COST EFFECTIVENESS OF A PHYSICIAN-PHARMACIST COLLABORATION INTERVENTION TO IMPROVE BLOOD PRESSURE CONTROL

Abstract

Previous studies have demonstrated the cost-effectiveness of physician-pharmacist collaborations to improve hypertension control. However, most studies have limited generalizability: lacking minority and low-income populations.

The Collaboration Among Pharmacist and Physicians to Improve Blood Pressure Now trial randomized 625 patients from 32 medical offices in 15 states. Each office had an existing clinical pharmacist on staff. Pharmacists in intervention offices communicated with patients and made recommendations to physicians about changes in therapy. Demographic information, blood pressure, medications and physician visits were recorded. In addition, pharmacists tracked time spent with each patient. Costs were assigned to medications, and pharmacist and physician time. Cost-effectiveness ratios were calculated based on changes in blood pressure measurements and hypertension-control rates.

Thirty-eight percent of patients were black, 14% were Hispanic, and 49% had annual income <$25,000. At 9 months, average systolic blood pressure was 6.1 mm Hg lower (+/− 3.5), diastolic was 2.9 mm Hg lower (+/− 1.9), and the percentage of patients with controlled hypertension was 43% in the intervention group and 34% in the control group. Total costs for the intervention group were $1462.87 (+/− 132.51), and $1259.94 (+/− 183.30) for the control group, a difference of $202.93. The cost to lower blood pressure by 1 mmHg was $33.27 for systolic blood pressure and $69.98 for diastolic blood pressure. The cost to increase the rate of hypertension control by one percentage point in the study population was $22.55. Our results highlight the cost-effectiveness of a clinical pharmacy intervention for hypertension control in primary care settings.

Background

An estimated 29% of adults are hypertensive. (1, 2) In the United States, hypertension has the greatest risk for all-cause mortality of any modifiable risk factor (3) and is the most common cause of cardiovascular deaths worldwide. (4) Anti-hypertensive therapies reduce the risk of strokes, kidney and heart disease and mortality (5) Furthermore, these therapies are cost effective. (6, 7)

Lifelong therapy for hypertension is usually required and represents one of the most common reasons patients take medications chronically. (8) However, the initiation of therapy is often not sufficient to establish effective blood pressure (BP) control. Patients need to be monitored at regular intervals over time, and titration of medication is often needed. (1) Despite widespread treatment of hypertension, only 50% of patients with hypertension achieve BP control, and poor control has been documented for the past several decades. (2, 9-11)

Patients diagnosed with hypertension are not optimally treated for a variety of reasons. First, inadequate control can be caused by clinical inertia: physicians may be reluctant to add drugs or increase dosages. (12-14) For example, one study showed that in patients with documented evidence of more than 6 months of uncontrolled hypertension, BP medications were started or changed only 38% of the time. (15) In addition, physicians’ busy clinical workloads and patients with multiple other symptomatic complaints can distract the physician and patient, preventing achievement of effective BP control. (16) Finally, poor hypertension control can result from poor adherence to prescribed therapy on the part of patients who may fail to take the medication or take it intermittently. (17)

Team-based care has been shown to be effective for improving BP control. (18-22) Teams with nurses or pharmacists have improved BP control, but the strongest evidence is with pharmacists. (20, 23) A recent review showed that interventions involving pharmacists resulted in average decreases of 7.6 mm Hg in systolic and 3.9 mm Hg in diastolic BP. (21) However, questions about the generalizability, and cost-effectiveness of these interventions remain. The Collaboration Among Pharmacist and Physicians to Improve Blood Pressure Now (CAPTION) trial was designed to implement pharmacist-physician collaboration in primary care offices among diverse populations of patients. (22) The purpose of the present study is to examine the cost effectiveness of the intervention implemented in the CAPTION Trial.

Methods

Data Sources

The main results from the CAPTION trial have previously been published. (22) The study included 32 medical offices in 15 states. Randomization occurred at the medical-office level (i.e., all subjects in each medical office were in the same study arm), and offices were stratified based on the number of minority patients and their score on our pharmacy-structure survey. (24) Offices were randomized to one of 3 groups: a 9-month BP intervention, a 24-month BP intervention, or usual care. The intervention was designed to be identical in the two intervention groups for the first 9-months. The primary endpoint was BP control at 9-months, and this period was used in the present analyses. The study was approved by the University of Iowa Institutional Review Board (IRB) and the IRBs affiliated with each medical office. All subjects provided written informed consent.

Patients were included if they spoke either English or Spanish, were at least 18 years old, and had uncontrolled hypertension defined as BP ≥ 140 mmHg systolic or ≥ 90 mmHg diastolic. For patients with diabetes or chronic kidney disease, these cut offs were ≥ 130 and ≥ 80 mmHg, respectively, based on JNC-7 recommendations. (25)

A study coordinator (SC) employed or affiliated with each office recruited subjects and collected data. The SC measured BP in the sitting position after appropriate rest using standard research technique at baseline, 6 and 9 months.(19, 26, 27) At least 3 BP measurements were taken using an automated device. The initial value was not used. Two subsequent BPs were obtained a minute apart and were averaged if they were within 4 mmHg. If more than 4 mmHg different, a fourth BP was obtained and the two closest values were averaged using the African American Study of Kidney Disease (AASK) trial procedures. (27) The SC collected the following at baseline: height, weight, pulse, the duration of hypertension, presence of other cardiovascular risk factors, symptoms and adverse drug reactions, socio-demographics, co-morbidities, current anti-hypertensive medications and dose.

The intervention involved a medical record review by the pharmacist and an initial patient interview, including: 1) a medication history; 2) an assessment of knowledge of BP medications, dosages and timing, and potential side effects; 3) and other barriers to BP control (e.g., side effects, non-adherence).Lifestyle modifications were discussed as well when appropriate. (28) The remaining intervention included a telephone call from the pharmacist at 2 weeks, structured face-to-face visits between patients and pharmacists at baseline, 1, 2, 4, 6 and 8 months, and additional visits if BP remained uncontrolled. This implementation trial did not require strict adherence to this protocol, but all pharmacist visits were tracked. The pharmacist created a care plan and made recommendations to the physician to adjust therapy. (19, 29, 30) Because the pharmacists worked in the medical offices, most pharmacist-physician communication was face-to-face. Physicians were free to accept, reject or modify any recommendation. Recommendations to patients focused on medication education, improving adherence, and strategies to implement lifestyle modifications.

After 9 months of this intervention, average systolic blood pressure was 6.1 mm Hg lower (+/− 3.5), diastolic was 2.9 mm Hg lower (+/− 1.9), and hypertension control was 43% in the intervention group and 34% in the control group.. Adverse events were uncommon, and there were no overall differences in the frequency of subjects reporting any serious events among the interventions and control groups.

Measures

Provider Time

The perspective of this cost analysis is society in general. For each patient encounter, pharmacists recorded the number of minutes spent in the following activities: medical record review, consultation, patient assessment, ordering laboratory tests, ordering medications, medical education, lifestyle education, BP measurement education, making recommendations, and documentation in the medical record. In addition, the beginning and ending time of the encounter was recorded. Each patient visit with a physician in the office was documented, but not the specific length of the visit nor the services provided, and these were billed by the office.

Provider Costs

Patient-specific pharmacist costs were estimated by multiplying patient-specific pharmacist time (ending time – beginning time) by the average compensation rate ($56.01/hour). All pharmacist encounters were included. Physician costs were estimated by multiplying the average amount of time spent in physician consultation with patients (20.3 minutes or 0.338 hours over 9 months) (31) by the average compensation rate for family practice physicians ($88.43/hour). Unlike pharmacist time, we did not measure physician time specifically but obtained these times from another survey (31). The pharmacist and physician compensation rates were obtained from the 2013 Occupational Employment Statistics (OES) Survey, Bureau of Labor Statistics (32).

Drug Costs

Anti-hypertensive drug costs were estimated using the sum of costs for baseline antihypertensive agents, and changed antihypertensive agents, specifically, those prescribed after baseline.

Generic prices were used when available. The drug cost per prescription for each patient was first calculated by multiplying the Average Wholesale Price (AWP) and/or Average AWP (AAWP) price with frequency and dose, and all prescription costs were added to obtain the total anti-hypertensive drug cost per patient. Drug prices were obtained from Lexicomp Online™ (https://online.lexi.com), which provides AWP prices for drugs. Drug names, unit strengths, and doses were used to determine drug prices. Laboratory tests were not recorded for this study, but in our previous study, the costs for these tests did not significantly differ between the intervention and control groups.(31)

Cost Calculations

We determined the cost of the intervention by subtracting the average costs for the intervention groups from the average costs for the control group. The incremental cost-effectiveness ratio was calculated as the additional costs of the intervention divided by the change in both systolic and diastolic BP related to the intervention. We also examined BP control. We divided the additional costs of the intervention by the percentage of subjects who achieved BP control as a result of the intervention.

We conducted two sensitivity analyses. First, some patients did not finish the 9-month intervention. So, as a sensitivity analysis, costs and incremental cost effectiveness ratios were also estimated for only those who finished at least 9 months of the intervention. Second, AWP often overstates the cost of drugs. The Kaiser Family Foundation estimates that AWP are, on average 17% higher than weighted acquisition costs for brand-name drugs, and 80% higher for generic drugs.(32) As a second sensitivity analysis, we deflated our drug costs and recalculated the incremental cost effectiveness ratios.

Statistical Analysis

For continuous variables, a generalized linear regression model was used to compare the means across groups. For categorical variables a generalized estimating equation model was used to compare the proportions across groups. Both analyses accounted for the correlation within a center. There was one exception: insurance status had four categories so a fisher's exact chi-squared test was used. SAS software, version 9.3 (SAS Institute Inc., Cary, NC) was used for all data analysis. More specific information can be found elsewhere (22)

Results

Figure 1 displays the distribution of subjects in this study. Analysis includes 625 patients: there were 401 patients in the intervention group, and 224 in the control group. The mean age was 61 years(+/− 1.01). A majority of patients were female (60.3%), but minorities were well-represented: 38.4% were black or African-American, and 14.2% were Hispanic or Latino. Nearly half of patients were diagnosed with diabetes (47.7%) and one fourth of patients were diagnosed with hyperlipidemia (26.1%). Nearly half of patients received a post-high school education (47.0%) and were married (46.9%). The majority of patients (84.2%) had insurance coverage: the most common sources of coverage were private insurance (38.6%), Medicare (30.2%) and Medicaid (13.9%). Low-income patients were well-represented, with 22.9% of patients earning less than $10,000 in household income annually. The only statistically significant differences between groups were seen for percent married (p=0.030) and type of insurance coverage (p<0.0001).

Figure 1.

Distribution of Study Subjects

We previously reported an average of 4.9 BP-medication changes over 9 months for intervention patients. (28) Patients in the usual care group averaged 1.1 BP medication changes (p=0.0003).

Table 1 presents descriptive statistics of patient-specific costs. There were 3,302 prescriptions for baseline antihypertensive agents where dose, strength, and drug name information were available. Patient-specific drug costs for the entire 9-month period ranged from $2.12 to $12,341.19 (mean: $1196.08 +/− 101.67), and 9 patients were taking no medications. The mean total drug cost per patient was slightly higher in the intervention groups, but not significantly so: $1223.91 (+/− 6.18) per patient on average in the intervention group and $1146.27 (+/− 11.85) in the control group (p=0.7848). However the patients in the intervention group had more of their medications changed during the course of the study, and this resulted in significantly higher costs for changed hypertension medications: $272.45 (+/− 64.39) per patient in the intervention group and $170.75 (+/− 51.63) in the control group (p=0.0352). Therefore, the cost differential for medications to provide this intervention for 9 months was $101.70 compared to the usual care group.

Table 1.

Cost comparison^*

	All Subjects (n=625)		Intervention (n=401)		Control (n=224)
Variable	Mean	Std. Dev.	Mean	Std. Dev.	Mean	Std. Dev.	p-value
Changed Hypertension Medications	236	579.12	272.45	657.9	170.75	394.23	0.0352
Hypertension Medications	960.08	1055.27	951.46	933.53	975.52	1245.99	0.7848
Total Drug Cost	1196.08	1296.85	1223.91	63.14	1146.27	90.51	0.4715
Pharmacist Cost	90.22	110.28	140.62	108.93	0	0	<0.001
Physician Cost	103.83	136.16	98.34	140.54	113.67	127.66	0.1774
Total Cost	1390.14	1372.81	1462.87	1353.86	1259.94	1399.70	0.0759

^*All costs are in U.S. dollars

The pharmacists had 2,036 encounters for 360 patients over a 9-month period. The total time each patient spent with the pharmacist ranged from 15 to 1,044 minutes, with an average of 155 minutes, resulting in an average of $140.62 (+/− 10.66) in pharmacist costs per patient. There were 439 patients who had at least one visit with a physician in the medical office, and the maximum number of physician visits per patient was 36. A larger percentage of subjects in the control group had a physician visit (82% vs. 64%), and the median number of visits for the control group was higher (3 vs. 2 per subject). This resulted in lower average costs for physician visits in the intervention group ($98.34 +/− 13.76 vs. $113.67 +/− 16.72), but these costs are not statistically significant (p = 0.1774).

Total costs, which are the sum of drug costs, physician time and pharmacist costs, ranged from $8.48 to $13074.64, and 5 patients had no costs during the study period. Mean total costs were $1462.87 (+/− 132.51) in the intervention group and $1259.94 (+/− 183.30) in the control group. Thus, the total additional cost to provide the intervention for 9 months compared to usual care was $202.93.

Compared to the control group, the cost to lower BP by 1 mmHg was $202.93 / 6.1 = $33.27 for systolic BP and $202.93 / 2.9 = $69.98 for diastolic BP. Comparing rates in the intervention and control groups, the cost to increase BP control by one percentage point was $202.93 / 9.0 = $22.55.

We performed a sensitivity analysis, and included only those patients who completed the 9-month intervention. The results of this analysis is given in Table 2. Five-hundred thirty-nine patients were included, and these costs were slightly lower than those for the full sample. (A few patients with very high drug costs did not have a 9-month visit, for example.) However, the difference between the costs in the intervention groups and the control group were slightly higher than in the full sample. For those with a 9-month visit, the cost to lower BP by 1 mmHg was $236.8/ 6.1 = $38.82 for systolic and $236.8/ 2.9 = $81.66 for diastolic. Comparing rates in the intervention and control groups, the cost to increase BP control by one percentage point was $236.8/ 9.0 = $26.31.

Table 2.

Cost comparison, sensitivity analysis – only patients who completed the 9-month intervention^*

	All Subjects (n=539)		Intervention (n=345)		Control (n=194)
Variable	Mean	Std. Dev.	Mean	Std. Dev.	Mean	Std. Dev.	p-value
Changed Hypertension Medications	218.54	460.89	251.3	493.2	160.3	391.5	0.028
Hypertension Medications	849.95	886.10	856.9	828.8	837.5	981.9	0.808
Total Drug Cost	1068.49	1045.14	1108.2	1013.1	997.8	1098.9	0.250
Pharmacist Cost	91.86	106.70	143.5	101.8	0	0	<0.001
Physician Cost	94.20	99.65	88.02	105.1	105.2	88.34	0.055
Total Cost	1254.55	1089.34	1339.8	1064.8	1103.0	1118.4	0.017

^*All costs are in U.S. dollars. This analysis only includes those subjects who had complete data at 9-months.

Because AWP tend to overstate drug prices, our second sensitivity analysis used deflated drug costs, and the results are shown in Table 3. Total drug costs fell to $328.27 (+/− 29.3) in the intervention group and $291.97 (+/− 41.33) in the control group. Subsequently, total costs fell to $559.80 (+/− 36.40) in the intervention group and $397.15 (+/− 45.01) in the control group. The resulting cost to lower BP by 1 mm Hg was $161.88/ 6.1 = $26.54 for systolic and $161.88/ 2.9 = $55.82 for diastolic. The cost to increase BP control by one percentage point was $161.88/ 9.0 = $17.99.

Table 3.

Cost comparison, sensitivity analysis using deflated drug costs^*

	All Subjects (n=539)		Intervention (n=345)		Control (n=194)
Variable	Mean	Std. Dev.	Mean	Std. Dev.	Mean	Std. Dev.	p-value
Changed Hypertension Medications	45.13	95.43	51.74	102.24	33.39	80.87	0.032
Hypertension Medications	270.06	250.12	276.52	240.50	258.58	266.62	0.425
Total Drug Cost	315.20	283.80	328.27	277.65	291.97	293.72	0.152
Pharmacist Cost	91.86	106.70	143.5	101.8	0	0	<0.001
Physician Cost	94.20	99.65	88.02	105.1	105.2	88.34	0.055
Total Cost	501.26	344.79	559.80	344.92	397.15	319.87	<0.001

^*All costs are in U.S. dollars. This analysis only includes those subjects who had complete data at 9-months.

Discussion

Our results demonstrate that a pharmacist-physician collaborative intervention specifically designed to use existing clinical pharmacists located in primary care offices was effective. Pharmacists spent approximately 2 hours with patients over a 9-month period of time. This additional pharmacist care resulted in statistically and clinically significant reductions in BP compared to the control group.

A criticism of previous physician-pharmacist BP interventions relates to their generalizability to non-research practice settings. A major strength of the CAPTION Trial was that patients with complex medical problems, and primary care providers caring for patients with lower socio-economic status were specifically recruited, to help ensure that the results were more generalizable. In fact, 38% of the patients were African American, and 23% of all patients had annual household income less than $10,000. Almost half of the population was diabetic, and many patients had multiple morbidities. Prior physician-pharmacist BP interventions have relied upon small numbers of intervention providers and few clinics or medical offices (33). In contrast, the present study included 26 intervention pharmacists, many of whom had worked in the medical offices for over 10 years. Our results will help health systems determine cost-effective strategies to implement value-based population-health strategies encouraged by the Affordable Care Act, insurers and employers.

In this study we found the cost of the intervention to be relatively modest. With a small amount of time spent by the pharmacist – around 2 hours over a 9-month period of time, significant reductions in BP were observed. Specifically, we observed a decrease in BP of 6.1 mm Hg systolic (+/− 3.5) and 2.9 mm Hg diastolic (+/−1.9). Without considering economies of scale or positive spillovers, if we extrapolate these findings, adding a pharmacist to a medical office that does not currently employ one could result in improvements for many hypertensive patients. Specifically, a pharmacist who works 40 hours per week and sees each patient for 2 hours every 9 months could manage 720 hypertensive patients in collaboration with physicians. An unexpected finding was that the intervention group had fewer physician visits, thus saving physician time for other, more complex patient health problems. With insufficient primary care providers to care for the population, this study demonstrates that pharmacist-physician collaboration can improve care while allowing primary care physicians to focus on other important medical issues.

Our study has several limitations. First, our results may not be generalizable to other practice settings. But, given the pragmatic nature of our study (using existing medical office personnel) and the oversampling of subjects in groups traditionally underrepresented in trials, the results of the CAPTION Trial are more generalizable than previous physician-pharmacist BP interventions. Not all primary care medical offices have an onsite clinical pharmacist, and this limits the extent to which this work can be implemented. Second, our cost estimation did not account for overhead or interruptions to clinic workflow, and we did not measure the opportunity cost of the pharmacist. If the time spent with patients was easy to incorporate into the pharmacist's day, the costs of this study would be lower than estimated. If pharmacists neglected other duties while attending to the patients in this study, the costs could be higher. In busy practices, space for meeting with patients or making confidential calls may be scarce, making this intervention more difficult to implement. In addition, because all the medical offices in this study employed pharmacists, patients in the control group might have spent time with the pharmacist as well. These costs were not included. Also, we did not calculate patient time in this study: the number of visits per patient was non-trivial and would have resulted in substantial costs for some patients. Third, patients were enrolled in this study for 24 months. We chose to use only 9 months for this study because the two intervention arms were the same for the first 9 months and because this time point was used for the primary outcome in the CAPTION Trial. (22) Also, this time point allowed us to maximize the number of subjects that we could include in our analysis. Although costs continued to accrue for these patients, some patients’ BP did not continue to improve. (22) Thus, costs to sustain this type of intervention would probably be higher than reported here. Fourth, we used retail drug prices, which, due to insurance, are probably not the costs for most patients in this study (out-of-pocket payments plus insurance payments). However, this leads to overestimates in both the intervention and control groups and should not bias our results. Fifth, our investigation focused only on hypertension and did not investigate other health outcomes. However, the effect of our intervention could result in positive spillover effects: pharmacist attention to hypertension could lead to improvements in other areas of a patient's health. Finally, our outcome was cost per decrease in mm Hg of BP, not quality-adjusted-life years (QALYs), which is the standard outcome in cost effectiveness analyses. Unfortunately, we could not calculate QALYs for this study because we did not collect long-term outcomes for the patients in this study.

Despite our limitations, we are confident that the effect of the intervention is clinically meaningful. Although a 6.1 mm Hg reduction may seem small from a numeric standpoint, it is clinically meaningful if it can be sustained. An analysis of 30 clinical trials demonstrated that a 5mm Hg difference in SBP over 3 to 5 years reduces the risk of all cardiovascular complications and stroke by 25% to 30%. (34) Hypertension is responsible for an estimated 395,000 deaths per year in the US, (35) many of which could be prevented with better BP control. Future efforts should consider reduction in morbidity and mortality that follow the treatment effects we observed in the CAPTION Trial. Such estimates must, of course, make assumptions regarding future BP control following the cessation of the intervention.

Perspectives

This study demonstrated the low cost of expanding a pharmacist-physician collaborative hypertension intervention. Prior studies have shown that similar interventions using research personnel are cost effective. However, the CAPTION trial is more pragmatic in nature and demonstrates cost effectiveness in a broader patient population. Our results highlight the cost-effectiveness of clinical pharmacists in primary care settings when increased attention is being focused on value-based care.

Table 4.

BP Results

Cost category	Intervention Mean (Std. Dev.) N = 401	Control Mean (Std. Dev.) N = 224	Difference Between Intervention and Control Groups	P-value
Baseline systolic BP	148.9 (14.8)	149.7 (15.3)	−0.8	0.5135
Baseline diastolic BP	85.1 (12.1)	84.3 (12.6)	+0.8	0.4341
9-month systolic BP	131.6 (15.8)	138.2 (19.7)	−6.1	0.0005
9-month diastolic BP	76.3 (11.1)	78.0 (14.5)	−2.9	0.0026
9-month BP control	43%	34%	+9%	0.052

Novelty and Significance:

• 1)
  What is New?

  • This study used clinic-based pharmacists rather than research personnel
  • The study population had a relatively large percentage of minority and low-income patients
• 2)
  What is Relevant?

  • This study demonstrated the cost effectiveness of using onsite clinic-based pharmacists to help control hypertension.
• 3)Summary: Our results highlight the cost-effectiveness of clinical pharmacists for hypertension control in primary care settings: the cost to increase hypertension control by one percentage point was $22.55. The cost to lower blood pressure by 1 mmHg was $38.82 for systolic and $81.66 for diastolic.