Share this article on:

Adherence to Highly Active Antiretroviral Therapy Assessed by Pharmacy Claims Predicts Survival in HIV-Infected South African Adults

Nachega, Jean B. MD, MPH*; Hislop, Michael MSc; Dowdy, David W. ScM; Lo, Melanie MHS*; Omer, Saad B. MBBS, MPH*; Regensberg, Leon MBChB, MRCP; Chaisson, Richard E. MD§; Maartens, Gary MBChB, FCP

JAIDS Journal of Acquired Immune Deficiency Syndromes: September 2006 - Volume 43 - Issue 1 - p 78-84
doi: 10.1097/01.qai.0000225015.43266.46
Epidemiology and Social Science

Summary: It is unclear how adherence to highly active antiretroviral therapy (HAART) may best be monitored in large HIV programs in sub-Saharan Africa where it is being scaled up. We aimed to evaluate the association between HAART adherence, as estimated by pharmacy claims, and survival in HIV-1-infected South African adults enrolled in a private-sector AIDS management program. Of the 6288 patients who began HAART between January 1999 and August 2004, 3805 (61%) were female and 6094 (97%) were black African. HAART adherence was ≥80% for 3298 patients (52%) and 100% for 1916 patients (30%). Women were significantly more likely to have adherence ≥80% than men (54% vs 49%, P < 0.001). The median (interquartile range) follow-up time was 1.8 (1.37-2.5) years. As of 1 September 2004, 222 patients had died-a crude mortality rate of 3.5%. In a multivariate Cox regression model, adherence <80% was associated with lower survival (relative hazard 3.23; 95% confidence interval: 2.37-4.39). When medication adherence was divided into 5 strata with a width of 20% each, each stratum had lower survival rates than the adjacent, higher-adherence stratum. Among other variables tested, only baseline CD4+ T-cell count was significantly associated with decreased survival in multivariate analysis (relative hazard 5.13; 95% confidence interval: 3.42-7.72, for CD4+ T-cell count ≤50 cells/μL vs >200 cells/μL). Pharmacy-based records may be a simple and effective population-level tool for monitoring adherence as HAART programs in Africa are scaled up.

From the Departments of *International Health and ‡Epidemiology, Johns Hopkins University, Bloomberg School of Public Health; §Department of Medicine, Division of Infectious Diseases, Johns Hopkins University, School of Medicine, Baltimore, MD; †Aid for AIDS Disease Management Programme (Pty) Ltd, Cape Town, South Africa; and the ∥Department of Medicine, Division of Clinical Pharmacology, University of Cape Town, Cape Town, South Africa.

Received for publication December 18, 2005; accepted April 17, 2006.

Drs J. Nachega, R.E. Chaisson, and G. Maartens acknowledge research support from the US National Institutes of Health, AI 5535901 and AI 016137. Dr J. Nachega is recipient of an NIH Mentored-Patient Oriented Research Career Award K23 AI068582-01. The funders had no input on the results or presentation of the results reported in the present article.

Presented as an oral presentation at the 12th Conference on Retroviruses and Opportunistic Infections, February 22 to 25, 2005, Boston, MA (WedsOrAb#25).

Reprints: Jean B. Nachega, MD, MPH, Department of International Health, Johns Hopkins University, Bloomberg School of Public Health, 615 N. Wolfe St, Suite W5031, Baltimore, MD 21205. E-mail:

The major determinants of survival in HIV-1-infected patients on highly active antiretroviral therapy (HAART) are the baseline CD4+ T-cell count, the presence of an AIDS-defining illness, HAART adherence, and drug-resistance status.1-3 Poor adherence significantly increases the risk of death in HIV-1-infected individuals.4,5

Antiretroviral therapy rollout programs through the US President's Emergency Plan for AIDS Relief (PEPFAR), the United Nations Global Fund to Fight AIDS, TB and Malaria (GFATM), and other national and international initiatives are increasingly providing access to HAART for eligible HIV-1-infected patients in sub-Saharan Africa.6,7 Although preliminary studies from these settings have shown that high levels of adherence and viral suppression are achievable, it is still unclear whether these results can be sustained when scaling up HAART programs for large numbers of patients in both the private and public sectors. Most of these earlier studies measured adherence by self-report alone, followed small numbers of patients for short periods, or provided no longitudinal data to shed light on sustained adherence rates.8-12 Furthermore, concerns have been raised that the inability to monitor adherence may ultimately undermine efforts to treat HIV/AIDS in high-burden areas.13,14

To date, there is no established "gold standard" for measuring HAART adherence. The various methods now used by researchers-including patient self-report, patient attendance at scheduled visits, clinician assessment, pill counts, plasma drug levels, electronic bottle monitors, and pharmacy-based records-each have advantages and disadvantages within different settings.15

As a measure of adherence, pharmacy-based data (refill and claims) are relatively simple to collect and therefore may be suitable as an adherence-monitoring tool in large HIV programs. Pharmacy refill prescriptions have previously been validated as a measure of adherence in HIV-1-infected adults, but only in developed-country settings.4,16

In this study, we evaluated the association between HAART adherence, as estimated by pharmacy claims in a private HIV/AIDS management program, and survival in HIV-1-infected South African adults.

Back to Top | Article Outline


Data Source

We evaluated records from HIV-1-infected adults enrolled in Aid for AIDS (AfA), a private-sector disease management program available to beneficiaries of contracted medical insurance funds (subsidized by employers) in southern Africa. Patient data and pharmacy claims have been recorded by AfA since June 1998. Pharmacy refill is defined here as the number of times that medicines have been dispensed, whereas pharmacy claims refer to the number of times that medicines were dispensed and subsequently claimed to gain reimbursement. With the patient's permission, baseline demographic and clinical data, including CD4+ T-cell count, HIV-1 RNA levels, and prior history of HAART, are captured in the AfA database upon application for the program. Acceptance is subject to confirmation of HIV-1 infection and proof of eligibility. Once enrolled, patients with a CD4+ T-cell count <350 cells/μL on 2 occasions or with an AIDS-defining condition were eligible for HAART. The regimen prescribed included at least 3 medications, including 2 nucleoside reverse transcriptase inhibitors (NRTIs) plus 1 nonnucleoside reverse transcriptase inhibitor, or 2 NRTIs plus 1 protease inhibitor (ritonavir-boosted saquinavir, indinavir or lopinavir).

Patients obtain authorization for reimbursement of HAART expenditure by their medical insurance fund, subject to (a) receipt of a prescription for HAART from their physician and (b) review and approval of the prescription by the AfA clinical staff in accordance with prespecified clinical guidelines.17 A 1-month supply of HAART was dispensed to patients, either at the pharmacy or via confidential mail-order pharmacy. For reimbursement, a claim must be submitted to the patient's health insurance fund. Each claim contains information about the dispense date, specific medication regimen, and quantity supplied. Nearly all claims received are reimbursed without any patient copayment.

Back to Top | Article Outline

Identification of Study Subjects

We identified all adult (≥18 years old) HAART-naive (medical records indicating no prior HAART use or no HIV-1 RNA <400 copies/mL or both) patients who qualified for and claimed at least 1 month of HAART between January 1999 and March 2003. For our primary analyses, we excluded patients who had died or had left the AfA program during their first 180 days of HAART (Fig. 1). The rationale was to minimize bias due to patients who entered with terminal disease and so had minimal chance of responding to therapy. Among these patients, early death may not be reflective of adherence levels. Secondary analyses were conducted in all patients and in patients with varying lag times after enrollment.



Back to Top | Article Outline

Operational Definitions, Outcome Measures, and Exploratory Variables

Adherence was expressed as a percentage, calculated as the number of months with claims submitted divided by the number of complete months between the date of HAART commencement and either (a) death, (b) withdrawal from the AfA program, or (c) the study end on 1 September 2004, and the result multiplied by 100. Patients were categorized into 6 groups based on calculated adherence: 0% to 19%, 20% to 39%, 40% to 59%, 60% to 79%, 80% to 99%, and 100%.

The primary outcome was all-cause mortality. During the follow-up period, deaths were identified by notification from either the attending medical practitioner, hospital case manager (for in-hospital deaths), or medical fund administrator. Patients who elected to leave their insurance fund or whose insurance fund changed to a different disease-management program were censored as lost to follow-up at the date of departure. Age, gender, race, baseline CD4+ T-cell count (surrogate for disease severity), and baseline plasma HIV-1 RNA levels were investigated in relation to both adherence and survival in the multivariate analyses. Secondary analyses included either death or loss to follow-up as the outcome. An additional secondary analysis included 12-month adherence, defined as the total months of HAART claimed during the first 12 months of therapy. By definition, the latter analysis was restricted to patients who survived for at least 12 months.

Back to Top | Article Outline

Statistical Analysis

Differences in baseline characteristics were assessed with contingency tables, agreement with the Pearson correlation coefficient (r), and statistical significance with the 2-sample Student t test for continuous variables and χ2 test for dichotomous variables. Kaplan-Meier plots were used to examine survival in the 6 strata of medication adherence defined earlier. These analyses were further stratified by baseline CD4+ T-cell count (≤50, 51-200, or >200 cells/μL). Cox proportional hazards regression18 and the log-rank test were used to model the individual and simultaneous effects of baseline variables and medication adherence on survival. Any baseline variable associated with survival (P < 0.1) in univariate analysis was included in the multivariate model. All variables were stratified into discrete categories, as follows: median adherence level (≥80%, <80%, or in the 6 strata defined earlier), gender (male, female), race (black, white, other), age (18-34, 35-44, 45-54, ≥55), HIV-1 RNA (<4, 4-4.99, ≥5 log10 copies/mL), and date of HAART initiation (in 6-month strata). Inasmuch as CD4+ T-cell count was found to modify the effect of adherence on survival, survival curves were stratified by CD4+ T-cell count. The assumption of proportional hazards was assessed by inspecting graphs of -log(-log[survival]) versus log(analysis time). All P values reported are nominal and 2-tailed, with a value of <0.05 considered statistically significant. Statistical analyses were performed using STATA Release 8.0 (Stata Corporation, College Station, TX).

Back to Top | Article Outline

Regulatory Approvals

This study was approved by the University of Cape Town Research Ethics Committee and the AfA Clinical Advisory Committee and Board, Cape Town, South Africa, and by the Johns Hopkins Bloomberg School of Public Health's Committee on Human Research, Baltimore, MD.

Back to Top | Article Outline


A total of 6288 patients were eligible for the study (Fig. 1). Of these, 82% were on 2 NRTIs plus 1 nonnucleoside reverse transcriptase inhibitor, and 18% were on 2 NRTIs plus 1 protease inhibitor. The total person-time contributed was 11,623 patient-years, and the median (interquartile range, IQR) follow-up time per participant was 1.8 (1.3-2.5) years; 3805 patients (60.5%) were female, 6094 (96.9%) were black Africans, and the mean age was 37 ± 8 years. The overall median (IQR) CD4+ T-cell count at HAART initiation in the population was 149 (65-227) cells/μL. The median (IQR) HIV-1 RNA level was 5.16 (4.63-5.62) log10 copies/mL. A total of 3298 (52.4%) had high (≥80%) adherence, which was associated with female gender (54% vs 49%, P < 0.001), older age (P < 0.04), and later enrollment in the study (P < 0.001). No other significant differences in baseline characteristics were observed between the 2 groups (Table 1).



During the study period, 222 patients (3.5%) died and 1155 (18%) were lost to follow-up (Fig. 1). In univariate analysis, patients with adherence <80% had significantly poorer survival than those with ≥80% adherence [relative hazard (RH) 3.01; 95% confidence interval (CI): 2.24-4.06]. However, no difference in survival was seen between patients with 80% to 99% versus 100% adherence (RH 0.85; 95% CI: 0.51-1.42). Below 80%, the patient's adherence level had a dose-response effect on survival, with each stratum of medication adherence having lower survival rates than the adjacent, higher-adherence stratum. This effect remained significant, although attenuated, when "death or loss to follow-up" as used as the outcome (Table 2).



Upon varying the lag period (the period following HAART initiation during which patients who died or were lost to follow-up were excluded from analysis), the association between adherence of <80% and decreased survival remained significant using any lag period of >90 days (Table 3). Furthermore, compared with 100% adherence, 80% to 99% adherence was significantly associated with decreased survival (RH 1.38; 95% CI: 1.08-1.80) using a lag period of 455 days (15 months). At longer lag periods, the point estimate of the association between adherence and survival continued to increase, but statistical significance was attenuated by small sample size.



Other variables significantly associated with decreased survival in univariate analysis included male gender (RH 1.50; 95% CI: 1.15-1.95), CD4+ T-cell count ≤200 cells/μL (RH 3.00; 95% CI: 2.12−4.28), and HIV-1 RNA >105 copies/mL (RH 2.93; 95% CI: 1.44-5.95). After adjusting for other variables, only adherence <80% (RH 3.23; 95% CI: 2.37-4.39) and CD4+ T-cell count ≤200 cells/μL (RH 2.54; 95% CI: 1.77-3.62) remained significantly and independently associated with decreased survival (Table 4).



A total of 3267 patients had data available with which to calculate adherence at 12 months after HAART initiation. In this population, the median adherence was 83% at 12 months and 80% at the end of the study. Adherence of ≥80% at 12 months was predictive of the same level of adherence at the end of study (Pearson r = 0.94, P < 0.001). Furthermore, 12-month adherence remained highly significant as a predictor of subsequent mortality (crude RH 3.61; 95% CI: 2.29-5.68; adjusted RH 3.64; 95% CI: 2.30-5.75).

No significant interactions were detected between adherence (at <80% vs ≥80% threshold) and any baseline variable as predictors of survival. Although the interaction between adherence and CD4+ T-cell count did not achieve statistical significance (P = 0.09), the effect of poor (<80%) adherence on survival was greatest in those with low CD4+ T-cell count, with a crude relative hazard of 4.54 (95% CI: 2.83-7.29) in the most immunosuppressed (≤50 cells/μL) versus 2.39 (95% CI: 1.51-3.78) and 2.08 (95% CI: 1.00- 4.31) in patients with CD4+ T-cell count of 51 to 200 and >200 cells/μL, respectively (Fig. 2). Treating CD4+ T-cell count as continuous variable rather than a categorical one did not change the findings. In addition, inclusion of patients with missing baseline data did not change any univariate point estimate of relative hazard by more than 5%.



Back to Top | Article Outline


To our knowledge, this study of HAART adherence in a cohort of 6288 HIV-1-infected South African adults is the largest to document HAART adherence as measured by pharmacy claims as a predictor of survival. Strengths of this study include its sample size, use of a simple adherence measure, direct measurement of adherence and survival in a developing-country setting, and a striking dose-response relationship between adherence and survival.

We found that patients on HAART claiming <80% of their prescription refills have a risk of mortality that is more than 3 times greater than that of patients who are ≥80% adherent, after adjusting for baseline variables. These findings corroborate prior results from studies in developed countries that have used pharmacy-refill data, including a hazard ratio of 2.90 (95% CI: 1.93-4.36) for adherence <75% in Canada4 and a hazard ratio of 3.87 (95% CI: 1.77-8.47) for adherence <90% in Spain.5 Furthermore, the effect of adherence is strongly modified by baseline CD4+ T-cell count; mortality risk increases significantly as adherence drops below 80% in patients with a CD4+ T-cell count ≤50 cells/μL, while remaining relatively constant in other patients unless adherence drops below 60% (Fig. 2). The latter finding suggests that intensive adherence interventions should target HIV-1-infected patients with the lowest CD4+ T-cell count.

Each 20% decrease in pharmacy-claim adherence below 80% was associated with a decrease in survival in dose-response fashion, even when death or loss to follow-up was used as the outcome. Indeed, in some cases, the reason for loss to follow-up may be due to imminent death, and in some cases, not. Both extremes of these scenarios were investigated in the survival analysis, one in which patients were censored from the analysis on leaving the insurance fund or a more conservative scenario where these patients were considered deceased. Compared with those with 100% adherence, patients with adherence of 80% to 99% had significantly better survival when using a lag period of <90 days (Table 3). This counter-intuitive finding does not persist when using a lag period of >90 days, suggesting that it reflects deaths among patients enrolling with advanced disease, for whom 100% adherence corresponds to a small number of claims made and in whom death is less likely to be a consequence of nonadherence. In addition, our finding of no significant difference in survival comparing those with 80% to 99% versus 100% adherence does not refute the need for nearly perfect (at least 95%) adherence for optimal outcomes. Rather, this finding reflects the inadequacy of even a 6-month lag period in excluding all patients with advanced disease, coupled with the inability of 100% monthly pharmacy claims to distinguish between those patients who are truly adherent from those patients who fill claims but do not actually ingest all of their medication. Finally, these results suggest that pharmacy claims may be a valid marker of adherence in AIDS patients on HAART for at least 3 to 6 months, but that claims should not be used to predict individual outcomes, particularly in the months immediately after HAART initiation.

In our study cohort, half of enrolled HIV-1-infected patients claimed more than 80% of their prescriptions, and one third claimed <60%. For patients who had been on HAART for at least 12 months, 62% claimed >75% of their prescriptions. These rates are relatively lower than those reported in studies from developed countries that used pharmacy records. Hogg et al4 reported that among 1282 HIV-1-infected individuals, 74% had adherence >75% and 57% had adherence >95% in their first year on HAART, based on pharmacy refill data. In addition, in a 1-year retrospective review of pharmacy refill data for 100 patients in Canada, Ostrop et al reported >80% adherence in 75% of patients.19 The lower rate of adherence found in this study also differs from previous studies in sub-Saharan Africa that reported a high proportion (>80%) of patients with excellent (>95%) adherence11-14 using different adherence measures and in developed countries (mean adherence range: 54%-84%) using electronic bottle monitors.3,20-23 This discrepancy is likely to reflect difference in adherence measurement tools used or a selection bias leading to overestimation of preliminary studies reported from Africa,8-12 most of which were cross-sectional, measured adherence by self-report, were with short-term follow-up, and included patients participating in ongoing randomized trials with rigorous exclusion criteria and substantial structural support.

Our study has several limitations. First, there is no evidence from our study that the pharmacy claims data reflect the number of pills taken correctly by a patient. Claims data may underestimate adherence if patients acquire their medications from other sources and fail to submit the claims. More likely, however, claims will overestimate actual adherence, as patients may not actually take all of their claimed medications. However, a number of studies have found that refill compliance correlates well with other compliance behaviors such as appointment keeping, medication consumption,24 or viral load suppression.16 It may be reasonably assumed that patients would not continue to refill a prescription (or, in this case, to claim medication) without intending to adhere.25 In any case, the dose-response relationship between adherence and survival found in our study suggests that pharmacy claims may be appropriately used as a program-level HAART adherence measure, irrespective of whether pharmacy claims directly correlate with consumption.

A second limitation is that the relatively coarse outcome measurement (in that pharmacy claims could only be assessed monthly) prevents us from analyzing adherence in strata finer than 20% intervals. As such, caution must be exercised in interpreting our results, particularly the implicit assumption that mortality is homogeneous within the 80% to 99% adherence stratum. For example, in our data set, no patient with 95% to 99% adherence experienced a fatal outcome. However, mathematically, such patients would be required to live for at least 20 months and to miss exactly 1 month of HAART to have an adherence rate between 95% and 99%, and thus, the sample size of such patients is too small to analyze appropriately.

Third, our findings are subject to the potential for "reverse causation" (eg, patients may stop taking medication for reasons related to poorer survival) and misclassification of adherence (eg, patients who fall within a given adherence stratum at the end of follow-up might not fall within that stratum throughout follow-up). These concerns, however, are unlikely to account for our primary findings. An analysis in a subset of patients with 12-month adherence data, in whom adherence was assessed before becoming eligible for any outcome, showed that HAART adherence remained strongly predictive of both survival and adherence at the end of the study follow-up.

The much-anticipated influx of funds and resources under initiatives such as the World Health Organization's 3-by-5 campaign and the GFATM6 and PEPFAR7 has dramatically increased access to HAART. However, it is still unclear how adherence to HAART should be monitored, especially in large HAART programs.26 Our results originate from a private HIV/AIDS managed health care program, but only 18% of South Africans have private medical insurance.27 These findings, however, have important policy implications because lessons learned here could be adapted to the public sector by using pharmacy refill records (as opposed to pharmacy claims) to monitor adherence in large HIV programs. Therefore, we recommend that HIV/AIDS programs, such as those sponsored by public agencies (GFATM, PEPFAR, Médecins Sans Frontières), consider tracking pharmacy data as a measure of adherence. Such public-private partnerships are essential in countries such as South Africa, where the private sector has greater resources.28 Further studies are needed to assess the utility of pharmacy-refill information in the public sector.

In summary, HAART adherence of ≥80%, as measured by pharmacy claims, is associated with a 3-fold decrease in mortality hazard among HIV-1-infected South African adults participating in a private insurance program. These data suggest that pharmacy claims records are valid indicators of HAART adherence in settings where other more labor-intensive or expensive methods are impractical. Measures of medication adherence based on pharmacy records may be useful as a simple and low-cost population-level tool to monitor adherence to antiretroviral therapy and to assess the impact of specific interventions targeting improved adherence in HIV-1-infected patients in sub-Saharan Africa.

Back to Top | Article Outline


1. Hogg R, Yip S, Chan B, et al. Rates of disease progression by baseline CD4 cell count and viral load after initiating triple-drug therapy. JAMA. 2001;286:2568-2577.
2. Bangsberg D, Hetch R, Charlebois FM. Adherence to protease inhibitors, HIV-1 viral load, and development of drug resistance in an indigent population. AIDS. 2000;14:357-366.
3. Paterson D, Swindells L, Mohr G, et al. Adherence to protease inhibitor therapy and outcomes in patients with HIV infection. Ann Intern Med. 2000;133:21-30.
4. Hogg R, Heath S, Bangsberg K, et al. Intermittent use of triple-combination therapy is predictive of mortality at baseline and after 1 year of follow-up. AIDS. 2002;16:1051-1058.
5. de Olalla P, Knobel G, Carmona H. Impact of adherence and highly active antiretroviral therapy on survival in HIV-infected patients. J Acquir Immune Defic Syndr. 2002;30:105-110.
6. WHO. Scaling up antiretroviral therapy in resource-limited settings: treatment guidelines for a public health approach [online]. 2003. Available at: Accessed on June 28, 2005.
7. US Department of State. The President's emergency plan for AIDS relief. [online]. 2003. Available at: Accessed on June 28, 2005.
8. Weidle P, Malamba J, Mwabaza S, et al. Assessment of a pilot antiretroviral drug therapy programme in Uganda: patients' response, survival, and drug resistance. Lancet. 2002;360:34-40.
9. Orrell C, Bangsberg DR, Badri M, et al. Adherence is not a barrier to successful antiretroviral therapy in South Africa. AIDS. 2003;17:1369-1375.
10. Nachega J, Stein B, Lehman DM, et al. Adherence to antiretroviral therapy in HIV-infected adults in Soweto, South Africa. Aids Res Hum Retroviruses. 2004;20:1053-1056.
11. Laurent C, Diakathe N, Gueye NFN, et al. The Senegalese government's highly active antiretroviral therapy initiative: an 18-month follow-up study. AIDS. 2002;16:1363-1370.
12. Oyugi JH, Byakika-Tusiime J, Charlebois ED, et al. Multiple validated measures of adherence indicate high levels of adherence to generic HIV antiretroviral therapy in a resource-limited setting. J Acquir Immune Defic Syndr. 2004;36:1100-1102.
13. Harries A, Nyangulu D, Hargreaves DS, et al. Preventing antiretroviral anarchy in sub-Saharan Africa. Lancet. 2001;358:410-414.
14. Gill CJ, Hamer DH, Simon JL, et al. No room for complacency about adherence to antiretroviral therapy in sub-Saharan Africa. AIDS. 2005;19:1243-1249.
15. Chesney MA, Morin M, Sherr L. Adherence to HIV combination therapy. Soc Sci Med. 2000;50:1599-1605.
16. Grossberg R, Zhang W, Gross R. A time-to-prescription-refill measure of antiretroviral adherence predicted changes in viral load in HIV. J Clin Epidemiol. 2004;57:1107-1110.
17. AfA Clinical Guidelines [online]. Available at: Accessed on June 28, 2005.
18. Cox DR. Regression models and life tables. J R Stat Soc B. 1972;34:187-202.
19. Ostrop NJ, Hallett KA, Gill MJ. Long-term patient adherence to antiretroviral therapy. Ann Pharmacother. 2000;34:703-709.
20. Walsh J, Mandalia C, Gazzard S. Responses to a 1 month self-report on adherence to antiretroviral therapy are consistent with electronic data and virological treatment outcome. AIDS. 2002;16:269-277.
21. Arnsten J, Demas H, Farzadegan PA, et al. 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.
22. McNabb J, Nicolau J, Stoner DP, et al. Patterns of adherence to antiretroviral medications: the value of electronic monitoring. AIDS. 2003;17:1763-1767.
23. Sethi AK, Celentano DD, Gange SJ, et al. Association between adherence to antiretroviral therapy and human immunodeficiency virus drug resistance. Clin Infect Dis. 2003;37:1112-1118.
24. Steiner JF, Prochazka AV. The assessment of refill compliance using pharmacy records: methods, validity, and applications. J Clin Epidemiol. 1997;50:105-116.
25. Dezii CM. Persistence with drug therapy: a practical approach using administrative claims data. Manag Care. 2001;10:42-45.
26. South Africa Department of Health. September 2004. Monitoring and evaluation framework for the comprehensive HIV and AIDS treatment programme for South Africa [online]. Available at: Accessed on June 28, 2005.
27. Benatar S. Health care reform and the crisis of HIV and AIDS in South Africa. N Engl J Med. 2004;351:81-92.
28. England R. HIV/AIDS: the private sector is vital. Lancet. 2004;364:1033-1034.

HIV; HAART; adherence; pharmacy claims; survival; South Africa

© 2006 Lippincott Williams & Wilkins, Inc.