Association of Polypharmacy and Statin New-User Adherence in a Veterans Health Administration Population

Abstract

Background: The relationship between multiple medication consumption and medication adherence is not well understood. Objective: To determine the association between the number of active medications on the patient medication profile at baseline and adherence in new users of statins. Methods: This was a retrospective cohort study of new users of statin medications from the Veterans Health Administration. We explored the correlation between the number of baseline medications and adherence, grouping patients by number of active medications on the study index date via Cochran-Armitage trend test and multiple linear regression. The adherence metric calculated for each patient was the medication possession ratio (MPR). Adherence was defined as achieving a 0.8 MPR or greater in primary analysis and a 0.9 MPR or greater in the secondary analysis. Results: There was a statistically significant trend of increasing proportion of adherent participants as baseline medication count grew (P value < .001). The regression further demonstrated that statin MPR was increased by 0.04, 0.07, 0.10, and 0.14 for the 6 to 10 medication count, 11 to 15 medication count, 16 to 20 medication count, and >20 medication count groups, respectively, in comparison with the reference 1 to 5 medication count group (P < .001 for all comparisons). An MPR threshold of 0.9 provided consistent evidence of improved adherence as number of medications increased (P < .001). Conclusions: Increased medication count at baseline was associated with improved adherence for new users of statins.

Keywords

medication possession ratio dyslipidemia polypharmacy

Introduction

The influence of polypharmacy, the concurrent use of multiple medications, on patient medication adherence is poorly understood.^1,2 In published studies, polypharmacy has been associated with reductions, improvements, and null effects on adherence.^3
-11 Although polypharmacy necessarily contributes to medication regimen complexity, it does not always diminish adherence. Investigators have previously described improved cardiovascular medication adherence for patients taking multiple medications in separate studies.^12,13 Shalansky and Levy⁶ found in a survey study of patients in British Columbia taking either angiotensin-converting enzyme inhibitors or lipid-lowering therapies, that patients prescribed a larger number of medications were more likely to be adherent than those prescribed fewer medications (5.9 vs 4.1, respectively, with P = .001).⁶ Previous investigators have hypothesized that patients prescribed a large number of medications may be more cognizant of their compromised health status and may therefore comply with a larger and more complex medication regimen.¹² A previous analysis of Veterans Health Administration (VHA) patients found no association between number of chronic medications and adherence but a significant increase in adjusted odds of adherence for patients with increased comorbidities.¹⁰ This comports with previous findings that relate medication adherence to the patient’s belief that the medication is important for managing current symptoms or protecting future health.⁵

Isolating the influence of polypharmacy on adherence is further hindered by limitations in the measurement of adherence. Directly measuring medication adherence through observation is generally not possible in large samples.¹³ For this reason, many of the studies have implemented surveys of adherence to explore the relationship with polypharmacy. However, survey studies rely on patient reporting, which is subject to pronounced recall and nonresponse biases.^14
-16 Moreover, other investigators have described improvements in patient adherence behavior when these same patients are aware that they are included in prospective studies, and exaggeration by respondents of their timely and appropriate medication consumption to their providers has also been previously described in the literature.^17,18 For these reasons, a retrospective claims analysis becomes an attractive option because such an analysis eliminates confounding caused by respondent reporting and removes patient modification of adherence behavior that simply results from study inclusion. Our goal was to determine the association of starting number of active medications on the patient profile and statin adherence using a cohort of new users of statin medications in the VHA population, applying a validated measure for secondary medication adherence: the medication possession ratio (MPR).

Methods

Design Overview

This was a retrospective cohort study of new statin users from the VHA. Participants were required to be eligible for VHA medical and pharmacy services throughout the 1-year study period from index date and to have complete data for exposure, outcome, and regression adjustment variables. MPR was defined as number of days supplied with prescription medication divided by days of observation.¹⁹

Setting and Participants

Study participants were from the Department of Veteran Affairs (VHA) Veterans Integrated System Network 22, a region that includes sites in Southern California (Los Angeles, Long Beach, San Diego, and Loma Linda) and Nevada (Las Vegas), with a system enrollment of approximately 1.4 million members. The VHA database has a comprehensive capture of pharmacy utilization, medical records, demographic characteristics, and health plan coverage elements, allowing robust analysis of health outcomes. Patients were included if they had begun a statin between November 30, 2006, and December 2, 2007, with no active statin prescription in the 6 months prior to prescription. Participants were required to be eligible for VHA medical and pharmacy services 6 months prior to index date and throughout the study period and to have complete data for exposure, outcome, and adjustment regression variables.

Interventions

We explored the differential correlation of the number of active baseline medications included on the patient medication profile at index date of statin use and statin adherence, grouping patients by number of active medications on the index event date as the study exposure in this retrospective cohort study. Baseline medication count groups were delineated as 1 to 5, 6 to 10, 11 to 15, 16 to 20, and >20.

Outcomes and Follow-up

The outcome of interest was the MPR, a standardized secondary adherence metric. MPR is the number of days supplied of prescription medication actually received divided by days of observation.¹⁹ The MPR calculation numerator excludes the final prescription days’ supply because it supports therapy beyond the study observation period that serves as the denominator. Patients were adherent in the primary analysis if their MPR was 0.8 or greater for the new statin. Patients were followed for a 1-year observation period from index date, counting the medication supply filled from the new statin prescription and refills of this prescription. To reduce measurement error for MPR, patients were excluded if they switched statins or experienced an admission for more than 30 consecutive days. Over-the-counter medications are not included in the VHA medication patient profile.

Statistical Analysis

The primary aim was to ascertain if the proportion of statin adherence was greater among the higher starting medication groups. We applied the Cochran-Armitage test for trend among the ordinal categories. The Cochran-Armitage test for trend is a modification of the χ² test, which allows for statistical assessment of the presence of increased proportions of an outcome as ordered exposures are increased.^20,21 We also assessed the categorical relationship between starting medication count groups and adherence using a χ² test for association. We reported percentages of adherent patients by starting medication count category.

We performed an adjusted analysis using a generalized linear regression model to observe the changes in MPR for the different starting medication count categories versus the reference group of 1 to 5 starting medications. Adjustment variables included age, gender, race, copayment status, income category, statin consumed, baseline body mass index, and baseline lipid levels: total cholesterol (TC), triglycerides (TG), high-density lipoproteins (HDL), and low-density lipoproteins (LDL). We also adjusted for the presence of the following comorbidities at baseline: hypertension, diabetes, congestive heart failure, peripheral vascular disease, history of myocardial infarction, mood disorder, history of angina, and chronic obstructive pulmonary disease.

Evidence has suggested that additional clinical benefits are imparted with achievement of an 0.9 MPR for statin medications versus 0.8 MPR.²² An additional secondary aim invoked the Cochran-Armitage trend test using the MPR of 0.9 as the boundary for adherence to assess the relationship between medication count category and adherence.

We assessed sensitivity of statin adherence to changes in medication count incurred during the study period to ensure that adherence was not inflated by a patient’s reduction in number of medications during the study period. To achieve this, we conducted the multiple regression using change in medication count from baseline to study completion as the exposure variable with the same MPR response variable. A significance level of .05 was set for all hypothesis tests used. Statistical analyses were executed using SAS 9.3 (SAS Inst, Inc, Cary, NC). This retrospective study was approved by the San Diego VHA Institutional Review Board and met all criteria for protection of human subjects.

Results

A total of 4886 patients met inclusion criteria. The mean age of the study sample was 63.6 years. The mean number of medications at baseline was 7.0. The mean baseline levels for TC, LDL, HDL, and TG were 213.7, 137.7, 42.4, and 171.0 mg/dL, respectively. The most common comorbidity reported was hypertension (73.1% of study participants). Simvastatin was the most common statin consumed, with 84.8% of study participants taking this medication. The study sample was 95.3% male (Table 1).

Table 1.

Summary Statistics of VHA Data Set (n = 4886).

Characteristic	Value
Age, mean (SD), years	63.6 (11.3)
Body mass index at baseline, mean (SD)	30.2 (5.7)
Number of medications at baseline, mean (SD)	7.0 (3.8)
Lipids
Total cholesterol at baseline, mean (SD), mg/dL	213.7 (48.5)
High-density lipoprotein at baseline, mean (SD), mg/dL	42.4 (12.3)
Low-density lipoprotein at baseline, mean (SD), mg/dL	137.7 (40.1)
Triglycerides at baseline, mean (SD), mg/dL	171.1 (154.7)
Male, n (%)	4449 (95.3)
Race
White, n (%)	2219 (47.5)
Unspecified, n (%)	1008 (21.6)
Black, n (%)	672 (14.4)
Hispanic, n (%)	531 (11.4)
Asian, n (%)	179 (3.8)
American Indian or Native, n (%)	59 (1.3)
New statin consumed
Simvastatin, n (%)	3959 (84.8)
Lovastatin, n (%)	322 (6.9)
Rosuvastatin, n (%)	265 (5.7)
Fluvastatin, n (%)	53 (1.1)
Pravastatin, n (%)	45 (1.0)
Atorvastatin, n (%)	24 (0.5)
Copayment category
Copayment non–service connected category, n (%)	2183 (46.8)
No copayment, n (%)	1526 (32.7)
Copayment service-connected category, n (%)	959 (20.5)
Comorbidities
Hypertension, n (%)	3412 (73.1)
Diabetes mellitus, n (%)	1751 (37.5)
Peripheral vascular disease, n (%)	1474 (31.6)
Chronic obstructive pulmonary disease, n (%)	441 (9.5)
Congestive heart failure, n (%)	202 (4.3)
History of myocardial infarction, n (%)	137 (2.9)
Angina, n (%)	104 (2.2)
Mood disorder, n (%)	76 (1.6)

Abbreviation: VHA, Veterans Health Administration.

There was a statistically significant trend of increasing proportion of adherent study participants with greater medication count via the Cochran-Armitage trend test (P < .001). The findings were consistent with a χ² test for association of increasing medication count category correlated with percentage of adherent participants (P < .001). The percentage of adherent participants increased at every progression of medication count category. The lowest percentage was found in the 1 to 5 medication count category at 39.4% adherent, increasing to 48.0% in the 6 to 10 medication category, 54.9% in the 11 to 15 medication count category, and 59.6% in the 16 to 20 count category and reaching a maximum of 66.1% in the >20 medication count category (Figure 1).

Figure 1.

Percentage of participants adherent to statin therapy (MPR ≥ 0.8) by baseline medication count category.

The multiple linear regression demonstrated improved statin adherence for new users when medication count was higher at baseline in all comparisons versus reference (P < .01 for all comparisons). Statin MPR was 0.04 higher for the 6 to 10 medication count group versus the reference 1 to 5 medication count group. The increase in adherence in MPR versus reference enhanced as the number of medications grew categorically, with the largest medication count category (>20 medications) associated with a 0.14 improvement in statin MPR versus the 1 to 5 medication count group (Table 2).

Table 2.

Increases in New-User Statin MPR by Medication Count Category Adjusted for Baseline Characteristics.

Starting Medication Count Category (n)	Increase in statin MPR (95% Confidence Interval)
1 To 5 (1867)	Reference
6 To 10 medications (2023)	0.04 (0.02, 0.06)
11 To 15 medications (722)	0.07 (0.05, 0.10)
16 To 20 medications (218)	0.10 (0.06, 0.14)
>20 Medications (56)	0.14 (0.07, 0.22)

Abbreviation: MPR, medication possession ratio.

In the secondary analysis, the use of a more rigorous adherence MPR threshold of 0.9 provided consistent evidence of a trend similar to that demonstrated in the primary aim, with adherence threshold at 0.8 MPR or greater. Increasing medication count category yielded a statistically significant trend (P < .001) of increasing adherence as the category escalated, with the lowest percentage adherence found in the 1 to 5 medication count group (29.4%) and the highest observed in the >20 medication count group (60.7%; Figure 2).

Figure 2.

The percentage of participants adherent to statin therapy (MPR ≥ 0.9) by baseline medication count category.

We also conducted regression using increase in medication count during the study period as the independent variable. This multiple regression demonstrated that each additional medication from baseline was associated with a 0.02 increase in statin MPR (P < .01).

Discussion

Implications for Practice

Clinicians must carefully balance the necessity of prescribing additional medications with the collateral burden of patient consumption of an expanded regimen. Prescribers may hesitate to recommend additional medications lest adherence be compromised. We explored the relationship of polypharmacy with adherence to a newly prescribed medication for chronic disease management. The analysis revealed that increasing medication count at baseline was associated with elevated adherence to statins for new users. This bolsters prescribing of indicated medications for chronic syndromes even when the medications are additions to an extensive multiple drug regimen. Our findings also suggest that enthusiasm by clinicians to cull large drug regimens to gain a predicted boost in patient adherence may be inappropriate. Rather, medication reconciliation efforts should continue to focus on appropriate use of agents and not necessarily on the quantity of use of agents.

Implications for Research

Our work demonstrated that a greater number of active medications on the patient profile at baseline was correlated with elevated adherence in terms of MPR for new users of statin medications. Trend analysis demonstrated an escalation in statin adherence with increase in medications at baseline of statin use. The findings were confirmed when we evaluated the relationship using a more stringent adherence threshold of 0.9 MPR.

To adjust for characteristics influencing the outcome, we executed multiple regression and witnessed statistically significant increases in adherence for those patients who had more medications at baseline. The smallest increase of 0.04 MPR versus the reference group (patients consuming 1 to 5 medications) was found with patients using 6 to 10 medications, and the maximum of 0.14 was found for the >20 baseline medication count versus the reference group.

Previous work has demonstrated that patient perception of the necessity of medications may be the prevailing impetus for appropriate consumption.⁵ Billups et al¹⁰ postulated that an improvement in compliance for patients taking more medications could be related to the Health Belief Model, in which patients are more likely to commit to a medical care regimen when they are confident that they are truly ill.²³ Similarly, Horne and Weinman¹² demonstrated that increased adherence correlated with patient perception of the “necessity of the medication.” In this framework, additional medications are commensurate with worsening comorbidity. Patient awareness of a more fragile health state elevates the appreciation of additional medications and promotes adherent consumption. Our findings are consistent with this notion. Patients who are prescribed a larger suite of medications at baseline may see the statin medication as a necessity to preclude worsening illness, fueling invigorated tenacity for on-time consumption. Recent studies have also found a link between the existence of social support and improved medication adherence.^24,25 In the context of polypharmacy, this may manifest as increasing number of medications being associated with de facto social support caregivers aiding appropriate consumption of medication. More research is needed to understand this important component of adherence behavior.

Our findings are in accord with other previous research efforts to understand the ability of patients to follow drug regimens as directed. Work by George et al²⁶ on the development of the Medication Regimen Complexity Index (MRCI) linked the difficulty of appropriate consumption of a medication regimen to the following: type of dosage forms included in the regimen (section A), frequency of dosing of the components of the regimen (section B), and additional directions in following the regimen (section C). Medication count was not included as a component of the MRCI score because the authors contended that medication count does not simply equate to complexity. Our findings support the design approach of the MRCI instrument. It would be an informative future investigation to follow a cohort of patients with a distribution of MRCI scores and assess adherence over several years.

Our study has several advantages over previous work. Many of the prior studies evaluated the connection between polypharmacy and adherence using surveys or patient interviews. These methods introduce recall bias that may obscure estimates. By using a pharmacy claims analysis, we eliminate reporter biases from the analysis. Pharmacy data allow a larger sample size (n = 4886) than those used in prior investigations of this phenomenon. With statistical trend analysis for adherence by category, we demonstrate a growth in adherence as medication count increases. We confirmed our findings with a second, more stringent adherence threshold of 0.9 MPR. The regression analysis refines these improvements with precise estimates of the differences by medication count category, which adjusted for patient demographic characteristics and biometrics that could influence results. The improvement in statin adherence was not explained by a change in prescribed medications from baseline, but an increase in the number of medications from baseline during the study period was associated with an improvement in statin adherence. By measuring statin adherence in new users, we compared consumption in those with similar statin use experience. This removes bias that is sometimes found when prevalent users of medications are compared.²⁷ Moreover, statins are well studied molecules, where the benefits of long-term consumption and elevated adherence goals are understood.^28,29

This study has limitations The VHA provides a greater level of integrated care from clinicians and pharmacies than commonly afforded to nonveterans, with more opportunities to reinforce the need for medication adherence through medication reconciliation, use of an integrated electronic health care record, and active pharmacist interventions. However, these measures vary depending on the Veterans Integrated System Network of interest. Prescriptions are not automatically filled for veterans. Patients are required to refill their own medications by the following mechanisms: refill line, go through the online resource www.myhealth.va.gov, medication refill slips, or request a health care provider to assist with refills. These refill options are fairly consistent with those found in most commercial plans. This should enhance the generalizability of our study findings. Our study population was 95% male. Women may exhibit different adherence behaviors in the context of varying medication counts. The analysis included residents of the Southern California and Las Vegas, Nevada, area only. These conditions may affect the generalizability to non-VHA patients or residents outside of the study region. Although MPR is a secondary adherence measure based on medication fill frequency and does not measure physical consumption of the medication by the patient, published evidence has demonstrated that MPR is correlated with primary adherence and that improved MPR is associated with augmented health outcomes.³⁰ Patients cannot take prescription medications that go unfilled. There is a possibility that patients aggressively fill their medications but fail to take them.³¹ In terms of our analysis, this “hoarding behavior” would only introduce systematic bias in adherence estimation if a differential likelihood of overfilling of the new statin medication across medication count groups existed. This phenomenon is not likely. That said, it remains important for clinicians to ensure that prescriptions are not only appropriately filled but appropriately consumed. So-called brown bag events, in which pharmacists review all of the patients’ medication bottles with the patient offers a reasonable means of assessing consumption.³² Consistent follow-up of patients’ clinical lab markers provides another mechanism of triangulating if appropriate consumption is taking place because aberrant surrogate markers may indicate failure to consume medication after filling.³³ We used the number of medications taken by the participant at baseline, but the data set did not include the name of the medications prescribed. Hence, we cannot explore the possible relationship between therapeutic category and statin adherence. The observation time was limited to 1 year. Longer-term adherence could shift for the patients taking a large number of medications. Future studies should evaluate this association over a more extensive interval because the majority of patients will be taking statin medications for many years. During the study period, patients in the VHA system had a simple copayment framework of either no copayment or an $8 copayment. Patients in commercial plans may experience more varied copayment schedules and may therefore respond to additional prescribed medications differently depending on prescription drug plan design.

As the US population ages and the number of prescription medications available grows, clinicians routinely decide how best to optimize care. The clinical necessity of the additional medication should not be jeopardized by simply assuming that there will be a reduction in adherence with a greater pill burden. These patients may be the most likely to use the medication to its full benefit.

Footnotes

Authors’ Note

This research has not been presented as an abstract at any public forum.

Declaration of Conflicting Interests

The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Funding

The author(s) received no financial support for the research, authorship, and/or publication of this article.

References

Viktil

Blix

Moger

Reikvam

. Polypharmacy as commonly defined is an indicator of limited value in the assessment of drug-related problems. Br J Clin Pharmacol. 2007;63:187-195. doi:10.1111/j.1365-2125.2006.02744.x.

Kroenke

. Polypharmacy: causes, consequences, and cure. Am J Med. 1985;79:149-152.

Steinman

Hanlon

. Managing medications in clinically complex elders: “There’s got to be a happy medium.” JAMA. 2010;304:1592-1601. doi:10.1001/jama.2010.1482.

Steinman

Landefeld

Rosenthal

Berthenthal

Sen

Kaboli

. Polypharmacy and prescribing quality in older people. J Am Geriatr Soc. 2006;54:1516-1523. doi:10.1111/j.1532-5415.2006.00889.x.

Grant

Devita

Singer

Meigs

. Polypharmacy and medication adherence in patients with type 2 diabetes. Diabetes Care. 2003;26:1408-1412.

Shalansky

Levy

. Effect of number of medications on cardiovascular therapy adherence. Ann Pharmacother. 2002;36:1532-1539. doi:10.1345/aph.1C044.

George

Shalansky

. Predictors of refill non-adherence in patients with heart failure. Br J Clin Pharmacol. 2007;63:488-493. doi:10.1111/j.1365-2125.2006.02800.x.

Coons

Sheahan

Martin

Hendricks

Robbins

Johnson

. Predictors of medication noncompliance in a sample of older adults. Clin Ther. 1994;16:110-117.

Sharkness

Snow

. The patient’s view of hypertension and compliance. Am J Prev Med. 1992;8:141-146.

10.

Billups

Malone

Carter

. The relationship between drug therapy noncompliance and patient characteristics, health-related quality of life, and health care costs. Pharmacotherapy. 2000;20:941-949. doi:10.1592/phco.20.11.941.35266.

11.

Col

Fanale

Kronholm

. The role of medication noncompliance and adverse drug reactions in hospitalizations of the elderly. Arch Intern Med. 1990;150:841-845.

12.

Horne

Weinman

. Patients’ beliefs about prescribed medicines and their role in adherence to treatment in chronic physical illness. J Psychosom Res. 1999;47:555-567.

13.

Osterberg

Blaschke

. Adherence to medication. N Engl J Med. 2005;353:487-497. doi:10.1056/NEJMra050100.

14.

Schmier

Halpern

. Patient recall and recall bias of health state and health status. Expert Rev Pharmacoecon Outcomes Res. 2004;4:159-163. doi:10.1586/14737167.4.2.159.

15.

Stull

Leidy

Parasuraman

Chassany

. Optimal recall periods for patient-reported outcomes: challenges and potential solutions. Curr Med Res Opin. 2009;25:929-942. doi:10.1185/03007990902774765.

16.

Berk

. An introduction to sample selection bias in sociological data. Am Sociol Rev. 1983;48:386. doi:10.2307/2095230.

17.

Golin

Liu

Hays

. A prospective study of predictors of adherence to combination antiretroviral medication. J Gen Intern Med. 2002;17:756-765. doi:10.1046/j.1525-1497.2002.11214.x.

18.

Arnsten

Demas

Farzadegan

. Antiretroviral therapy adherence and viral suppression in HIV-infected drug users: comparison of self-report and electronic monitoring. Clin Infect Dis. 2001;33:1417-1423. doi:10.1086/323201.

19.

Halpern

Khan

Schmier

. Recommendations for evaluating compliance and persistence with hypertension therapy using retrospective data. Hypertension. 2006;47:1039-1048. doi:10.1161/01.HYP.0000222373.59104.3d.

20.

Cochran

. Some methods for strengthening the common; 2 tests. Biometrics. 1954;10:417. doi:10.2307/3001616.

21.

Armitage

. Tests for linear trends in proportions and frequencies. Biometrics. 1955;11:375. doi:10.2307/3001775.

22.

Oleen-Burkey

Dor

Castelli-Haley

Lage

. The relationship between alternative medication possession ratio thresholds and outcomes: evidence from the use of glatiramer acetate. J Med Econ. 2011;14:739-747. doi:10.3111/13696998.2011.618517.

23.

Rosenstock

. Why people use health services. Milbank Mem Fund Q. 1966;44:94. doi:10.2307/3348967.

24.

Scheurer

Choudhry

Swanton

Matlin

Shrank

. Association between different types of social support and medication adherence. Am J Manag Care. 2012;18:e461-e467.

25.

Lehavot

Huh

Walters

King

Andrasik

Simoni

. Buffering effects of general and medication-specific social support on the association between substance use and HIV medication adherence. AIDS Patient Care STDS. 2011;25:181-189. doi:10.1089/apc.2010.0314.

26.

George

Phun

Y-T

Bailey

Kong

Stewart

. Development and validation of the Medication Regimen Complexity Index. Ann Pharmacother. 2004;38:1369-1376. doi:10.1345/aph.1D479.

27.

Ray

. Evaluating medication effects outside of clinical trials: new-user designs. Am J Epidemiol. 2003;158:915-920.

28.

Grabowski

Lakdawalla

Goldman

. The large social value resulting from use of statins warrants steps to improve adherence and broaden treatment. Health Aff (Millwood). 2012;31:2276-2285. doi:10.1377/hlthaff.2011.1120.

29.

Kastelein

JJP

van der Steeg

Holme

. Lipids, apolipoproteins, and their ratios in relation to cardiovascular events with statin treatment. Circulation. 2008;117:3002-3009. doi:10.1161/CIRCULATIONAHA.107.713438.

30.

Degli Esposti

Saragoni

Batacchi

. Adherence to statin treatment and health outcomes in an Italian cohort of newly treated patients: results from an administrative database analysis. Clin Ther. 2012;34:190-199. doi:10.1016/j.clinthera.2011.12.011.

31.

Martinez

Vande Griend

Linnebur

. Medication management: a case of brown bag-identified medication hoarding. Consult Pharm. 2012;27:729-736. doi:10.4140/TCP.n.2012.729.

32.

Nathan

Goodyer

Lovejoy

Rashid

. “Brown bag” medication reviews as a means of optimizing patients’ use of medication and of identifying potential clinical problems. Fam Pract. 1999;16:278-282.

33.

Howard

Arnsten

. A prospective study of adherence and viral load in a large multi-center cohort of HIV-infected women. AIDS. 2002;16:2175-2182.