Sage Journals: Discover world-class research

Abstract

Background:

Adherence is a major determinant of the effectiveness of antiretroviral therapy (ART). We determined the association between self-reported adherence (SRA) and medication possession ratio (MPR), a pharmacy-based adherence measure, and their respective associations with viral load.

Methods:

Adherence to ART was assessed by MPR over 6 months and by self-report which included a question with a Likert-type scale response, a visual analogue scale (VAS), and an inquiry about the last time the patients skipped any prescribed medications.

Results:

Taking MPR as the “gold standard,” all 3 SRA measures displayed high specificity but low sensitivity. The prevalence ratio (95% confidence interval) for viral load ≥50 copies/mL was 2.19 (1.07-4.50) for MPR <90%, 1.98 (1.04-3.78) for poor/fair/good versus excellent/very good ability to take antiretroviral drugs, 1.47 (0.79-2.75) for skipping medications within the past 2 weeks, and 2.51 (1.39-4.53) for VAS <95%.

Conclusion:

These data suggest various SRA measures hold clinical value in screening for poor ART adherence.

Keywords

HIV adherence medication possession ratio antiretroviral therapy

Introduction

Adherence to a medication regimen has been defined as the extent to which patients take the medication as prescribed.¹ High levels of antiretroviral therapy (ART) adherence are needed to reach long-lasting viral suppression, and many studies have underscored the important role of adherence to ART medications in the treatment of HIV infection.^1

–6 Furthermore, inadequate adherence can lead to the selection of drug-resistant strains of HIV that could be transmitted to uninfected or treatment-naive individuals.^7

–10

Although there is no gold standard to measure adherence, several methods have been described.^1,11
–13 Two of the most common approaches for ascertaining adherence in clinical settings are patient self-report and pharmacy-based metrics. Self-reported adherence (SRA) measures are easily administered but less sensitive than pharmacy-based measures and tend to overestimate true adherence; however, SRA for assessing adherence is simple, correlated with virologic outcomes, and may be more efficiently implemented in routine clinical practice.^14,15 In a closed pharmacy system, prescription records can be used to calculate surrogate measures of adherence, such as medication possession ratio (MPR), which have been correlated with biological and clinical outcomes.¹ However, calculation of pharmacy-based measures of adherence, such as MPR, is not feasible in many clinical settings. Therefore, it is important to determine the degree to which SRA measures capture adherence behavioral information relative to pharmacy-based measures. This study aimed to gain new insights into the associations between SRA and MPR, and between both approaches to measurement of adherence and viral suppression, including the novel evaluation of the use of successive questionnaires in enhancing the sensitivity to adherence assessment.

Methods

Study Setting

The University of Alabama at Birmingham (UAB) 1917 Clinic provides comprehensive outpatient medical and supportive health care services to individuals living with HIV. Clinic patients participate in an institutional review board–approved prospective and retrospective clinical cohort study. Established in 1992, the UAB 1917 Clinic Cohort captures comprehensive clinical, behavioral, and administrative data.^16,17 Additionally, UAB 1917 Clinic patients can participate in the Center for AIDS Research Network of Integrated Clinical Systems (CNICS) research protocol, providing written informed consent to contribute additional behavioral data via a battery of validated instruments and banked biological specimens (plasma and peripheral blood mononuclear cells) during routine care visits for research purposes.

Study Design

This was a cross-sectional study within the UAB 1917 Clinic Cohort.

Study Population and Sampling

The study population was composed of HIV-infected patients receiving care at the UAB 1917 Clinic, who met the following inclusion criteria: First, patients had to be on ART 6 months before their index date (i.e., the date of the most recent visit in the database); we further restricted inclusion to patients whose index date fell within the period ranging from July 1, 2011, to July 29, 2012. Second, patients must have completed all 3 self-report measures of adherence as captured through the CNICS protocol (see study variables section). Third, patients were required to have viral load results available within 30 days of their index date. Fourth, patients had to be receiving ART medication exclusively from the 1917 Clinic’s Pharmacy through the federally funded AIDS Drug Assistance Program (ADAP), which provides HIV medications to qualifying HIV-infected individuals¹⁸; this requirement allowed for the estimation of MPR based upon pharmacy refill records. Finally, medication refill data had to be available from the clinic’s pharmacy records to calculate patients’ MPR.

Study Variables

The study variables included demographics (age, sex, and race), medical history (study entry viral load and CD4 count values closest to study entry [−180 days to + 90 days], HIV risk factors), current medical information (viral load, CD4 counts, ART drugs, and substance use), and adherence information.

Self-Reported Adherence

The SRA assessment was captured via electronically administered patient-reported outcome questionnaires completed on touch screen computers during regular HIV medical provider visits through the CNICS protocol.¹⁷ We used the following 3 SRA measures: (1) An item with a Likert-type response, “In the past 4 weeks, how was your ability to take all of your anti-HIV medications that were prescribed by your doctor?” with possible answers very poor, poor, fair, good, very good and excellent. In a recent study, this measure performed favorably in the prediction of adherence-related clinical outcomes, such as HIV viral load and CD4 count, when compared to other SRA tools.¹⁹ (2) A last skipped dose inquiry: “When was the last time you missed any of your medications?” with possible answers within the past week, 1 to 2 weeks ago, 2 to 4 weeks ago, 1 to 3 months ago, more than 3 months ago, and never skip medication. This question is part of the Adult AIDS Clinical Trials Group adherence assessment questionnaire and has been used in prior studies.^17,20 (3) A visual analogue scale (VAS): “Mark the point along the line that most closely reflects how much of your HIV medications you have taken in the last month,” on a 0 to 100 scale.

Pharmacy Refills

The co-located pharmacy at the UAB 1917 Clinic dispenses medications with a 30-day supply. For each patient in the sample, data on pharmacy refills over the 6-month period preceding the index date were used to calculate MPR expressed as a percentage. This adherence indicator corresponds to the percentage of days during which patients are in possession of their medications. MPR is thus calculated by the formula MPR = total number of days’ supply obtained between the first and last fill, not including the last fill, divided by number of days between first and last fills.^13,21,22 Based on this definition, MPR values may be greater than 100% (eg, if patients refill before their medication supply runs out). In the present analyses, MPR values greater than 100% were truncated to 100%.

Data Collection

We manually extracted medication refill dates from the pharmacy dispensation records. All other variables, including demographics, medical information, and SRA, were extracted from the electronic database of the 1917 Clinic Cohort. The administrative cutoff date for data collection was July 29, 2012.

Data Analyses

Descriptive statistics including means, medians, ranges, and visual depictions of study variables were calculated and distributions were evaluated to confirm that they conformed to the assumptions of statistical tests employed. The MPR was used as the adherence “gold standard” and the screening performance (sensitivity, specificity, and positive and negative predictive values) of each of the 3 SRA measures was determined. We also performed a 3-step sequential screening for poor adherence using first the Likert-type item, then the “last skipped dose,” and finally the VAS. Patients who test negative on one measure (good adherers) were further screened by the next test, in order to increase the sensitivity of SRA. When all 3 tests were applied, only patients with high adherence to all measures were considered “negative” (ie, good adherers), whereas patients who test positive for at least one test were considered “positive” (poor adherers; Figure 1).

Figure 1.

Sequential screening for poor adherence.

Based on data distribution in the current population and previously published studies,^{1,2,17,18,23,24} we categorized adherence variables as follows: MPR was categorized as <90% (poor adherence) or ≥90% (good adherence); the Likert-type item (ability to take HIV medications in the past 4 weeks) was categorized into 2 groups (1) excellent, very good and (2) good, fair, poor, very poor; the last skipped dose was also categorized into 2 groups (1) at least 2 weeks ago and (2) within the past 2 weeks; and the VAS (SRA on a 0-100 scale) was also categorized into 2 groups (1) “VAS ≥95%” (good adherence) and (2) “VAS <95%” (poor adherence). We used the chi-square to test for independence between SRA variables and dichotomized MPR.

The crude and adjusted associations between measures of adherence and viral load were assessed using log binomial regression to compute prevalence ratios (PRs). Prevalence ratios are recommended in lieu of odds ratios for cross-sectional studies with common outcomes. Viral load was categorized as “undetectable” (<50 copies/mL) or “detectable” (≥50 copies/mL). Our first multivariate model adjusted for race, gender, age, depression, substance abuse, alcohol use, and HIV risk factors. Baseline viral load and CD4 values were not included in the model because these variables are likely in the causal pathway of the relationships between adherence to antiretroviral (ARV) drugs (exposure) and viral suppression (outcome). We then ran a second model whereby baseline viral load and CD4 were adjusted for. The results of these 2 models are provided in Table 1 and Supplemental Table 1, respectively. All analyses were performed in SAS 9.3 (SAS Institute Inc., Cary, North Carolina). P values ≤.05 were considered statistically significant.

Table 1.

Prevalence Ratios for Detectable Viral Load Associated with Poor Adherence to Antiretroviral Drugs.

	Adherence		P Value
	Medication Possession Ratio
	≥90% (n = 130)	<90% (n = 139)
Viral load detectable, n (%)	10 (7.7)	24 (17.3)
Prevalence ratio (95% CI)
Crude	1 (ref)	2.24 (1.12-4.51)	.02
Adjusted^a	1 (ref)	2.19 (1.07-4.50)	.03
	Ability to take ARVs during the last month
	Excellent, very good (n = 237)	Poor, fair, good (n = 32)
Viral load detectable, n (%)	25 (10.6)	9 (28.1)
Prevalence ratio (95% CI)
Crude	1 (ref)	2.67 (1.37-5.19)	.004
Adjusted^a	1 (ref)	1.98 (1.04-3.78)	.04
	Last skipped dose
	≥2 weeks (n = 212)	< 2 weeks (n = 57)
Viral load detectable, n (%)	23 (10.8%)	11 (19.3%)
Prevalence ratio (95% CI)
Crude	1 (ref)	1.78 (0.92-3.43)	.08
Adjusted^a	1 (ref)	1.47 (0.79-2.75)	.22
	Visual analogue scale for adherence to ARVs during the last month
	≥ 95% (n = 203)	< 95% (n = 66)
Viral load detectable, n (%)	18 (8.9)	16 (24.2)
Prevalence ratio (95% CI)
Crude	1 (ref)	2.73 (1.48-5.05)	.001
Adjusted^a	1 (ref)	2.51 (1.39-4.53)	.002
	All self-report combined^b
	Good (n = 180)	Poor (n = 89)
Viral load detectable, n (%)	17 (9.4)	17 (19.1)
Prevalence ratio (95% CI)
Crude	1(ref)	2.02 (1.08-3.77)	.03
Adjusted^a	1(ref)	1.80 (0.98-3.31)	.06

Abbreviations: ARV, antiretroviral drugs; CI, confidence interval.

^a Adjusted for age, race, gender, age, depression, substance abuse, alcohol use, and HIV risk factors.

^b For “All self-report combined,” Poor was defined as poor adherence for at least 1 self-report measure, whereas Good adherence corresponds to patients who have good adherence for all self-report measures.

Ethics

This study obtained separate approval from the UAB Institutional Review Board to review cohort records from the parent protocol.

Results

Overall, 269 patients met study eligibility criteria and were included in the analyses. The mean age was 42.3 years (standard deviation = 9.3). Most patients were males (87.4%) and 48% were white. A majority (69.0%) of patients had a baseline CD4 count greater than 350; 85.5% had a baseline viral load of less than 50 copies/mL. Apart from “current substance abuse,” which was more common among patients with MPR of less than 90%, there were no statistically significant differences in the distributions of demographic and clinical characteristics between the 2 categories of MPR (<90% versus ≥90%; Table 2).

Table 2.

Demographics and Clinical Characteristics of Patients by Medication Possession Ratio for Antiretroviral Drugs.

Characteristic	Medication Possession Ratio		P Value
	<90%	≥90%
	Mean (SD) or n (%)^a	Mean (SD) or n (%)
Age, years	42.2 (9.3)	43.6 (10.0)	.23
Male	118 (84.9)	117 (90.0)	.21
White	64 (46.0)	65 (50.0)	.52
At risk of alcohol use	18 (13.3)	10 (7.8)	.14
Current depression	58 (42.0)	43 (33.3)	.14
Current substance abuse	15 (10.9)	2 (1.6)	.002
HIV risk factors
Heterosexual	37 (27.0)	31 (24.0)	.72
IVDU	14 (10.2)	11 (8.5)
MSM	86 (62.8)	87 (67.4)
Study^b entry CD4 count, cells/mm³
<200	14 (10.8)	12 (9.6)	.92
200-350	26 (20.0)	27 (21.6)
>350	90 (69.2)	86 (68.8)
Study entry viral load^b <50 copies/mL	112 (83.0)	112 (88.2)	.23

Abbreviations: IVDU, intravenous drug use; MSM, men who have sex with men; SD, standard deviation.

^a Column percentage.

^b Value closest to study entry (−180 days to +90 days).

The median MPR was 88.7% (25th-75th percentiles 71.0%-98.9%); 130 patients (48.3%) had an MPR value greater or equal to 90%. For the Likert-type item, 88.1% of patients (n = 237) reported an excellent or very good ability to take their HIV medications during the last 4 weeks. For the last skipped dose, 78.8% of patients (n = 212) had not skipped their medications in the past 2 weeks by self-report. On the VAS, 75.5% of patients (n = 203) had an SRA of 95% or greater. There were significant associations between all 3 SRA measures and MPR (Table 3). Furthermore, the median MPR values were significantly higher for “good adherers” compared with “poor adherers” on each of the SRA measures (Supplemental Table 2). When MPR was used as the “gold standard,” the sensitivity was 0.19, 0.27, and 0.34 for the Likert-type item, the last skipped dose, and the VAS, respectively. The corresponding specificity was 0.95, 0.85, and 0.85, respectively. The sequential testing using the 3 SRA measures led to a sensitivity of 0.43 and a specificity of 0.78 (Figure 1).

Table 3.

Cross Tabulation of Self-reported Adherence and Medication Possession Ratio for Antiretroviral Drugs.

	Pharmacy Fill		P Value	Performance of Self-Report Adherence in Detecting Poor Adherence by Medication Possession Ratio
	MPR < 90%, n (%)	MPR ≥ 90%, n (%)^a	P Value
Self-reported adherence				Sensitivity	Specificity	PPV	NPV
Ability to take ARV drugs during the last month			<.001	0.19	0.95	0.81	0.52
Poor, fair, good	26 (81.2)	6 (18.8)
Excellent, very good	113 (47.7)	124 (52.3)
Last time medication skipped			.01	0.27	0.85	0.67	0.48
<2 weeks	38 (66.7)	19 (33.3)
≥2 weeks	101 (47.6)	111 (52.4)
Visual analogue scale for adherence to ARV drugs during the last month			<.001	0.34	0.85	0.71	0.55
<95%	47 (71.2)	19 (28.8)
≥95%	92 (45.3)	111 (54.7)

Abbreviations: MPR, medication possession ratio; ARV, antiretroviral drugs; PPV, positive predictive value; NPV, negative predictive value.

^a Row percentage.

The proportion of patients with a detectable viral load (>50 copies/mL) on their index date was 12.6% (n = 34). The adjusted PR for detectable viral load was 2.19 (95% confidence interval [CI]: 1.07-4.50), P = .03, for an MPR of less than 90% versus MPR greater than or equal to 90%. For the Likert-type item, the adjusted PR for detectable viral load for very poor, poor, fair, and good versus excellent and very good was 1.98 (95% CI: 1.04-3.78), P = .04. For the last skipped dose, the adjusted PR for <2 weeks” versus ≥2 weeks was 1.47 (95% CI: 0.79-2.75), P = .22. For the VAS, the adjusted PR for <95% versus ≥95% was 2.51 (95% CI: 1.39-4.53), P = .002. For sequential testing using all 3 SRA measures, the adjusted PR was 1.80 (95% CI: 0.98-3.31), P = .06 (Table 1). Supplemental Table 3 provides the PR described previously when adjusted variables include baseline viral load and baseline CD4 count.

Discussion

In the current analysis, we assessed the associations between pharmacy-based adherence and 3 SRA measures as well as their respective associations with virologic outcomes. The SRA was significantly associated with the MPR. However, each of the 3 SRA measures had a low sensitivity when MPR was used as gold standard but high specificity. The sequential combination of the 3 SRA measures led to an improvement in sensitivity but a decline in specificity. High adherence values by MPR, the Likert-type item, and the VAS were each associated with undetectable viral load.

The associations between an objective metric (MPR) and subjective (SRA) adherence measures suggest that the latter has clinical value in monitoring ART adherence. Similar findings have been reported for antihypertensive medications in the elderly individuals.²⁵ Compared to MPR, SRA yielded many false negatives in the screening for poor adherence, hence the low sensitivity. This finding is consistent with numerous findings of the propensity of patients to overreport their adherence to medications due to social desirability.^26,27,28 However, the specificity of self-reported nonadherence was relatively high, suggesting that any report of nonadherence by SRA is clinically meaningful and should lead to further adherence discussions between care providers and patients.

The MPR was significantly associated with viral load after adjustment for potential confounders; an MPR of less than 90% was associated with 2-fold higher prevalence of a detectable viral load (≥50 copies/ mL). This finding is consistent with previous studies reporting the value of MPR as a determinant of virologic outcomes.^2,22,23 Among SRA measures, the Likert-type item and the VAS were significantly associated with a detectable viral load. These findings reiterate associations between SRA and MPR that were noted previously and corroborate the value of SRA in clinical settings. As many HIV care settings lack a closed pharmacy system, SRA represents a relatively simple approach to assessing ART adherence in clinical care settings, with immediate feedback available to providers to facilitate education and counseling with patients.

Using adherence measures to predict virologic failure has been under consideration for use in low-resource settings where the measurement of viral load is not frequently performed.²⁴ However, even in venues where viral load testing is available, our findings provide evidence for the utilization of the Likert-type item and the VAS in routine clinical practice; patients may be screened for adherence interventions using poor reported adherence with these measures. It is essential to screen for poor adherence and implement interventions prior to the development of viral load failure, which represents the biological correlate of the nonadherence behavior. As a novel contribution, we demonstrate the combination of several SRA measures enhances sensitivity, which may have added clinical value.

Our study has limitations that should be taken into account when interpreting study findings. The sample was composed of ART treatment experienced patients for whom we assessed adherence over the 6 months prior to an arbitrary selected index date. Most of these patients (85%) had a suppressed viral load (<50 copies/mL) at study entry, suggestive of a high level of baseline adherence. Thus, the consequences of nonadherence may be less tangible, compared to patients initiating ART, or among a sample with a lower baseline prevalence of viral suppression. Moreover, adherence to ART has been shown to change over time,^26,27 so a longer observation period with repeated adherence assessment may be warranted in future studies. Patients had to renew their ADAP benefit eligibility every 6 months to ensure uninterrupted coverage. In our sample, patients received medications through the ADAP program during the 6 months assessed; we were not able to ascertain that patients received uninterrupted benefits of ADAP enrollment during the entire period. This may indicate an underestimation of MPR because of loss of ADAP coverage (not a deliberate nonadherence behavior). However, because patients without prescription medication access are likely to discontinue ART, the MPR metric remains appropriate in our assessment. This observational study is unsuited to establish a causal relationship, and reverse causality cannot be ruled out. Although we adjusted for potential confounders in our multivariate models, residual confounding may still affect our results.

We also note several study strengths. Our sample was drawn from persons completing adherence questionnaires in the context of routine primary HIV medical care, whereas most prior studies recruited more selected samples for participation in distinct research protocols. Our assessment of adherence embedded in routine care speaks to the generalizability and applicability of our findings to real-world treatment settings. Moreover, we focused our evaluation on ADAP clients, an understudied group with important policy implications due to the number of persons living with HIV infection who receive medications through this program. Finally, most prior studies have evaluated adherence questionnaires in isolation, recognizing the high specificity of self-reported tools. This was corroborated by our study. As a novel aspect, we demonstrated that the successive use of SRA measures could enhance sensitivity, albeit modestly, and may open the door for future approaches to using such tools in clinical care settings to enhance the identification of patients with suboptimal adherence who may benefit from enhanced counseling or available adherence interventions.

In conclusion, we found significant associations between SRA and MPR approaches to measuring ART adherence. The sensitivity of the SRA measures in detecting poor adherence compared to MPR was low but the specificity high. The sequential combination of the 3 self-report measures led to a modest increase in sensitivity. Consistent with previous studies, low MPR values (<90%) were associated with detectable viral load. Among SRA measures, significant associations with virologic outcomes were observed for the Likert-type item and the VAS. These results suggest that SRA measures may be useful in clinical practice, given their association with the objective MPR, their good specificity, and their association with viral load. In particular, any self-reported nonadherence is an important clinical indicator that should prompt education, counseling, and adherence intervention, in an effort to enhance ART adherence behavior prior to the advent of the biological sequelae of detectable viral load with potential for ARV resistance.

Footnotes

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.

Supplemental Material

The online [tables] are available at

References

Osterberg

Blaschke

. Adherence to medication. N Engl J Med. 2005;353 (5):487–497.

Messou

Chaix

M-L

Gabillard

. Association between medication possession ratio, virologic failure and drug resistance in HIV-1-infected adults on antiretroviral therapy in Côte d’Ivoire. J Acquir Immune Defic Syndr. 2011;56 (4):356–364.

Fielding

Charalambous

Stenson

. Risk factors for poor virological outcome at 12 months in a workplace-based antiretroviral therapy programme in South Africa: a cohort study. BMC Infect Dis. 2008;8:93.

Chesney

Ickovics

Chambers

. Self-reported adherence to antiretroviral medications among participants in HIV clinical trials: the AACTG adherence instruments. Patient Care Committee & Adherence Working Group of the Outcomes Committee of the Adult AIDS Clinical Trials Group (AACTG). AIDS Care. 2000;12 (3):255–266.

Machtinger

Bangsberg

. Adherence to HIV Antiretroviral Therapy [Internet]. 2006. Web site. http://hivinsite.ucsf.edu/insite?page=kb-03-02-09. Accessed August 26, 2012.

Ickovics

Cameron

Zackin

. Consequences and determinants of adherence to antiretroviral medication: results from Adult AIDS Clinical Trials Group protocol 370. Antivir Ther. 2002;7 (3):185–193.

Wainberg

Friedland

. Public health implications of antiretroviral therapy and HIV drug resistance. JAMA. 1998;279 (24):1977–1983.

Bangsberg

. Preventing HIV antiretroviral resistance through better monitoring of treatment adherence. J Infect Dis. 2008;197 (suppl 3):S272–S278.

Bershteyn

Eckhoff

. A model of HIV drug resistance driven by heterogeneities in host immunity and adherence patterns. BMC Syst Biol. 2013;7:11.

10.

Parienti

Massari

Descamps

. Predictors of virologic failure and resistance in HIV-infected patients treated with nevirapine- or efavirenz-based antiretroviral therapy. Clin Infect Dis. 2004;38 (9):1311–1316.

11.

Miller

Hays

. Adherence to combination antiretroviral therapy: synthesis of the literature and clinical implications. AIDS Read. 2000;10 (3):177–185.

12.

Farmer

. Methods for measuring and monitoring medication regimen adherence in clinical trials and clinical practice. Clin Ther. 1999;21(6):1074–1090; discussion 1073.

13.

Fairman

Motheral

. Evaluating medication adherence: which measure is right for your program? J Manag Care Pharm. 2000;6 (6):499–504.

14.

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 (8):1417–1423.

15.

Haubrich

Little

Currier

. The value of patient-reported adherence to antiretroviral therapy in predicting virologic and immunologic response. California Collaborative Treatment Group. AIDS Lond Engl. 1999;13 (9):1099–1107.

16.

UAB 1917 Clinic COHORT [Internet]. Web site. http://www.uab1917cliniccohort.org/. Accessed August 26, 2012.

17.

Kozak

Mugavero

. Patient reported outcomes in routine care: advancing data capture for HIV cohort research. Clin Infect Dis. 2012;54 (1):141–147.

18.

Godwin

Willig

Nevin

. Underutilization of the AIDS Drug Assistance Program: associated factors and policy implications. Health Serv Res. 2011;46 (3):982–995.

19.

Feldman

Fredericksen

Crane

. Evaluation of the single-item self-rating adherence scale for use in routine clinical care of people living with HIV. AIDS Behav. 2013;17 (1):307–318.

20.

Mugavero

Ostermann

Whetten

. Barriers to antiretroviral adherence: the importance of depression, abuse, and other traumatic events. AIDS Patient Care STDs. 2006;20 (6):418–428.

21.

Grossberg

Zhang

Gross

. A time-to-prescription-refill measure of antiretroviral adherence predicted changes in viral load in HIV. J Clin Epidemiol. 2004;57 (10):1107–1110.

22.

McMahon

Jordan

Kelley

. Pharmacy adherence measures to assess adherence to antiretroviral therapy: review of the literature and implications for treatment monitoring. Clin Infect Dis. 2011;52 (4):493–506.

23.

Hong

Jerger

Jonas

. Medication possession ratio associated with short-term virologic response in individuals initiating antiretroviral therapy in namibia. PloS One. 2013;8 (2):e56307.

24.

Goldman

Cantrell

Mulenga

Tambatamba

Reid

Levy

. Simple adherence assessments to predict virologic failure among HIV-infected adults with discordant immunologic and clinical responses to antiretroviral therapy. AIDS Res Hum Retroviruses. 2008;24 (8):1031–1035.

25.

Krousel-Wood

Islam

Webber

Morisky

Muntner

. New medication adherence scale versus pharmacy fill rates in seniors with hypertension. Am J Manag Care. 2009;15 (1):59–66.

26.

Fairley

Permana

Read

TRH

. Long-term utility of measuring adherence by self-report compared with pharmacy record in a routine clinic setting. HIV Med. 2005;6 (5):366–369.

27.

Simoni

Kurth

Pearson

Pantalone

Merrill

Frick

. Self-report measures of antiretroviral therapy adherence: A review with recommendations for HIV research and clinical management. AIDS Behav. 2006;10 (3):227–245.

28.

Liu

Golin