Alcohol-Related Discussions in Primary Care

Brief physician intervention with problem drinkers can be effective in primary care practice: Patients who receive it are twice as likely to moderate their alcohol intake as patients in a control group.^1-4 Most problem drinkers, however, go unrecognized and untreated in medical encounters.^5,6 Two separate surveys found that only 39% of patients reported being asked by their physician about their alcohol use.^7,8 Several barriers to widespread adoption of brief physician-based intervention techniques have been hypothesized, including time constraints and physician reluctance to impose a new agenda on the patient;^9-11 but the quantitative effects of those barriers on current physician practice have not been empirically explored.

In research trials brief interventions with problem drinkers have required one or more office visits, each lasting from 5 to 15 minutes.^2,3 In those studies, patients were screened at one visit, and those identified as problem drinkers were recruited for the study by research personnel. If assigned to the intervention group, they returned for another office visit. This 2-stage approach (screening and intervention at separate visits) may miss some patients. In the Wisconsin study,² 30% of those patients willing to participate failed to complete the baseline interview, and 22% of those assigned to the intervention group did not keep the subsequent appointment. Furthermore, physicians may find the requirement of a second visit too burdensome. In a study in Scotland, half the general practitioners who were invited to participate did not join the study because they felt that a 10-minute intervention was too time consuming.¹² These problems could be partially addressed by opportunistic intervention (screening and doing the brief intervention at the same office visit), but an Australian study¹³ found even that approach unsuccessful.

A better understanding of current alcohol-related discussions in primary care may facilitate adoption of brief interventions with problem drinkers, and we know of no study that has described them. We designed our study to describe the frequency, duration, intensity, and triggers of alcohol-related discussions in primary care routine practice. Then in a before-after design, we investigated changes in alcohol-related discussions with the addition of systematic screening of all adult patients for problem drinking, and then again after the clinicians received training in brief interventions with problem drinkers.

Methods

The Ambulatory Sentinel Practice Network (ASPN) included primary care clinicians in the United States and Canada; 86% were family physicians, and most of the rest were nurse practitioners and physician assistants. ASPN conducted more than 40 studies that spanned a broad spectrum of clinical and health services research that informed both clinical practice and health care policy.¹⁴ A comparison of key characteristics of ASPN patients with a probability sample of US family physicians participating in the 1991 National Ambulatory Care Survey suggests minimal selection bias associated with voluntary participation in ASPN.¹⁵

We recruited volunteer clinicians to collect data about alcohol-involved visits for 1 week without change in their routine practice (Phase 1). Clinicians recorded data about any patient they knew had an alcohol problem or if they discussed alcohol for any reason. On a pocket-sized card,¹⁶ they noted the patient’s age and sex, the type of visit (acute self-limited illness, acute serious, chronic illness, obstetrical care, or health maintenance), whether the presenting complaint seemed related to alcohol, whether the clinician had any previous knowledge or suspicion of an alcohol problem, what triggered the alcohol-related discussion, how long it lasted (<2, 2 to 4, 4 to 8, or >8 minutes), and how intense it was for the clinician and for the patient (coded as no, mild to moderate, or marked intensity).

Clinicians who completed Phase 1 were invited to participate in a second week of data collection, this time with the addition of routine screening of all adult patients presenting for care (Phase 2). As in Phase 1, data were recorded for any patient who had a known alcohol problem, any patient with whom the clinician discussed alcohol, and any patient who had a positive screening result for problem drinking. The screening question—“When was the last time you had more than 5 drinks on one occasion?”¹⁷—was asked between questions regarding tobacco and seat belt use. An answer indicating any time within the past 3 months was considered a positive screening result for problem drinking.

We invited clinicians who participated in Phase 2 to receive training in brief interventions with problem drinkers. We mailed them a copy of The Physicians’ Guide to Helping Patients with Alcohol Problems¹⁸ and a videotaped lecture illustrating the steps in that guide. We also telephoned each participating clinician to address any concerns or perceived barriers and to practice brief intervention skills. Following training, each clinician collected card data for a third week (Phase 3), that included routine screening of all adult patients as in Phase 2. The 3 weeks of data collection were not consecutive in any practice, and each phase was not done in the same week across practices.

In comparing data from clinicians participating in more than one phase of card data collection, we assessed statistical significance with confidence interval analysis and with nonparametric tests (chi square, Fisher exact, and Mann-Whitney U) because most distributions were skewed.¹⁹

Results

A total of 114 clinicians collected data in Phase 1. During that week (which varied among practices), they saw a total of 7695 patients and had an alcohol-related discussion with 732 of them (9.5% of all visits). Among those 732 patients, the average age was 40 years (standard deviation = 18); 52.1% were women. Of the visits at which alcohol was discussed, 40% were for health maintenance, 28% for chronic illness, and 23% for acute self-limited illness. The clinician had previously seen the patient in 69% of the cases.

Nineteen clinicians participated in both Phases 1 and 2. They saw 1685 patients in Phase 1 and 1719 in Phase 2. Compared with clinicians who did not participate, more of the participants’ Phase 1 alcohol-related discussions were with patients they knew or suspected had an alcohol problem (34% vs 15%, P <.001 by chi square). Their alcohol-related discussions were longer (P = .008 by Mann-Whitney U), more likely to be prompted by their own concern (29% vs 20%) and were perceived as having greater intensity for the patient (6% perceived as having marked intensity vs 1%). However, the proportion of visits during which those 19 clinicians discussed alcohol was not significantly different from clinicians in Phase 1 who did not participate in Phase 2.

Table 1 shows comparisons of Phase 1 and Phase 2 for the 19 clinicians who completed data collection for both. Although we had anticipated a significant increase in the frequency, duration, and perceived intensity of alcohol-related discussions, the results from the first 2 phases were surprisingly similar. The addition of routine screening was associated with a shift in the triggers noted, especially from “clinician screening” to “primary prevention,” a decrease in the proportion that occurred in health maintenance visits, and an increase in the proportion in acute-care visits. Notably, of the 168 patients who had a positive screening result for problem drinking, the clinician did not discuss alcohol with 61, usually because of discussions at previous visits or a lack of time. The clinician had no previous knowledge or suspicion of an alcohol problem in 41 of those 61 instances; approximately one fourth (41 of 168) of the positive screening results were unexpected but were not addressed at the screening visit.

Of the 19 clinicians who participated in Phases 1 and 2, 15 participated in Phase 3 Table 2. We anticipated that training would increase the frequency and duration of alcohol-related discussions with patients who had a positive screening result and decrease the clinicians’ perceived discomfort with the intervention. Although these 15 clinicians were slightly more likely to address a positive screening result in Phase 3 than in Phase 2, the difference was not statistically significant. The discussions, however, were significantly longer. This was especially evident if the discussion was in response to a positive screening result: only 32% of those discussions lasted longer than 2 minutes in Phase 2, 58% in Phase 3 (P <.001). However, only 26% lasted longer than 4 minutes, and “not enough time” was checked more often in Phase 3 than in Phase 2 as a reason for not addressing a positive screening result.

In Phase 2, 70% of the screening was done by office assistants, compared with 57% in Phase 3 (P = .03). However, the screener did not significantly affect the characteristics of alcohol-related discussions, such as duration and intensity. Although 2 of the participating clinicians reported missing the screening of some patients, the clinicians reported only one patient who refused screening of the 3391 patients seen during these 2 phases of our study. Data were recorded on the pocket-sized card after each visit in approximately 40%, at the end of each day in another 40%, and at the end of the week in about 20% in both Phase 1 and 2; data were recorded somewhat closer to the visit in Phase 3 (60% after each visit, 33% each day).

Discussion

In all phases of this study, the primary care clinicians discussed alcohol with approximately 10% of their patients, usually at their own initiation. Relatively few discussions (approximately 1 out of 5) were prompted by the clinician’s concern about a possible alcohol problem, and even fewer were initiated by a patient or concerned family member. Unlike the successful 5- to 15-minute interventions published in the literature, these alcohol-related discussions were remarkably short. In Phase 1, considering only the 19 clinicians who also participated in Phase 2, 93% of the alcohol-related discussions were shorter than 4 minutes, even when they were prompted by the clinician’s concern.

We thought routine screening would increase the frequency of alcohol-related discussions, shift the focus from clinicians doing their own screening to intervening with identified problem drinkers, and make the discussions longer and more uncomfortable. But routine screening made almost no difference in alcohol-related discussions. Clinicians reported shifts in the triggers for alcohol-related discussions and the types of patient visits during which alcohol was discussed, but we found no significant changes in the duration or intensity of alcohol-related discussions.

In Phase 3, after the physicians received training in identifying and intervening with problem drinkers, alcohol-related discussions were significantly longer. When initiated in response to a positive screening result, most discussions were longer than 2 minutes, but only 26% lasted longer than 4 minutes, still shorter than effective brief interventions.^1,2

Our findings are, in some respects, generalizable to other primary care clinicians. The proportion of visits in which alcohol was discussed in our study was similar to the 9% reported by the Direct Observation of Primary Care study.²⁰ This was the first study of problem drinking by ASPN, and the patients were comparable with those seen in the National Ambulatory Medical Care Survey.¹⁵

The reasons for the lack of change with the addition of routine screening remain unclear; we suggest 3 possibilities. First, these clinicians may have already known their patients well; adding routine screening gave them little additional information. However, the most commonly cited reason for not discussing alcohol with patients who had a positive screening result was a lack of time. Second, most clinicians had their office assistant screen patients, and more discussions might have taken place had the clinicians done the screening and used a positive response as an opportunity to engage the patient in a conversation about drinking. The duration of discussions prompted by clinician concern, however, was similar across all 3 phases of our study. Third, time constraints may have limited clinicians’ ability to respond to an unexpectedly positive screening result. A lack of time was given as a reason for not discussing a positive screening result with 22 of 61 patients in Phase 2, 17 of 30 in Phase 3. Notably and also contrary to our expectations, clinician discomfort with screening-prompted discussions was no greater than with those in Phase 1, suggesting that adding routine screening seldom creates an awkward situation for the clinician.

Although training was associated with longer discussions, the changes from Phase 2 (before training) to Phase 3 (after training) were smaller than we anticipated. Further exploration is warranted, but as documented by Stange and colleagues,²⁰ an ordinary office visit deals with a multitude of issues. The lack of substantial change in alcohol-related discussions may simply be because of the crowded agenda of the primary care office visit.

We do not know the content or outcome of the alcohol-related discussions during the 3 phases of data collection or how many of those who screened positive actually had at-risk drinking, an alcohol-use disorder, both, or neither. Of those who had a negative screening result yet had an alcohol-related discussion, 5 of the 34 discussions in Phase 2 and 15 of the 44 in Phase 3 were prompted by clinician concern. We suspect some of these patients had false-negative screening results, and we do not know how many other problem drinkers were missed by the screening question.

Limitations

The 19 clinicians who participated in both Phase 1 and Phase 2 were demonstrably different from the Phase 1—only participants. Their alcohol-related discussions were longer, and more of those discussions were triggered by clinician concern. Assuming that the physicians who did not participate in Phase 2 were similar to family physicians in general, our findings are likely to show greater clinician concern about problem drinking and longer alcohol-related discussions than would be true nationally.

Also, all our data were from physician self-reports. Patient report, chart review, or observation by a third person would probably demonstrate differences. Given the lag between encounters and data recording for some clinicians, recall of alcohol-related discussions may have faded, and even those clinicians who recorded data after each visit might have missed some. However, this would be true of data collection in all 3 phases of our study, and the magnitude of bias would be similar. Furthermore, the phases were not contiguous for any practice, and we doubt there were any substantial learning or fatigue effects that might have changed the accuracy of data collection from one phase to another.

Conclusions

Alcohol use is a common topic of discussion in primary care, and the clinicians in this study usually felt comfortable addressing it. The duration of these discussions, however, was almost always shorter than the effective brief interventions with problem drinkers in all published clinical trials. Whether a 2-minute intervention would be successful is a matter for future empirical research, but it is clear that either brief interventions need to be substantially shorter, or the approach to alcohol problems in primary care needs to be changed. Furthermore, it is clear that simply adding routine screening to primary care practice is not sufficient to change clinician behavior significantly. Screening with a single question, however, was almost universally acceptable to patients and did not increase discomfort for the clinicians. And training the clinicians in recognition of and brief intervention with problem drinkers was effective in increasing the duration of the discussions.

Acknowledgments

We received funding from the Alcoholic Beverage Medical Research Foundation. During the study, Dr. Vinson was supported by a Generalist Physician Faculty Scholars Program grant from the Robert Wood Johnson Foundation.

Brief physician intervention with problem drinkers can be effective in primary care practice: Patients who receive it are twice as likely to moderate their alcohol intake as patients in a control group.^1-4 Most problem drinkers, however, go unrecognized and untreated in medical encounters.^5,6 Two separate surveys found that only 39% of patients reported being asked by their physician about their alcohol use.^7,8 Several barriers to widespread adoption of brief physician-based intervention techniques have been hypothesized, including time constraints and physician reluctance to impose a new agenda on the patient;^9-11 but the quantitative effects of those barriers on current physician practice have not been empirically explored.

In research trials brief interventions with problem drinkers have required one or more office visits, each lasting from 5 to 15 minutes.^2,3 In those studies, patients were screened at one visit, and those identified as problem drinkers were recruited for the study by research personnel. If assigned to the intervention group, they returned for another office visit. This 2-stage approach (screening and intervention at separate visits) may miss some patients. In the Wisconsin study,² 30% of those patients willing to participate failed to complete the baseline interview, and 22% of those assigned to the intervention group did not keep the subsequent appointment. Furthermore, physicians may find the requirement of a second visit too burdensome. In a study in Scotland, half the general practitioners who were invited to participate did not join the study because they felt that a 10-minute intervention was too time consuming.¹² These problems could be partially addressed by opportunistic intervention (screening and doing the brief intervention at the same office visit), but an Australian study¹³ found even that approach unsuccessful.

A better understanding of current alcohol-related discussions in primary care may facilitate adoption of brief interventions with problem drinkers, and we know of no study that has described them. We designed our study to describe the frequency, duration, intensity, and triggers of alcohol-related discussions in primary care routine practice. Then in a before-after design, we investigated changes in alcohol-related discussions with the addition of systematic screening of all adult patients for problem drinking, and then again after the clinicians received training in brief interventions with problem drinkers.

Methods

The Ambulatory Sentinel Practice Network (ASPN) included primary care clinicians in the United States and Canada; 86% were family physicians, and most of the rest were nurse practitioners and physician assistants. ASPN conducted more than 40 studies that spanned a broad spectrum of clinical and health services research that informed both clinical practice and health care policy.¹⁴ A comparison of key characteristics of ASPN patients with a probability sample of US family physicians participating in the 1991 National Ambulatory Care Survey suggests minimal selection bias associated with voluntary participation in ASPN.¹⁵

We recruited volunteer clinicians to collect data about alcohol-involved visits for 1 week without change in their routine practice (Phase 1). Clinicians recorded data about any patient they knew had an alcohol problem or if they discussed alcohol for any reason. On a pocket-sized card,¹⁶ they noted the patient’s age and sex, the type of visit (acute self-limited illness, acute serious, chronic illness, obstetrical care, or health maintenance), whether the presenting complaint seemed related to alcohol, whether the clinician had any previous knowledge or suspicion of an alcohol problem, what triggered the alcohol-related discussion, how long it lasted (<2, 2 to 4, 4 to 8, or >8 minutes), and how intense it was for the clinician and for the patient (coded as no, mild to moderate, or marked intensity).

Clinicians who completed Phase 1 were invited to participate in a second week of data collection, this time with the addition of routine screening of all adult patients presenting for care (Phase 2). As in Phase 1, data were recorded for any patient who had a known alcohol problem, any patient with whom the clinician discussed alcohol, and any patient who had a positive screening result for problem drinking. The screening question—“When was the last time you had more than 5 drinks on one occasion?”¹⁷—was asked between questions regarding tobacco and seat belt use. An answer indicating any time within the past 3 months was considered a positive screening result for problem drinking.

We invited clinicians who participated in Phase 2 to receive training in brief interventions with problem drinkers. We mailed them a copy of The Physicians’ Guide to Helping Patients with Alcohol Problems¹⁸ and a videotaped lecture illustrating the steps in that guide. We also telephoned each participating clinician to address any concerns or perceived barriers and to practice brief intervention skills. Following training, each clinician collected card data for a third week (Phase 3), that included routine screening of all adult patients as in Phase 2. The 3 weeks of data collection were not consecutive in any practice, and each phase was not done in the same week across practices.

In comparing data from clinicians participating in more than one phase of card data collection, we assessed statistical significance with confidence interval analysis and with nonparametric tests (chi square, Fisher exact, and Mann-Whitney U) because most distributions were skewed.¹⁹

Results

A total of 114 clinicians collected data in Phase 1. During that week (which varied among practices), they saw a total of 7695 patients and had an alcohol-related discussion with 732 of them (9.5% of all visits). Among those 732 patients, the average age was 40 years (standard deviation = 18); 52.1% were women. Of the visits at which alcohol was discussed, 40% were for health maintenance, 28% for chronic illness, and 23% for acute self-limited illness. The clinician had previously seen the patient in 69% of the cases.

Nineteen clinicians participated in both Phases 1 and 2. They saw 1685 patients in Phase 1 and 1719 in Phase 2. Compared with clinicians who did not participate, more of the participants’ Phase 1 alcohol-related discussions were with patients they knew or suspected had an alcohol problem (34% vs 15%, P <.001 by chi square). Their alcohol-related discussions were longer (P = .008 by Mann-Whitney U), more likely to be prompted by their own concern (29% vs 20%) and were perceived as having greater intensity for the patient (6% perceived as having marked intensity vs 1%). However, the proportion of visits during which those 19 clinicians discussed alcohol was not significantly different from clinicians in Phase 1 who did not participate in Phase 2.

Table 1 shows comparisons of Phase 1 and Phase 2 for the 19 clinicians who completed data collection for both. Although we had anticipated a significant increase in the frequency, duration, and perceived intensity of alcohol-related discussions, the results from the first 2 phases were surprisingly similar. The addition of routine screening was associated with a shift in the triggers noted, especially from “clinician screening” to “primary prevention,” a decrease in the proportion that occurred in health maintenance visits, and an increase in the proportion in acute-care visits. Notably, of the 168 patients who had a positive screening result for problem drinking, the clinician did not discuss alcohol with 61, usually because of discussions at previous visits or a lack of time. The clinician had no previous knowledge or suspicion of an alcohol problem in 41 of those 61 instances; approximately one fourth (41 of 168) of the positive screening results were unexpected but were not addressed at the screening visit.

Of the 19 clinicians who participated in Phases 1 and 2, 15 participated in Phase 3 Table 2. We anticipated that training would increase the frequency and duration of alcohol-related discussions with patients who had a positive screening result and decrease the clinicians’ perceived discomfort with the intervention. Although these 15 clinicians were slightly more likely to address a positive screening result in Phase 3 than in Phase 2, the difference was not statistically significant. The discussions, however, were significantly longer. This was especially evident if the discussion was in response to a positive screening result: only 32% of those discussions lasted longer than 2 minutes in Phase 2, 58% in Phase 3 (P <.001). However, only 26% lasted longer than 4 minutes, and “not enough time” was checked more often in Phase 3 than in Phase 2 as a reason for not addressing a positive screening result.

In Phase 2, 70% of the screening was done by office assistants, compared with 57% in Phase 3 (P = .03). However, the screener did not significantly affect the characteristics of alcohol-related discussions, such as duration and intensity. Although 2 of the participating clinicians reported missing the screening of some patients, the clinicians reported only one patient who refused screening of the 3391 patients seen during these 2 phases of our study. Data were recorded on the pocket-sized card after each visit in approximately 40%, at the end of each day in another 40%, and at the end of the week in about 20% in both Phase 1 and 2; data were recorded somewhat closer to the visit in Phase 3 (60% after each visit, 33% each day).

Discussion

In all phases of this study, the primary care clinicians discussed alcohol with approximately 10% of their patients, usually at their own initiation. Relatively few discussions (approximately 1 out of 5) were prompted by the clinician’s concern about a possible alcohol problem, and even fewer were initiated by a patient or concerned family member. Unlike the successful 5- to 15-minute interventions published in the literature, these alcohol-related discussions were remarkably short. In Phase 1, considering only the 19 clinicians who also participated in Phase 2, 93% of the alcohol-related discussions were shorter than 4 minutes, even when they were prompted by the clinician’s concern.

We thought routine screening would increase the frequency of alcohol-related discussions, shift the focus from clinicians doing their own screening to intervening with identified problem drinkers, and make the discussions longer and more uncomfortable. But routine screening made almost no difference in alcohol-related discussions. Clinicians reported shifts in the triggers for alcohol-related discussions and the types of patient visits during which alcohol was discussed, but we found no significant changes in the duration or intensity of alcohol-related discussions.

In Phase 3, after the physicians received training in identifying and intervening with problem drinkers, alcohol-related discussions were significantly longer. When initiated in response to a positive screening result, most discussions were longer than 2 minutes, but only 26% lasted longer than 4 minutes, still shorter than effective brief interventions.^1,2

Our findings are, in some respects, generalizable to other primary care clinicians. The proportion of visits in which alcohol was discussed in our study was similar to the 9% reported by the Direct Observation of Primary Care study.²⁰ This was the first study of problem drinking by ASPN, and the patients were comparable with those seen in the National Ambulatory Medical Care Survey.¹⁵

The reasons for the lack of change with the addition of routine screening remain unclear; we suggest 3 possibilities. First, these clinicians may have already known their patients well; adding routine screening gave them little additional information. However, the most commonly cited reason for not discussing alcohol with patients who had a positive screening result was a lack of time. Second, most clinicians had their office assistant screen patients, and more discussions might have taken place had the clinicians done the screening and used a positive response as an opportunity to engage the patient in a conversation about drinking. The duration of discussions prompted by clinician concern, however, was similar across all 3 phases of our study. Third, time constraints may have limited clinicians’ ability to respond to an unexpectedly positive screening result. A lack of time was given as a reason for not discussing a positive screening result with 22 of 61 patients in Phase 2, 17 of 30 in Phase 3. Notably and also contrary to our expectations, clinician discomfort with screening-prompted discussions was no greater than with those in Phase 1, suggesting that adding routine screening seldom creates an awkward situation for the clinician.

Although training was associated with longer discussions, the changes from Phase 2 (before training) to Phase 3 (after training) were smaller than we anticipated. Further exploration is warranted, but as documented by Stange and colleagues,²⁰ an ordinary office visit deals with a multitude of issues. The lack of substantial change in alcohol-related discussions may simply be because of the crowded agenda of the primary care office visit.

We do not know the content or outcome of the alcohol-related discussions during the 3 phases of data collection or how many of those who screened positive actually had at-risk drinking, an alcohol-use disorder, both, or neither. Of those who had a negative screening result yet had an alcohol-related discussion, 5 of the 34 discussions in Phase 2 and 15 of the 44 in Phase 3 were prompted by clinician concern. We suspect some of these patients had false-negative screening results, and we do not know how many other problem drinkers were missed by the screening question.

Limitations

The 19 clinicians who participated in both Phase 1 and Phase 2 were demonstrably different from the Phase 1—only participants. Their alcohol-related discussions were longer, and more of those discussions were triggered by clinician concern. Assuming that the physicians who did not participate in Phase 2 were similar to family physicians in general, our findings are likely to show greater clinician concern about problem drinking and longer alcohol-related discussions than would be true nationally.

Also, all our data were from physician self-reports. Patient report, chart review, or observation by a third person would probably demonstrate differences. Given the lag between encounters and data recording for some clinicians, recall of alcohol-related discussions may have faded, and even those clinicians who recorded data after each visit might have missed some. However, this would be true of data collection in all 3 phases of our study, and the magnitude of bias would be similar. Furthermore, the phases were not contiguous for any practice, and we doubt there were any substantial learning or fatigue effects that might have changed the accuracy of data collection from one phase to another.

Conclusions

Alcohol use is a common topic of discussion in primary care, and the clinicians in this study usually felt comfortable addressing it. The duration of these discussions, however, was almost always shorter than the effective brief interventions with problem drinkers in all published clinical trials. Whether a 2-minute intervention would be successful is a matter for future empirical research, but it is clear that either brief interventions need to be substantially shorter, or the approach to alcohol problems in primary care needs to be changed. Furthermore, it is clear that simply adding routine screening to primary care practice is not sufficient to change clinician behavior significantly. Screening with a single question, however, was almost universally acceptable to patients and did not increase discomfort for the clinicians. And training the clinicians in recognition of and brief intervention with problem drinkers was effective in increasing the duration of the discussions.

Acknowledgments

We received funding from the Alcoholic Beverage Medical Research Foundation. During the study, Dr. Vinson was supported by a Generalist Physician Faculty Scholars Program grant from the Robert Wood Johnson Foundation.

Brief physician intervention with problem drinkers can be effective in primary care practice: Patients who receive it are twice as likely to moderate their alcohol intake as patients in a control group.^1-4 Most problem drinkers, however, go unrecognized and untreated in medical encounters.^5,6 Two separate surveys found that only 39% of patients reported being asked by their physician about their alcohol use.^7,8 Several barriers to widespread adoption of brief physician-based intervention techniques have been hypothesized, including time constraints and physician reluctance to impose a new agenda on the patient;^9-11 but the quantitative effects of those barriers on current physician practice have not been empirically explored.

In research trials brief interventions with problem drinkers have required one or more office visits, each lasting from 5 to 15 minutes.^2,3 In those studies, patients were screened at one visit, and those identified as problem drinkers were recruited for the study by research personnel. If assigned to the intervention group, they returned for another office visit. This 2-stage approach (screening and intervention at separate visits) may miss some patients. In the Wisconsin study,² 30% of those patients willing to participate failed to complete the baseline interview, and 22% of those assigned to the intervention group did not keep the subsequent appointment. Furthermore, physicians may find the requirement of a second visit too burdensome. In a study in Scotland, half the general practitioners who were invited to participate did not join the study because they felt that a 10-minute intervention was too time consuming.¹² These problems could be partially addressed by opportunistic intervention (screening and doing the brief intervention at the same office visit), but an Australian study¹³ found even that approach unsuccessful.

A better understanding of current alcohol-related discussions in primary care may facilitate adoption of brief interventions with problem drinkers, and we know of no study that has described them. We designed our study to describe the frequency, duration, intensity, and triggers of alcohol-related discussions in primary care routine practice. Then in a before-after design, we investigated changes in alcohol-related discussions with the addition of systematic screening of all adult patients for problem drinking, and then again after the clinicians received training in brief interventions with problem drinkers.

Methods

The Ambulatory Sentinel Practice Network (ASPN) included primary care clinicians in the United States and Canada; 86% were family physicians, and most of the rest were nurse practitioners and physician assistants. ASPN conducted more than 40 studies that spanned a broad spectrum of clinical and health services research that informed both clinical practice and health care policy.¹⁴ A comparison of key characteristics of ASPN patients with a probability sample of US family physicians participating in the 1991 National Ambulatory Care Survey suggests minimal selection bias associated with voluntary participation in ASPN.¹⁵

We recruited volunteer clinicians to collect data about alcohol-involved visits for 1 week without change in their routine practice (Phase 1). Clinicians recorded data about any patient they knew had an alcohol problem or if they discussed alcohol for any reason. On a pocket-sized card,¹⁶ they noted the patient’s age and sex, the type of visit (acute self-limited illness, acute serious, chronic illness, obstetrical care, or health maintenance), whether the presenting complaint seemed related to alcohol, whether the clinician had any previous knowledge or suspicion of an alcohol problem, what triggered the alcohol-related discussion, how long it lasted (<2, 2 to 4, 4 to 8, or >8 minutes), and how intense it was for the clinician and for the patient (coded as no, mild to moderate, or marked intensity).

Clinicians who completed Phase 1 were invited to participate in a second week of data collection, this time with the addition of routine screening of all adult patients presenting for care (Phase 2). As in Phase 1, data were recorded for any patient who had a known alcohol problem, any patient with whom the clinician discussed alcohol, and any patient who had a positive screening result for problem drinking. The screening question—“When was the last time you had more than 5 drinks on one occasion?”¹⁷—was asked between questions regarding tobacco and seat belt use. An answer indicating any time within the past 3 months was considered a positive screening result for problem drinking.

We invited clinicians who participated in Phase 2 to receive training in brief interventions with problem drinkers. We mailed them a copy of The Physicians’ Guide to Helping Patients with Alcohol Problems¹⁸ and a videotaped lecture illustrating the steps in that guide. We also telephoned each participating clinician to address any concerns or perceived barriers and to practice brief intervention skills. Following training, each clinician collected card data for a third week (Phase 3), that included routine screening of all adult patients as in Phase 2. The 3 weeks of data collection were not consecutive in any practice, and each phase was not done in the same week across practices.

In comparing data from clinicians participating in more than one phase of card data collection, we assessed statistical significance with confidence interval analysis and with nonparametric tests (chi square, Fisher exact, and Mann-Whitney U) because most distributions were skewed.¹⁹

Results

A total of 114 clinicians collected data in Phase 1. During that week (which varied among practices), they saw a total of 7695 patients and had an alcohol-related discussion with 732 of them (9.5% of all visits). Among those 732 patients, the average age was 40 years (standard deviation = 18); 52.1% were women. Of the visits at which alcohol was discussed, 40% were for health maintenance, 28% for chronic illness, and 23% for acute self-limited illness. The clinician had previously seen the patient in 69% of the cases.

Nineteen clinicians participated in both Phases 1 and 2. They saw 1685 patients in Phase 1 and 1719 in Phase 2. Compared with clinicians who did not participate, more of the participants’ Phase 1 alcohol-related discussions were with patients they knew or suspected had an alcohol problem (34% vs 15%, P <.001 by chi square). Their alcohol-related discussions were longer (P = .008 by Mann-Whitney U), more likely to be prompted by their own concern (29% vs 20%) and were perceived as having greater intensity for the patient (6% perceived as having marked intensity vs 1%). However, the proportion of visits during which those 19 clinicians discussed alcohol was not significantly different from clinicians in Phase 1 who did not participate in Phase 2.

Table 1 shows comparisons of Phase 1 and Phase 2 for the 19 clinicians who completed data collection for both. Although we had anticipated a significant increase in the frequency, duration, and perceived intensity of alcohol-related discussions, the results from the first 2 phases were surprisingly similar. The addition of routine screening was associated with a shift in the triggers noted, especially from “clinician screening” to “primary prevention,” a decrease in the proportion that occurred in health maintenance visits, and an increase in the proportion in acute-care visits. Notably, of the 168 patients who had a positive screening result for problem drinking, the clinician did not discuss alcohol with 61, usually because of discussions at previous visits or a lack of time. The clinician had no previous knowledge or suspicion of an alcohol problem in 41 of those 61 instances; approximately one fourth (41 of 168) of the positive screening results were unexpected but were not addressed at the screening visit.

Of the 19 clinicians who participated in Phases 1 and 2, 15 participated in Phase 3 Table 2. We anticipated that training would increase the frequency and duration of alcohol-related discussions with patients who had a positive screening result and decrease the clinicians’ perceived discomfort with the intervention. Although these 15 clinicians were slightly more likely to address a positive screening result in Phase 3 than in Phase 2, the difference was not statistically significant. The discussions, however, were significantly longer. This was especially evident if the discussion was in response to a positive screening result: only 32% of those discussions lasted longer than 2 minutes in Phase 2, 58% in Phase 3 (P <.001). However, only 26% lasted longer than 4 minutes, and “not enough time” was checked more often in Phase 3 than in Phase 2 as a reason for not addressing a positive screening result.

In Phase 2, 70% of the screening was done by office assistants, compared with 57% in Phase 3 (P = .03). However, the screener did not significantly affect the characteristics of alcohol-related discussions, such as duration and intensity. Although 2 of the participating clinicians reported missing the screening of some patients, the clinicians reported only one patient who refused screening of the 3391 patients seen during these 2 phases of our study. Data were recorded on the pocket-sized card after each visit in approximately 40%, at the end of each day in another 40%, and at the end of the week in about 20% in both Phase 1 and 2; data were recorded somewhat closer to the visit in Phase 3 (60% after each visit, 33% each day).

Discussion

In all phases of this study, the primary care clinicians discussed alcohol with approximately 10% of their patients, usually at their own initiation. Relatively few discussions (approximately 1 out of 5) were prompted by the clinician’s concern about a possible alcohol problem, and even fewer were initiated by a patient or concerned family member. Unlike the successful 5- to 15-minute interventions published in the literature, these alcohol-related discussions were remarkably short. In Phase 1, considering only the 19 clinicians who also participated in Phase 2, 93% of the alcohol-related discussions were shorter than 4 minutes, even when they were prompted by the clinician’s concern.

We thought routine screening would increase the frequency of alcohol-related discussions, shift the focus from clinicians doing their own screening to intervening with identified problem drinkers, and make the discussions longer and more uncomfortable. But routine screening made almost no difference in alcohol-related discussions. Clinicians reported shifts in the triggers for alcohol-related discussions and the types of patient visits during which alcohol was discussed, but we found no significant changes in the duration or intensity of alcohol-related discussions.

In Phase 3, after the physicians received training in identifying and intervening with problem drinkers, alcohol-related discussions were significantly longer. When initiated in response to a positive screening result, most discussions were longer than 2 minutes, but only 26% lasted longer than 4 minutes, still shorter than effective brief interventions.^1,2

Our findings are, in some respects, generalizable to other primary care clinicians. The proportion of visits in which alcohol was discussed in our study was similar to the 9% reported by the Direct Observation of Primary Care study.²⁰ This was the first study of problem drinking by ASPN, and the patients were comparable with those seen in the National Ambulatory Medical Care Survey.¹⁵

The reasons for the lack of change with the addition of routine screening remain unclear; we suggest 3 possibilities. First, these clinicians may have already known their patients well; adding routine screening gave them little additional information. However, the most commonly cited reason for not discussing alcohol with patients who had a positive screening result was a lack of time. Second, most clinicians had their office assistant screen patients, and more discussions might have taken place had the clinicians done the screening and used a positive response as an opportunity to engage the patient in a conversation about drinking. The duration of discussions prompted by clinician concern, however, was similar across all 3 phases of our study. Third, time constraints may have limited clinicians’ ability to respond to an unexpectedly positive screening result. A lack of time was given as a reason for not discussing a positive screening result with 22 of 61 patients in Phase 2, 17 of 30 in Phase 3. Notably and also contrary to our expectations, clinician discomfort with screening-prompted discussions was no greater than with those in Phase 1, suggesting that adding routine screening seldom creates an awkward situation for the clinician.

Although training was associated with longer discussions, the changes from Phase 2 (before training) to Phase 3 (after training) were smaller than we anticipated. Further exploration is warranted, but as documented by Stange and colleagues,²⁰ an ordinary office visit deals with a multitude of issues. The lack of substantial change in alcohol-related discussions may simply be because of the crowded agenda of the primary care office visit.

We do not know the content or outcome of the alcohol-related discussions during the 3 phases of data collection or how many of those who screened positive actually had at-risk drinking, an alcohol-use disorder, both, or neither. Of those who had a negative screening result yet had an alcohol-related discussion, 5 of the 34 discussions in Phase 2 and 15 of the 44 in Phase 3 were prompted by clinician concern. We suspect some of these patients had false-negative screening results, and we do not know how many other problem drinkers were missed by the screening question.

Limitations

The 19 clinicians who participated in both Phase 1 and Phase 2 were demonstrably different from the Phase 1—only participants. Their alcohol-related discussions were longer, and more of those discussions were triggered by clinician concern. Assuming that the physicians who did not participate in Phase 2 were similar to family physicians in general, our findings are likely to show greater clinician concern about problem drinking and longer alcohol-related discussions than would be true nationally.

Also, all our data were from physician self-reports. Patient report, chart review, or observation by a third person would probably demonstrate differences. Given the lag between encounters and data recording for some clinicians, recall of alcohol-related discussions may have faded, and even those clinicians who recorded data after each visit might have missed some. However, this would be true of data collection in all 3 phases of our study, and the magnitude of bias would be similar. Furthermore, the phases were not contiguous for any practice, and we doubt there were any substantial learning or fatigue effects that might have changed the accuracy of data collection from one phase to another.

Conclusions

Alcohol use is a common topic of discussion in primary care, and the clinicians in this study usually felt comfortable addressing it. The duration of these discussions, however, was almost always shorter than the effective brief interventions with problem drinkers in all published clinical trials. Whether a 2-minute intervention would be successful is a matter for future empirical research, but it is clear that either brief interventions need to be substantially shorter, or the approach to alcohol problems in primary care needs to be changed. Furthermore, it is clear that simply adding routine screening to primary care practice is not sufficient to change clinician behavior significantly. Screening with a single question, however, was almost universally acceptable to patients and did not increase discomfort for the clinicians. And training the clinicians in recognition of and brief intervention with problem drinkers was effective in increasing the duration of the discussions.

Acknowledgments

We received funding from the Alcoholic Beverage Medical Research Foundation. During the study, Dr. Vinson was supported by a Generalist Physician Faculty Scholars Program grant from the Robert Wood Johnson Foundation.