger than 16 years of age.

A brief standardized assessment form was completed by the physician for each patient and a throat swab was obtained. The throat swab was submitted to the physician’s local laboratory. A copy of the culture result was forwarded to the study center. Treatment decisions and the management of subsequent culture results were the responsibility of the treating physician.

In the intervention group, physicians were provided with a sticker to apply to the encounter form that listed the score management approach. The sticker contained boxes to be checked by the physicians to calculate the score total and determine appropriate management. Physicians not wishing to use the sticker were prompted on the form to write the score total in a space provided. As a result, physicians in the intervention group received repeated prompts that reminded them to use the score approach each time they completed a clinical encounter form. The control group completed a similar form but without either the sticker or the chart prompts.

The details of the clinical score approach have been previously published.^19,25 Briefly, 4 clinical findings (fever > 38°C, absence of cough, tender anterior cervical adenopathy, tonsillar swelling or exudate) and age < 15 years are each assigned 1 point and totaled. One point is subtracted for age 45 years or more. Explicit recommendations for management are linked to score totals. If the score total is 1 or less, no throat swab or antibiotic is indicated. A throat swab is recommended for a score of 2 or 3 and an antibiotic only if the culture is positive. Either initiating treatment with an antibiotic or taking a throat swab is appropriate for a score of 4 or more.

The main outcome for the study was the prescription of unnecessary antibiotics, defined as a prescription for antibiotic medication given to a patient whose subsequent throat culture was negative for group A streptococcus. The secondary outcome was overall antibiotic use. The sample size was calculated to detect a 30% decrease in unnecessary antibiotic use (2-sided = 0.05, 1- = 0.90), assuming a 40% baseline prescription rate⁹ and a 70% negative culture rate.¹⁷¹⁹ Because groups of patients were treated by the same physician, the sample size was adjusted to take the clustered sampling design into account.³⁶ The intraclass correlation coefficient for prescribing estimated from an earlier study was 0.07.¹⁹ Assuming an average of 5 patients assessed per physician, the sample size was estimated to be 85 physicians and 425 patients in each group.

The clinical characteristics of patients in the intervention and control groups were compared with a chi-square test for categorical variables and a t-test for continuous variables. Associations between prescription rates and the practice and demographic characteristics of the physicians were assessed and adjusted for the clustered sampling with Stata Statistical Software (Release 6, Stata Corp., College Station, Tex.). While clustering improves the efficiency of sampling by requiring participation by fewer physicians, confidence intervals that do not account for the design effect are too narrow. Multiple logistic regression was used to adjust for differences in patient and physician characteristics, taking into account the patient clusters by physician in estimating the effect of the intervention.

Results

One hundred sixty-four physicians agreed to participate and were randomized. Of these, only 97 (59.1%) completed the study and provided patient data (Figure). An equal proportion of physicians in the intervention group (40.2%) and control group (41.5%, P = .87) failed to complete the study. No significant differences were identified between the sex or age of physicians who participated and those who did not participate. Of the participating physicians, 86 (84.3%) returned surveys describing their practice settings.

Patients assessed included 692 children and adults. Of these, 71 (10.3%) were excluded because of a diagnosis of bronchitis (35), sinusitis (16), otitis media (11), or pneumonia (4) or because the patient was less than 3 years old (5). The score approach did not apply to the 4 conditions of exclusion because they involve organisms other than GAS. The remaining 621 patients in the control and intervention groups were similar in demographic and clinical characteristics as well as regarding the prevalence of GAS as documented by throat culture (Table 1). However, a diagnosis of tonsillitis, strep throat, or pharyngitis was more likely to occur in the intervention group (38.6%) than in the control group (28.9%, P = .01). These diagnoses were associated with a higher rate of antibiotic prescribing (54.8%) than were situations in which physicians recorded a URTI or other diagnosis (14.2%, P < .001).

Differences were noted in the characteristics of the treating physicians in each group when considered by patient encounter. Although there were no differences in the age or sex of individual physicians in each group, the participating physicians did not contribute equal numbers of patient encounters. The average number of patients assessed per physician was 3, ranging from a low of 1 patient contributed by some physicians to a high of 8 for others. More patient encounters in the intervention group were contributed by male physicians who had been in practice longer, who worked in smaller communities, and who reported larger practice volumes (Table 1). Physicians from small communities were more likely to diagnose strep throat, tonsillitis, or pharyngitis than were those in larger communities (45.1% vs 28.1%, respectively, P < .001), as were those with higher patient volumes (46.5% vs 30.2%, P = .003).

Certain physician practice characteristics were associated with a patient’s being more likely to receive a prescription for an unnecessary antibiotic (Table 2). For example, physicians were more likely to prescribe unnecessary antibiotics if they saw more than 150 patients per week than if they saw fewer and if they had been in practice for 20 or more years than if they had practiced for a shorter time. In addition, higher overall antibiotic use was associated with higher patient volume and with practicing in a smaller community.

There were no differences between the intervention and control groups in either unnecessary antibiotic prescriptions (20.4% vs 16.1%, respectively, P = .17) or overall antibiotic use (28.1% vs 27.9%, P = .96) (Table 1). However, while the culture reports that were needed to classify prescriptions as unnecessary were available for most (600) patients (96.6%), significantly more culture reports were missing in the control group (5.4%) than in the intervention group (1.2%, P = .007). Antibiotics were prescribed in 59% of the 17 cases with missing culture reports in the control group but for none of the 4 cases with missing culture reports in the intervention group.

Because intervention patients were more likely than controls to have been treated by physicians with higher prescribing characteristics, adjustments were made for the differing physician characteristics and diagnostic practices and for the clustering of patients by physician, using multiple logistic regression (Table 3). After adjustment, the intervention was associated with a nonsignificant reduction in unnecessary antibiotic prescriptions (odds ratio [OR] = 0.76, 95% confidence interval [CI] = 0.42, 1.40) and in overall antibiotic use (OR = 0.57, 95% CI = 0.27, 1.17).

TABLE 1
COMPARISON OF PATIENTS IN CONTROL AND INTERVENTION GROUPS

TABLE 2
ASSOCIATION BETWEEN INDIVIDUAL PHYSICIAN FACTORS* AND ANTIBIOTIC PRESCRIBING, ADJUSTING FOR THE CLUSTERING OF PATIENTS BY PHYSICIAN

TABLE 3
EFFECT OF REPEATED CHART PROMPTS ON PRESCRIBING RATES, ADJUSTING FOR PHYSICIAN FACTORS AND CLUSTERING* OF PATIENTS BY PHYSICIAN (N = 453†)

Figure
FAMILY PHYSICIANS WHO WERE CONTACTED AND WHO COMPLETED THE STUDY

Discussion

The use of repeated chart prompt reminders to family physicians to use a clinical scoring approach in the management of children and adults presenting with URTI and a sore throat did not affect unnecessary antibiotic prescriptions or overall antibiotic use. Problems encountered in conducting this community-based trial may have contributed to the negative result.

Sixty-seven (41%) physicians agreed to be randomized but failed to complete the study. These losses after randomization and the differing sizes of the patient clusters per physician led to differences in the characteristics of the treating physician between the 2 groups. Characteristics associated with higher antibiotic prescribing rates were more common in the intervention group. As a result, despite the randomized design, the 2 patient groups were not initially similar in terms of their likelihood to receive a prescription for an antibiotic. To compensate for these differences, we controlled for the different physician characteristics in the analysis. However, the large number of physician dropouts also resulted in a failure to achieve the planned sample size. As a result, the study had insufficient power to detect the effect size that had been hypothesized.

We had planned the sample size to detect a 30% decrease in unnecessary antibiotic use. The adjusted analysis produced a point estimate of a 23% decrease in unnecessary antibiotic use and a 43% decrease in overall antibiotic use. These point estimates are the same whether or not the clustering is taken into effect; however, the more appropriate clustered analysis increases the estimate for the sample variance, resulting in wider confidence intervals. Examination of the lower 95% confidence interval reveals that the study lacked sufficient power to rule out as much as a 58% reduction in unnecessary antibiotic use. Therefore, while the study failed to find a statistically significant effect from the intervention, it also did not have the power to rule out a clinically important reduction in unnecessary antibiotic use.

We gave information about the clinical scoring approach to physicians in the control group. Doing so may have reduced the study’s ability to detect an effect of the intervention. We did not include a group that had been not exposed to information because we believed that mailed information was the equivalent of “standard” care in terms of changing physician behavior. Mailed information is a common method of informing physicians about new clinical information but has a limited ability to influence clinical behavior.³⁴ However, the rate of antibiotic prescribing in the control group was indeed somewhat lower than is generally reported in the literature.⁹ This finding may be compatible with volunteer bias or the Hawthorne effect. More likely, perhaps, asking the control group to complete encounter forms for multiple patients may have inadvertently reminded them about the score. As a result, the control group may have been contaminated from repeated clinical prompts.

Some problems encountered in this study have been noted by other investigators conducting community-based research in primary care.³⁷ The difficulty of retaining community-based physicians resulted in significant losses after randomization. This situation occurred even though qualifying to be randomized required physicians to mail back a reply card indicating that they wished to participate, suggesting that they were motivated to some degree.³⁷ In addition, they received a modest cash honorarium. Some physicians returned the package stating that circumstances had changed and they would be unable to participate. Many who initially agreed to participate failed to reply despite 3 mailed reminders. The level of dropouts did not become apparent until late in the study. In retrospect, it might have been advisable to phone physicians directly soon after randomization in order to detect problems early. Other physicians could then have been randomly selected from the general membership listing to replace those who had dropped out.

This study found that repeated reminders to physicians to use a clinical score in the management of their patients with a sore throat did not reduce unnecessary antibiotic use. The problems encountered in this community-based intervention trial may have contributed to the negative result. Studies of prescribing behavior may need to stratify physicians before randomization by characteristics, such as patient volume and experience, that are related to prescribing behavior. Including a group that received no information is probably necessary to allow the greatest chance of detecting an effect. Particular attention and resources need to be available to ensure the retention, and replacement if needed, of community-based family physicians participating in research studies.

Acknowledgments

This study was supported by a grant from the Medical Research Council of Canada, Grant No. MA-15088. Dr McIsaac’s work is supported by the Mt. Sinai Hospital and the Family Healthcare Research Unit of the Department of Family and Community Medicine, University of Toronto, Toronto, Canada. This study was conducted in conjunction with the National Research System of the College of Family Physicians of Canada. The cooperation of the Ontario Association of Medical Laboratories is gratefully acknowledged.

In the past decade, bacterial resistance to commonly used antibiotics has risen dramatically.^1,2 While a number of factors have contributed to this problem, overuse of antibiotics by physicians has been implicated.^3-6 An association has been demonstrated between the volume of antibiotic prescriptions and bacterial resistance at both a national^4,5 and a local level.³ Where prescribing by physicians has been reduced, rates of antibiotic resistance have subsequently been observed to decline.^5,6 As a result, physicians have been urged to reduce their use of antibiotics.^7,8 Respiratory infections are the most common reason for the prescribing of antibiotics.⁹ Upper respiratory tract infections (URTIs) and pharyngitis account for 19% to 28% of all antibiotic prescriptions written by family physicians.^9-11

While the use of antibiotics for URTI with sore throat is frequently debated,^12-14 experts continue to recommend such treatment for group A streptococcus (GAS) infections to prevent rheumatic fever.^15,16 However, only 10% to 20% of patients with a sore throat who visit a family physician have a GAS infection,^17-19 whereas antibiotics are prescribed for 50% of URTIs¹⁰ and 90% of cases of tonsillitis.²⁰ Uncertainty as to whether or not a bacterial infection is present and clinical error in estimating the likelihood of a GAS infection are associated with the unnecessary prescription of antibiotics.^21,22 To address clinical uncertainty, a number of prediction rules and clinical scores have been proposed.^23-30 However, physicians taught simply to generate more accurate estimates of the likelihood of a strep infection in this manner do not necessarily lower their use of antibiotics.³¹

We have previously shown that linking score estimates for the likelihood of a GAS infection to explicit management recommendations to take a throat swab or prescribe an antibiotic has the potential to lower antibiotic use significantly.^19,32 In an observational study involving 621 children and adults, this approach would have reduced unnecessary antibiotic prescriptions by 63%.³² We also found a trend toward reduced antibiotic use when physicians were provided with an explicit reminder about the score approach.³³ As a result, we hypothesized that this might also help physicians to learn to adopt the sore throat score approach. Reminders have been found to improve the delivery of preventive health services.^34,35 The objective of this study was to determine whether repeated clinical prompts to community-based family physicians about the score approach could reduce unnecessary antibiotic prescriptions and lower overall antibiotic use for patients with a sore throat.

Methods

In the fall of 1998, a sample of family physicians in the province of Ontario were invited to participate in a trial to reduce antibiotic use in patients with a sore throat. Physicians who had previously participated in practice-based research projects for the College of Family Physicians of Canada and a random sample from the College’s general membership listing were contacted. Those who mailed back a reply card indicating that they wished to participate were randomized to either an intervention or a control group. The study was approved by the University of Toronto Ethics Review Committee.

Both groups of physicians received, by mail, an article describing the clinical score management approach¹⁹; a laminated pocket card summarizing the method; clinical encounter and patient consent forms; and a 1-page survey of practice characteristics. Each physician was asked to enroll 8 patients aged 3 years or older whom they believed to have a new URTI with a sore throat. No attempt was made to further define an eligible presentation to encourage physicians to enroll cases representative of their usual practice. Patients were ineligible if they had taken antibiotics during the previous week, were immunocompromised, or could not understand English. Parents were asked to provide consent for children younger than 16 years of age.

A brief standardized assessment form was completed by the physician for each patient and a throat swab was obtained. The throat swab was submitted to the physician’s local laboratory. A copy of the culture result was forwarded to the study center. Treatment decisions and the management of subsequent culture results were the responsibility of the treating physician.

In the intervention group, physicians were provided with a sticker to apply to the encounter form that listed the score management approach. The sticker contained boxes to be checked by the physicians to calculate the score total and determine appropriate management. Physicians not wishing to use the sticker were prompted on the form to write the score total in a space provided. As a result, physicians in the intervention group received repeated prompts that reminded them to use the score approach each time they completed a clinical encounter form. The control group completed a similar form but without either the sticker or the chart prompts.

The details of the clinical score approach have been previously published.^19,25 Briefly, 4 clinical findings (fever > 38°C, absence of cough, tender anterior cervical adenopathy, tonsillar swelling or exudate) and age < 15 years are each assigned 1 point and totaled. One point is subtracted for age 45 years or more. Explicit recommendations for management are linked to score totals. If the score total is 1 or less, no throat swab or antibiotic is indicated. A throat swab is recommended for a score of 2 or 3 and an antibiotic only if the culture is positive. Either initiating treatment with an antibiotic or taking a throat swab is appropriate for a score of 4 or more.

The main outcome for the study was the prescription of unnecessary antibiotics, defined as a prescription for antibiotic medication given to a patient whose subsequent throat culture was negative for group A streptococcus. The secondary outcome was overall antibiotic use. The sample size was calculated to detect a 30% decrease in unnecessary antibiotic use (2-sided = 0.05, 1- = 0.90), assuming a 40% baseline prescription rate⁹ and a 70% negative culture rate.¹⁷¹⁹ Because groups of patients were treated by the same physician, the sample size was adjusted to take the clustered sampling design into account.³⁶ The intraclass correlation coefficient for prescribing estimated from an earlier study was 0.07.¹⁹ Assuming an average of 5 patients assessed per physician, the sample size was estimated to be 85 physicians and 425 patients in each group.

The clinical characteristics of patients in the intervention and control groups were compared with a chi-square test for categorical variables and a t-test for continuous variables. Associations between prescription rates and the practice and demographic characteristics of the physicians were assessed and adjusted for the clustered sampling with Stata Statistical Software (Release 6, Stata Corp., College Station, Tex.). While clustering improves the efficiency of sampling by requiring participation by fewer physicians, confidence intervals that do not account for the design effect are too narrow. Multiple logistic regression was used to adjust for differences in patient and physician characteristics, taking into account the patient clusters by physician in estimating the effect of the intervention.

Results

One hundred sixty-four physicians agreed to participate and were randomized. Of these, only 97 (59.1%) completed the study and provided patient data (Figure). An equal proportion of physicians in the intervention group (40.2%) and control group (41.5%, P = .87) failed to complete the study. No significant differences were identified between the sex or age of physicians who participated and those who did not participate. Of the participating physicians, 86 (84.3%) returned surveys describing their practice settings.

Patients assessed included 692 children and adults. Of these, 71 (10.3%) were excluded because of a diagnosis of bronchitis (35), sinusitis (16), otitis media (11), or pneumonia (4) or because the patient was less than 3 years old (5). The score approach did not apply to the 4 conditions of exclusion because they involve organisms other than GAS. The remaining 621 patients in the control and intervention groups were similar in demographic and clinical characteristics as well as regarding the prevalence of GAS as documented by throat culture (Table 1). However, a diagnosis of tonsillitis, strep throat, or pharyngitis was more likely to occur in the intervention group (38.6%) than in the control group (28.9%, P = .01). These diagnoses were associated with a higher rate of antibiotic prescribing (54.8%) than were situations in which physicians recorded a URTI or other diagnosis (14.2%, P < .001).

Differences were noted in the characteristics of the treating physicians in each group when considered by patient encounter. Although there were no differences in the age or sex of individual physicians in each group, the participating physicians did not contribute equal numbers of patient encounters. The average number of patients assessed per physician was 3, ranging from a low of 1 patient contributed by some physicians to a high of 8 for others. More patient encounters in the intervention group were contributed by male physicians who had been in practice longer, who worked in smaller communities, and who reported larger practice volumes (Table 1). Physicians from small communities were more likely to diagnose strep throat, tonsillitis, or pharyngitis than were those in larger communities (45.1% vs 28.1%, respectively, P < .001), as were those with higher patient volumes (46.5% vs 30.2%, P = .003).

Certain physician practice characteristics were associated with a patient’s being more likely to receive a prescription for an unnecessary antibiotic (Table 2). For example, physicians were more likely to prescribe unnecessary antibiotics if they saw more than 150 patients per week than if they saw fewer and if they had been in practice for 20 or more years than if they had practiced for a shorter time. In addition, higher overall antibiotic use was associated with higher patient volume and with practicing in a smaller community.

There were no differences between the intervention and control groups in either unnecessary antibiotic prescriptions (20.4% vs 16.1%, respectively, P = .17) or overall antibiotic use (28.1% vs 27.9%, P = .96) (Table 1). However, while the culture reports that were needed to classify prescriptions as unnecessary were available for most (600) patients (96.6%), significantly more culture reports were missing in the control group (5.4%) than in the intervention group (1.2%, P = .007). Antibiotics were prescribed in 59% of the 17 cases with missing culture reports in the control group but for none of the 4 cases with missing culture reports in the intervention group.

Because intervention patients were more likely than controls to have been treated by physicians with higher prescribing characteristics, adjustments were made for the differing physician characteristics and diagnostic practices and for the clustering of patients by physician, using multiple logistic regression (Table 3). After adjustment, the intervention was associated with a nonsignificant reduction in unnecessary antibiotic prescriptions (odds ratio [OR] = 0.76, 95% confidence interval [CI] = 0.42, 1.40) and in overall antibiotic use (OR = 0.57, 95% CI = 0.27, 1.17).

TABLE 1
COMPARISON OF PATIENTS IN CONTROL AND INTERVENTION GROUPS

TABLE 2
ASSOCIATION BETWEEN INDIVIDUAL PHYSICIAN FACTORS* AND ANTIBIOTIC PRESCRIBING, ADJUSTING FOR THE CLUSTERING OF PATIENTS BY PHYSICIAN

TABLE 3
EFFECT OF REPEATED CHART PROMPTS ON PRESCRIBING RATES, ADJUSTING FOR PHYSICIAN FACTORS AND CLUSTERING* OF PATIENTS BY PHYSICIAN (N = 453†)

Figure
FAMILY PHYSICIANS WHO WERE CONTACTED AND WHO COMPLETED THE STUDY

Discussion

The use of repeated chart prompt reminders to family physicians to use a clinical scoring approach in the management of children and adults presenting with URTI and a sore throat did not affect unnecessary antibiotic prescriptions or overall antibiotic use. Problems encountered in conducting this community-based trial may have contributed to the negative result.

Sixty-seven (41%) physicians agreed to be randomized but failed to complete the study. These losses after randomization and the differing sizes of the patient clusters per physician led to differences in the characteristics of the treating physician between the 2 groups. Characteristics associated with higher antibiotic prescribing rates were more common in the intervention group. As a result, despite the randomized design, the 2 patient groups were not initially similar in terms of their likelihood to receive a prescription for an antibiotic. To compensate for these differences, we controlled for the different physician characteristics in the analysis. However, the large number of physician dropouts also resulted in a failure to achieve the planned sample size. As a result, the study had insufficient power to detect the effect size that had been hypothesized.

We had planned the sample size to detect a 30% decrease in unnecessary antibiotic use. The adjusted analysis produced a point estimate of a 23% decrease in unnecessary antibiotic use and a 43% decrease in overall antibiotic use. These point estimates are the same whether or not the clustering is taken into effect; however, the more appropriate clustered analysis increases the estimate for the sample variance, resulting in wider confidence intervals. Examination of the lower 95% confidence interval reveals that the study lacked sufficient power to rule out as much as a 58% reduction in unnecessary antibiotic use. Therefore, while the study failed to find a statistically significant effect from the intervention, it also did not have the power to rule out a clinically important reduction in unnecessary antibiotic use.

We gave information about the clinical scoring approach to physicians in the control group. Doing so may have reduced the study’s ability to detect an effect of the intervention. We did not include a group that had been not exposed to information because we believed that mailed information was the equivalent of “standard” care in terms of changing physician behavior. Mailed information is a common method of informing physicians about new clinical information but has a limited ability to influence clinical behavior.³⁴ However, the rate of antibiotic prescribing in the control group was indeed somewhat lower than is generally reported in the literature.⁹ This finding may be compatible with volunteer bias or the Hawthorne effect. More likely, perhaps, asking the control group to complete encounter forms for multiple patients may have inadvertently reminded them about the score. As a result, the control group may have been contaminated from repeated clinical prompts.

Some problems encountered in this study have been noted by other investigators conducting community-based research in primary care.³⁷ The difficulty of retaining community-based physicians resulted in significant losses after randomization. This situation occurred even though qualifying to be randomized required physicians to mail back a reply card indicating that they wished to participate, suggesting that they were motivated to some degree.³⁷ In addition, they received a modest cash honorarium. Some physicians returned the package stating that circumstances had changed and they would be unable to participate. Many who initially agreed to participate failed to reply despite 3 mailed reminders. The level of dropouts did not become apparent until late in the study. In retrospect, it might have been advisable to phone physicians directly soon after randomization in order to detect problems early. Other physicians could then have been randomly selected from the general membership listing to replace those who had dropped out.

This study found that repeated reminders to physicians to use a clinical score in the management of their patients with a sore throat did not reduce unnecessary antibiotic use. The problems encountered in this community-based intervention trial may have contributed to the negative result. Studies of prescribing behavior may need to stratify physicians before randomization by characteristics, such as patient volume and experience, that are related to prescribing behavior. Including a group that received no information is probably necessary to allow the greatest chance of detecting an effect. Particular attention and resources need to be available to ensure the retention, and replacement if needed, of community-based family physicians participating in research studies.

Acknowledgments

This study was supported by a grant from the Medical Research Council of Canada, Grant No. MA-15088. Dr McIsaac’s work is supported by the Mt. Sinai Hospital and the Family Healthcare Research Unit of the Department of Family and Community Medicine, University of Toronto, Toronto, Canada. This study was conducted in conjunction with the National Research System of the College of Family Physicians of Canada. The cooperation of the Ontario Association of Medical Laboratories is gratefully acknowledged.