Federal Practitioner

Hypertension is one of the most common cardiovascular disease (CVD) states, affecting nearly half of all adults in the United States.¹ Numerous classes of antihypertensives are available for blood pressure (BP) management, including thiazide diuretics, which contain both thiazide and thiazide-like agents. Thiazide diuretics available in the US include hydrochlorothiazide (HCTZ), chlorthalidone, metolazone, and indapamide. These agents are commonly used and recommended as first-line treatment in the current 2017 American College of Cardiology/American Heart Association (ACC/AHA) guideline for the prevention, detection, evaluation, and management of high BP in adults.²

The ACC/AHA guideline recommends chlorthalidone as the preferred thiazide diuretic.² This recommendation is based on its prolonged half-life compared with other thiazide agents, as well as the reduction of CVD seen with chlorthalidone in previous trials. The main evidence supporting chlorthalidone use comes from the ALLHAT trial, which compared chlorthalidone, amlodipine, and lisinopril in patients with hypertension. The primary composite outcome of fatal coronary artery disease or nonfatal myocardial infarction was not significantly different between groups. However, when looking at the incidence of heart failure, chlorthalidone was superior to both amlodipine and lisinopril.³ In the TOMHS trial, chlorthalidone was more effective in reducing left ventricular hypertrophy than amlodipine, enalapril, doxazosin, or acebutolol.⁴ Furthermore, both a systematic review and a retrospective cohort analysis suggested that chlorthalidone may be associated with improved CVD outcomes compared with HCTZ.^5,6 However, prospective randomized trial data is needed to confirm the superiority of chlorthalidone over other thiazide diuretics.

HCTZ has historically been the most common thiazide diuretic.⁷ However, with the available evidence and 2017 ACC/AHA BP guideline recommendations, it is unclear whether this trend continues and what impact it may have on CVD outcomes. It is unclear which thiazide diuretic is most commonly used in the US Department of Veterans Affairs (VA) health care system. The purpose of this project was to evaluate current thiazide diuretic utilization within the VA.

Methods

This retrospective, observational study evaluated the prescribing pattern of thiazide diuretics from all VA health care systems from January 1, 2016, to January 21, 2022. Thiazide diuretic agents included in this study were HCTZ, chlorthalidone, indapamide, and any combination antihypertensive products that included these 3 thiazide diuretics. Metolazone was excluded as it is commonly used in the setting of diuretic resistance with heart failure. Data was obtained from the VA Corporate Data Warehouse (CDW) and divided into 2 cohorts: the active and historic cohorts. The active cohort was of primary interest and included any active VA thiazide diuretic prescriptions on January 21, 2022. The historic cohort included thiazide prescriptions assessed at yearly intervals from January 1, 2016, to December 31, 2021. This date range was selected to assess what impact the 2017 ACC/AHA BP guideline had on clinician preferences and thiazide diuretic prescribing rates.

Within the active cohort, demographic data, vital information, and concomitant potassium or magnesium supplementation were collected. Baseline characteristics included were age, sex, race and ethnicity, and BP. Patients with > 1 race or ethnicity reported were categorized as other. The first BP reading documented after the active thiazide diuretic initiation date was included for analysis to capture on-therapy BPs while limiting confounding factors due to other potential antihypertensive changes. This project was ruled exempt from institutional review board review by the West Palm Beach VA Healthcare System Research and Development Committee.

The primary outcome was the evaluation of utilization rates of each thiazide in the active cohort, reported as a proportion of overall thiazide class utilization within the VA. Secondary outcomes in the active thiazide cohort included concomitant potassium or magnesium supplement utilization rates in each of the thiazide groups, BP values, and BP control rates. BP control was defined as a systolic BP < 130 mm Hg and a diastolic BP < 80 mm Hg. Finally, the change in thiazide diuretic utilization patterns from January 1, 2016, to December 31, 2021, was evaluated in the historic cohort.

Statistical Analysis

Data collection and analysis were completed using the CDW analyzed with Microsoft SQL Server Management Studio 18 and Microsoft Excel. All exported data to Microsoft Excel was kept in a secure network drive that was only accessible to the authors. Protected health information remained confidential per VA policy and the Health Insurance Portability and Accountability Act.

Baseline demographics were evaluated across thiazide arms using descriptive statistics. The primary outcome was assessed and a χ² test with a single comparison α level of 0.05 with Bonferroni correction to adjust for multiple comparisons when appropriate. For the secondary outcomes, analysis of continuous data was assessed using analysis of variance (ANOVA), and nominal data were assessed with a χ² test with a single comparison α level of 0.05 and Bonferroni correction to adjust for multiple comparisons where appropriate. When comparing all 3 thiazide groups, after the Bonferroni correction, P < .01667 was considered statistically significant to avoid a type 1 error in a family of statistical tests.

Results

As of January 21, 2022, the active thiazide cohort yielded 628,994 thiazide prescriptions within the VA nationwide. Most patients were male, with female patients representing 8.4%, 6.6%, and 5.6% of the HCTZ, chlorthalidone, and indapamide arms, respectively (Table 1). Utilization rates were significantly different between thiazide groups (P < .001). HCTZ was the most prescribed thiazide diuretic (84.6%) followed by chlorthalidone (14.9%) and indapamide (0.5%) (Table 2).

BP values documented after prescription initiation date were available for few individuals in the HCTZ, chlorthalidone, and indapamide groups (0.3%, 0.2%, and 0.5%, respectively). Overall, the mean BP values were similar among thiazide groups: 135/79 mm Hg for HCTZ, 137/78 mm Hg for chlorthalidone, and 133/79 mm Hg for indapamide (P = .32). BP control was also similar with control rates of 26.0%, 27.1%, and 33.3% for those on HCTZ, chlorthalidone, and indapamide, respectively (P = .75). The use of concomitant potassium or magnesium supplementation was significantly different between thiazide groups with rates of 12.4%, 22.6%, and 27.1% for HCTZ, chlorthalidone, and indapamide, respectively (P < .001). When comparing chlorthalidone to HCTZ, there was a significantly higher rate of concomitant supplementation with chlorthalidone (P < .001) (Table 3).

In the historic cohort, HCTZ utilization decreased from 90.2% to 83.5% (P < .001) and chlorthalidone utilization increased significantly from 9.3% to 16.0% (P < .001) (Figure). There was no significant change in the use of indapamide during this period (P = .73). Yearly trends from 2016 to 2021 are listed in Table 4.

Discussion

The findings of our evaluation demonstrate that despite the 2017 ACC/AHA BP guideline recommendations for using chlorthalidone, HCTZ predominates as the most prescribed thiazide diuretic within the VA. However, since the publication of this guideline, there has been an increase in chlorthalidone prescribing and a decrease in HCTZ prescribing within the VA.

A 2010 study by Ernst and colleagues revealed a similar trend to what was seen in our study. At that time, HCTZ was the most prescribed thiazide encompassing 95% of total thiazide utilization; however, chlorthalidone utilization increased from 1.1% in 2003 to 2.4% in 2008.⁸ In comparing our chlorthalidone utilization rates with these results, 9.3% in 2016 and 16.0% in 2021, the change in chlorthalidone prescribing from 2003 to 2016 represents a more than linear increase. This trend continued in our study from 2016 to 2021; the expected chlorthalidone utilization would be 21.2% in 2021 if it followed the 2003 to 2016 rate of change. Thus the trend in increasing chlorthalidone use predated the 2017 guideline recommendation. Nonetheless, this change in the thiazide prescribing pattern represents a positive shift in practice.

Our evaluation found a significantly higher rate of concomitant potassium or magnesium supplementation with chlorthalidone and indapamide compared with HCTZ in the active cohort. Electrolyte abnormalities are well documented adverse effects associated with thiazide diuretic use.⁹ A cross-sectional analysis by Ravioli and colleagues revealed thiazide diuretic use was an independent predictor of both hyponatremia (22.1% incidence) and hypokalemia (19% incidence) and that chlorthalidone was associated with the highest risk of electrolyte abnormalities whereas HCTZ was associated with the lowest risk. Their study also found these electrolyte abnormalities to have a dose-dependent relationship with the thiazide diuretic prescribed.¹⁰

While Ravioli and colleagues did not address the incidence of hypomagnesemia with thiazide diuretic use, a cross-sectional analysis by Kieboom and colleagues reported a significant increase in hypomagnesemia in patients prescribed thiazide diuretics.¹¹ Although rates of electrolyte abnormalities are reported in the literature, the rates of concomitant supplementation are unclear, especially when compared across thiazide agents. Our study provides insight into the use of concomitant potassium and magnesium supplementation compared between HCTZ, chlorthalidone, and indapamide. In our active cohort, potassium was more commonly prescribed than magnesium. Interestingly, magnesium supplementation accounted for 25.9% of the total supplement use for HCTZ compared with rates of 22.4% and 21.0% for chlorthalidone and indapamide, respectively. It is unclear if this trend highlights a greater incidence of hypomagnesemia with HCTZ or greater clinician awareness to monitor this agent, but this finding may warrant further investigation. In addition, when considering the overall lower rate of supplementation seen with HCTZ in our study, the use of potassium-sparing diuretics should be considered. These agents, including triamterene, amiloride, eplerenone, and spironolactone, can be supplement-sparing and are available in combination products only with HCTZ.

Low chlorthalidone utilization rates are concerning especially given the literature demonstrating CVD benefit with chlorthalidone and the lack of compelling outcomes data to support HCTZ as the preferred agent.^3,4 There are several reasons why HCTZ use may be higher in practice. First is clinical inertia, which is defined as a lack of treatment intensification or lack of changing practice patterns, despite evidence-based goals of care.¹² HCTZ has been the most widely prescribed thiazide diuretic for years.⁷ As a result, converting HCTZ to chlorthalidone for a patient with suboptimal BP control may not be considered and instead clinicians may add on another antihypertensive or titrate doses of current antihypertensives.

There is also a consideration for patient adherence. HCTZ has many more combination products available than chlorthalidone and indapamide. If switching a patient from an HCTZ-containing combination product to chlorthalidone, adherence and patient willingness to take another capsule or tablet must be considered. Finally, there may be clinical controversy and questions around switching patients from HCTZ to chlorthalidone. Although the guidelines do not explicitly recommend switching to chlorthalidone, it may be reasonable in most patients unless they have or are at significant risk of electrolyte or metabolic disturbances that may be exacerbated or triggered with conversion.

When converting from HCTZ to chlorthalidone, it is important to consider dosing. Previous studies have demonstrated that chlorthalidone is 1.5 to 2 times more potent than HCTZ.^13,14 Therefore, the conversion from HCTZ to chlorthalidone is not 1:1, but instead 50 mg of HCTZ is approximately equal to25 to 37.5 mg of chlorthalidone.¹⁴

Limitations

This study was limited by its retrospective design, gaps in data, duplicate active prescription data, and the assessment of concomitant electrolyte supplementation. As with any retrospective study, there is a potential for confounding and a concern for information bias with missing information. This study relied on proper documentation of prescription and demographic information in the Veterans Health Information Systems and Technology Architecture (VistA), as the CDW compiles information from this electronic health record. Strengths of the VistA include ease in clinical functions, documentation, and the ability for records to be updated from any VA facility nationally. However, there is always the possibility of user error and information to be omitted.

In our study, the documentation of BP values and subsequent analysis of overall BP control were limited. For BP values to be included in this study, they had to be recorded after the active thiazide prescription was written and from an in-person encounter documented in VistA. The COVID-19 pandemic shifted the clinical landscape and many primary care appointments during the active cohort evaluation period were conducted virtually. Therefore, patients may not have had formal vitals recorded. There may also be an aspect of selection bias regarding the chlorthalidone group. Although rates of thiazide switching were not assessed, some patients may have been switched from HCTZ or indapamide to chlorthalidone to achieve additional BP control. Thus, patients receiving chlorthalidone may represent a more difficult-to-control hypertensive population, making a finding of similar BP control rates between HCTZ and chlorthalidone an actual positive finding regarding chlorthalidone. Finally, this study did not assess adherence to medications. As the intent of the study was to analyze prescribing patterns, it is impossible to know if the patient was actively taking the medication at the time of assessment. When considering the rates of BP control, there were limited BP values, a potential for selection bias, and neither adherence nor patient self-reported home BP values were assessed. Therefore, the interpretation of overall BP control must be done with caution.

Additionally, duplicate prescriptions were noted in the active cohort. Rates of duplication were 0.2%, 0.08%, and 0.09% for HCTZ, chlorthalidone, and indapamide, respectively. With these small percentages, we felt this would not have a significant impact on the overall thiazide use trends seen in our study. Patients can receive prescriptions from multiple VA facilities and may have > 1 active prescriptions. This has been mitigated in recent years with the introduction of the OneVA program, allowing pharmacists to access any prescription on file from any VA facility and refill if needed (except controlled substance prescriptions). However, there are certain instances in which duplicate prescriptions may be necessary. These include patients enrolled and receiving care at another VA facility (eg, traveling for part of a year) and patients hospitalized at a different facility and given medications on discharge.

With the overall low rate of duplication prescriptions seen in each thiazide group, we determined that this was not large enough to cause substantial variation in the results of this evaluation and was unlikely to alter the results. This study also does not inform on the incidence of switching between thiazide diuretics. If a patient was switched from HCTZ to chlorthalidone in 2017, for example, a prescription for HCTZ and chlorthalidone would have been reported in this study. We felt that the change in chlorthalidone prescribing from January 1, 2016, to December 31, 2021, would reflect overall utilization rates, which may include switching from HCTZ or indapamide to chlorthalidone in addition to new chlorthalidone prescriptions.

Finally, there are confounders and trends in concomitant potassium or magnesium supplementation that were not accounted for in our study. These include concomitant loop diuretics or other medications that may cause electrolyte abnormalities and the dose-dependent relationship between thiazide diuretics and electrolyte abnormalities.¹⁰ Actual laboratory values were not included in this analysis and thus we cannot assess whether supplementation or management of electrolyte disturbances was clinically appropriate.

Conclusions

Thiazide utilization patterns have shifted possibly due to the 2017 ACC/AHA BP guideline recommendations. However, HCTZ continues to be the most widely prescribed thiazide diuretic within the VA. There is a need for future projects and clinician education to increase the implementation of guideline-recommended therapy within the VA, particularly regarding chlorthalidone use.

Hypertension is one of the most common cardiovascular disease (CVD) states, affecting nearly half of all adults in the United States.¹ Numerous classes of antihypertensives are available for blood pressure (BP) management, including thiazide diuretics, which contain both thiazide and thiazide-like agents. Thiazide diuretics available in the US include hydrochlorothiazide (HCTZ), chlorthalidone, metolazone, and indapamide. These agents are commonly used and recommended as first-line treatment in the current 2017 American College of Cardiology/American Heart Association (ACC/AHA) guideline for the prevention, detection, evaluation, and management of high BP in adults.²

The ACC/AHA guideline recommends chlorthalidone as the preferred thiazide diuretic.² This recommendation is based on its prolonged half-life compared with other thiazide agents, as well as the reduction of CVD seen with chlorthalidone in previous trials. The main evidence supporting chlorthalidone use comes from the ALLHAT trial, which compared chlorthalidone, amlodipine, and lisinopril in patients with hypertension. The primary composite outcome of fatal coronary artery disease or nonfatal myocardial infarction was not significantly different between groups. However, when looking at the incidence of heart failure, chlorthalidone was superior to both amlodipine and lisinopril.³ In the TOMHS trial, chlorthalidone was more effective in reducing left ventricular hypertrophy than amlodipine, enalapril, doxazosin, or acebutolol.⁴ Furthermore, both a systematic review and a retrospective cohort analysis suggested that chlorthalidone may be associated with improved CVD outcomes compared with HCTZ.^5,6 However, prospective randomized trial data is needed to confirm the superiority of chlorthalidone over other thiazide diuretics.

HCTZ has historically been the most common thiazide diuretic.⁷ However, with the available evidence and 2017 ACC/AHA BP guideline recommendations, it is unclear whether this trend continues and what impact it may have on CVD outcomes. It is unclear which thiazide diuretic is most commonly used in the US Department of Veterans Affairs (VA) health care system. The purpose of this project was to evaluate current thiazide diuretic utilization within the VA.

Methods

This retrospective, observational study evaluated the prescribing pattern of thiazide diuretics from all VA health care systems from January 1, 2016, to January 21, 2022. Thiazide diuretic agents included in this study were HCTZ, chlorthalidone, indapamide, and any combination antihypertensive products that included these 3 thiazide diuretics. Metolazone was excluded as it is commonly used in the setting of diuretic resistance with heart failure. Data was obtained from the VA Corporate Data Warehouse (CDW) and divided into 2 cohorts: the active and historic cohorts. The active cohort was of primary interest and included any active VA thiazide diuretic prescriptions on January 21, 2022. The historic cohort included thiazide prescriptions assessed at yearly intervals from January 1, 2016, to December 31, 2021. This date range was selected to assess what impact the 2017 ACC/AHA BP guideline had on clinician preferences and thiazide diuretic prescribing rates.

Within the active cohort, demographic data, vital information, and concomitant potassium or magnesium supplementation were collected. Baseline characteristics included were age, sex, race and ethnicity, and BP. Patients with > 1 race or ethnicity reported were categorized as other. The first BP reading documented after the active thiazide diuretic initiation date was included for analysis to capture on-therapy BPs while limiting confounding factors due to other potential antihypertensive changes. This project was ruled exempt from institutional review board review by the West Palm Beach VA Healthcare System Research and Development Committee.

The primary outcome was the evaluation of utilization rates of each thiazide in the active cohort, reported as a proportion of overall thiazide class utilization within the VA. Secondary outcomes in the active thiazide cohort included concomitant potassium or magnesium supplement utilization rates in each of the thiazide groups, BP values, and BP control rates. BP control was defined as a systolic BP < 130 mm Hg and a diastolic BP < 80 mm Hg. Finally, the change in thiazide diuretic utilization patterns from January 1, 2016, to December 31, 2021, was evaluated in the historic cohort.

Statistical Analysis

Data collection and analysis were completed using the CDW analyzed with Microsoft SQL Server Management Studio 18 and Microsoft Excel. All exported data to Microsoft Excel was kept in a secure network drive that was only accessible to the authors. Protected health information remained confidential per VA policy and the Health Insurance Portability and Accountability Act.

Baseline demographics were evaluated across thiazide arms using descriptive statistics. The primary outcome was assessed and a χ² test with a single comparison α level of 0.05 with Bonferroni correction to adjust for multiple comparisons when appropriate. For the secondary outcomes, analysis of continuous data was assessed using analysis of variance (ANOVA), and nominal data were assessed with a χ² test with a single comparison α level of 0.05 and Bonferroni correction to adjust for multiple comparisons where appropriate. When comparing all 3 thiazide groups, after the Bonferroni correction, P < .01667 was considered statistically significant to avoid a type 1 error in a family of statistical tests.

Results

As of January 21, 2022, the active thiazide cohort yielded 628,994 thiazide prescriptions within the VA nationwide. Most patients were male, with female patients representing 8.4%, 6.6%, and 5.6% of the HCTZ, chlorthalidone, and indapamide arms, respectively (Table 1). Utilization rates were significantly different between thiazide groups (P < .001). HCTZ was the most prescribed thiazide diuretic (84.6%) followed by chlorthalidone (14.9%) and indapamide (0.5%) (Table 2).

BP values documented after prescription initiation date were available for few individuals in the HCTZ, chlorthalidone, and indapamide groups (0.3%, 0.2%, and 0.5%, respectively). Overall, the mean BP values were similar among thiazide groups: 135/79 mm Hg for HCTZ, 137/78 mm Hg for chlorthalidone, and 133/79 mm Hg for indapamide (P = .32). BP control was also similar with control rates of 26.0%, 27.1%, and 33.3% for those on HCTZ, chlorthalidone, and indapamide, respectively (P = .75). The use of concomitant potassium or magnesium supplementation was significantly different between thiazide groups with rates of 12.4%, 22.6%, and 27.1% for HCTZ, chlorthalidone, and indapamide, respectively (P < .001). When comparing chlorthalidone to HCTZ, there was a significantly higher rate of concomitant supplementation with chlorthalidone (P < .001) (Table 3).

In the historic cohort, HCTZ utilization decreased from 90.2% to 83.5% (P < .001) and chlorthalidone utilization increased significantly from 9.3% to 16.0% (P < .001) (Figure). There was no significant change in the use of indapamide during this period (P = .73). Yearly trends from 2016 to 2021 are listed in Table 4.

Discussion

The findings of our evaluation demonstrate that despite the 2017 ACC/AHA BP guideline recommendations for using chlorthalidone, HCTZ predominates as the most prescribed thiazide diuretic within the VA. However, since the publication of this guideline, there has been an increase in chlorthalidone prescribing and a decrease in HCTZ prescribing within the VA.

A 2010 study by Ernst and colleagues revealed a similar trend to what was seen in our study. At that time, HCTZ was the most prescribed thiazide encompassing 95% of total thiazide utilization; however, chlorthalidone utilization increased from 1.1% in 2003 to 2.4% in 2008.⁸ In comparing our chlorthalidone utilization rates with these results, 9.3% in 2016 and 16.0% in 2021, the change in chlorthalidone prescribing from 2003 to 2016 represents a more than linear increase. This trend continued in our study from 2016 to 2021; the expected chlorthalidone utilization would be 21.2% in 2021 if it followed the 2003 to 2016 rate of change. Thus the trend in increasing chlorthalidone use predated the 2017 guideline recommendation. Nonetheless, this change in the thiazide prescribing pattern represents a positive shift in practice.

Our evaluation found a significantly higher rate of concomitant potassium or magnesium supplementation with chlorthalidone and indapamide compared with HCTZ in the active cohort. Electrolyte abnormalities are well documented adverse effects associated with thiazide diuretic use.⁹ A cross-sectional analysis by Ravioli and colleagues revealed thiazide diuretic use was an independent predictor of both hyponatremia (22.1% incidence) and hypokalemia (19% incidence) and that chlorthalidone was associated with the highest risk of electrolyte abnormalities whereas HCTZ was associated with the lowest risk. Their study also found these electrolyte abnormalities to have a dose-dependent relationship with the thiazide diuretic prescribed.¹⁰

While Ravioli and colleagues did not address the incidence of hypomagnesemia with thiazide diuretic use, a cross-sectional analysis by Kieboom and colleagues reported a significant increase in hypomagnesemia in patients prescribed thiazide diuretics.¹¹ Although rates of electrolyte abnormalities are reported in the literature, the rates of concomitant supplementation are unclear, especially when compared across thiazide agents. Our study provides insight into the use of concomitant potassium and magnesium supplementation compared between HCTZ, chlorthalidone, and indapamide. In our active cohort, potassium was more commonly prescribed than magnesium. Interestingly, magnesium supplementation accounted for 25.9% of the total supplement use for HCTZ compared with rates of 22.4% and 21.0% for chlorthalidone and indapamide, respectively. It is unclear if this trend highlights a greater incidence of hypomagnesemia with HCTZ or greater clinician awareness to monitor this agent, but this finding may warrant further investigation. In addition, when considering the overall lower rate of supplementation seen with HCTZ in our study, the use of potassium-sparing diuretics should be considered. These agents, including triamterene, amiloride, eplerenone, and spironolactone, can be supplement-sparing and are available in combination products only with HCTZ.

Low chlorthalidone utilization rates are concerning especially given the literature demonstrating CVD benefit with chlorthalidone and the lack of compelling outcomes data to support HCTZ as the preferred agent.^3,4 There are several reasons why HCTZ use may be higher in practice. First is clinical inertia, which is defined as a lack of treatment intensification or lack of changing practice patterns, despite evidence-based goals of care.¹² HCTZ has been the most widely prescribed thiazide diuretic for years.⁷ As a result, converting HCTZ to chlorthalidone for a patient with suboptimal BP control may not be considered and instead clinicians may add on another antihypertensive or titrate doses of current antihypertensives.

There is also a consideration for patient adherence. HCTZ has many more combination products available than chlorthalidone and indapamide. If switching a patient from an HCTZ-containing combination product to chlorthalidone, adherence and patient willingness to take another capsule or tablet must be considered. Finally, there may be clinical controversy and questions around switching patients from HCTZ to chlorthalidone. Although the guidelines do not explicitly recommend switching to chlorthalidone, it may be reasonable in most patients unless they have or are at significant risk of electrolyte or metabolic disturbances that may be exacerbated or triggered with conversion.

When converting from HCTZ to chlorthalidone, it is important to consider dosing. Previous studies have demonstrated that chlorthalidone is 1.5 to 2 times more potent than HCTZ.^13,14 Therefore, the conversion from HCTZ to chlorthalidone is not 1:1, but instead 50 mg of HCTZ is approximately equal to25 to 37.5 mg of chlorthalidone.¹⁴

Limitations

This study was limited by its retrospective design, gaps in data, duplicate active prescription data, and the assessment of concomitant electrolyte supplementation. As with any retrospective study, there is a potential for confounding and a concern for information bias with missing information. This study relied on proper documentation of prescription and demographic information in the Veterans Health Information Systems and Technology Architecture (VistA), as the CDW compiles information from this electronic health record. Strengths of the VistA include ease in clinical functions, documentation, and the ability for records to be updated from any VA facility nationally. However, there is always the possibility of user error and information to be omitted.

In our study, the documentation of BP values and subsequent analysis of overall BP control were limited. For BP values to be included in this study, they had to be recorded after the active thiazide prescription was written and from an in-person encounter documented in VistA. The COVID-19 pandemic shifted the clinical landscape and many primary care appointments during the active cohort evaluation period were conducted virtually. Therefore, patients may not have had formal vitals recorded. There may also be an aspect of selection bias regarding the chlorthalidone group. Although rates of thiazide switching were not assessed, some patients may have been switched from HCTZ or indapamide to chlorthalidone to achieve additional BP control. Thus, patients receiving chlorthalidone may represent a more difficult-to-control hypertensive population, making a finding of similar BP control rates between HCTZ and chlorthalidone an actual positive finding regarding chlorthalidone. Finally, this study did not assess adherence to medications. As the intent of the study was to analyze prescribing patterns, it is impossible to know if the patient was actively taking the medication at the time of assessment. When considering the rates of BP control, there were limited BP values, a potential for selection bias, and neither adherence nor patient self-reported home BP values were assessed. Therefore, the interpretation of overall BP control must be done with caution.

Additionally, duplicate prescriptions were noted in the active cohort. Rates of duplication were 0.2%, 0.08%, and 0.09% for HCTZ, chlorthalidone, and indapamide, respectively. With these small percentages, we felt this would not have a significant impact on the overall thiazide use trends seen in our study. Patients can receive prescriptions from multiple VA facilities and may have > 1 active prescriptions. This has been mitigated in recent years with the introduction of the OneVA program, allowing pharmacists to access any prescription on file from any VA facility and refill if needed (except controlled substance prescriptions). However, there are certain instances in which duplicate prescriptions may be necessary. These include patients enrolled and receiving care at another VA facility (eg, traveling for part of a year) and patients hospitalized at a different facility and given medications on discharge.

With the overall low rate of duplication prescriptions seen in each thiazide group, we determined that this was not large enough to cause substantial variation in the results of this evaluation and was unlikely to alter the results. This study also does not inform on the incidence of switching between thiazide diuretics. If a patient was switched from HCTZ to chlorthalidone in 2017, for example, a prescription for HCTZ and chlorthalidone would have been reported in this study. We felt that the change in chlorthalidone prescribing from January 1, 2016, to December 31, 2021, would reflect overall utilization rates, which may include switching from HCTZ or indapamide to chlorthalidone in addition to new chlorthalidone prescriptions.

Finally, there are confounders and trends in concomitant potassium or magnesium supplementation that were not accounted for in our study. These include concomitant loop diuretics or other medications that may cause electrolyte abnormalities and the dose-dependent relationship between thiazide diuretics and electrolyte abnormalities.¹⁰ Actual laboratory values were not included in this analysis and thus we cannot assess whether supplementation or management of electrolyte disturbances was clinically appropriate.

Conclusions

Thiazide utilization patterns have shifted possibly due to the 2017 ACC/AHA BP guideline recommendations. However, HCTZ continues to be the most widely prescribed thiazide diuretic within the VA. There is a need for future projects and clinician education to increase the implementation of guideline-recommended therapy within the VA, particularly regarding chlorthalidone use.

Hypertension is one of the most common cardiovascular disease (CVD) states, affecting nearly half of all adults in the United States.¹ Numerous classes of antihypertensives are available for blood pressure (BP) management, including thiazide diuretics, which contain both thiazide and thiazide-like agents. Thiazide diuretics available in the US include hydrochlorothiazide (HCTZ), chlorthalidone, metolazone, and indapamide. These agents are commonly used and recommended as first-line treatment in the current 2017 American College of Cardiology/American Heart Association (ACC/AHA) guideline for the prevention, detection, evaluation, and management of high BP in adults.²

The ACC/AHA guideline recommends chlorthalidone as the preferred thiazide diuretic.² This recommendation is based on its prolonged half-life compared with other thiazide agents, as well as the reduction of CVD seen with chlorthalidone in previous trials. The main evidence supporting chlorthalidone use comes from the ALLHAT trial, which compared chlorthalidone, amlodipine, and lisinopril in patients with hypertension. The primary composite outcome of fatal coronary artery disease or nonfatal myocardial infarction was not significantly different between groups. However, when looking at the incidence of heart failure, chlorthalidone was superior to both amlodipine and lisinopril.³ In the TOMHS trial, chlorthalidone was more effective in reducing left ventricular hypertrophy than amlodipine, enalapril, doxazosin, or acebutolol.⁴ Furthermore, both a systematic review and a retrospective cohort analysis suggested that chlorthalidone may be associated with improved CVD outcomes compared with HCTZ.^5,6 However, prospective randomized trial data is needed to confirm the superiority of chlorthalidone over other thiazide diuretics.

HCTZ has historically been the most common thiazide diuretic.⁷ However, with the available evidence and 2017 ACC/AHA BP guideline recommendations, it is unclear whether this trend continues and what impact it may have on CVD outcomes. It is unclear which thiazide diuretic is most commonly used in the US Department of Veterans Affairs (VA) health care system. The purpose of this project was to evaluate current thiazide diuretic utilization within the VA.

Methods

This retrospective, observational study evaluated the prescribing pattern of thiazide diuretics from all VA health care systems from January 1, 2016, to January 21, 2022. Thiazide diuretic agents included in this study were HCTZ, chlorthalidone, indapamide, and any combination antihypertensive products that included these 3 thiazide diuretics. Metolazone was excluded as it is commonly used in the setting of diuretic resistance with heart failure. Data was obtained from the VA Corporate Data Warehouse (CDW) and divided into 2 cohorts: the active and historic cohorts. The active cohort was of primary interest and included any active VA thiazide diuretic prescriptions on January 21, 2022. The historic cohort included thiazide prescriptions assessed at yearly intervals from January 1, 2016, to December 31, 2021. This date range was selected to assess what impact the 2017 ACC/AHA BP guideline had on clinician preferences and thiazide diuretic prescribing rates.

Within the active cohort, demographic data, vital information, and concomitant potassium or magnesium supplementation were collected. Baseline characteristics included were age, sex, race and ethnicity, and BP. Patients with > 1 race or ethnicity reported were categorized as other. The first BP reading documented after the active thiazide diuretic initiation date was included for analysis to capture on-therapy BPs while limiting confounding factors due to other potential antihypertensive changes. This project was ruled exempt from institutional review board review by the West Palm Beach VA Healthcare System Research and Development Committee.

The primary outcome was the evaluation of utilization rates of each thiazide in the active cohort, reported as a proportion of overall thiazide class utilization within the VA. Secondary outcomes in the active thiazide cohort included concomitant potassium or magnesium supplement utilization rates in each of the thiazide groups, BP values, and BP control rates. BP control was defined as a systolic BP < 130 mm Hg and a diastolic BP < 80 mm Hg. Finally, the change in thiazide diuretic utilization patterns from January 1, 2016, to December 31, 2021, was evaluated in the historic cohort.

Statistical Analysis

Data collection and analysis were completed using the CDW analyzed with Microsoft SQL Server Management Studio 18 and Microsoft Excel. All exported data to Microsoft Excel was kept in a secure network drive that was only accessible to the authors. Protected health information remained confidential per VA policy and the Health Insurance Portability and Accountability Act.

Baseline demographics were evaluated across thiazide arms using descriptive statistics. The primary outcome was assessed and a χ² test with a single comparison α level of 0.05 with Bonferroni correction to adjust for multiple comparisons when appropriate. For the secondary outcomes, analysis of continuous data was assessed using analysis of variance (ANOVA), and nominal data were assessed with a χ² test with a single comparison α level of 0.05 and Bonferroni correction to adjust for multiple comparisons where appropriate. When comparing all 3 thiazide groups, after the Bonferroni correction, P < .01667 was considered statistically significant to avoid a type 1 error in a family of statistical tests.

Results

As of January 21, 2022, the active thiazide cohort yielded 628,994 thiazide prescriptions within the VA nationwide. Most patients were male, with female patients representing 8.4%, 6.6%, and 5.6% of the HCTZ, chlorthalidone, and indapamide arms, respectively (Table 1). Utilization rates were significantly different between thiazide groups (P < .001). HCTZ was the most prescribed thiazide diuretic (84.6%) followed by chlorthalidone (14.9%) and indapamide (0.5%) (Table 2).

BP values documented after prescription initiation date were available for few individuals in the HCTZ, chlorthalidone, and indapamide groups (0.3%, 0.2%, and 0.5%, respectively). Overall, the mean BP values were similar among thiazide groups: 135/79 mm Hg for HCTZ, 137/78 mm Hg for chlorthalidone, and 133/79 mm Hg for indapamide (P = .32). BP control was also similar with control rates of 26.0%, 27.1%, and 33.3% for those on HCTZ, chlorthalidone, and indapamide, respectively (P = .75). The use of concomitant potassium or magnesium supplementation was significantly different between thiazide groups with rates of 12.4%, 22.6%, and 27.1% for HCTZ, chlorthalidone, and indapamide, respectively (P < .001). When comparing chlorthalidone to HCTZ, there was a significantly higher rate of concomitant supplementation with chlorthalidone (P < .001) (Table 3).

In the historic cohort, HCTZ utilization decreased from 90.2% to 83.5% (P < .001) and chlorthalidone utilization increased significantly from 9.3% to 16.0% (P < .001) (Figure). There was no significant change in the use of indapamide during this period (P = .73). Yearly trends from 2016 to 2021 are listed in Table 4.

Discussion

The findings of our evaluation demonstrate that despite the 2017 ACC/AHA BP guideline recommendations for using chlorthalidone, HCTZ predominates as the most prescribed thiazide diuretic within the VA. However, since the publication of this guideline, there has been an increase in chlorthalidone prescribing and a decrease in HCTZ prescribing within the VA.

A 2010 study by Ernst and colleagues revealed a similar trend to what was seen in our study. At that time, HCTZ was the most prescribed thiazide encompassing 95% of total thiazide utilization; however, chlorthalidone utilization increased from 1.1% in 2003 to 2.4% in 2008.⁸ In comparing our chlorthalidone utilization rates with these results, 9.3% in 2016 and 16.0% in 2021, the change in chlorthalidone prescribing from 2003 to 2016 represents a more than linear increase. This trend continued in our study from 2016 to 2021; the expected chlorthalidone utilization would be 21.2% in 2021 if it followed the 2003 to 2016 rate of change. Thus the trend in increasing chlorthalidone use predated the 2017 guideline recommendation. Nonetheless, this change in the thiazide prescribing pattern represents a positive shift in practice.

Our evaluation found a significantly higher rate of concomitant potassium or magnesium supplementation with chlorthalidone and indapamide compared with HCTZ in the active cohort. Electrolyte abnormalities are well documented adverse effects associated with thiazide diuretic use.⁹ A cross-sectional analysis by Ravioli and colleagues revealed thiazide diuretic use was an independent predictor of both hyponatremia (22.1% incidence) and hypokalemia (19% incidence) and that chlorthalidone was associated with the highest risk of electrolyte abnormalities whereas HCTZ was associated with the lowest risk. Their study also found these electrolyte abnormalities to have a dose-dependent relationship with the thiazide diuretic prescribed.¹⁰

While Ravioli and colleagues did not address the incidence of hypomagnesemia with thiazide diuretic use, a cross-sectional analysis by Kieboom and colleagues reported a significant increase in hypomagnesemia in patients prescribed thiazide diuretics.¹¹ Although rates of electrolyte abnormalities are reported in the literature, the rates of concomitant supplementation are unclear, especially when compared across thiazide agents. Our study provides insight into the use of concomitant potassium and magnesium supplementation compared between HCTZ, chlorthalidone, and indapamide. In our active cohort, potassium was more commonly prescribed than magnesium. Interestingly, magnesium supplementation accounted for 25.9% of the total supplement use for HCTZ compared with rates of 22.4% and 21.0% for chlorthalidone and indapamide, respectively. It is unclear if this trend highlights a greater incidence of hypomagnesemia with HCTZ or greater clinician awareness to monitor this agent, but this finding may warrant further investigation. In addition, when considering the overall lower rate of supplementation seen with HCTZ in our study, the use of potassium-sparing diuretics should be considered. These agents, including triamterene, amiloride, eplerenone, and spironolactone, can be supplement-sparing and are available in combination products only with HCTZ.

Low chlorthalidone utilization rates are concerning especially given the literature demonstrating CVD benefit with chlorthalidone and the lack of compelling outcomes data to support HCTZ as the preferred agent.^3,4 There are several reasons why HCTZ use may be higher in practice. First is clinical inertia, which is defined as a lack of treatment intensification or lack of changing practice patterns, despite evidence-based goals of care.¹² HCTZ has been the most widely prescribed thiazide diuretic for years.⁷ As a result, converting HCTZ to chlorthalidone for a patient with suboptimal BP control may not be considered and instead clinicians may add on another antihypertensive or titrate doses of current antihypertensives.

There is also a consideration for patient adherence. HCTZ has many more combination products available than chlorthalidone and indapamide. If switching a patient from an HCTZ-containing combination product to chlorthalidone, adherence and patient willingness to take another capsule or tablet must be considered. Finally, there may be clinical controversy and questions around switching patients from HCTZ to chlorthalidone. Although the guidelines do not explicitly recommend switching to chlorthalidone, it may be reasonable in most patients unless they have or are at significant risk of electrolyte or metabolic disturbances that may be exacerbated or triggered with conversion.

When converting from HCTZ to chlorthalidone, it is important to consider dosing. Previous studies have demonstrated that chlorthalidone is 1.5 to 2 times more potent than HCTZ.^13,14 Therefore, the conversion from HCTZ to chlorthalidone is not 1:1, but instead 50 mg of HCTZ is approximately equal to25 to 37.5 mg of chlorthalidone.¹⁴

Limitations

This study was limited by its retrospective design, gaps in data, duplicate active prescription data, and the assessment of concomitant electrolyte supplementation. As with any retrospective study, there is a potential for confounding and a concern for information bias with missing information. This study relied on proper documentation of prescription and demographic information in the Veterans Health Information Systems and Technology Architecture (VistA), as the CDW compiles information from this electronic health record. Strengths of the VistA include ease in clinical functions, documentation, and the ability for records to be updated from any VA facility nationally. However, there is always the possibility of user error and information to be omitted.

In our study, the documentation of BP values and subsequent analysis of overall BP control were limited. For BP values to be included in this study, they had to be recorded after the active thiazide prescription was written and from an in-person encounter documented in VistA. The COVID-19 pandemic shifted the clinical landscape and many primary care appointments during the active cohort evaluation period were conducted virtually. Therefore, patients may not have had formal vitals recorded. There may also be an aspect of selection bias regarding the chlorthalidone group. Although rates of thiazide switching were not assessed, some patients may have been switched from HCTZ or indapamide to chlorthalidone to achieve additional BP control. Thus, patients receiving chlorthalidone may represent a more difficult-to-control hypertensive population, making a finding of similar BP control rates between HCTZ and chlorthalidone an actual positive finding regarding chlorthalidone. Finally, this study did not assess adherence to medications. As the intent of the study was to analyze prescribing patterns, it is impossible to know if the patient was actively taking the medication at the time of assessment. When considering the rates of BP control, there were limited BP values, a potential for selection bias, and neither adherence nor patient self-reported home BP values were assessed. Therefore, the interpretation of overall BP control must be done with caution.

Additionally, duplicate prescriptions were noted in the active cohort. Rates of duplication were 0.2%, 0.08%, and 0.09% for HCTZ, chlorthalidone, and indapamide, respectively. With these small percentages, we felt this would not have a significant impact on the overall thiazide use trends seen in our study. Patients can receive prescriptions from multiple VA facilities and may have > 1 active prescriptions. This has been mitigated in recent years with the introduction of the OneVA program, allowing pharmacists to access any prescription on file from any VA facility and refill if needed (except controlled substance prescriptions). However, there are certain instances in which duplicate prescriptions may be necessary. These include patients enrolled and receiving care at another VA facility (eg, traveling for part of a year) and patients hospitalized at a different facility and given medications on discharge.

With the overall low rate of duplication prescriptions seen in each thiazide group, we determined that this was not large enough to cause substantial variation in the results of this evaluation and was unlikely to alter the results. This study also does not inform on the incidence of switching between thiazide diuretics. If a patient was switched from HCTZ to chlorthalidone in 2017, for example, a prescription for HCTZ and chlorthalidone would have been reported in this study. We felt that the change in chlorthalidone prescribing from January 1, 2016, to December 31, 2021, would reflect overall utilization rates, which may include switching from HCTZ or indapamide to chlorthalidone in addition to new chlorthalidone prescriptions.

Finally, there are confounders and trends in concomitant potassium or magnesium supplementation that were not accounted for in our study. These include concomitant loop diuretics or other medications that may cause electrolyte abnormalities and the dose-dependent relationship between thiazide diuretics and electrolyte abnormalities.¹⁰ Actual laboratory values were not included in this analysis and thus we cannot assess whether supplementation or management of electrolyte disturbances was clinically appropriate.

Conclusions

Thiazide utilization patterns have shifted possibly due to the 2017 ACC/AHA BP guideline recommendations. However, HCTZ continues to be the most widely prescribed thiazide diuretic within the VA. There is a need for future projects and clinician education to increase the implementation of guideline-recommended therapy within the VA, particularly regarding chlorthalidone use.

The American Diabetes Association recommends that patients on intensive insulin regimens self-monitor blood glucose (SMBG) to assist in therapy optimization.¹ To be useful, SMBG data must be captured by patients, shared with care teams, and used and interpreted by patients and practitioners.^2,3 Communication of SMBG data from the patient to practitioner can be challenging. Although technology can help in this process, limitations exist, such as manual data entry into systems, patient and/or practitioner technological challenges (eg, accessing interface), and compatibility and integration between SMBG devices and electronic health record (EHR) systems.⁴

The Boise Veterans Affairs Medical Center (BVAMC) in Idaho serves more than 100,000 veterans. It includes a main site, community-based outpatient clinics, and a clinical resource hub that provides telehealth services to veterans residing in rural neighboring states. The BVAMC pharmacy department provides both inpatient and outpatient services. At the BVAMC, clinical pharmacist practitioners (CPPs) are independent practitioners who support their care teams in comprehensive medication management and have the ability to initiate, modify, and discontinue drug therapy for referred patients.⁵ A prominent role of CPPs in primary care teams is to manage patients with uncontrolled diabetes and intensive insulin regimens in which SMBG data are vital to therapy optimization. As collecting SMBG data from patients is seen anecdotally as time intensive, we determined the mean time spent by CPPs collecting patient SMBG data and its potential implications.

Methods

Pharmacists at BVAMC were asked to estimate and record the following: SMBG data collection method, time spent collecting data, extra time spent documenting or formatting SMBG readings, total patient visit time, and visit type. Time was collected in minutes. Extra time spent documenting or formatting SMBG readings included any additional time formatting or entering data in the clinical note after talking to the patient; if this was done while multitasking and talking to the patient, it was not considered extra time. For total patient visit time, pharmacists were asked to estimate only time spent discussing diabetes care and collecting SMBG data. Visit types were categorized as in-person/face-to-face, telephone, and telehealth using clinical video telehealth (CVT)/VA Video Connect (VVC). Data were collected using a standardized spreadsheet. The spreadsheet was pilot tested by a CPP before distribution to all pharmacists.

CPPs were educated about the project in March 2021 and were asked to record data for a 1-week period between April 5, 2021, and April 30, 2021. One CPP also provided delayed data collected from May 17 to 21, 2021, and these data were included in our analysis.

Descriptive statistics were used to determine the mean time spent by CPPs collecting SMBG data. Unpaired t tests were used to compare time spent collecting SMBG data by different collection methods and patient visit types. A P value of ≤ .05 was considered statistically significant. Data were organized in Microsoft Excel, and statistics were completed with JMP Pro v15.

Results

Eight CPPs provided data from 120 patient encounters. For all patient encounter types, the mean time spent collecting SMBG data was 3.3 minutes, and completing additional documentation/formatting was 1.3 minutes (Table 1). Total mean time for SMBG collection and documentation was 4.6 minutes in visits that had a mean length of 20.1 minutes. Twenty-three percent of the visit was devoted to SMBG data, 16% for data collection, and 6% for documentation. In 23 encounters, at least half the time was spent collecting and documenting/formatting data.

When compared by the SMBG collection method, the longest time spent collecting SMBG data was with patient report (3.7 minutes), and the longest time spent documenting/formatting time was with meter download/home telehealth (2 minutes). There was no statistically significant difference in the time to collect SMBG data between patient report and other methods (3.7 minutes vs 2.8 minutes; P = .07).

When compared by visit type, there was not a statistically significant difference between time spent collecting in person vs telephone or video SMBG data (3.8 minutes vs 3.2 minutes; P = .39) (Table 2). The most common SMBG collection method for in-person/face-to-face visits was continuous glucose monitor (CGM) (n = 10), followed by meter download/home telehealth (n = 5), patient report (n = 3), and directly from log/meter (n = 1). For telephone or video visits, the most common collection method was patient report (n = 72), followed by directly from log/meter (n = 18), CGM (n = 5), meter download/home telehealth (n = 4), and secure message (n = 2).

Discussion

We found that the mean amount of time spent collecting and documenting/formatting SMBG data was only 4.6 minutes; however, this still represented a substantial portion of visit time. For telephone and CVT/VVC appointments, this represented > 25% of total visit time. While CPPs make important contributions to interprofessional team management of patients with diabetes, their cost is not trivial.^6-8 It is worth exploring the most effective and efficient ways to use CPPs. Our results indicate that streamlining SMBG data collection may be beneficial.

Pharmacy technicians, licensed practical nurses/clinical associates, registered nurses/nurse care managers, or other team members could help improve SMBG data collection. Using other team members is also an opportunity for comanagement, for team collaboration, and for more patients to be seen. For example, if a CPP currently has 12 patient encounters that last 20 minutes each, this results in about 240 minutes of direct patient care. If patient encounters were 16 minutes, CPPs could have 15 patient encounters in 240 minutes. Saved time could be used for other clinical tasks involved in disease management or clinical reminder reviews. While there are benefits to CPPs collecting SMBG data, such as further inquiry about patient-reported values, other team members could be trained to ask appropriate follow-up questions for abnormal blood glucose readings. In addition, leveraging current team members and optimizing their roles could prevent the need to acquire additional full-time equivalent employees.

Another opportunity to increase efficiency in SMBG data collection is with SMBG devices and EHR integration.^4,9 However, integration can be difficult with different types of SMBG devices and EHR platforms. Education for patients and practitioners could help to ensure accurate and reliable data uploads; patient internet availability; data protection, privacy, and sharing; workflow management; and clear patient-practitioner expectations.¹⁰ For example, if patient SMBG data are automatically uploaded to practitioners, patients’ expectations for practitioner review of data and follow-up need to be determined.

We found a subset of 23 patient encounters where data collection and documenting/formatting represented more than half of the total visit time. In this subset, 13 SMBG reports were pulled from a log or meter, 8 were patient reported, and 3 were meter download or home telehealth.

Limitations

A potential reason for the lack of statistically significant differences in SMBG collection method or visit type in this study includes the small sample size. Participation in this work was voluntary, and all participating CPPs had ≥ 3 years of practice in their current setting, which includes a heavy workload of diabetes management. These pharmacists noted self-established procedures/systems for SMBG data collection, including the use of Excel spreadsheets with pregenerated formulas. For less experienced CPPs, SMBG data collection time may be even longer. Pharmacists also noted that they may limit time spent collecting SMBG data depending on the patient encounter and whether they have gathered sufficient data to guide clinical care. Other limitations of this work include data collection from a single institution and that the time documented represented estimates; there was no external monitor.

Conclusions

In this analysis, we found that CPPs spend about 3 minutes collecting SMBG data from patients and about an additional 1 minute documenting and formatting data. While 4 to 5 minutes may not represent a substantial amount of time for 1 patient, it can be when multiplied by several patient encounters. The time spent collecting SMBG data did not significantly differ by collection method or visit type. Opportunities to increase efficiency in SMBG data collection, such as the use of nonpharmacist team members, are worth exploring.

Acknowledgments

Thank you to the pharmacists at the Boise Veterans Affairs Medical Center for their time and support of this work: Danielle Ahlstrom, Paul Black, Robyn Cruz, Sarah Naidoo, Anthony Nelson, Laura Spoutz, Eileen Twomey, Donovan Victorine, and Michelle Wilkin.

The American Diabetes Association recommends that patients on intensive insulin regimens self-monitor blood glucose (SMBG) to assist in therapy optimization.¹ To be useful, SMBG data must be captured by patients, shared with care teams, and used and interpreted by patients and practitioners.^2,3 Communication of SMBG data from the patient to practitioner can be challenging. Although technology can help in this process, limitations exist, such as manual data entry into systems, patient and/or practitioner technological challenges (eg, accessing interface), and compatibility and integration between SMBG devices and electronic health record (EHR) systems.⁴

The Boise Veterans Affairs Medical Center (BVAMC) in Idaho serves more than 100,000 veterans. It includes a main site, community-based outpatient clinics, and a clinical resource hub that provides telehealth services to veterans residing in rural neighboring states. The BVAMC pharmacy department provides both inpatient and outpatient services. At the BVAMC, clinical pharmacist practitioners (CPPs) are independent practitioners who support their care teams in comprehensive medication management and have the ability to initiate, modify, and discontinue drug therapy for referred patients.⁵ A prominent role of CPPs in primary care teams is to manage patients with uncontrolled diabetes and intensive insulin regimens in which SMBG data are vital to therapy optimization. As collecting SMBG data from patients is seen anecdotally as time intensive, we determined the mean time spent by CPPs collecting patient SMBG data and its potential implications.

Methods

Pharmacists at BVAMC were asked to estimate and record the following: SMBG data collection method, time spent collecting data, extra time spent documenting or formatting SMBG readings, total patient visit time, and visit type. Time was collected in minutes. Extra time spent documenting or formatting SMBG readings included any additional time formatting or entering data in the clinical note after talking to the patient; if this was done while multitasking and talking to the patient, it was not considered extra time. For total patient visit time, pharmacists were asked to estimate only time spent discussing diabetes care and collecting SMBG data. Visit types were categorized as in-person/face-to-face, telephone, and telehealth using clinical video telehealth (CVT)/VA Video Connect (VVC). Data were collected using a standardized spreadsheet. The spreadsheet was pilot tested by a CPP before distribution to all pharmacists.

CPPs were educated about the project in March 2021 and were asked to record data for a 1-week period between April 5, 2021, and April 30, 2021. One CPP also provided delayed data collected from May 17 to 21, 2021, and these data were included in our analysis.

Descriptive statistics were used to determine the mean time spent by CPPs collecting SMBG data. Unpaired t tests were used to compare time spent collecting SMBG data by different collection methods and patient visit types. A P value of ≤ .05 was considered statistically significant. Data were organized in Microsoft Excel, and statistics were completed with JMP Pro v15.

Results

Eight CPPs provided data from 120 patient encounters. For all patient encounter types, the mean time spent collecting SMBG data was 3.3 minutes, and completing additional documentation/formatting was 1.3 minutes (Table 1). Total mean time for SMBG collection and documentation was 4.6 minutes in visits that had a mean length of 20.1 minutes. Twenty-three percent of the visit was devoted to SMBG data, 16% for data collection, and 6% for documentation. In 23 encounters, at least half the time was spent collecting and documenting/formatting data.

When compared by the SMBG collection method, the longest time spent collecting SMBG data was with patient report (3.7 minutes), and the longest time spent documenting/formatting time was with meter download/home telehealth (2 minutes). There was no statistically significant difference in the time to collect SMBG data between patient report and other methods (3.7 minutes vs 2.8 minutes; P = .07).

When compared by visit type, there was not a statistically significant difference between time spent collecting in person vs telephone or video SMBG data (3.8 minutes vs 3.2 minutes; P = .39) (Table 2). The most common SMBG collection method for in-person/face-to-face visits was continuous glucose monitor (CGM) (n = 10), followed by meter download/home telehealth (n = 5), patient report (n = 3), and directly from log/meter (n = 1). For telephone or video visits, the most common collection method was patient report (n = 72), followed by directly from log/meter (n = 18), CGM (n = 5), meter download/home telehealth (n = 4), and secure message (n = 2).

Discussion

We found that the mean amount of time spent collecting and documenting/formatting SMBG data was only 4.6 minutes; however, this still represented a substantial portion of visit time. For telephone and CVT/VVC appointments, this represented > 25% of total visit time. While CPPs make important contributions to interprofessional team management of patients with diabetes, their cost is not trivial.^6-8 It is worth exploring the most effective and efficient ways to use CPPs. Our results indicate that streamlining SMBG data collection may be beneficial.

Pharmacy technicians, licensed practical nurses/clinical associates, registered nurses/nurse care managers, or other team members could help improve SMBG data collection. Using other team members is also an opportunity for comanagement, for team collaboration, and for more patients to be seen. For example, if a CPP currently has 12 patient encounters that last 20 minutes each, this results in about 240 minutes of direct patient care. If patient encounters were 16 minutes, CPPs could have 15 patient encounters in 240 minutes. Saved time could be used for other clinical tasks involved in disease management or clinical reminder reviews. While there are benefits to CPPs collecting SMBG data, such as further inquiry about patient-reported values, other team members could be trained to ask appropriate follow-up questions for abnormal blood glucose readings. In addition, leveraging current team members and optimizing their roles could prevent the need to acquire additional full-time equivalent employees.

Another opportunity to increase efficiency in SMBG data collection is with SMBG devices and EHR integration.^4,9 However, integration can be difficult with different types of SMBG devices and EHR platforms. Education for patients and practitioners could help to ensure accurate and reliable data uploads; patient internet availability; data protection, privacy, and sharing; workflow management; and clear patient-practitioner expectations.¹⁰ For example, if patient SMBG data are automatically uploaded to practitioners, patients’ expectations for practitioner review of data and follow-up need to be determined.

We found a subset of 23 patient encounters where data collection and documenting/formatting represented more than half of the total visit time. In this subset, 13 SMBG reports were pulled from a log or meter, 8 were patient reported, and 3 were meter download or home telehealth.

Limitations

A potential reason for the lack of statistically significant differences in SMBG collection method or visit type in this study includes the small sample size. Participation in this work was voluntary, and all participating CPPs had ≥ 3 years of practice in their current setting, which includes a heavy workload of diabetes management. These pharmacists noted self-established procedures/systems for SMBG data collection, including the use of Excel spreadsheets with pregenerated formulas. For less experienced CPPs, SMBG data collection time may be even longer. Pharmacists also noted that they may limit time spent collecting SMBG data depending on the patient encounter and whether they have gathered sufficient data to guide clinical care. Other limitations of this work include data collection from a single institution and that the time documented represented estimates; there was no external monitor.

Conclusions

In this analysis, we found that CPPs spend about 3 minutes collecting SMBG data from patients and about an additional 1 minute documenting and formatting data. While 4 to 5 minutes may not represent a substantial amount of time for 1 patient, it can be when multiplied by several patient encounters. The time spent collecting SMBG data did not significantly differ by collection method or visit type. Opportunities to increase efficiency in SMBG data collection, such as the use of nonpharmacist team members, are worth exploring.

Acknowledgments

Thank you to the pharmacists at the Boise Veterans Affairs Medical Center for their time and support of this work: Danielle Ahlstrom, Paul Black, Robyn Cruz, Sarah Naidoo, Anthony Nelson, Laura Spoutz, Eileen Twomey, Donovan Victorine, and Michelle Wilkin.

The American Diabetes Association recommends that patients on intensive insulin regimens self-monitor blood glucose (SMBG) to assist in therapy optimization.¹ To be useful, SMBG data must be captured by patients, shared with care teams, and used and interpreted by patients and practitioners.^2,3 Communication of SMBG data from the patient to practitioner can be challenging. Although technology can help in this process, limitations exist, such as manual data entry into systems, patient and/or practitioner technological challenges (eg, accessing interface), and compatibility and integration between SMBG devices and electronic health record (EHR) systems.⁴

The Boise Veterans Affairs Medical Center (BVAMC) in Idaho serves more than 100,000 veterans. It includes a main site, community-based outpatient clinics, and a clinical resource hub that provides telehealth services to veterans residing in rural neighboring states. The BVAMC pharmacy department provides both inpatient and outpatient services. At the BVAMC, clinical pharmacist practitioners (CPPs) are independent practitioners who support their care teams in comprehensive medication management and have the ability to initiate, modify, and discontinue drug therapy for referred patients.⁵ A prominent role of CPPs in primary care teams is to manage patients with uncontrolled diabetes and intensive insulin regimens in which SMBG data are vital to therapy optimization. As collecting SMBG data from patients is seen anecdotally as time intensive, we determined the mean time spent by CPPs collecting patient SMBG data and its potential implications.

Methods

Pharmacists at BVAMC were asked to estimate and record the following: SMBG data collection method, time spent collecting data, extra time spent documenting or formatting SMBG readings, total patient visit time, and visit type. Time was collected in minutes. Extra time spent documenting or formatting SMBG readings included any additional time formatting or entering data in the clinical note after talking to the patient; if this was done while multitasking and talking to the patient, it was not considered extra time. For total patient visit time, pharmacists were asked to estimate only time spent discussing diabetes care and collecting SMBG data. Visit types were categorized as in-person/face-to-face, telephone, and telehealth using clinical video telehealth (CVT)/VA Video Connect (VVC). Data were collected using a standardized spreadsheet. The spreadsheet was pilot tested by a CPP before distribution to all pharmacists.

CPPs were educated about the project in March 2021 and were asked to record data for a 1-week period between April 5, 2021, and April 30, 2021. One CPP also provided delayed data collected from May 17 to 21, 2021, and these data were included in our analysis.

Descriptive statistics were used to determine the mean time spent by CPPs collecting SMBG data. Unpaired t tests were used to compare time spent collecting SMBG data by different collection methods and patient visit types. A P value of ≤ .05 was considered statistically significant. Data were organized in Microsoft Excel, and statistics were completed with JMP Pro v15.

Results

Eight CPPs provided data from 120 patient encounters. For all patient encounter types, the mean time spent collecting SMBG data was 3.3 minutes, and completing additional documentation/formatting was 1.3 minutes (Table 1). Total mean time for SMBG collection and documentation was 4.6 minutes in visits that had a mean length of 20.1 minutes. Twenty-three percent of the visit was devoted to SMBG data, 16% for data collection, and 6% for documentation. In 23 encounters, at least half the time was spent collecting and documenting/formatting data.

When compared by the SMBG collection method, the longest time spent collecting SMBG data was with patient report (3.7 minutes), and the longest time spent documenting/formatting time was with meter download/home telehealth (2 minutes). There was no statistically significant difference in the time to collect SMBG data between patient report and other methods (3.7 minutes vs 2.8 minutes; P = .07).

When compared by visit type, there was not a statistically significant difference between time spent collecting in person vs telephone or video SMBG data (3.8 minutes vs 3.2 minutes; P = .39) (Table 2). The most common SMBG collection method for in-person/face-to-face visits was continuous glucose monitor (CGM) (n = 10), followed by meter download/home telehealth (n = 5), patient report (n = 3), and directly from log/meter (n = 1). For telephone or video visits, the most common collection method was patient report (n = 72), followed by directly from log/meter (n = 18), CGM (n = 5), meter download/home telehealth (n = 4), and secure message (n = 2).

Discussion

We found that the mean amount of time spent collecting and documenting/formatting SMBG data was only 4.6 minutes; however, this still represented a substantial portion of visit time. For telephone and CVT/VVC appointments, this represented > 25% of total visit time. While CPPs make important contributions to interprofessional team management of patients with diabetes, their cost is not trivial.^6-8 It is worth exploring the most effective and efficient ways to use CPPs. Our results indicate that streamlining SMBG data collection may be beneficial.

Pharmacy technicians, licensed practical nurses/clinical associates, registered nurses/nurse care managers, or other team members could help improve SMBG data collection. Using other team members is also an opportunity for comanagement, for team collaboration, and for more patients to be seen. For example, if a CPP currently has 12 patient encounters that last 20 minutes each, this results in about 240 minutes of direct patient care. If patient encounters were 16 minutes, CPPs could have 15 patient encounters in 240 minutes. Saved time could be used for other clinical tasks involved in disease management or clinical reminder reviews. While there are benefits to CPPs collecting SMBG data, such as further inquiry about patient-reported values, other team members could be trained to ask appropriate follow-up questions for abnormal blood glucose readings. In addition, leveraging current team members and optimizing their roles could prevent the need to acquire additional full-time equivalent employees.

Another opportunity to increase efficiency in SMBG data collection is with SMBG devices and EHR integration.^4,9 However, integration can be difficult with different types of SMBG devices and EHR platforms. Education for patients and practitioners could help to ensure accurate and reliable data uploads; patient internet availability; data protection, privacy, and sharing; workflow management; and clear patient-practitioner expectations.¹⁰ For example, if patient SMBG data are automatically uploaded to practitioners, patients’ expectations for practitioner review of data and follow-up need to be determined.

We found a subset of 23 patient encounters where data collection and documenting/formatting represented more than half of the total visit time. In this subset, 13 SMBG reports were pulled from a log or meter, 8 were patient reported, and 3 were meter download or home telehealth.

Limitations

A potential reason for the lack of statistically significant differences in SMBG collection method or visit type in this study includes the small sample size. Participation in this work was voluntary, and all participating CPPs had ≥ 3 years of practice in their current setting, which includes a heavy workload of diabetes management. These pharmacists noted self-established procedures/systems for SMBG data collection, including the use of Excel spreadsheets with pregenerated formulas. For less experienced CPPs, SMBG data collection time may be even longer. Pharmacists also noted that they may limit time spent collecting SMBG data depending on the patient encounter and whether they have gathered sufficient data to guide clinical care. Other limitations of this work include data collection from a single institution and that the time documented represented estimates; there was no external monitor.

Conclusions

In this analysis, we found that CPPs spend about 3 minutes collecting SMBG data from patients and about an additional 1 minute documenting and formatting data. While 4 to 5 minutes may not represent a substantial amount of time for 1 patient, it can be when multiplied by several patient encounters. The time spent collecting SMBG data did not significantly differ by collection method or visit type. Opportunities to increase efficiency in SMBG data collection, such as the use of nonpharmacist team members, are worth exploring.

Acknowledgments

Thank you to the pharmacists at the Boise Veterans Affairs Medical Center for their time and support of this work: Danielle Ahlstrom, Paul Black, Robyn Cruz, Sarah Naidoo, Anthony Nelson, Laura Spoutz, Eileen Twomey, Donovan Victorine, and Michelle Wilkin.

Heart failure (HF) is a chronic, progressive condition that is characterized by the heart’s inability to effectively pump blood throughout the body. In 2018, approximately 6.2 million US adults had HF, and 13.4% of all death certificates noted HF as a precipitating factor.¹ Patients not receiving appropriate guideline-directed medical therapy (GDMT) face a 29% excess mortality risk over a 2-year period.² Each additional GDMT included in a patient’s regimen significantly reduces all-cause mortality.³

The Change the Management of Patients with Heart Failure (CHAMP) registry reports that only about 1% of patients with HF are prescribed 3 agents from contemporary GDMT at target doses, highlighting the need for optimizing clinicians’ approaches to GDMT.⁴ Similarly, The Get With The Guidelines Heart Failure Registry has noted that only 20.2% of patients with HF with reduced ejection fraction (HFrEF) are prescribed a sodium-glucose cotransporter 2 inhibitor (SGLT2i) following hospital discharge for HFrEF exacerbation.⁵ Overall, treatment rates with GDMT saw limited improvement between 2013 and 2019, with no significant difference between groups in mortality, indicating the need for optimized methods to encourage the initiation of GDMT.⁶

Remote monitoring and telecare are novel ways to improve GDMT rates in those with HFrEF. However, data are inconsistent regarding the impact of remote HF monitoring and improvements in GDMT or HF-related outcomes.^6-10 The modalities of remote monitoring for GDMT vary among studies, but the potential for telehealth monitoring to improve GDMT, thereby potentially reducing HF-related hospitalizations, is clear.

Telemonitoring has demonstrated improved participant adherence with weight monitoring, although the withdrawal rate was high, and has the potential to reduce all-cause mortality and HF-related hospitalizations.^11,12 Telemonitoring for GDMT optimization led to an increase in the proportion of patients who achieved optimal GDMT doses, a decrease in the time to dose optimization, and a reduction in the number of clinic visits.¹³ Remote GDMT titration was accomplished in the general patient population with HFrEF; however, in populations already followed by cardiologists or HF specialists, remote optimization strategies did not yield different proportions of GDMT use.¹⁴ The aim of this study was to assess the impact of the home telehealth (HT) monitoring program on the initiation and optimization of HF GDMT among veterans with HFrEF at the Veterans Affairs Ann Arbor Healthcare System (VAAAHS) in Michigan.

Methods

This was a single-center retrospective study of Computerized Patient Record System (CPRS)data. Patients at the VAAAHS were evaluated if they were diagnosed with HFrEF and were eligible for enrollment in the HT monitoring program. Eligibility criteria included a diagnosis of stage C HF, irrespective of EF, and a history of any HF-related hospitalization. We focused on patients with HFrEF due to stronger guideline-based recommendations for certain pharmacotherapies as compared with HF with mildly reduced ejection fraction (HFmrEF) and preserved ejection fraction (HFpEF). Initial patient data for HT enrolling were accessed using the Heart Failure Dashboard via the US Department of Veterans Affairs (VA) Academic Detailing Service. The target daily doses of typical agents used in HFrEF GDMT are listed in the Appendix.

The HT program is an embedded model in which HT nurses receive remote data from the patient and triage that with the VAAAHS cardiology team. Patients’ questions, concerns, and/or vital signs are recovered remotely. In this model, nurses are embedded in the cardiology team, working with the cardiologists, cardiology clinical pharmacist, and/or cardiology nurse practitioners to make medication interventions. Data are recorded with an HT device, including weight, blood pressure (BP), heart rate, and pulse oximetry. HT nurses are also available to the patient via phone or video. The program uses a 180-day disease management protocol for HF via remote device, enabling the patient to answer questions and receive education on their disease daily. Responses to questions and data are then reviewed by an HT nurse remotely during business hours and triaged as appropriate with the cardiology team. Data can be communicated to the cardiology team via the patient record, eliminating the need for the cardiology team to use the proprietary portal affiliated with the HT device.

Study Sample

Patient information was obtained from a list of 417 patients eligible for enrollment in the HT program; the list was sent to the HT program for review and enrollment. Patient data were extracted from the VAAAHS HF Dashboard and included all patients with HFrEF and available data on the platform. The sample for the retrospective chart review included 40 adults who had HFrEF, defined as a left ventricular EF (LVEF) of ≤ 40% as evidenced by a transthoracic echocardiogram or cardiac magnetic resonance imaging. These patients were contacted and agreed to enroll in the HT monitoring program. The HT program population was compared against a control group of 33 patients who were ineligible for the HT program. Patients were deemed ineligible for HT if they resided in a nursing home, lacked a VAAAHS primary care clinician, or declined participation in the HT program.

Procedures

Patients enrolled in the HT program were assigned a nurse who monitored and responded to changes in vital signs, such as an increase in weight of ≥ 3 lb in 24 hours or ≥ 5 lb in 7 days. Each participant in HT received a home BP machine to check BP and heart rates and a scale for daily weights. The patient also responded to questions on a tablet device with the BP machine and scale. The vital signs were put into the tablet device, and the patient answered questions about their symptoms for that day. If the patient endorsed any symptoms of HF, such as pedal edema, abdominal distention, orthopnea, paroxysmal nocturnal dyspnea, dyspnea on exertion, or orthopnea, they received a call from the nurse to discuss symptoms and make appropriate adjustments in the diuretic or HF regimen. The HF-trained pharmacist served as a resource for drug therapy management and provided recommendations for diuretic adjustments on an as-needed basis.

Patients who declined participation in the HT program followed the standard of care, which was limited to visits with primary care clinicians and/or cardiologists as per the follow-up plan. Patient data were collected over 12 months. The study was approved by the VAAAHS Institutional Review Board (reference number, 1703034), Research and Development Committee, and Research Administration.

The primary goal of the study was to assess the impact of the HT program on drug interventions, specifically initiating and titrating HFrEF pharmacotherapies. Interventions were based on GDMT with known mortality- and morbidity-reducing properties when used at their maximum tolerated doses, including angiotensin-converting enzyme inhibitors (ACEi), angiotensin receptor-neprilysin inhibitor (ARNi), or angiotensin receptor blockers (ARB), with a preference for ARNi, β-blockers for HFrEF (metoprolol succinate, bisoprolol, or carvedilol), aldosterone antagonists, and SGLT2is.

Secondary goals included HF-related hospitalizations, medication adherence, time to enrollment in HT, time to laboratory analysis after the initiation or titration of an ACEi/ARB/ARNi or aldosterone antagonist, and time enrolled in the HT program. Patients were considered adherent if their drug refill history showed consistent fills of their medications. The χ² test was used for total interventions made during the study period and Fisher exact test for all others.

Results

Patient data were collected between July 2022 and June 2023. All 73 patients were male, and the mean age in the HT group (n = 40) was 72.6 years and 75.2 years for the control group (n = 33). Overall, the baseline demographics were similar between the groups (Table 1). Of 40 patients screened for enrollment in the HT program, 33 were included in the analysis (Figure 1).

At baseline, the HT group included more individuals than the control group on ACEi/ARB/ARNi (24 vs 19, respectively), β-blocker (28 vs 24, respectively), SGLT2i (14 vs 11, respectively), and aldosterone antagonist (15 vs 9, respectively) (Figure 2). There were 20 interventions made in the HT group compared with 11 therapy changes in the control arm during the study (odds ratio, 1.43; P = .23) (Table 2). In the HT group, 1 patient achieved an ACEi target dose, 3 patients achieved a β-blocker target dose, and 7 achieved a target dose of spironolactone (titration is not required for SGLT2i therapy and is counted as target dose). In the HT group, 17 patients were on ≥ 3 recommended agents, while 9 patients were taking 4 agents. Seven of 20 HT group interventions resulted in titration to the target dose. In the control group, no patients achieved an ARNi target dose, 3 patients achieved a β-blocker target dose, and 2 patients achieved a spironolactone antagonist target dose. In the control arm, 7 patients were on ≥ 3 GDMTs, and 2 were taking 4 agents. No patient in either group achieved a target dose of 4 agents. Five of 11 control group interventions resulted in initiation or titration of GDMT to the target dose.

We analyzed the overall percentage of VAAAHS patients with HFrEF who were on appropriate GDMT. Given the ongoing drive to improve HF-related outcomes, HT interventions could not be compared to a static population, so the HT and control patients were compared with the rates of GDMT at VAAAHS, which was available in the Academic Detailing Service Heart Failure Dashboard (Figure 3). Titration and optimization rates were also compared (Figure 4). From July 2022 to June 2023, ARNi use increased by 16.6%, aldosterone antagonist by 6.8%, and β-blockers by 2.4%. Target doses of GDMTs were more difficult to achieve in the hospital system. There was an increase in aldosterone antagonist target dose achievement by 4.7%, but overall there were decreases in target doses in other GDMTs: ACEi/ARB/ARNi target dose use decreased by 3.2%, ARNi target dose use decreased by 2.7% and target β-blocker use decreased by 0.9%.

Discussion

Telehealth yielded clinically important interventions, with some titrations bringing patients to their target doses of medications for HFrEF. The 20 interventions made in the HT group can be largely attributed to the nurses’ efforts to alert clinicians to drug titrations or ACEi/ARB to ARNi transitions. Although the findings were not statistically significant, the difference in the number of drug therapy changes supports the use of the HT program for a GDMT optimization strategy. Patients may be difficult to titrate secondary to adverse effects that make medication initiation or titration inappropriate, such as hypotension and hyperkalemia, although this was not observed in this small sample size. Considering a mean HT enrollment of 5.3 months, many patients had adequate disease assessment and medication titration. Given that patients are discharged from the service once deemed appropriate, this decreases the burden on the patient and increases the utility and implementation of the HT program for other patients.

A surprising finding of this study was the lower rate of HF-related hospitalizations in the HT group. Although not the primary subject of interest in the study, it reinforced the importance of close health care professional follow-up for patients living with HF. Telehealth may improve communication and shared decision making over medication use. Because the finding was unanticipated, the rate of diuretic adjustments was not tracked.

Patients were reevaluated every 6 months for willingness to continue the program, adherence, and clinical needs. These results are similar to those of other trials that demonstrated improved rates of GDMT in the setting of pharmacist- or nurse-led HF treatment optimization.^15,16 These positive results differ from other trials incorporating remote monitoring regarding patient continuation in HT programs. For example, in a study by Ding and colleagues, the withdrawal rate from their monitoring service was about 22%, while in our study only 1 patient withdrew from the HT program.¹¹

The HT program resulted in fewer hospitalizations than the control arm. There were 6 HF-related hospitalizations in the control group, although 5 involved a single patient. Typically, such a patient would be encouraged to follow HT monitoring after just 1 HF-related hospitalization; however, the patient declined to participate.

Early optimization of GDMT in patients who were recently discharged from the hospital for an HF-related hospitalization yields a reduction in hospital rehospitalization.¹⁷ GDMT optimization has unequivocal benefits in HF outcomes. Unfortunately, the issues surrounding methodologies on how to best optimize GDMT are lacking. While HT has been found to be feasible in the aid of optimizing medical therapy, the TIM-HF trial concluded that remote monitoring services had no significant benefit in reducing mortality.^7,8 On the other hand, the OptiLink HF Study showed that when clinicians respond to remote monitoring prompts from fluid index threshold crossing alerts, these interventions are associated with significantly improved clinical outcomes in patients with implantable cardioverter-defibrillators and advanced HF.⁹ In contrast to previous trials, the AMULET trial showed that remote monitoring compared with standard care significantly reduced the risk of HF hospitalization or cardiovascular death during the 12-month follow-up among patients with HF and LVEF ≤ 49% after an episode of acute exacerbation.¹⁰ Additionally, patients who received skilled home health services and participated in remote monitoring for their chronic HF experienced a reduction in all-cause 30-day readmission.¹⁸

Given the contrasting evidence regarding remote monitoring and variable modalities of implementing interventions, we investigated whether HT monitoring yields improvements in GDMT optimization. We found that HT nurses were able to nearly double the rate of interventions for patients with HFrEF. The HT program in providing expanded services will require adequate staffing responsibilities and support. The HT program is geared toward following a large, diverse patient population, such as those with chronic obstructive pulmonary disease, hypertension, and HF. We only evaluated services for patients with HFrEF, but the program also follows patients with HfmrEF and HfpEF. These patients were not included as GDMT optimization differs for patients with an LVEF > 40%.^19,20

The lower rates of achieving target doses of GDMTs were likely obstructed by continuous use of initial drug doses and further limited by lack of follow-up. When compared with the rest of the VAAAHS, there was a greater effort to increase ARNi use in the HT group as 7 of 33 patients (21%) were started on ARNi compared with a background increase of ARNi use of 17%. There was a lower mortality rate observed in the HT group compared with the control group. One patient in each group died of unrelated causes, while 2 deaths in the control group were due to worsening HF. The difference in mortality is likely multifactorial, possibly related to the control group’s greater disease burden or higher mean age (75.2 years vs 72.6 years).

Limitations

This was an observational cohort design, which is subject to bias. Thus, the findings of this study are entirely hypothesis-generating and a randomized controlled trial would be necessary for clearer results. Second, low numbers of participants may have skewed the data set. Given the observational nature of the study, this nonetheless is a positive finding to support the HT program for assisting with HF monitoring and prompting drug interventions. Due to the low number of participants, a single patient may have skewed the results with 5 hospitalizations.

Conclusions

This pilot study demonstrates the applicability of HT monitoring to optimize veteran HFrEF GDMT. The HT program yielded numerically relevant interventions and fewer HF-related hospitalizations compared with the control arm. The study shows the feasibility of the program to safely optimize GDMT toward their target doses and may serve as an additional catalyst to further develop HT programs specifically targeted toward HF monitoring and management. Cost-savings analyses would likely need to demonstrate the cost utility of such a service.

Acknowledgments

We thank the home telehealth nursing staff for their assistance in data collection and enrollment of patients into the monitoring program.

Heart failure (HF) is a chronic, progressive condition that is characterized by the heart’s inability to effectively pump blood throughout the body. In 2018, approximately 6.2 million US adults had HF, and 13.4% of all death certificates noted HF as a precipitating factor.¹ Patients not receiving appropriate guideline-directed medical therapy (GDMT) face a 29% excess mortality risk over a 2-year period.² Each additional GDMT included in a patient’s regimen significantly reduces all-cause mortality.³

The Change the Management of Patients with Heart Failure (CHAMP) registry reports that only about 1% of patients with HF are prescribed 3 agents from contemporary GDMT at target doses, highlighting the need for optimizing clinicians’ approaches to GDMT.⁴ Similarly, The Get With The Guidelines Heart Failure Registry has noted that only 20.2% of patients with HF with reduced ejection fraction (HFrEF) are prescribed a sodium-glucose cotransporter 2 inhibitor (SGLT2i) following hospital discharge for HFrEF exacerbation.⁵ Overall, treatment rates with GDMT saw limited improvement between 2013 and 2019, with no significant difference between groups in mortality, indicating the need for optimized methods to encourage the initiation of GDMT.⁶

Remote monitoring and telecare are novel ways to improve GDMT rates in those with HFrEF. However, data are inconsistent regarding the impact of remote HF monitoring and improvements in GDMT or HF-related outcomes.^6-10 The modalities of remote monitoring for GDMT vary among studies, but the potential for telehealth monitoring to improve GDMT, thereby potentially reducing HF-related hospitalizations, is clear.

Telemonitoring has demonstrated improved participant adherence with weight monitoring, although the withdrawal rate was high, and has the potential to reduce all-cause mortality and HF-related hospitalizations.^11,12 Telemonitoring for GDMT optimization led to an increase in the proportion of patients who achieved optimal GDMT doses, a decrease in the time to dose optimization, and a reduction in the number of clinic visits.¹³ Remote GDMT titration was accomplished in the general patient population with HFrEF; however, in populations already followed by cardiologists or HF specialists, remote optimization strategies did not yield different proportions of GDMT use.¹⁴ The aim of this study was to assess the impact of the home telehealth (HT) monitoring program on the initiation and optimization of HF GDMT among veterans with HFrEF at the Veterans Affairs Ann Arbor Healthcare System (VAAAHS) in Michigan.

Methods

This was a single-center retrospective study of Computerized Patient Record System (CPRS)data. Patients at the VAAAHS were evaluated if they were diagnosed with HFrEF and were eligible for enrollment in the HT monitoring program. Eligibility criteria included a diagnosis of stage C HF, irrespective of EF, and a history of any HF-related hospitalization. We focused on patients with HFrEF due to stronger guideline-based recommendations for certain pharmacotherapies as compared with HF with mildly reduced ejection fraction (HFmrEF) and preserved ejection fraction (HFpEF). Initial patient data for HT enrolling were accessed using the Heart Failure Dashboard via the US Department of Veterans Affairs (VA) Academic Detailing Service. The target daily doses of typical agents used in HFrEF GDMT are listed in the Appendix.

The HT program is an embedded model in which HT nurses receive remote data from the patient and triage that with the VAAAHS cardiology team. Patients’ questions, concerns, and/or vital signs are recovered remotely. In this model, nurses are embedded in the cardiology team, working with the cardiologists, cardiology clinical pharmacist, and/or cardiology nurse practitioners to make medication interventions. Data are recorded with an HT device, including weight, blood pressure (BP), heart rate, and pulse oximetry. HT nurses are also available to the patient via phone or video. The program uses a 180-day disease management protocol for HF via remote device, enabling the patient to answer questions and receive education on their disease daily. Responses to questions and data are then reviewed by an HT nurse remotely during business hours and triaged as appropriate with the cardiology team. Data can be communicated to the cardiology team via the patient record, eliminating the need for the cardiology team to use the proprietary portal affiliated with the HT device.

Study Sample

Patient information was obtained from a list of 417 patients eligible for enrollment in the HT program; the list was sent to the HT program for review and enrollment. Patient data were extracted from the VAAAHS HF Dashboard and included all patients with HFrEF and available data on the platform. The sample for the retrospective chart review included 40 adults who had HFrEF, defined as a left ventricular EF (LVEF) of ≤ 40% as evidenced by a transthoracic echocardiogram or cardiac magnetic resonance imaging. These patients were contacted and agreed to enroll in the HT monitoring program. The HT program population was compared against a control group of 33 patients who were ineligible for the HT program. Patients were deemed ineligible for HT if they resided in a nursing home, lacked a VAAAHS primary care clinician, or declined participation in the HT program.

Procedures

Patients enrolled in the HT program were assigned a nurse who monitored and responded to changes in vital signs, such as an increase in weight of ≥ 3 lb in 24 hours or ≥ 5 lb in 7 days. Each participant in HT received a home BP machine to check BP and heart rates and a scale for daily weights. The patient also responded to questions on a tablet device with the BP machine and scale. The vital signs were put into the tablet device, and the patient answered questions about their symptoms for that day. If the patient endorsed any symptoms of HF, such as pedal edema, abdominal distention, orthopnea, paroxysmal nocturnal dyspnea, dyspnea on exertion, or orthopnea, they received a call from the nurse to discuss symptoms and make appropriate adjustments in the diuretic or HF regimen. The HF-trained pharmacist served as a resource for drug therapy management and provided recommendations for diuretic adjustments on an as-needed basis.

Patients who declined participation in the HT program followed the standard of care, which was limited to visits with primary care clinicians and/or cardiologists as per the follow-up plan. Patient data were collected over 12 months. The study was approved by the VAAAHS Institutional Review Board (reference number, 1703034), Research and Development Committee, and Research Administration.

The primary goal of the study was to assess the impact of the HT program on drug interventions, specifically initiating and titrating HFrEF pharmacotherapies. Interventions were based on GDMT with known mortality- and morbidity-reducing properties when used at their maximum tolerated doses, including angiotensin-converting enzyme inhibitors (ACEi), angiotensin receptor-neprilysin inhibitor (ARNi), or angiotensin receptor blockers (ARB), with a preference for ARNi, β-blockers for HFrEF (metoprolol succinate, bisoprolol, or carvedilol), aldosterone antagonists, and SGLT2is.

Secondary goals included HF-related hospitalizations, medication adherence, time to enrollment in HT, time to laboratory analysis after the initiation or titration of an ACEi/ARB/ARNi or aldosterone antagonist, and time enrolled in the HT program. Patients were considered adherent if their drug refill history showed consistent fills of their medications. The χ² test was used for total interventions made during the study period and Fisher exact test for all others.

Results

Patient data were collected between July 2022 and June 2023. All 73 patients were male, and the mean age in the HT group (n = 40) was 72.6 years and 75.2 years for the control group (n = 33). Overall, the baseline demographics were similar between the groups (Table 1). Of 40 patients screened for enrollment in the HT program, 33 were included in the analysis (Figure 1).

At baseline, the HT group included more individuals than the control group on ACEi/ARB/ARNi (24 vs 19, respectively), β-blocker (28 vs 24, respectively), SGLT2i (14 vs 11, respectively), and aldosterone antagonist (15 vs 9, respectively) (Figure 2). There were 20 interventions made in the HT group compared with 11 therapy changes in the control arm during the study (odds ratio, 1.43; P = .23) (Table 2). In the HT group, 1 patient achieved an ACEi target dose, 3 patients achieved a β-blocker target dose, and 7 achieved a target dose of spironolactone (titration is not required for SGLT2i therapy and is counted as target dose). In the HT group, 17 patients were on ≥ 3 recommended agents, while 9 patients were taking 4 agents. Seven of 20 HT group interventions resulted in titration to the target dose. In the control group, no patients achieved an ARNi target dose, 3 patients achieved a β-blocker target dose, and 2 patients achieved a spironolactone antagonist target dose. In the control arm, 7 patients were on ≥ 3 GDMTs, and 2 were taking 4 agents. No patient in either group achieved a target dose of 4 agents. Five of 11 control group interventions resulted in initiation or titration of GDMT to the target dose.

We analyzed the overall percentage of VAAAHS patients with HFrEF who were on appropriate GDMT. Given the ongoing drive to improve HF-related outcomes, HT interventions could not be compared to a static population, so the HT and control patients were compared with the rates of GDMT at VAAAHS, which was available in the Academic Detailing Service Heart Failure Dashboard (Figure 3). Titration and optimization rates were also compared (Figure 4). From July 2022 to June 2023, ARNi use increased by 16.6%, aldosterone antagonist by 6.8%, and β-blockers by 2.4%. Target doses of GDMTs were more difficult to achieve in the hospital system. There was an increase in aldosterone antagonist target dose achievement by 4.7%, but overall there were decreases in target doses in other GDMTs: ACEi/ARB/ARNi target dose use decreased by 3.2%, ARNi target dose use decreased by 2.7% and target β-blocker use decreased by 0.9%.

Discussion

Telehealth yielded clinically important interventions, with some titrations bringing patients to their target doses of medications for HFrEF. The 20 interventions made in the HT group can be largely attributed to the nurses’ efforts to alert clinicians to drug titrations or ACEi/ARB to ARNi transitions. Although the findings were not statistically significant, the difference in the number of drug therapy changes supports the use of the HT program for a GDMT optimization strategy. Patients may be difficult to titrate secondary to adverse effects that make medication initiation or titration inappropriate, such as hypotension and hyperkalemia, although this was not observed in this small sample size. Considering a mean HT enrollment of 5.3 months, many patients had adequate disease assessment and medication titration. Given that patients are discharged from the service once deemed appropriate, this decreases the burden on the patient and increases the utility and implementation of the HT program for other patients.

A surprising finding of this study was the lower rate of HF-related hospitalizations in the HT group. Although not the primary subject of interest in the study, it reinforced the importance of close health care professional follow-up for patients living with HF. Telehealth may improve communication and shared decision making over medication use. Because the finding was unanticipated, the rate of diuretic adjustments was not tracked.

Patients were reevaluated every 6 months for willingness to continue the program, adherence, and clinical needs. These results are similar to those of other trials that demonstrated improved rates of GDMT in the setting of pharmacist- or nurse-led HF treatment optimization.^15,16 These positive results differ from other trials incorporating remote monitoring regarding patient continuation in HT programs. For example, in a study by Ding and colleagues, the withdrawal rate from their monitoring service was about 22%, while in our study only 1 patient withdrew from the HT program.¹¹

The HT program resulted in fewer hospitalizations than the control arm. There were 6 HF-related hospitalizations in the control group, although 5 involved a single patient. Typically, such a patient would be encouraged to follow HT monitoring after just 1 HF-related hospitalization; however, the patient declined to participate.

Early optimization of GDMT in patients who were recently discharged from the hospital for an HF-related hospitalization yields a reduction in hospital rehospitalization.¹⁷ GDMT optimization has unequivocal benefits in HF outcomes. Unfortunately, the issues surrounding methodologies on how to best optimize GDMT are lacking. While HT has been found to be feasible in the aid of optimizing medical therapy, the TIM-HF trial concluded that remote monitoring services had no significant benefit in reducing mortality.^7,8 On the other hand, the OptiLink HF Study showed that when clinicians respond to remote monitoring prompts from fluid index threshold crossing alerts, these interventions are associated with significantly improved clinical outcomes in patients with implantable cardioverter-defibrillators and advanced HF.⁹ In contrast to previous trials, the AMULET trial showed that remote monitoring compared with standard care significantly reduced the risk of HF hospitalization or cardiovascular death during the 12-month follow-up among patients with HF and LVEF ≤ 49% after an episode of acute exacerbation.¹⁰ Additionally, patients who received skilled home health services and participated in remote monitoring for their chronic HF experienced a reduction in all-cause 30-day readmission.¹⁸

Given the contrasting evidence regarding remote monitoring and variable modalities of implementing interventions, we investigated whether HT monitoring yields improvements in GDMT optimization. We found that HT nurses were able to nearly double the rate of interventions for patients with HFrEF. The HT program in providing expanded services will require adequate staffing responsibilities and support. The HT program is geared toward following a large, diverse patient population, such as those with chronic obstructive pulmonary disease, hypertension, and HF. We only evaluated services for patients with HFrEF, but the program also follows patients with HfmrEF and HfpEF. These patients were not included as GDMT optimization differs for patients with an LVEF > 40%.^19,20

The lower rates of achieving target doses of GDMTs were likely obstructed by continuous use of initial drug doses and further limited by lack of follow-up. When compared with the rest of the VAAAHS, there was a greater effort to increase ARNi use in the HT group as 7 of 33 patients (21%) were started on ARNi compared with a background increase of ARNi use of 17%. There was a lower mortality rate observed in the HT group compared with the control group. One patient in each group died of unrelated causes, while 2 deaths in the control group were due to worsening HF. The difference in mortality is likely multifactorial, possibly related to the control group’s greater disease burden or higher mean age (75.2 years vs 72.6 years).

Limitations

This was an observational cohort design, which is subject to bias. Thus, the findings of this study are entirely hypothesis-generating and a randomized controlled trial would be necessary for clearer results. Second, low numbers of participants may have skewed the data set. Given the observational nature of the study, this nonetheless is a positive finding to support the HT program for assisting with HF monitoring and prompting drug interventions. Due to the low number of participants, a single patient may have skewed the results with 5 hospitalizations.

Conclusions

This pilot study demonstrates the applicability of HT monitoring to optimize veteran HFrEF GDMT. The HT program yielded numerically relevant interventions and fewer HF-related hospitalizations compared with the control arm. The study shows the feasibility of the program to safely optimize GDMT toward their target doses and may serve as an additional catalyst to further develop HT programs specifically targeted toward HF monitoring and management. Cost-savings analyses would likely need to demonstrate the cost utility of such a service.

Acknowledgments

We thank the home telehealth nursing staff for their assistance in data collection and enrollment of patients into the monitoring program.

Heart failure (HF) is a chronic, progressive condition that is characterized by the heart’s inability to effectively pump blood throughout the body. In 2018, approximately 6.2 million US adults had HF, and 13.4% of all death certificates noted HF as a precipitating factor.¹ Patients not receiving appropriate guideline-directed medical therapy (GDMT) face a 29% excess mortality risk over a 2-year period.² Each additional GDMT included in a patient’s regimen significantly reduces all-cause mortality.³

The Change the Management of Patients with Heart Failure (CHAMP) registry reports that only about 1% of patients with HF are prescribed 3 agents from contemporary GDMT at target doses, highlighting the need for optimizing clinicians’ approaches to GDMT.⁴ Similarly, The Get With The Guidelines Heart Failure Registry has noted that only 20.2% of patients with HF with reduced ejection fraction (HFrEF) are prescribed a sodium-glucose cotransporter 2 inhibitor (SGLT2i) following hospital discharge for HFrEF exacerbation.⁵ Overall, treatment rates with GDMT saw limited improvement between 2013 and 2019, with no significant difference between groups in mortality, indicating the need for optimized methods to encourage the initiation of GDMT.⁶

Remote monitoring and telecare are novel ways to improve GDMT rates in those with HFrEF. However, data are inconsistent regarding the impact of remote HF monitoring and improvements in GDMT or HF-related outcomes.^6-10 The modalities of remote monitoring for GDMT vary among studies, but the potential for telehealth monitoring to improve GDMT, thereby potentially reducing HF-related hospitalizations, is clear.

Telemonitoring has demonstrated improved participant adherence with weight monitoring, although the withdrawal rate was high, and has the potential to reduce all-cause mortality and HF-related hospitalizations.^11,12 Telemonitoring for GDMT optimization led to an increase in the proportion of patients who achieved optimal GDMT doses, a decrease in the time to dose optimization, and a reduction in the number of clinic visits.¹³ Remote GDMT titration was accomplished in the general patient population with HFrEF; however, in populations already followed by cardiologists or HF specialists, remote optimization strategies did not yield different proportions of GDMT use.¹⁴ The aim of this study was to assess the impact of the home telehealth (HT) monitoring program on the initiation and optimization of HF GDMT among veterans with HFrEF at the Veterans Affairs Ann Arbor Healthcare System (VAAAHS) in Michigan.

Methods

This was a single-center retrospective study of Computerized Patient Record System (CPRS)data. Patients at the VAAAHS were evaluated if they were diagnosed with HFrEF and were eligible for enrollment in the HT monitoring program. Eligibility criteria included a diagnosis of stage C HF, irrespective of EF, and a history of any HF-related hospitalization. We focused on patients with HFrEF due to stronger guideline-based recommendations for certain pharmacotherapies as compared with HF with mildly reduced ejection fraction (HFmrEF) and preserved ejection fraction (HFpEF). Initial patient data for HT enrolling were accessed using the Heart Failure Dashboard via the US Department of Veterans Affairs (VA) Academic Detailing Service. The target daily doses of typical agents used in HFrEF GDMT are listed in the Appendix.

The HT program is an embedded model in which HT nurses receive remote data from the patient and triage that with the VAAAHS cardiology team. Patients’ questions, concerns, and/or vital signs are recovered remotely. In this model, nurses are embedded in the cardiology team, working with the cardiologists, cardiology clinical pharmacist, and/or cardiology nurse practitioners to make medication interventions. Data are recorded with an HT device, including weight, blood pressure (BP), heart rate, and pulse oximetry. HT nurses are also available to the patient via phone or video. The program uses a 180-day disease management protocol for HF via remote device, enabling the patient to answer questions and receive education on their disease daily. Responses to questions and data are then reviewed by an HT nurse remotely during business hours and triaged as appropriate with the cardiology team. Data can be communicated to the cardiology team via the patient record, eliminating the need for the cardiology team to use the proprietary portal affiliated with the HT device.

Study Sample

Patient information was obtained from a list of 417 patients eligible for enrollment in the HT program; the list was sent to the HT program for review and enrollment. Patient data were extracted from the VAAAHS HF Dashboard and included all patients with HFrEF and available data on the platform. The sample for the retrospective chart review included 40 adults who had HFrEF, defined as a left ventricular EF (LVEF) of ≤ 40% as evidenced by a transthoracic echocardiogram or cardiac magnetic resonance imaging. These patients were contacted and agreed to enroll in the HT monitoring program. The HT program population was compared against a control group of 33 patients who were ineligible for the HT program. Patients were deemed ineligible for HT if they resided in a nursing home, lacked a VAAAHS primary care clinician, or declined participation in the HT program.

Procedures

Patients enrolled in the HT program were assigned a nurse who monitored and responded to changes in vital signs, such as an increase in weight of ≥ 3 lb in 24 hours or ≥ 5 lb in 7 days. Each participant in HT received a home BP machine to check BP and heart rates and a scale for daily weights. The patient also responded to questions on a tablet device with the BP machine and scale. The vital signs were put into the tablet device, and the patient answered questions about their symptoms for that day. If the patient endorsed any symptoms of HF, such as pedal edema, abdominal distention, orthopnea, paroxysmal nocturnal dyspnea, dyspnea on exertion, or orthopnea, they received a call from the nurse to discuss symptoms and make appropriate adjustments in the diuretic or HF regimen. The HF-trained pharmacist served as a resource for drug therapy management and provided recommendations for diuretic adjustments on an as-needed basis.

Patients who declined participation in the HT program followed the standard of care, which was limited to visits with primary care clinicians and/or cardiologists as per the follow-up plan. Patient data were collected over 12 months. The study was approved by the VAAAHS Institutional Review Board (reference number, 1703034), Research and Development Committee, and Research Administration.

The primary goal of the study was to assess the impact of the HT program on drug interventions, specifically initiating and titrating HFrEF pharmacotherapies. Interventions were based on GDMT with known mortality- and morbidity-reducing properties when used at their maximum tolerated doses, including angiotensin-converting enzyme inhibitors (ACEi), angiotensin receptor-neprilysin inhibitor (ARNi), or angiotensin receptor blockers (ARB), with a preference for ARNi, β-blockers for HFrEF (metoprolol succinate, bisoprolol, or carvedilol), aldosterone antagonists, and SGLT2is.

Secondary goals included HF-related hospitalizations, medication adherence, time to enrollment in HT, time to laboratory analysis after the initiation or titration of an ACEi/ARB/ARNi or aldosterone antagonist, and time enrolled in the HT program. Patients were considered adherent if their drug refill history showed consistent fills of their medications. The χ² test was used for total interventions made during the study period and Fisher exact test for all others.

Results

Patient data were collected between July 2022 and June 2023. All 73 patients were male, and the mean age in the HT group (n = 40) was 72.6 years and 75.2 years for the control group (n = 33). Overall, the baseline demographics were similar between the groups (Table 1). Of 40 patients screened for enrollment in the HT program, 33 were included in the analysis (Figure 1).

At baseline, the HT group included more individuals than the control group on ACEi/ARB/ARNi (24 vs 19, respectively), β-blocker (28 vs 24, respectively), SGLT2i (14 vs 11, respectively), and aldosterone antagonist (15 vs 9, respectively) (Figure 2). There were 20 interventions made in the HT group compared with 11 therapy changes in the control arm during the study (odds ratio, 1.43; P = .23) (Table 2). In the HT group, 1 patient achieved an ACEi target dose, 3 patients achieved a β-blocker target dose, and 7 achieved a target dose of spironolactone (titration is not required for SGLT2i therapy and is counted as target dose). In the HT group, 17 patients were on ≥ 3 recommended agents, while 9 patients were taking 4 agents. Seven of 20 HT group interventions resulted in titration to the target dose. In the control group, no patients achieved an ARNi target dose, 3 patients achieved a β-blocker target dose, and 2 patients achieved a spironolactone antagonist target dose. In the control arm, 7 patients were on ≥ 3 GDMTs, and 2 were taking 4 agents. No patient in either group achieved a target dose of 4 agents. Five of 11 control group interventions resulted in initiation or titration of GDMT to the target dose.

We analyzed the overall percentage of VAAAHS patients with HFrEF who were on appropriate GDMT. Given the ongoing drive to improve HF-related outcomes, HT interventions could not be compared to a static population, so the HT and control patients were compared with the rates of GDMT at VAAAHS, which was available in the Academic Detailing Service Heart Failure Dashboard (Figure 3). Titration and optimization rates were also compared (Figure 4). From July 2022 to June 2023, ARNi use increased by 16.6%, aldosterone antagonist by 6.8%, and β-blockers by 2.4%. Target doses of GDMTs were more difficult to achieve in the hospital system. There was an increase in aldosterone antagonist target dose achievement by 4.7%, but overall there were decreases in target doses in other GDMTs: ACEi/ARB/ARNi target dose use decreased by 3.2%, ARNi target dose use decreased by 2.7% and target β-blocker use decreased by 0.9%.

Discussion

Telehealth yielded clinically important interventions, with some titrations bringing patients to their target doses of medications for HFrEF. The 20 interventions made in the HT group can be largely attributed to the nurses’ efforts to alert clinicians to drug titrations or ACEi/ARB to ARNi transitions. Although the findings were not statistically significant, the difference in the number of drug therapy changes supports the use of the HT program for a GDMT optimization strategy. Patients may be difficult to titrate secondary to adverse effects that make medication initiation or titration inappropriate, such as hypotension and hyperkalemia, although this was not observed in this small sample size. Considering a mean HT enrollment of 5.3 months, many patients had adequate disease assessment and medication titration. Given that patients are discharged from the service once deemed appropriate, this decreases the burden on the patient and increases the utility and implementation of the HT program for other patients.

A surprising finding of this study was the lower rate of HF-related hospitalizations in the HT group. Although not the primary subject of interest in the study, it reinforced the importance of close health care professional follow-up for patients living with HF. Telehealth may improve communication and shared decision making over medication use. Because the finding was unanticipated, the rate of diuretic adjustments was not tracked.

Patients were reevaluated every 6 months for willingness to continue the program, adherence, and clinical needs. These results are similar to those of other trials that demonstrated improved rates of GDMT in the setting of pharmacist- or nurse-led HF treatment optimization.^15,16 These positive results differ from other trials incorporating remote monitoring regarding patient continuation in HT programs. For example, in a study by Ding and colleagues, the withdrawal rate from their monitoring service was about 22%, while in our study only 1 patient withdrew from the HT program.¹¹

The HT program resulted in fewer hospitalizations than the control arm. There were 6 HF-related hospitalizations in the control group, although 5 involved a single patient. Typically, such a patient would be encouraged to follow HT monitoring after just 1 HF-related hospitalization; however, the patient declined to participate.

Early optimization of GDMT in patients who were recently discharged from the hospital for an HF-related hospitalization yields a reduction in hospital rehospitalization.¹⁷ GDMT optimization has unequivocal benefits in HF outcomes. Unfortunately, the issues surrounding methodologies on how to best optimize GDMT are lacking. While HT has been found to be feasible in the aid of optimizing medical therapy, the TIM-HF trial concluded that remote monitoring services had no significant benefit in reducing mortality.^7,8 On the other hand, the OptiLink HF Study showed that when clinicians respond to remote monitoring prompts from fluid index threshold crossing alerts, these interventions are associated with significantly improved clinical outcomes in patients with implantable cardioverter-defibrillators and advanced HF.⁹ In contrast to previous trials, the AMULET trial showed that remote monitoring compared with standard care significantly reduced the risk of HF hospitalization or cardiovascular death during the 12-month follow-up among patients with HF and LVEF ≤ 49% after an episode of acute exacerbation.¹⁰ Additionally, patients who received skilled home health services and participated in remote monitoring for their chronic HF experienced a reduction in all-cause 30-day readmission.¹⁸

Given the contrasting evidence regarding remote monitoring and variable modalities of implementing interventions, we investigated whether HT monitoring yields improvements in GDMT optimization. We found that HT nurses were able to nearly double the rate of interventions for patients with HFrEF. The HT program in providing expanded services will require adequate staffing responsibilities and support. The HT program is geared toward following a large, diverse patient population, such as those with chronic obstructive pulmonary disease, hypertension, and HF. We only evaluated services for patients with HFrEF, but the program also follows patients with HfmrEF and HfpEF. These patients were not included as GDMT optimization differs for patients with an LVEF > 40%.^19,20

The lower rates of achieving target doses of GDMTs were likely obstructed by continuous use of initial drug doses and further limited by lack of follow-up. When compared with the rest of the VAAAHS, there was a greater effort to increase ARNi use in the HT group as 7 of 33 patients (21%) were started on ARNi compared with a background increase of ARNi use of 17%. There was a lower mortality rate observed in the HT group compared with the control group. One patient in each group died of unrelated causes, while 2 deaths in the control group were due to worsening HF. The difference in mortality is likely multifactorial, possibly related to the control group’s greater disease burden or higher mean age (75.2 years vs 72.6 years).

Limitations

This was an observational cohort design, which is subject to bias. Thus, the findings of this study are entirely hypothesis-generating and a randomized controlled trial would be necessary for clearer results. Second, low numbers of participants may have skewed the data set. Given the observational nature of the study, this nonetheless is a positive finding to support the HT program for assisting with HF monitoring and prompting drug interventions. Due to the low number of participants, a single patient may have skewed the results with 5 hospitalizations.

Conclusions

This pilot study demonstrates the applicability of HT monitoring to optimize veteran HFrEF GDMT. The HT program yielded numerically relevant interventions and fewer HF-related hospitalizations compared with the control arm. The study shows the feasibility of the program to safely optimize GDMT toward their target doses and may serve as an additional catalyst to further develop HT programs specifically targeted toward HF monitoring and management. Cost-savings analyses would likely need to demonstrate the cost utility of such a service.

Acknowledgments

We thank the home telehealth nursing staff for their assistance in data collection and enrollment of patients into the monitoring program.

Opioid use disorder (OUD) is a significant cause of morbidity, mortality, and health care costs in the US.^1,2 Surgery can be the inciting cause for exposure to an opioid; as many as 23% of patients develop chronic OUD following surgery.^3,4 Patients with a history of substance use, mood disorders, anxiety, or previous chronic opioid use (COU) are at risk for relapse, dose escalation, and poor pain control after high-risk surgery, such as orthopedic joint procedures.⁵ Recently focus has been on identifying high-risk patients before orthopedic joint surgery and implementing evidence-based strategies that reduce the postoperative incidence of COU.

A transitional pain service (TPS) has been shown to reduce COU for high-risk surgical patients in different health care settings.^6-9 The TPS model bundles multiple interventions that can be applied to patients at high risk for COU within a health care system. This includes individually tailored programs for preoperative education or pain management planning, use of multimodal analgesia (including regional or neuraxial techniques or nonopioid systemic medications), application of nonpharmacologic modalities (such as cognitive-based intervention), and a coordinated approach to postdischarge instructions and transitions of care. These interventions are coordinated by a multidisciplinary clinical service consisting of anesthesiologists and advanced practice clinicians with specialization in acute pain management and opioid tapering, nurse care coordinators, and psychologists with expertise in cognitive behavioral therapy.

TPS has been shown to reduce the incidence of COU for patients undergoing orthopedic joint surgery, but its impact on health care use and costs is unknown.^6-9 The TPS intervention is resource intensive and increases the use of health care for preoperative education or pain management, which may increase the burden of costs. However, reducing long-term COU may reduce the use of health care for COU- and OUD-related complications, leading to cost savings. This study evaluated whether the TPS intervention influenced health care use and cost for inpatient, outpatient, or pharmacy services during the year following orthopedic joint surgery compared with that of the standard pain management care for procedures that place patients at high risk for COU. We used a difference-in-differences (DID) analysis to estimate this intervention effect, using multivariable regression models that controlled for unobserved time trends and cohort characteristics.

METHODS

This was a quasi-experimental study of patients who underwent orthopedic joint surgery and associated procedures at high risk for COU at the Veterans Affairs Salt Lake City Healthcare System (VASLCHS) between January 2016 through April 2020. The pre-TPS period between January 2016 through December 2017 was compared with the post-TPS period between January 2018 to September 2019. The control patient cohort was selected from 5 geographically diverse VA health care systems throughout the US: Eastern Colorado, Central Plains (Nebraska), White River Junction (Vermont), North Florida/South Georgia, and Portland (Oregon). By sampling health care costs from VA medical centers (VAMCs) across these different regions, our control group was generalizable to veterans receiving orthopedic joint surgery across the US. This study used data from the US Department of Veterans Affairs (VA) Corporate Data Warehouse, a repository of nearly all clinical and administrative data found in electronic health records for VA-provided care and fee-basis care paid for by the VA.¹⁰ All data were hosted and analyzed in the VA Informatics and Computing Infrastructure (VINCI) workspace. The University of Utah Institutional Review Board and the VASLCHS Office of Research and Development approved the protocol for this study.

TPS Intervention

The VASLCHS TPS has already been described in detail elsewhere.^6,7 Briefly, patients at high risk for COU at the VASLCHS were enrolled in the TPS program before surgery for total knee, hip, or shoulder arthroplasty or rotator cuff procedures. The TPS service consists of an anesthesiologist and advanced practice clinician with specialization in acute pain management and opioid tapering, a psychologist with expertise in cognitive behavioral therapy, and 3 nurse care coordinators. These TPS practitioners work together to provide preoperative education, including setting expectations regarding postoperative pain, recommending nonopioid pain management strategies, and providing guidance regarding the appropriate use of opioids for surgical pain. Individual pain plans were developed and implemented for the perioperative period. After surgery, the TPS provided recommendations and support for nonopioid pain therapies and opioid tapers. Patients were followed by the TPS team for at least 12 months after surgery. At a minimum, the goals set by TPS included cessation of all opioid use for prior nonopioid users (NOU) by 90 days after surgery and the return to baseline opioid use or lower for prior COU patients by 90 days after surgery. The TPS also encouraged and supported opioid tapering among COU patients to reduce or completely stop opioid use after surgery.

Patient Cohorts

Veterans having primary or revision total knee, hip, or shoulder arthroplasty or rotator cuff repair between January 1, 2016, and September 30, 2019, at the aforementioned VAMCs were included in the study. Patients who had any hospitalization within 90 days pre- or postindex surgery or who died within 8 months after surgery were excluded from analysis. Patients who had multiple surgeries during the index inpatient visit or within 90 days after the index surgery also were excluded. Comorbid conditions for mental health and substance use were identified using the International Classification of Diseases, 10th revision Clinical Modification (ICD-10) codes or 9th revision equivalent grouped by Clinical Classifications Software Refined (CCS-R).¹¹ Preoperative exposure to clinically relevant pharmacotherapy (ie, agents associated with prolonged opioid use and nonopioid adjuvants) was captured using VA outpatient prescription records (eAppendix 1).

The study patient cohort was stratified into either NOU or COU groups based on opioid use before surgery. Preoperative COU was defined as > 25% nonzero days (calculated using the cabinet supply method) in the 180 days before surgery admit date time (> 45 nonzero days) or ≥ 1 opioid prescription for ≥ 28-day supply released within 90 days before surgery date.^12,13 For NOU patients, the postoperative outcome of interest was new postoperative prolonged opioid use. This was defined by Page and colleagues as ≥ 1 opioid prescription released between surgery discharge date and postdischarge day 44, ≥ 1 opioid prescription released between postdischarge day 45 and 89, and ≥ 1 opioid prescription released between postdischarge day 90 and 180.¹⁴ For COU patients at the time of surgery, the postoperative outcome measure of interest was continued COU, defined as > 25% nonzero days between postdischarge day 90 and 240 (> 37 nonzero days) or ≥ 1 opioid prescription for ≥ 28-day supply released between postdischarge day 90 and 180.

Outcome Variables

Outcome variables included health care use and costs during 1-year pre- and postperiods from the date of surgery. VA health care costs for outpatient, inpatient, and pharmacy services for direct patient care were collected from the Managerial Cost Accounting System, an activity-based cost allocation system that generates estimates of the cost of individual VA hospital stays, health care encounters, and medications. Health care use was defined as the number of encounters for each visit type in the Managerial Cost Accounting System. All costs were adjusted to 2019 US dollars, using the Personal Consumption Expenditures price index for health care services.¹⁵

A set of sociodemographic variables including sex, age at surgery, race and ethnicity, rurality, military branch (Army, Air Force, Marine Corps, Navy, and other), and service connectivity were included as covariates in our regression models. In addition, ICD-9 and ICD-10 codes were used to define the indicators of whether veterans had a diagnosis of mental illness (including anxiety, bipolar disorder, depression, or trauma) or substance use (including alcohol, cannabis, opioids, or tobacco). Finally, pharmacy records were used to create indicators for prescribed opioid-relevant pharmacotherapy (including antidepressants, benzodiazepines [BZD], gabapentinoids, muscle relaxants, non-BZD sedative hypnotics) and active antidepressant drug use during the 1-year preindex period.

Statistical Analyses

Descriptive analyses were used to evaluate differences in baseline patient sociodemographic and clinical characteristics between pre- and postperiods for TPS intervention and control cohorts using 2-sample t tests for continuous variables and χ² tests for categorical variables. We summarized unadjusted health care use and costs for outpatient, inpatient, and pharmacy visits and compared the pre- and postintervention periods using the Mann-Whitney test. Both mean (SD) and median (IQR) were considered, reflecting the skewed distribution of the outcome variables.

We used a DID approach to assess the intervention effect while minimizing confounding from the nonrandom sample. The DID approach controls for unobserved differences between VAMCs that are related to both the intervention and outcomes while controlling for trends over time that could affect outcomes across clinics. To implement the DID approach, we included 3 key independent variables in our regression models: (1) an indicator for whether the observation occurred in the postintervention period; (2) an indicator for whether the patient was exposed to the TPS intervention; and (3) the interaction between these 2 variables.

For cost outcomes, we used multivariable generalized linear models with a log link and a Poisson or Υ family. We analyzed inpatient costs using a 2-part generalized linear model because only 17% to 20% of patients had ≥ 1 inpatient visit. We used multivariable negative binomial regression for health care use outcomes. Demographic and clinical covariates described earlier were included in the regression models to control for differences in the composition of patient groups and clinics that could lead to confounding bias.

RESULTS

Of the 4954 patients included in our study cohort, 3545 (71.6%) were in the NOU group and 1409 (28.4%) were in the COU group. Among the NOU cohort, 361 patients were in the intervention group and 3184 in the control group. Among the COU cohort, 149 patients were in the intervention group and 1260 in the control group (Table 1). Most patients were male, White race, with a mean (SD) age of 64 (11) years. The most common orthopedic procedure was total knee arthroplasty, followed by total hip arthroplasty. Among both NOU and COU cohorts, patients’ characteristics were similar between the pre- and postintervention period among either TPS or control cohort.

Figures 1 and 2 and eAppendix 2 depict unadjusted per-person average outpatient, inpatient, and pharmacy visits and costs incurred during the 1-year pre- and postintervention periods for the NOU and COU cohorts. Average total health care follow-up costs ranged from $40,000 to $53,000 for NOU and from $47,000 to $82,000 for COU cohort. Cost for outpatient visits accounted for about 70% of the average total costs, followed by costs for inpatient visits of about 20%, and costs for pharmacy for the remaining.

For the NOU cohort, the number of health care encounters remained fairly stable between periods except for the outpatient visits among the TPS group. The TPS group experienced an increase in mean outpatient visits in the postperiod: 30 vs 37 visits (23%) (P < .001). Mean outpatient and inpatient costs in the pre- and postintervention periods were not significantly different for either the TPS or control groups. Similarly, within the COU cohort, the TPS group experienced a 27% increase in the mean number of outpatient visits (41 vs 52 visits; P = .02) and a 39% increase in mean outpatient costs in the postintervention compared with the preintervention period ($44,682 vs $61,890; P = .02). Although the mean number of outpatient visits for control group remained at a similar level, average outpatient costs increased roughly 13% ($31,068 vs $35,148; P = .01) between the pre- and postintervention periods.

Table 2 summarizes the results from the multivariable DID analyses for the outpatient, inpatient, and pharmacy visit and cost outcomes. Here, the estimated effect of the TPS intervention is the coefficient from the interaction between the postintervention and TPS exposure indicator variables. This coefficient was calculated as the difference in the outcome before and after the TPS intervention among the TPS group minus the difference in the outcome before and after the TPS intervention among the control group. For the NOU cohort, TPS was associated with an increase in the use of outpatient health care (mean [SD] increase of 6.9 [2] visits; P < .001) after the surgery with no statistically significant effect on outpatient costs (mean [SD] increase of $2787 [$3749]; P = .55). There was no statistically significant effect of TPS on the use of inpatient visits or pharmacy, but a decrease in costs for inpatient visits among those who had at least 1 inpatient visit (mean [SD] decrease of $12,170 [$6100]; P = .02). For the COU cohort, TPS had no statistically significant impact on the use of outpatient, inpatient, or pharmacy or the corresponding costs.

DISCUSSION

TPS is a multidisciplinary approach to perioperative pain management that has been shown to reduce both the quantity and duration of opioid use among orthopedic surgery patients.^6,7 Although the cost burden of providing TPS services to prevent COU is borne by the individual health care system, it is unclear whether this expense is offset by lower long-term medical costs and health care use for COU- and OUD-related complications. In this study focused on a veteran population undergoing orthopedic joint procedures, a DID analysis of cost and health care use showed that TPS, which has been shown to reduce COU for high-risk surgical patients, can be implemented without increasing the overall costs to the VA health care system during the 1 year following surgery, even with increased outpatient visits. For NOU patients, there was no difference in outpatient visit costs or pharmacy costs over 12 months after surgery, although there was a significant reduction in subsequent inpatient costs over the same period. Further, there was no difference in outpatient, inpatient, or pharmacy costs after surgery for COU patients. These findings suggest that TPS can be a cost-effective approach to reduce opioid use among patients undergoing orthopedic joint surgery in VAMCs.

The costs of managing COU after surgery are substantial. Prior reports have shown that adjusted total health care costs are 1.6 to 2.5 times higher for previously NOU patients with new COU after major surgery than those for such patients without persistent use.¹⁶ The 1-year costs associated with new COU in this prior study ranged between $7944 and $17,702 after inpatient surgery and between $5598 and $12,834 after outpatient index surgery, depending on the payer, which are in line with the cost differences found in our current study. Another report among patients with COU following orthopedic joint replacement showed that they had higher use of inpatient, emergency department, and ambulance/paramedic services in the 12 months following their surgery than did those without persistent use.¹⁷ Although these results highlight the impact that COU plays in driving increased costs after major surgery, there have been limited studies focused on interventions that can neutralize the costs associated with opioid misuse after surgery. To our knowledge, our study is the first analysis to show the impact of using an intervention such as TPS to reduce postoperative opioid use on health care use and cost.

Although a rigorous and comprehensive return on investment analysis was beyond the scope of this analysis, these results may have several implications for other health care systems and hospitals that wish to invest in a multidisciplinary perioperative pain management program such as TPS but may be reluctant due to the upfront investment. First, the increased number of patient follow-up visits needed during TPS seems to be more than offset by the reduction in opioid use and associated complications that may occur after surgery. Second, TPS did not seem to be associated with an increase in overall health care costs during the 1-year follow-up period. Together, these results indicate that the return on investment for a TPS approach to perioperative pain management in which optimal patient-centered outcomes are achieved without increasing long-term costs to a health care system may be positive.

Limitations

This study has several limitations. First, this was a quasi-experimental observational study, and the associations we identified between intervention and outcomes should not be assumed to demonstrate causality. Although our DID analysis controlled for an array of demographic and clinical characteristics, differences in medical costs and health care use between the 2 cohorts might be driven by unobserved confounding variables.

Our study also was limited to veterans who received medical care at the VA, and results may not be generalizable to other non-VA health care systems or to veterans with Medicare insurance who have dual benefits. While our finding on health care use and costs may be incomplete because of the uncaptured health care use outside the VA, our DID analysis helped reduce unobserved bias because the absence of data outside of VA care applies to both TPS and control groups. Further, the total costs of operating a TPS program at any given institution will depend on the size of the hospital and volume of surgical patients who meet criteria for enrollment. However, the relative differences in health care use and costs may be extrapolated to patients undergoing orthopedic surgery in other types of academic and community-based health care systems.

Furthermore, this analysis focused primarily on COU and NOU patients undergoing orthopedic joint surgery. While this represents a high-risk population for OUD, the costs and health care use associated with delivering the TPS intervention to other types of surgical procedures may be significantly different. All costs in this analysis were based on 2019 estimates and do not account for the potential inflation over the past several years. Nonmonetary costs to the patient and per-person average total intervention costs were not included in the study. However, we assumed that costs associated with TPS and standard of care would have increased to an equivalent degree over the same period. Further, the average cost of TPS per patient (approximately $900) is relatively small compared with the average annual costs during 1-year pre- and postoperative periods and was not expected to have a significant effect on the analysis.

Conclusions

We found that the significant reduction in COU seen in previous studies following the implementation of TPS was not accompanied by increased health care costs.^6,7 When considering the other costs of long-term opioid use, such as abuse potential, overdose, death, and increased disability, implementation of a TPS service has the potential to improve patient quality of life while reducing other health-related costs. Health care systems should consider the implementation of similar multidisciplinary approaches to perioperative pain management to improve outcomes after orthopedic joint surgery and other high-risk procedures.

Opioid use disorder (OUD) is a significant cause of morbidity, mortality, and health care costs in the US.^1,2 Surgery can be the inciting cause for exposure to an opioid; as many as 23% of patients develop chronic OUD following surgery.^3,4 Patients with a history of substance use, mood disorders, anxiety, or previous chronic opioid use (COU) are at risk for relapse, dose escalation, and poor pain control after high-risk surgery, such as orthopedic joint procedures.⁵ Recently focus has been on identifying high-risk patients before orthopedic joint surgery and implementing evidence-based strategies that reduce the postoperative incidence of COU.

A transitional pain service (TPS) has been shown to reduce COU for high-risk surgical patients in different health care settings.^6-9 The TPS model bundles multiple interventions that can be applied to patients at high risk for COU within a health care system. This includes individually tailored programs for preoperative education or pain management planning, use of multimodal analgesia (including regional or neuraxial techniques or nonopioid systemic medications), application of nonpharmacologic modalities (such as cognitive-based intervention), and a coordinated approach to postdischarge instructions and transitions of care. These interventions are coordinated by a multidisciplinary clinical service consisting of anesthesiologists and advanced practice clinicians with specialization in acute pain management and opioid tapering, nurse care coordinators, and psychologists with expertise in cognitive behavioral therapy.

TPS has been shown to reduce the incidence of COU for patients undergoing orthopedic joint surgery, but its impact on health care use and costs is unknown.^6-9 The TPS intervention is resource intensive and increases the use of health care for preoperative education or pain management, which may increase the burden of costs. However, reducing long-term COU may reduce the use of health care for COU- and OUD-related complications, leading to cost savings. This study evaluated whether the TPS intervention influenced health care use and cost for inpatient, outpatient, or pharmacy services during the year following orthopedic joint surgery compared with that of the standard pain management care for procedures that place patients at high risk for COU. We used a difference-in-differences (DID) analysis to estimate this intervention effect, using multivariable regression models that controlled for unobserved time trends and cohort characteristics.

METHODS

This was a quasi-experimental study of patients who underwent orthopedic joint surgery and associated procedures at high risk for COU at the Veterans Affairs Salt Lake City Healthcare System (VASLCHS) between January 2016 through April 2020. The pre-TPS period between January 2016 through December 2017 was compared with the post-TPS period between January 2018 to September 2019. The control patient cohort was selected from 5 geographically diverse VA health care systems throughout the US: Eastern Colorado, Central Plains (Nebraska), White River Junction (Vermont), North Florida/South Georgia, and Portland (Oregon). By sampling health care costs from VA medical centers (VAMCs) across these different regions, our control group was generalizable to veterans receiving orthopedic joint surgery across the US. This study used data from the US Department of Veterans Affairs (VA) Corporate Data Warehouse, a repository of nearly all clinical and administrative data found in electronic health records for VA-provided care and fee-basis care paid for by the VA.¹⁰ All data were hosted and analyzed in the VA Informatics and Computing Infrastructure (VINCI) workspace. The University of Utah Institutional Review Board and the VASLCHS Office of Research and Development approved the protocol for this study.

TPS Intervention

The VASLCHS TPS has already been described in detail elsewhere.^6,7 Briefly, patients at high risk for COU at the VASLCHS were enrolled in the TPS program before surgery for total knee, hip, or shoulder arthroplasty or rotator cuff procedures. The TPS service consists of an anesthesiologist and advanced practice clinician with specialization in acute pain management and opioid tapering, a psychologist with expertise in cognitive behavioral therapy, and 3 nurse care coordinators. These TPS practitioners work together to provide preoperative education, including setting expectations regarding postoperative pain, recommending nonopioid pain management strategies, and providing guidance regarding the appropriate use of opioids for surgical pain. Individual pain plans were developed and implemented for the perioperative period. After surgery, the TPS provided recommendations and support for nonopioid pain therapies and opioid tapers. Patients were followed by the TPS team for at least 12 months after surgery. At a minimum, the goals set by TPS included cessation of all opioid use for prior nonopioid users (NOU) by 90 days after surgery and the return to baseline opioid use or lower for prior COU patients by 90 days after surgery. The TPS also encouraged and supported opioid tapering among COU patients to reduce or completely stop opioid use after surgery.

Patient Cohorts

Veterans having primary or revision total knee, hip, or shoulder arthroplasty or rotator cuff repair between January 1, 2016, and September 30, 2019, at the aforementioned VAMCs were included in the study. Patients who had any hospitalization within 90 days pre- or postindex surgery or who died within 8 months after surgery were excluded from analysis. Patients who had multiple surgeries during the index inpatient visit or within 90 days after the index surgery also were excluded. Comorbid conditions for mental health and substance use were identified using the International Classification of Diseases, 10th revision Clinical Modification (ICD-10) codes or 9th revision equivalent grouped by Clinical Classifications Software Refined (CCS-R).¹¹ Preoperative exposure to clinically relevant pharmacotherapy (ie, agents associated with prolonged opioid use and nonopioid adjuvants) was captured using VA outpatient prescription records (eAppendix 1).

The study patient cohort was stratified into either NOU or COU groups based on opioid use before surgery. Preoperative COU was defined as > 25% nonzero days (calculated using the cabinet supply method) in the 180 days before surgery admit date time (> 45 nonzero days) or ≥ 1 opioid prescription for ≥ 28-day supply released within 90 days before surgery date.^12,13 For NOU patients, the postoperative outcome of interest was new postoperative prolonged opioid use. This was defined by Page and colleagues as ≥ 1 opioid prescription released between surgery discharge date and postdischarge day 44, ≥ 1 opioid prescription released between postdischarge day 45 and 89, and ≥ 1 opioid prescription released between postdischarge day 90 and 180.¹⁴ For COU patients at the time of surgery, the postoperative outcome measure of interest was continued COU, defined as > 25% nonzero days between postdischarge day 90 and 240 (> 37 nonzero days) or ≥ 1 opioid prescription for ≥ 28-day supply released between postdischarge day 90 and 180.

Outcome Variables

Outcome variables included health care use and costs during 1-year pre- and postperiods from the date of surgery. VA health care costs for outpatient, inpatient, and pharmacy services for direct patient care were collected from the Managerial Cost Accounting System, an activity-based cost allocation system that generates estimates of the cost of individual VA hospital stays, health care encounters, and medications. Health care use was defined as the number of encounters for each visit type in the Managerial Cost Accounting System. All costs were adjusted to 2019 US dollars, using the Personal Consumption Expenditures price index for health care services.¹⁵

A set of sociodemographic variables including sex, age at surgery, race and ethnicity, rurality, military branch (Army, Air Force, Marine Corps, Navy, and other), and service connectivity were included as covariates in our regression models. In addition, ICD-9 and ICD-10 codes were used to define the indicators of whether veterans had a diagnosis of mental illness (including anxiety, bipolar disorder, depression, or trauma) or substance use (including alcohol, cannabis, opioids, or tobacco). Finally, pharmacy records were used to create indicators for prescribed opioid-relevant pharmacotherapy (including antidepressants, benzodiazepines [BZD], gabapentinoids, muscle relaxants, non-BZD sedative hypnotics) and active antidepressant drug use during the 1-year preindex period.

Statistical Analyses

Descriptive analyses were used to evaluate differences in baseline patient sociodemographic and clinical characteristics between pre- and postperiods for TPS intervention and control cohorts using 2-sample t tests for continuous variables and χ² tests for categorical variables. We summarized unadjusted health care use and costs for outpatient, inpatient, and pharmacy visits and compared the pre- and postintervention periods using the Mann-Whitney test. Both mean (SD) and median (IQR) were considered, reflecting the skewed distribution of the outcome variables.

We used a DID approach to assess the intervention effect while minimizing confounding from the nonrandom sample. The DID approach controls for unobserved differences between VAMCs that are related to both the intervention and outcomes while controlling for trends over time that could affect outcomes across clinics. To implement the DID approach, we included 3 key independent variables in our regression models: (1) an indicator for whether the observation occurred in the postintervention period; (2) an indicator for whether the patient was exposed to the TPS intervention; and (3) the interaction between these 2 variables.

For cost outcomes, we used multivariable generalized linear models with a log link and a Poisson or Υ family. We analyzed inpatient costs using a 2-part generalized linear model because only 17% to 20% of patients had ≥ 1 inpatient visit. We used multivariable negative binomial regression for health care use outcomes. Demographic and clinical covariates described earlier were included in the regression models to control for differences in the composition of patient groups and clinics that could lead to confounding bias.

RESULTS

Of the 4954 patients included in our study cohort, 3545 (71.6%) were in the NOU group and 1409 (28.4%) were in the COU group. Among the NOU cohort, 361 patients were in the intervention group and 3184 in the control group. Among the COU cohort, 149 patients were in the intervention group and 1260 in the control group (Table 1). Most patients were male, White race, with a mean (SD) age of 64 (11) years. The most common orthopedic procedure was total knee arthroplasty, followed by total hip arthroplasty. Among both NOU and COU cohorts, patients’ characteristics were similar between the pre- and postintervention period among either TPS or control cohort.

Figures 1 and 2 and eAppendix 2 depict unadjusted per-person average outpatient, inpatient, and pharmacy visits and costs incurred during the 1-year pre- and postintervention periods for the NOU and COU cohorts. Average total health care follow-up costs ranged from $40,000 to $53,000 for NOU and from $47,000 to $82,000 for COU cohort. Cost for outpatient visits accounted for about 70% of the average total costs, followed by costs for inpatient visits of about 20%, and costs for pharmacy for the remaining.

For the NOU cohort, the number of health care encounters remained fairly stable between periods except for the outpatient visits among the TPS group. The TPS group experienced an increase in mean outpatient visits in the postperiod: 30 vs 37 visits (23%) (P < .001). Mean outpatient and inpatient costs in the pre- and postintervention periods were not significantly different for either the TPS or control groups. Similarly, within the COU cohort, the TPS group experienced a 27% increase in the mean number of outpatient visits (41 vs 52 visits; P = .02) and a 39% increase in mean outpatient costs in the postintervention compared with the preintervention period ($44,682 vs $61,890; P = .02). Although the mean number of outpatient visits for control group remained at a similar level, average outpatient costs increased roughly 13% ($31,068 vs $35,148; P = .01) between the pre- and postintervention periods.

Table 2 summarizes the results from the multivariable DID analyses for the outpatient, inpatient, and pharmacy visit and cost outcomes. Here, the estimated effect of the TPS intervention is the coefficient from the interaction between the postintervention and TPS exposure indicator variables. This coefficient was calculated as the difference in the outcome before and after the TPS intervention among the TPS group minus the difference in the outcome before and after the TPS intervention among the control group. For the NOU cohort, TPS was associated with an increase in the use of outpatient health care (mean [SD] increase of 6.9 [2] visits; P < .001) after the surgery with no statistically significant effect on outpatient costs (mean [SD] increase of $2787 [$3749]; P = .55). There was no statistically significant effect of TPS on the use of inpatient visits or pharmacy, but a decrease in costs for inpatient visits among those who had at least 1 inpatient visit (mean [SD] decrease of $12,170 [$6100]; P = .02). For the COU cohort, TPS had no statistically significant impact on the use of outpatient, inpatient, or pharmacy or the corresponding costs.

DISCUSSION

TPS is a multidisciplinary approach to perioperative pain management that has been shown to reduce both the quantity and duration of opioid use among orthopedic surgery patients.^6,7 Although the cost burden of providing TPS services to prevent COU is borne by the individual health care system, it is unclear whether this expense is offset by lower long-term medical costs and health care use for COU- and OUD-related complications. In this study focused on a veteran population undergoing orthopedic joint procedures, a DID analysis of cost and health care use showed that TPS, which has been shown to reduce COU for high-risk surgical patients, can be implemented without increasing the overall costs to the VA health care system during the 1 year following surgery, even with increased outpatient visits. For NOU patients, there was no difference in outpatient visit costs or pharmacy costs over 12 months after surgery, although there was a significant reduction in subsequent inpatient costs over the same period. Further, there was no difference in outpatient, inpatient, or pharmacy costs after surgery for COU patients. These findings suggest that TPS can be a cost-effective approach to reduce opioid use among patients undergoing orthopedic joint surgery in VAMCs.

The costs of managing COU after surgery are substantial. Prior reports have shown that adjusted total health care costs are 1.6 to 2.5 times higher for previously NOU patients with new COU after major surgery than those for such patients without persistent use.¹⁶ The 1-year costs associated with new COU in this prior study ranged between $7944 and $17,702 after inpatient surgery and between $5598 and $12,834 after outpatient index surgery, depending on the payer, which are in line with the cost differences found in our current study. Another report among patients with COU following orthopedic joint replacement showed that they had higher use of inpatient, emergency department, and ambulance/paramedic services in the 12 months following their surgery than did those without persistent use.¹⁷ Although these results highlight the impact that COU plays in driving increased costs after major surgery, there have been limited studies focused on interventions that can neutralize the costs associated with opioid misuse after surgery. To our knowledge, our study is the first analysis to show the impact of using an intervention such as TPS to reduce postoperative opioid use on health care use and cost.

Although a rigorous and comprehensive return on investment analysis was beyond the scope of this analysis, these results may have several implications for other health care systems and hospitals that wish to invest in a multidisciplinary perioperative pain management program such as TPS but may be reluctant due to the upfront investment. First, the increased number of patient follow-up visits needed during TPS seems to be more than offset by the reduction in opioid use and associated complications that may occur after surgery. Second, TPS did not seem to be associated with an increase in overall health care costs during the 1-year follow-up period. Together, these results indicate that the return on investment for a TPS approach to perioperative pain management in which optimal patient-centered outcomes are achieved without increasing long-term costs to a health care system may be positive.

Limitations

This study has several limitations. First, this was a quasi-experimental observational study, and the associations we identified between intervention and outcomes should not be assumed to demonstrate causality. Although our DID analysis controlled for an array of demographic and clinical characteristics, differences in medical costs and health care use between the 2 cohorts might be driven by unobserved confounding variables.

Our study also was limited to veterans who received medical care at the VA, and results may not be generalizable to other non-VA health care systems or to veterans with Medicare insurance who have dual benefits. While our finding on health care use and costs may be incomplete because of the uncaptured health care use outside the VA, our DID analysis helped reduce unobserved bias because the absence of data outside of VA care applies to both TPS and control groups. Further, the total costs of operating a TPS program at any given institution will depend on the size of the hospital and volume of surgical patients who meet criteria for enrollment. However, the relative differences in health care use and costs may be extrapolated to patients undergoing orthopedic surgery in other types of academic and community-based health care systems.

Furthermore, this analysis focused primarily on COU and NOU patients undergoing orthopedic joint surgery. While this represents a high-risk population for OUD, the costs and health care use associated with delivering the TPS intervention to other types of surgical procedures may be significantly different. All costs in this analysis were based on 2019 estimates and do not account for the potential inflation over the past several years. Nonmonetary costs to the patient and per-person average total intervention costs were not included in the study. However, we assumed that costs associated with TPS and standard of care would have increased to an equivalent degree over the same period. Further, the average cost of TPS per patient (approximately $900) is relatively small compared with the average annual costs during 1-year pre- and postoperative periods and was not expected to have a significant effect on the analysis.

Conclusions

We found that the significant reduction in COU seen in previous studies following the implementation of TPS was not accompanied by increased health care costs.^6,7 When considering the other costs of long-term opioid use, such as abuse potential, overdose, death, and increased disability, implementation of a TPS service has the potential to improve patient quality of life while reducing other health-related costs. Health care systems should consider the implementation of similar multidisciplinary approaches to perioperative pain management to improve outcomes after orthopedic joint surgery and other high-risk procedures.

Opioid use disorder (OUD) is a significant cause of morbidity, mortality, and health care costs in the US.^1,2 Surgery can be the inciting cause for exposure to an opioid; as many as 23% of patients develop chronic OUD following surgery.^3,4 Patients with a history of substance use, mood disorders, anxiety, or previous chronic opioid use (COU) are at risk for relapse, dose escalation, and poor pain control after high-risk surgery, such as orthopedic joint procedures.⁵ Recently focus has been on identifying high-risk patients before orthopedic joint surgery and implementing evidence-based strategies that reduce the postoperative incidence of COU.

A transitional pain service (TPS) has been shown to reduce COU for high-risk surgical patients in different health care settings.^6-9 The TPS model bundles multiple interventions that can be applied to patients at high risk for COU within a health care system. This includes individually tailored programs for preoperative education or pain management planning, use of multimodal analgesia (including regional or neuraxial techniques or nonopioid systemic medications), application of nonpharmacologic modalities (such as cognitive-based intervention), and a coordinated approach to postdischarge instructions and transitions of care. These interventions are coordinated by a multidisciplinary clinical service consisting of anesthesiologists and advanced practice clinicians with specialization in acute pain management and opioid tapering, nurse care coordinators, and psychologists with expertise in cognitive behavioral therapy.

TPS has been shown to reduce the incidence of COU for patients undergoing orthopedic joint surgery, but its impact on health care use and costs is unknown.^6-9 The TPS intervention is resource intensive and increases the use of health care for preoperative education or pain management, which may increase the burden of costs. However, reducing long-term COU may reduce the use of health care for COU- and OUD-related complications, leading to cost savings. This study evaluated whether the TPS intervention influenced health care use and cost for inpatient, outpatient, or pharmacy services during the year following orthopedic joint surgery compared with that of the standard pain management care for procedures that place patients at high risk for COU. We used a difference-in-differences (DID) analysis to estimate this intervention effect, using multivariable regression models that controlled for unobserved time trends and cohort characteristics.

METHODS

This was a quasi-experimental study of patients who underwent orthopedic joint surgery and associated procedures at high risk for COU at the Veterans Affairs Salt Lake City Healthcare System (VASLCHS) between January 2016 through April 2020. The pre-TPS period between January 2016 through December 2017 was compared with the post-TPS period between January 2018 to September 2019. The control patient cohort was selected from 5 geographically diverse VA health care systems throughout the US: Eastern Colorado, Central Plains (Nebraska), White River Junction (Vermont), North Florida/South Georgia, and Portland (Oregon). By sampling health care costs from VA medical centers (VAMCs) across these different regions, our control group was generalizable to veterans receiving orthopedic joint surgery across the US. This study used data from the US Department of Veterans Affairs (VA) Corporate Data Warehouse, a repository of nearly all clinical and administrative data found in electronic health records for VA-provided care and fee-basis care paid for by the VA.¹⁰ All data were hosted and analyzed in the VA Informatics and Computing Infrastructure (VINCI) workspace. The University of Utah Institutional Review Board and the VASLCHS Office of Research and Development approved the protocol for this study.

TPS Intervention

The VASLCHS TPS has already been described in detail elsewhere.^6,7 Briefly, patients at high risk for COU at the VASLCHS were enrolled in the TPS program before surgery for total knee, hip, or shoulder arthroplasty or rotator cuff procedures. The TPS service consists of an anesthesiologist and advanced practice clinician with specialization in acute pain management and opioid tapering, a psychologist with expertise in cognitive behavioral therapy, and 3 nurse care coordinators. These TPS practitioners work together to provide preoperative education, including setting expectations regarding postoperative pain, recommending nonopioid pain management strategies, and providing guidance regarding the appropriate use of opioids for surgical pain. Individual pain plans were developed and implemented for the perioperative period. After surgery, the TPS provided recommendations and support for nonopioid pain therapies and opioid tapers. Patients were followed by the TPS team for at least 12 months after surgery. At a minimum, the goals set by TPS included cessation of all opioid use for prior nonopioid users (NOU) by 90 days after surgery and the return to baseline opioid use or lower for prior COU patients by 90 days after surgery. The TPS also encouraged and supported opioid tapering among COU patients to reduce or completely stop opioid use after surgery.

Patient Cohorts

Veterans having primary or revision total knee, hip, or shoulder arthroplasty or rotator cuff repair between January 1, 2016, and September 30, 2019, at the aforementioned VAMCs were included in the study. Patients who had any hospitalization within 90 days pre- or postindex surgery or who died within 8 months after surgery were excluded from analysis. Patients who had multiple surgeries during the index inpatient visit or within 90 days after the index surgery also were excluded. Comorbid conditions for mental health and substance use were identified using the International Classification of Diseases, 10th revision Clinical Modification (ICD-10) codes or 9th revision equivalent grouped by Clinical Classifications Software Refined (CCS-R).¹¹ Preoperative exposure to clinically relevant pharmacotherapy (ie, agents associated with prolonged opioid use and nonopioid adjuvants) was captured using VA outpatient prescription records (eAppendix 1).

The study patient cohort was stratified into either NOU or COU groups based on opioid use before surgery. Preoperative COU was defined as > 25% nonzero days (calculated using the cabinet supply method) in the 180 days before surgery admit date time (> 45 nonzero days) or ≥ 1 opioid prescription for ≥ 28-day supply released within 90 days before surgery date.^12,13 For NOU patients, the postoperative outcome of interest was new postoperative prolonged opioid use. This was defined by Page and colleagues as ≥ 1 opioid prescription released between surgery discharge date and postdischarge day 44, ≥ 1 opioid prescription released between postdischarge day 45 and 89, and ≥ 1 opioid prescription released between postdischarge day 90 and 180.¹⁴ For COU patients at the time of surgery, the postoperative outcome measure of interest was continued COU, defined as > 25% nonzero days between postdischarge day 90 and 240 (> 37 nonzero days) or ≥ 1 opioid prescription for ≥ 28-day supply released between postdischarge day 90 and 180.

Outcome Variables

Outcome variables included health care use and costs during 1-year pre- and postperiods from the date of surgery. VA health care costs for outpatient, inpatient, and pharmacy services for direct patient care were collected from the Managerial Cost Accounting System, an activity-based cost allocation system that generates estimates of the cost of individual VA hospital stays, health care encounters, and medications. Health care use was defined as the number of encounters for each visit type in the Managerial Cost Accounting System. All costs were adjusted to 2019 US dollars, using the Personal Consumption Expenditures price index for health care services.¹⁵

A set of sociodemographic variables including sex, age at surgery, race and ethnicity, rurality, military branch (Army, Air Force, Marine Corps, Navy, and other), and service connectivity were included as covariates in our regression models. In addition, ICD-9 and ICD-10 codes were used to define the indicators of whether veterans had a diagnosis of mental illness (including anxiety, bipolar disorder, depression, or trauma) or substance use (including alcohol, cannabis, opioids, or tobacco). Finally, pharmacy records were used to create indicators for prescribed opioid-relevant pharmacotherapy (including antidepressants, benzodiazepines [BZD], gabapentinoids, muscle relaxants, non-BZD sedative hypnotics) and active antidepressant drug use during the 1-year preindex period.

Statistical Analyses

Descriptive analyses were used to evaluate differences in baseline patient sociodemographic and clinical characteristics between pre- and postperiods for TPS intervention and control cohorts using 2-sample t tests for continuous variables and χ² tests for categorical variables. We summarized unadjusted health care use and costs for outpatient, inpatient, and pharmacy visits and compared the pre- and postintervention periods using the Mann-Whitney test. Both mean (SD) and median (IQR) were considered, reflecting the skewed distribution of the outcome variables.

We used a DID approach to assess the intervention effect while minimizing confounding from the nonrandom sample. The DID approach controls for unobserved differences between VAMCs that are related to both the intervention and outcomes while controlling for trends over time that could affect outcomes across clinics. To implement the DID approach, we included 3 key independent variables in our regression models: (1) an indicator for whether the observation occurred in the postintervention period; (2) an indicator for whether the patient was exposed to the TPS intervention; and (3) the interaction between these 2 variables.

For cost outcomes, we used multivariable generalized linear models with a log link and a Poisson or Υ family. We analyzed inpatient costs using a 2-part generalized linear model because only 17% to 20% of patients had ≥ 1 inpatient visit. We used multivariable negative binomial regression for health care use outcomes. Demographic and clinical covariates described earlier were included in the regression models to control for differences in the composition of patient groups and clinics that could lead to confounding bias.

RESULTS

Of the 4954 patients included in our study cohort, 3545 (71.6%) were in the NOU group and 1409 (28.4%) were in the COU group. Among the NOU cohort, 361 patients were in the intervention group and 3184 in the control group. Among the COU cohort, 149 patients were in the intervention group and 1260 in the control group (Table 1). Most patients were male, White race, with a mean (SD) age of 64 (11) years. The most common orthopedic procedure was total knee arthroplasty, followed by total hip arthroplasty. Among both NOU and COU cohorts, patients’ characteristics were similar between the pre- and postintervention period among either TPS or control cohort.

Figures 1 and 2 and eAppendix 2 depict unadjusted per-person average outpatient, inpatient, and pharmacy visits and costs incurred during the 1-year pre- and postintervention periods for the NOU and COU cohorts. Average total health care follow-up costs ranged from $40,000 to $53,000 for NOU and from $47,000 to $82,000 for COU cohort. Cost for outpatient visits accounted for about 70% of the average total costs, followed by costs for inpatient visits of about 20%, and costs for pharmacy for the remaining.

For the NOU cohort, the number of health care encounters remained fairly stable between periods except for the outpatient visits among the TPS group. The TPS group experienced an increase in mean outpatient visits in the postperiod: 30 vs 37 visits (23%) (P < .001). Mean outpatient and inpatient costs in the pre- and postintervention periods were not significantly different for either the TPS or control groups. Similarly, within the COU cohort, the TPS group experienced a 27% increase in the mean number of outpatient visits (41 vs 52 visits; P = .02) and a 39% increase in mean outpatient costs in the postintervention compared with the preintervention period ($44,682 vs $61,890; P = .02). Although the mean number of outpatient visits for control group remained at a similar level, average outpatient costs increased roughly 13% ($31,068 vs $35,148; P = .01) between the pre- and postintervention periods.

Table 2 summarizes the results from the multivariable DID analyses for the outpatient, inpatient, and pharmacy visit and cost outcomes. Here, the estimated effect of the TPS intervention is the coefficient from the interaction between the postintervention and TPS exposure indicator variables. This coefficient was calculated as the difference in the outcome before and after the TPS intervention among the TPS group minus the difference in the outcome before and after the TPS intervention among the control group. For the NOU cohort, TPS was associated with an increase in the use of outpatient health care (mean [SD] increase of 6.9 [2] visits; P < .001) after the surgery with no statistically significant effect on outpatient costs (mean [SD] increase of $2787 [$3749]; P = .55). There was no statistically significant effect of TPS on the use of inpatient visits or pharmacy, but a decrease in costs for inpatient visits among those who had at least 1 inpatient visit (mean [SD] decrease of $12,170 [$6100]; P = .02). For the COU cohort, TPS had no statistically significant impact on the use of outpatient, inpatient, or pharmacy or the corresponding costs.

DISCUSSION

TPS is a multidisciplinary approach to perioperative pain management that has been shown to reduce both the quantity and duration of opioid use among orthopedic surgery patients.^6,7 Although the cost burden of providing TPS services to prevent COU is borne by the individual health care system, it is unclear whether this expense is offset by lower long-term medical costs and health care use for COU- and OUD-related complications. In this study focused on a veteran population undergoing orthopedic joint procedures, a DID analysis of cost and health care use showed that TPS, which has been shown to reduce COU for high-risk surgical patients, can be implemented without increasing the overall costs to the VA health care system during the 1 year following surgery, even with increased outpatient visits. For NOU patients, there was no difference in outpatient visit costs or pharmacy costs over 12 months after surgery, although there was a significant reduction in subsequent inpatient costs over the same period. Further, there was no difference in outpatient, inpatient, or pharmacy costs after surgery for COU patients. These findings suggest that TPS can be a cost-effective approach to reduce opioid use among patients undergoing orthopedic joint surgery in VAMCs.

The costs of managing COU after surgery are substantial. Prior reports have shown that adjusted total health care costs are 1.6 to 2.5 times higher for previously NOU patients with new COU after major surgery than those for such patients without persistent use.¹⁶ The 1-year costs associated with new COU in this prior study ranged between $7944 and $17,702 after inpatient surgery and between $5598 and $12,834 after outpatient index surgery, depending on the payer, which are in line with the cost differences found in our current study. Another report among patients with COU following orthopedic joint replacement showed that they had higher use of inpatient, emergency department, and ambulance/paramedic services in the 12 months following their surgery than did those without persistent use.¹⁷ Although these results highlight the impact that COU plays in driving increased costs after major surgery, there have been limited studies focused on interventions that can neutralize the costs associated with opioid misuse after surgery. To our knowledge, our study is the first analysis to show the impact of using an intervention such as TPS to reduce postoperative opioid use on health care use and cost.

Although a rigorous and comprehensive return on investment analysis was beyond the scope of this analysis, these results may have several implications for other health care systems and hospitals that wish to invest in a multidisciplinary perioperative pain management program such as TPS but may be reluctant due to the upfront investment. First, the increased number of patient follow-up visits needed during TPS seems to be more than offset by the reduction in opioid use and associated complications that may occur after surgery. Second, TPS did not seem to be associated with an increase in overall health care costs during the 1-year follow-up period. Together, these results indicate that the return on investment for a TPS approach to perioperative pain management in which optimal patient-centered outcomes are achieved without increasing long-term costs to a health care system may be positive.

Limitations

This study has several limitations. First, this was a quasi-experimental observational study, and the associations we identified between intervention and outcomes should not be assumed to demonstrate causality. Although our DID analysis controlled for an array of demographic and clinical characteristics, differences in medical costs and health care use between the 2 cohorts might be driven by unobserved confounding variables.

Our study also was limited to veterans who received medical care at the VA, and results may not be generalizable to other non-VA health care systems or to veterans with Medicare insurance who have dual benefits. While our finding on health care use and costs may be incomplete because of the uncaptured health care use outside the VA, our DID analysis helped reduce unobserved bias because the absence of data outside of VA care applies to both TPS and control groups. Further, the total costs of operating a TPS program at any given institution will depend on the size of the hospital and volume of surgical patients who meet criteria for enrollment. However, the relative differences in health care use and costs may be extrapolated to patients undergoing orthopedic surgery in other types of academic and community-based health care systems.

Furthermore, this analysis focused primarily on COU and NOU patients undergoing orthopedic joint surgery. While this represents a high-risk population for OUD, the costs and health care use associated with delivering the TPS intervention to other types of surgical procedures may be significantly different. All costs in this analysis were based on 2019 estimates and do not account for the potential inflation over the past several years. Nonmonetary costs to the patient and per-person average total intervention costs were not included in the study. However, we assumed that costs associated with TPS and standard of care would have increased to an equivalent degree over the same period. Further, the average cost of TPS per patient (approximately $900) is relatively small compared with the average annual costs during 1-year pre- and postoperative periods and was not expected to have a significant effect on the analysis.

Conclusions

We found that the significant reduction in COU seen in previous studies following the implementation of TPS was not accompanied by increased health care costs.^6,7 When considering the other costs of long-term opioid use, such as abuse potential, overdose, death, and increased disability, implementation of a TPS service has the potential to improve patient quality of life while reducing other health-related costs. Health care systems should consider the implementation of similar multidisciplinary approaches to perioperative pain management to improve outcomes after orthopedic joint surgery and other high-risk procedures.

The importance of formalized antimicrobial stewardship programs (ASPs) has gained recognition over the past 2 decades. The increasing requirements for ASP programs from national entities often outpace the staffing, technology, and analytic support needed to meet these demands.^1,2 A multimodal approach to stewardship that includes education initiatives, audit-and-feedback methodology, and system support is effective in producing sustained change.³ However, this approach is resource intensive, and many ASPs must look outward for additional support.

Centralized ASP collaboratives and stewardship networks have been effective in positively impacting initiatives and outcomes through resource sharing.^3-5 These collaboratives can take on multiple forms ranging from centralized education distribution to individual sites coming together to set goals and develop strategies to address common issues.^5-8 Collaboratives can provide enhanced data analysis through data pooling, which may lead to shared dashboards or antibiotic use (AU) reports, allowing for robust benchmarking.^5-7 Productivity at individual centers is often measured by AU and antimicrobial resistance (AMR) rates, but these measures alone do not fully capture the benefits of collaborative participation.

The US Department of Veterans Affairs (VA), similar to other large health care systems, is uniquely positioned to promote the development of ASP collaboratives due to the use of the same electronic health record system and infrastructure for data. This centralized data lends itself more readily to data dashboards and interfacility comparison. In turn, the identification of facilities that have outlying data for specific measures can lead to a collaborative effort to identify aberrant processes or facility-specific problems and identify, implement, and track the progress of appropriate solutions with less effort and resources.⁷ The VA has a national stewardship group, the Antimicrobial Stewardship Task Force (ASTF), that identifies and disseminates best practices and advocates for stewardship resources.

VA facilities are heterogeneous with regard to patient population, services, availability of specialists, and antibiotic resistance patterns.⁹ Therefore, clinical practice and needs vary. The ASTF has spearheaded the development of regional collaboratives, recognizing the potential benefit of smaller groups with shared leadership.The Veterans Integrated Services Networks (VISNs) are geographically demarcated regions that lend themselves well to coordination among member facilities due to similar populations, challenges, and opportunities. The Veterans Affairs Midsouth Healthcare Network (VISN 9) includes 5 facilities across Tennessee, Kentucky, Mississippi, Arkansas, Georgia, Virginia, and Indiana and serves about 293,000 veterans, ranging from 35,000 to 105,000 per facility.

A VISN 9 stewardship collaborative (as described by Buckel and colleagues in 2022) was established to enhance member facility ASPs through shared goal setting.⁶ Initially, the collaborative met quarterly; however, with increased participation and the onset of COVID-19, the collaborative evolved to meet burgeoning ASP needs. While intrafacility multidisciplinary ASP collaboration has been previously published, few publications on interfacility collaborations exist.^3-6 To our knowledge, no previous publications have reported the impact of a VA ASP collaborative on the productivity and effectiveness of participating ASP facilities and the region. We aim to share the structure and processes of this ASP collaborative, demonstrate its impact through quantification of productivity, and aid others in developing similar collaboratives to further ASPs’ impact.

Methods

The regional VISN 9 ASP collaborative was formed in January 2020 to address common issues across facilities and optimize human capital and resources. The initial collaborative included ASP pharmacists but quickly expanded to include physicians and nurse practitioners. The collaborative is co-led by 2 members from different facilities that rotate.

In April 2021, clinical guidance and research/quality improvement (QI) subcommittees were created. The monthly research/QI subcommittee discusses current initiatives and barriers to ongoing research, adapt and disseminate successful interventions to other facilities, and develop new collaborative initiatives. The clinical guidance subcommittee creates and disseminates clinical expert recommendations regarding common issues or emerging needs.

Data Plan and Collection

To measure success and growth, we evaluated annual facility reports that convey the state of each facility’s ASP, outline its current initiatives and progress, highlight areas of need, and set a programmatic goal and strategy for the upcoming year. These reports, required by a VA directive, are submitted annually by each facility to local and VISN leadership and must address the following 7 areas: (1) ASP structure and fulfillment of national VA policy for ASP; (2) fulfillment of the Joint Commission ASP standards; (3) ASP metrics; (4) ASP activities and interventions; (5) ASP QI and research initiatives; (6) education; and (7) goals and priorities.

To standardize evaluation and accurately reflect ASP effort across heterogeneous reports, 4 core areas were identified from areas 1, 3, 4 and 5 listed previously. Area 2 was excluded for its similarity among all facilities, and areas 6 and 7 were excluded for significant differences in definitions and reporting across facilities.

The project team consisted of 5 members from the collaborative who initially discussed definitions and annual report review methodology. A subgroup was assigned to area 1 and another to areas 3, 4, and 5 for initial review and data extraction. Results were later reviewed to address discrepancies and finalize collation and presentation. 
The impact of the collaborative on individual facilities was measured by both quantitative and qualitative measures. Quantitative measures included: (1) designated ASP pharmacy, physician, or advanced practice provider (APP) full-time equivalents (FTE) at each facility compared with the recommended FTE for facility size; (2) the number of inpatient and outpatient ASP AU metrics for each facility and the VISN total; (3) reported improvement in annual ASP metrics calculated as frequency of improved metrics for each facility and the VISN; (4) the number of QI or research initiatives for each facility and the VISN, which included clinical pathways and order sets; and (5) the number of initiatives published as either abstract or manuscript.¹⁰ Additionally, the number of collaborative efforts involving more than 1 facility was tracked. Qualitative data included categories of metrics and QI and research initiatives. Data were collected by year and facility. Facilities are labeled A to E throughout this article.

Along with facility annual ASP reports, facility and VISN AU trends for fiscal years (FY) 2019-2022 were collected from existing VA dashboards tracking AU in both acute respiratory infections (ARI) and in patients with COVID-19. Quantitative data included facility and VISN quarterly AU rates for ARI, extracted from the national VA dashboard. Facility and VISN AU rates in patients with COVID-19 were extracted from a dashboard developed by the VISN 9 ASP collaborative. The VISN 9 Institutional Review Board deemed this work QI and approval was waived.

Results

In 2019, only 2 sites (A and C) reported dedicated FTE compared with recommended minimum staffing; neither met minimum requirements. In 2020, 1 facility (B) met the physician FTE recommendation, and 2 facilities met the pharmacy minimum FTE (D and E). In 2021 and 2022, 2 of 5 facilities (B and E) met the physician minimum FTE, and 2 of 5 (D and E) met the minimum pharmacy FTE recommendations. For the study years 2019 to 2022, 1 facility (E) met both pharmacy and physician FTE recommendations in 2021 and 2022, and 2 facilities (A and C) never met minimum FTE recommendations.

Regarding ASP metrics, all facilities tracked and reported inpatient AU; however, facility A did not document inpatient metrics for FY 2021. The number of individual inpatient metrics varied annually; however, FY 2022 saw the highest reported for the VISN (n = 40), with a more even distribution across facilities (Figure 1). Common metrics in 2022 included total AU, broad-spectrum gram-negative AU, anti–methicillin-resistant Staphylococcus aureus (MRSA) agent use, antibiotics with high risk for Clostridioides difficile infection (CDI), and AU in patients with COVID-19. The percentage of improved metrics for VISN 9 was consistent, ranging from 26.5% to 34.8%, throughout the study period.

From 2019 to 2022, facilities reporting outpatient AU increased from 3 to 5 and included fluoroquinolone use and AU in ARI. VISN 9 outpatient metrics increased every year except in 2021 with improved distribution across facilities. The number of total metrics with reported improvement in the outpatient setting overall increased from 3 of 11 (27%) in 2019 to 20 of 33 (60%) in 2022.

Antimicrobial Stewardship Initiatives

Quantitative and qualitative data regarding initiatives are reported in Figure 2 and the eAppendix respectively. Since the formation of the collaborative, total initiatives increased from 33 in 2019 to 41 in 2022. In 2019, before the collaborative, individual facilities were working on similar projects in parallel, which included MRSA decolonization (A and C), surgical prophylaxis (A and E), asymptomatic bacteriuria (A and C), and CDI (B, C, D, and E). The development of clinical pathways and order sets remained consistent, ranging from 15 to 19 throughout the study period except for 2020, when 33 clinical pathways and/or order sets were developed. Collaboration between sites also remained consistent, with 1 shared clinical pathway and/or order menu between at least 1 site reported yearly for 2020, 2021, and 2022. The number of publications from VISN 9 grew from 2 in 2019 to 17 in 2022. In 2019, there were no collaborative research or QI publications, but in 2022 there were 2 joint publications, 1 between 2 facilities (A and C) and 1 including all facilities.

ARI and COVID-19 were identified by the collaborative as VISN priorities, leading to shared metrics and benchmarking across facilities. From 2019 to 2022, increased collaboration on these initiatives was noted at all facilities. The ARI goal was established to reduce inappropriate prescribing for ARI/bronchitis to under 20% across VISN 9. Rates dropped from 50.3% (range, 35.4%-77.6%) in FY 2019 quarter (Q) 1 to 15% (range, 8%-18.3%) in FY 2022 Q4. The clinical guidance subcommittee developed a guideline for AU in patients with COVID-19 that was approved by the VISN 9 Pharmacy & Therapeutics Committee. A VISN 9 dashboard was developed to track inpatient and outpatient AU for COVID-19. Antibiotic prescribing in the first 4 days of hospitalization decreased from 62.2% at the start of the COVID-19 pandemic to 48.7% after dissemination of COVID-19 guidance.

Discussion

This study demonstrates the benefit of participating in a regional ASP collaborative for individual facilities and the region. Some products from the collaborative include the development of regionwide guidance for the use of antimicrobials in COVID-19, interfacility collaborative initiatives, a COVID-19 dashboard, improvement in metrics, and several publications. Importantly, this expansion occurred during the COVID-19 pandemic when many ASP members were spread thin. Moreover, despite 4 sites not meeting VA-recommended ASP staffing requirements for both pharmacists and physicians, productivity increased within the VISN as facilities worked together sharing local challenges and successful paths in removing ASP barriers.The collaborative shared QI strategies, advocated for technological support (ie, Theradoc and dashboards) to maximize available ASP human capital, standardized metric reporting, and made continued efforts sustainable. VA ASTF disseminates evidence-based practice but is not designed to develop tailored site-specific interventions, which has led to the support of VISN-level collaboratives to serve local facilities’ needs. We postulate the use of a collaborative as a beneficial strategy to increase productivity and achieve local goals with limited resources.

Previous reports in the literature have found ASP collaboratives to be an effective model for long-term program growth.³ Two collaboratives found improved adherence to the Centers for Disease Control and Prevention core elements for ASP.^4,5Similar to our findings, other collaboratives noted a reduction in AU after implementation, although statistical analysis of improvement over time was not performed to verify significance.^3-5,7 One VA study reviewed the use of dashboards with a monthly learning collaborative and identified a reduction in AU.⁷ However, the structure of our ASP collaborative was through joint meetings and projects, as defined by Buckel and colleagues.⁶

Our findings highlight that ASP collaboratives can help answer the recent call to action from McGregor, Fitzpatrick, and Suda who advocated for ASPs to take the next steps in stewardship, which include standardization of evaluating metrics and the use of robust QI frameworks.¹¹ Moving forward, an area for research could include a comparison of ASP collaborative infrastructures and productivity to identify optimal fit dependent on facility structure and setting. Parallel to our experience, other reports cite heterogeneous ASP metrics and a lack of benchmarking, spotlighting the need for standardization.^8,11,12The VA and other health care facilities would benefit from national benchmarking of AU metrics to make comparisons across sites beneficial.

Limitations

Using annual reports was a limitation for analyzing and reporting the full impact of the collaborative. Local facility-level discretion of content inclusion led to many facilities only reporting on the forefront of new initiatives that they had developed and may have led to the omission of other ongoing work. Further, time invested into the ASP regional collaborative was not captured within annual reports; therefore, the opportunity cost cannot be determined.

Conclusions

The VA has an advantage that many private health care facilities do not: the ability to work across systems to ease the burden of duplicative work and more readily disseminate effective strategies. The regional ASP collaborative bred innovation and the tearing down of silos. The implementation of the collaborative aided in robust QI infrastructure, standardization of reporting and metrics, and greater support through facility alignments with regional guidance. ASP interfacility collaboratives provide a sustainable solution in a resource-limited landscape.

Acknowledgments

This work was made possible by the resources provided through the Antimicrobial Stewardship Programs in the Veterans Integrated Services Network (VISN) 9.