The expansion of antiretroviral therapy (ART) services to provide treatment to more than 10 million people living in resource-limited settings over the past decade has been an unprecedented development in global public health.1 Despite this, HIV treatment coverage in adults is estimated at only 38%,1 and there is a particularly urgent need for expansion of ART services across sub-Saharan Africa and other resource-limited settings. Alongside the challenge of accelerating ART expansion, attrition among patients on ART represents a substantial hurdle1,2 with levels of loss to follow-up (LTFU) ranging between 30% and 35% after 3 years.3 In South Africa and many other countries, the risk of LTFU frequently seems higher in men4,5 and younger patients,5 and there is an ongoing need to identify interventions that can help minimize LTFU in patients on ART.
To facilitate this expansion and improve retention in care, reorientation of ART delivery through decentralization of services into communities and the shifting of specific tasks related to ART provision to different cadres of health care workers have been proposed.6 Recent systematic reviews from models with task shifting concluded that patients supported by nurses or community health workers (CHWs) have comparable outcomes to those supported by doctors.7,8 There is also burgeoning interest in community-based ART delivery with pilot studies reporting favorable outcomes.9,10 Community-based adherence clubs (CACs) facilitated by CHWs have been implemented to provide effective ART support to stable ART patients. These were designed both to decongest local primary health care facilities and to reduce attrition by decreasing the frequency and intensity of patient visits and providing support within the community.
There is an urgent need for data on the outcomes of patients managed by community-based ART delivery models. However, comparing patient outcomes between community-versus facility-based models of care is a challenge within operational research given that the clinical and/or demographic criteria used to identify patients to be referred to decentralized services may be associated with a positive prognosis.11,12 This study describes outcomes [LTFU and viral rebound] over the first 18 months of CAC implementation in Cape Town, South Africa and compares patient outcomes under the CAC model of care to those of patients managed in facility-based primary care.
We conducted a cohort study of patients receiving ART at the Gugulethu Community Health Centre (CHC) and their CACs.
The Gugulethu CHC is a large primary health care facility typical of urban public sector ART services across the region and has been described in detail previously.13,14 The service began providing ART in 2002 based on South African national program criteria for ART initiation and regimens. CD4 cell count and viral load were monitored before ART initiation, after 4 months on ART and then repeated annually. At each visit patients had a clinical consultation with either a medical doctor or professional nurse followed by adherence support from a counselor before proceeding to the pharmacy to collect their ART. Each patient was assigned to a peer counselor who conducted pill counts at each clinic visit and provided individual and group sessions to promote high levels of adherence. For patients who defaulted, their counselor followed up through phone or home visit. The ratio of counselors to ART patients increased from 2006 onward from when the total number of counselors was limited to approximately 30.15
Community-Based Adherence Clubs
The adherence club model of care16 and details of the implementation of CACs in Gugulethu have been presented previously.17 From June 2012, patients classified as stable on ART were recruited from the CHC for down-referral to the CAC program. Specific eligibility criteria were: self-reported adherence to ART, on ART for >12 months, 2 consecutive suppressed viral loads (<400 copies/mL), and no active opportunistic infections. A CAC was a community-based, CHW led- and nurse-supported model of care supporting groups of 25 to 30 patients. CACs met every 2 months for group counseling, a brief symptom screening, and distribution of prepacked ART. CAC patients could send a patient-nominated treatment supporter or “buddy” to collect their ART at alternating CAC visits. Over a 12-month period, each CAC would meet 5 times including 1 clinical consultation. There was a strong emphasis on patient self-management, and patients who developed clinical complications would be up-referred to the CHC. Beginning in May 2013, adherence clubs were relocated to a community venue where all CAC visits, including clinical consultations and blood collection, took place. A team of 4 CHWs supported CACs with a professional nurse available at the community venue for annual monitoring bloods (CD4, viral load, and creatinine as necessary) and clinical consultation.
Data were obtained from the prospectively collected CHC clinical database, the CAC database, and the National Health Laboratory Service database.
Definitions of Outcomes
Patients were eligible for inclusion in the analysis if they initiated ART before the end of 2012. Patients entered the analysis on the date of ART initiation in the CHC, contributing person time to the facility-based model of care. Patients decentralized to the CACs contributed person time to the CAC model from their first CAC visit. Patient outcomes were assessed as of the end of 2013 (analysis closure); the database closure was the 21st of March 2014.18
The outcomes of interest in this analysis were LTFU and viral rebound. Mortality was excluded as an outcome because only 9 patients who accessed a CAC died during the study period.17 LTFU was defined as having no visit in the first 12 weeks of 2014, and patients were censored at the date of last contact with either health care service. Viral rebound was defined as a single viral load measurement >1000 copies per milliliter after previous suppression (<1000 copies/mL). Outcome definitions were the same for all patients irrespective of their exposure to a CAC.
Pre-ART characteristics of all patients by model of delivery were summarized; continuous variables are reported with medians and interquartile ranges and categorical variables as proportions with 95% confidence intervals (CIs). Time-updated CD4 and viral load values were available at each patient contact, defined as the closest measure within a 2-month window of each patient contact. P-values from Pearson χ2 and Wilcoxon rank-sum tests were calculated to investigate differences between patients in facility-based care at the CHC versus those attending CACs.
Time to LTFU and viral rebound from first CAC visit were analyzed by sex and age [youth (16–24 years) and adults (≥25 years)] and differences investigated with the log-rank test. A series of proportional hazards models were used to model relative hazards of different patient outcomes adjusting for demographic, programmatic, and clinical variables including time-updated laboratory values. Results are presented as hazard ratios (HR) or adjusted hazard ratios (aHR) with 95% CI. The reference categories used in the models were the age group 25–34 years old at ART initiation, female, 2008–2010 ART initiation, CD4 cell count of 100–199 cells per microliter at ART initiation, and suppressed viral load. The reference categories used for age, year of ART initiation, and CD4 count were to facilitate examination of possible trends at lower and higher values. All available demographic and clinical variables were included in the multivariate models if they were considered potential confounders or associated with the outcome. To preserve observations, models were built by adding variables with progressively less complete data.
Because allocation to the CAC model of care was not randomly assigned, a number of approaches were undertaken to compare LTFU with the CHC. In the primary results (approach 1), we used proportional hazards models with the treatment (being in a CAC versus CHC) and time-updated laboratory values as time-varying covariates to model time to LTFU.14 In addition, we modeled the probability of CAC participation using inverse probability weighting (IPW)19 and incorporated this weight into the proportional hazards models. To calculate the IPW we used a dataset restricted to patients for whom a CAC was available after 12 months on ART (ie patients initiating after June 2011). Stratified HR by age, sex, year of initiation, and CD4 cell count at ART initiation were also generated to assess if LTFU differed by subgroup.
We ran a series of secondary analyses to examine different methods of accounting for selection biases when examining the association between model of care and LTFU. In approach 2, proportional hazards models without IPW adjusting for the probability of being in a CAC were used. Furthermore, Cox models without (approach 3) and with the IPW (approach 4) were restricted to patients who were in care 12 months after ART initiation. We also used the IPW with logistic regression to model LTFU (approach 5).11 Approach 6 generated propensity scores and the average effect of the treatment on the treated was assessed.12,20
Ethical approval for collection and analysis of routine clinical data and data from CACs was obtained from the Human Research Ethics Committee at the University of Cape Town Faculty of Health Sciences. Data were analyzed using STATA 13.0 (STATA Corporation, College Station, TX).
A total of 2113 patients were decentralized to a CAC between May 2012 and December 2013 contributing a total of 2098 person-years of follow-up to CACs (median 1.1 years, interquartile range 0.75–1.26). CAC patients represented a quarter of all patients ever initiated onto ART at the site (26%). The time in the CAC was 6.7% of the total person-years of follow-up in the analysis. Patients referred to CACs had similar pre-ART demographic and clinical characteristics to patients retained in facility-based care (Table 1). Differences between the two groups were apparent after ART initiation: before down-referral, CAC patients were more likely to achieve viral suppression (viral load >1000 copies/mL) after 4 and 12 months on ART compared with patients managed in facility-based care at the CHC (P-values <0.001 [Table 1]). After 12 months on ART, the median CD4 cell count was higher in CAC patients versus CHC patients (320 cells/μL versus 295 cells/μL, P-value <0.001).
Outcomes of Patients Referred to CACs
LTFU among CAC patients was 5.6% and 6.4% at 12 months after the first CAC visit, for men and women, respectively (Fig. 1A; P-value 0.961). LTFU was higher for youth (ages 16–24) than adults (patients ≥25), with 9.1% and 5.9% LTFU at 12 months after referral to CACs, respectively (Fig. 1B, P-value 0.022). When stratified by sex and age, youth had higher LTFU than adults for both men and women (12-month LTFU: 7.7% versus 5.5% for men and 9.2% versus 6.1% for women) (results not shown).
In proportional hazards models of patients referred to CACs, the hazard of LTFU was higher in youth versus adults (HR: 1.67, 95% CI: 1.00 to 2.79), and in patients who initiated ART in 2002–2004 versus after 2008 (HR: 1.64, 95% CI: 1.00 to 2.69) (Table 2). In adjusted models of CACs, youth were twice as likely to be LTFU compared with adults (aHR: 2.17, 95% CI: 1.26 to 3.73). An increased risk of LTFU was also found for patients who initiated ART with a CD4 cell count ≥200 cells per microliter (aHR 2.25, 95% CI: 1.36 to 3.72) and a decreased risk in patients with a CD4 cell count <50 cells per microliter (aHR: 0.52, 95% CI: 0.29 to 0.93) compared with patients initiating with a CD4 cell count between 50 and 199 cells per microliter in CACs. Patients in CACs who initiated ART between 2011 and 2012 had a 58% decrease in the risk of LTFU (aHR: 0.42, 0.22–0.86) compared with those initiating between 2008 and 2010. No difference in the risk of LTFU was found by sex (aHR: 0.90, 95% CI: 0.61 to 1.34) or in patients who sent a buddy to collect medications (aHR 0.89, 95% CI: 0.59 to 1.36).
During the study period, 3.0% of CAC patients experienced viral rebound. No difference in viral rebound was found by sex (Fig. 1C, P-value 0.173) or age (Fig. 1D, P-value 0.194). No associations were found between pre-ART characteristics and risk of viral rebound in univariate models (Table 2). In final models of CACs, youth were twice as likely to experience viral rebound compared with adults (aHR: 2.24, 95% CI: 1.00 to 5.04).
Comparison of LTFU in CACs Versus CHC
CACs were associated with a decreased risk of LTFU compared with the CHC in all crude and adjusted models (Table 3). We report on approach 1 using inverse probability weights to estimate CAC participation and included in proportional hazard models with time-varying covariates (see Appendix, Supplemental Digital Content, http://links.lww.com/QAI/A758). The IPW model correctly classified 91% of patients (not shown). In the final model, also adjusting for time-updated CD4 and viral load, CACs were associated with a CAC 67% reduction in LTFU compared with the CHC (aHR: 0.33, 95% CI: 0.27 to 0.40, Table 3).
In stratified models, CACs were associated with a reduced risk of LTFU for all patients compared with the CHC irrespective of age, sex, year of ART initiation, or CD4 cell count at ART initiation (Table 4). The exception was for youth 16–24 years of age whose adjusted hazard of LTFU not significantly different comparing CACs and the CHC (aHR: 0.68, 95% CI: 0.37 to 1.22, P-value 0.197).
In all secondary analyses, CACs were associated with a decreased risk of LTFU compared with the CHC (Table 3). Under approach 3, the most conservative scenario, models were restricted to patients on ART after 12 months and the hazard of LTFU in CACs compared with the CHC was 0.25 (95% CI: 0.21 to 0.31) (Table 3). When inverse probability weights to estimate CAC participation were incorporated into this model with further restriction (approach 4), the adjusted hazard of LTFU was 0.47 (95% CI: 0.39 to 0.56). Propensity scores were also generated adjusting for age, sex, rate of scale-up, and program size. In all matching methods, the average treatment on the treated effect was reduced, ranging from −0.141 to −0.641 (results not shown).
This analysis suggests that the CAC model may achieve favorable programmatic outcomes for stable patients on ART in resource-limited settings. For most patient groups, CACs were associated with a substantial decrease in the risk of LTFU compared with facility-based care in the CHC. The exception was youth: patients 16–24 years of age at ART initiation experienced no difference in the risk of LTFU in CACs compared with the CHC. Within CACs, being a youth at ART initiation was associated with an increased risk of LTFU and viral rebound compared with adults. Differences in the risk of LTFU were observed by CD4 cell count and year of ART initiation within CACs, although men and women had comparable LTFU and viral suppression.
The CACs were associated with a reduction in the risk of LTFU compared with facility-based care with a two-thirds reduction in the hazard of LTFU. The HR in the final adjusted model of 0.33 is similar to that from facility-based adherence clubs reporting 0.43 in modeling the hazard of LTFU or death.11 These are novel data as previous results for models of care labeled as “community-based” are focused primarily on models of home-based care for individual patients.21
There is a particular concern about how to effectively provide ART to men, a group that represents more than 40% of the HIV epidemic in SA22 but have inequitable access to ART.23 Community-based models of ART delivery may be uniquely positioned to support men on ART.9 Among CAC patients, outcomes did not differ by sex, a promising finding given that men consistently have worse outcomes on ART in primary care.24,25 Interestingly, these data are in keeping with recent evidence from community-based drug dispensing programs suggesting no difference in retention between men and women when care is provided outside facilities.26 By contrast, men in the community-based adherence group model in Mozambique had twice the risk of attrition compared with women.10 Clearly further research into what components of community models of care can support the retention of men is warranted.
We found that young people were the only subset of patients who did not experience a decrease risk of LTFU in CACs compared with the CHC. This finding is crucial as 40% of new infections occur in this age group and young people are a special concern within ART services.27 A renewed focus on strategies to retain youth in care, both for their own health and given the potential benefits of ART as prevention in this age group, is required.
CAC patients initiating ART with CD4 cell counts ≥200 cells per microliter had an increased risk of LTFU compared with patients with CD4 cell counts between 100 and 199 cells per microliter at ART initiation. However, rates of LTFU in CACs among patients with CD4 cell counts ≥200 cells per microliter at initiation were still less than a third of those observed under facility-based care. These findings highlight that retention of patients initiating ART at higher CD4 cells is a challenge facing both models of care. As the guidelines encourage healthier patients to be initiated onto ART, there must be a concurrent shift away from the common perception that ART is for sick patients.28 Our findings for decreased LTFU in CACs regardless of CD4 cell count at baseline points to the potential for community-based models of care as a vehicle to assist in reaching optimistic global treatment targets while supporting long-term retention in care.1,9
A number of strengths and limitations must be considered in the interpretation of these data. Our data are from a large cohort that decentralized a previously described model into the community.14 Data were available for comparing the CAC with facility-based care including data on time-varying laboratory values.5 We were limited to reporting short-term outcomes over the 18 months of the program. Furthermore, we acknowledge that the outcome LTFU represents all patients with an unknown outcome including unascertained deaths and transfers,29 administrative LTFU, and those currently interrupting treatment.30 Moreover, our data are from a health service with an established CHW program, and results from any health services research should be generalized with caution. To assess data quality during the recent scale-up at the CHC, a random sample of patients identified to be LTFU among those initiating ART in 2011 and 2012 were selected for a folder review. Of the patients classified as LTFU, two-thirds were confirmed to be LTFU, whereas approximately 20% were administrative LTFU and 15% were unascertained transfers. The effect of CACs on reducing LTFU is therefore likely overestimated but not entirely explained by issues of data quality.
These observational data do not provide a direct comparison of outcomes between the CACs and facility-based models of care. The major concern here is selection bias into the intervention (CACs). We used different methods to attempt to adjust for this bias.11,12,14 However in the absence of randomization, we cannot rule out residual confounding because of unmeasured covariates or covariates measured with imprecision. This problem is ubiquitous in observational evaluations of health services interventions. In this light, it may be more appropriate to compare novel models of care, such as CACs, to predetermined targets for outcomes rather than attempting to overcome the methodological challenges of comparing them to increasingly heterogeneous patients in facility-based care. Here, LTFU from CACs at 12 months was 6% and 98% of the retained patients were virally suppressed. These figures, even without comparison to other services, may be taken to represent successful ART provision under this model of care.
Our findings on the potential benefits of CACs are relevant as national programs look to expand ART eligibility in line with the 2015 World Health organization guidelines.6 Nevertheless, questions remain about how community-based models of ART delivery can best support service expansion and retention in care. Patients were only eligible to join this program after a year on ART, but the model is of possible benefit to patients sooner after treatment initiation, especially as healthier patients access ART. Expansion of the model to include groups of relatively health HIV-infected individuals with inequitable access to and outcomes on ART, such as women who initiate ART in pregnancy, should be considered. In addition, because sending a buddy reduced the visit frequency for the patient and we found no difference in the outcomes for use of a buddy, further reducing the visit frequency may be possible without negatively impacting outcomes.31 More data are needed to determine the long-term outcomes of CACs. In addition, documenting patient experiences and preferences and conducting cost-effectiveness analyses would be beneficial to scaling up the model. Further investigations into how community-based programs could support ART at different stages of the treatment cascade, such as testing and ART initiation, are also needed.
From a policy perspective, community-based models raise questions about how to integrate ART provision allowing for patients to return to normal life and minimize the disruptions from ART delivery.32,33 In the context of increased availability of fixed-dose combination therapy and point of care viral load testing, policies regarding who can distribute ART and the frequency of rescripting and clinical consultations need to be reconsidered.
In summary, we found that stable primary care patients were successfully managed by CHWs within a community-based model of ART delivery. Higher rates of retention and viral suppression were maintained in both men and women. More long-term data are needed to understand how best community-based models of care can improve retention in care with particular attention given to supporting youth. Given that ART programs in high-prevalence, resource-limited settings are continuing to expand, community-based models of care represent a potential alternative to traditional facility-based models of ART delivery.
The authors thank the Desmond Tutu HIV Foundation and Dr. Richard Kaplan for their ongoing support of the Hannan Crusaid Treatment Centre and the Sizophila Counsellors. Thank you to Joseph Sharp for comments on an earlier version of this article. The authors would like to acknowledge the staff of the Provincial Government of the Western Cape, particularly Dr. Patti Olckers and the Clubs Steering Committee. Finally, thank you to the patients and staff of the Gugulethu CHC and the Adherence Club program.
1. UNAIDS. Fast-track: Ending the AIDS Epidemic by 2030. Geneva, Switzerland: UNAIDS; 2014.
2. World Health Organization. Retention in HIV Programmes Defining the Challenges and Identifying Solutions: Meeting Report, 13–15 September 2011. Geneva, Switzerland: WHO; 2012.
3. Fox MP, Rosen S. Retention of adult patients on antiretroviral therapy in low- and middle-income countries: systematic review and meta-analysis 2008–2013. J Acquir Immune Defic Syndr. 2015;69:98–108.
4. Wandeler G, Keiser O, Pfeiffer K, et al.. Outcomes of antiretroviral treatment programs in rural Southern Africa. J Acquir Immune Defic Syndr. 2012;59:e9–16.
5. Meloni ST, Chang C, Chaplin B, et al.. Time-dependent predictors of loss to follow-up
in a large HIV treatment cohort in Nigeria. Open Forum Infect Dis. 2014;1:1–11.
7. Kredo T, Adeniyi FB, Bateganya M, et al.. Task shifting
from doctors to non-doctors for initiation and maintenance of antiretroviral therapy. Cochrane Database Syst Rev. 2014;7:CD007331.
8. Suthar AB, Rutherford GW, Horvath T, et al.. Improving antiretroviral therapy scale-up and effectiveness through service integration and decentralization
. AIDS. 2014;28:S175–S185.
9. World Health Organization. March 2014 Supplement to the 2013 Consolidated Guidelines on the Use of Antiretroviral Drugs for Treating and Preventing HIV Infection. Recommendations for a Public Health Approach. Geneva, Switzerland: WHO; 2014.
10. Decroo T, Koole O, Remartinez D, et al.. Four-year retention and risk factors for attrition among members of community ART groups in Tete, Mozambique. Trop Med Int Health. 2014;19:514–521.
11. Luque-Fernandez MA, Van Cutsem G, Goemaere E, et al.. Effectiveness of patient adherence groups as a model of care for stable patients on antiretroviral therapy in Khayelitsha, Cape Town, South Africa. PLoS One. 2013;8:e56088.
12. Brennan AT, Long L, Maskew M, et al.. Outcomes of stable HIV-positive patients down-referred from a doctor-managed antiretroviral therapy clinic to a nurse-managed primary health clinic for monitoring and treatment. AIDS. 2011;25:2027–2036.
13. Bekker L-G, Myer L, Orrell C, et al.. Rapid scale-up of a community-based HIV treatment service: programme performance over 3 consecutive years in Guguletu, South Africa. South Afr Med J. 2006;96:315–320.
14. Grimsrud A, Kaplan R, Bekker LG, et al.. Outcomes of a nurse-managed service for stable HIV-positive patients in a large South African public sector antiretroviral therapy programme. Trop Med Int Health. 2014;19:1029–1039.
15. Nglazi MD, Lawn SD, Kaplan R, et al.. Changes in programmatic outcomes during 7 years of scale-up at a community-based antiretroviral treatment service in South Africa. J Acquir Immune Defic Syndr. 2011;56:e1–8.
16. Wilkinson LS. ART adherence clubs: a long-term retention strategy for clinically stable patients receiving antiretroviral therapy. S Afr J HIV Med. 2013;14:48–50.
17. Grimsrud A, Sharp J, Kalombo C, et al.. Implementation of community-based adherence clubs for stable ART patients. J Int AIDS Soc. 2015;18:19984.
18. Grimsrud AT, Cornell M, Egger M, et al.. Impact of definitions of loss to follow-up
(LTFU) in antiretroviral therapy program evaluation: variation in the definition can have an appreciable impact on estimated proportions of LTFU. J Clin Epidemiol. 2013;66:1006–1013.
19. Cole SR, Hernan MA. Constructing inverse probability weights for marginal structural models. Am J Epidemiol. 2008;168:656–664.
20. Rosenbaum PR, Rubin DB. The central role of the propensity score in observational studies for causal effects. Biometrika. 1983;70:41–55.
21. Kredo T, Ford N, Adeniyi FB, et al.. Decentralising HIV treatment in lower- and middle-income countries. Cochrane Database Syst Rev. 2013;6:CD009987.
22. Shisana O, Rehle T, Simbayi L, et al.. South African National HIV Prevalence, Incidence and Behaviour Survey, 2012. Cape Town, South Africa: HSRC Press; 2014.
23. Cornell M, Schomaker M, Garone DB, et al.. Gender differences in survival among adult patients starting antiretroviral therapy in South Africa: a multicentre cohort study. PLoS Med. 2012;9:e1001304.
24. Cornell M, Grimsrud A, Fairall L, et al.. Temporal changes in programme outcomes among adult patients initiating antiretroviral therapy across South Africa, 2002–2007. AIDS. 2010;24:2263–2270.
25. Grimsrud A, Balkan S, Casas EC, et al.. Outcomes of antiretroviral therapy over a 10-year period of expansion: a multicohort analysis of African and Asian HIV programs. J Acquir Immune Defic Syndr. 2014;67:e55–e66.
26. Koole O, Tsui S, Wabwire-Mangen F, et al.. Retention and risk factors for attrition among adults in antiretroviral treatment programmes in Tanzania, Uganda and Zambia. Trop Med Int Health. 2014;19:1397–1410.
27. Bekker LG, Slack C, Lee S, et al.. Ethical issues in adolescent HIV research in resource-limited countries. J Acquir Immune Defic Syndr. 2014;65(suppl 1):S24–S28.
28. Katz IT, Dietrich J, Tshabalala G, et al.. Understanding treatment refusal among adults presenting for HIV-testing in Soweto, South Africa: a qualitative study. AIDS Behav. 2014;19:704–714.
29. Geng EH, Nash D, Kambugu A, et al.. Retention in care among HIV-infected patients in resource-limited settings: emerging insights and new directions. Curr HIV/AIDS Rep. 2010;7:234–244.
30. Johnson L, Estill J, Keiser O, et al.. Do increasing rates of loss to follow-up
in antiretroviral therapy programs imply deteriorating patient retention? Am J Epidemiol. 2014;180:1208–1212.
31. Grimsrud A, Patten G, Sharp J, et al.. Extending dispensing intervals for stable patients on ART. J Acquir Immune Defic Syndr. 2014;66:e58–60.
32. Alamo ST, Colebunders R, Ouma J, et al.. Return to normal life after AIDS as a reason for lost to follow-up in a community-based antiretroviral treatment program. J Acquir Immune Defic Syndr. 2012;60:e36–e45.
33. Ford N, Mills EJ. Simplified ART delivery
models are needed for the next phase of scale up. PLoS Med. 2011;8:e1001060.