Individual factors associated with time to non-adherence to ART pick-up within HIV care and treatment services in three health facilities of Zambézia Province, Mozambique

Introduction Mozambique has made significant gains in addressing its HIV epidemic, yet adherence to visit schedules remains a challenge. HIV programmatic gains to date could be impaired if adherence and retention to ART remains low. We investigate individual factors associated with non-adherence to ART pick-up in Mozambique. Methods This was a retrospective cohort of patients initiating ART between January 2013 and June 2014. Non-adherence to ART pick-up was defined as a delay in pick-up 15 days. Descriptive statistics were used to calculate socio-demographic and clinical characteristics. Adherence to ART pick-up was assessed using Kaplan Meier estimates. Cox proportional hazards model was used to determine factors associated with non-adherence. Results 1,413 participants were included (77% female). Median age was 30.4 years. 19% of patients remained adherent to ART pick-up during the evaluation period, while 81% of patients were non-adherent to ART pick-up. Probability of being non-adherent to ART pick-up by 166 days following initiation was 50%. In univariate analysis, being widowed was associated with higher adherence to ART pick-up than other marital status groups (p = 0.01). After adjusting, being 35 years (aHR: 0.843, 95% CI: 0.738–0.964, p = 0.012); receiving efavirenz (aHR: 0.932, 95% CI: 0.875–0.992, p = 0.026); and being urban (aHR: 0.754, 95% CI: 0.661–0.861, p<0.0001) were associated with improved adherence. Non-participation in a Community ART Support Group (CASG) was associated with a 43% increased hazard of non-adherence to ART pick-up (aHR 1.431, 1.192–1.717, p<0.0001) Conclusions Interventions should focus on the first 6 months following ARV initiation for improvements. Younger persons and widows are two target groups for better understanding facilitators and barriers to visit schedule adherence. Future strategies should explore the benefits of joining CASGs earlier in one´s treatment course. Finally, greater efforts should be made to accelerate the scale-up of viral load capacity and HIV resistance monitoring.

This retrospective cohort study was conducted in three health facilities of Zambézia Province, Mozambique, representing three different urban-rural zones: 24th of July, an urban health facility in the provincial capital city of Quelimane; the District Hospital of Maganja da Costa, a rural district capital hospital; and Nante Health Facility, an even more rural peripheral health facility in the district of Maganja da Costa (Fig 2). We identified HIV infected individuals at these three sites, aged 15 years or older who started ART between January 1, 2013 and June 30, 2014. For purposes of service delivery, the Ministry of Health (MOH) considers adults to be ≥15 years of age. Each patient was followed from the point of initiation of ART treatment, until June 30, 2016 for a minimum 24-month follow-up (Fig 3). *Charlotte Buehler Cherry, November 08, 2017, WGS_1984 Vanderbilt Institute for Global Health. *Patients LTFU, transferred, or who had died were censored at time of becoming LTFU, transfer, or death. Data were extracted from patients' paper medical records, filled out by clinicians, counselors and pharmacists, and aggregated into an open source HIV electronic patient tracking system called Open Medical Record System (OpenMRS); including demographic data, clinical information, laboratory results, and pharmacy data. Indicators evaluated included age, sex, education level, profession, marital status, whether urban or rural, health facility, CD4+ T cell count, World Health Organization (WHO) clinical staging, body mass index (BMI), ART regimen, and whether involved or not in a Community Adherence Support Group (CASG). Data collection comprised two phases. First, authorization was requested to access the database of each of the three health facilities. After obtaining approvals, data was extracted based on eligibility criteria and the variables to be analyzed were exported into Microsoft Excel. Second, data was linked with pharmacy records (called FILA, Formulário de Levantamento dos ARVs), comparing unique patient identification numbers (NID) of the patient record in OpenMRS and the NID of each FILA. Data were further linked with the NID of patient registers from the tuberculosis services, maternal child health services, and the health facilities death registry. FILAs are a paper-based record that is filled out prospectively at each pharmacy pick-up visit. From each FILA record that the study team could find and then confirm as being a study patient, data were extracted on both scheduled and actual ART pick-up dates for the defined study period. Data systems and patient NIDs are facility specific without ability to track patients across facilities. Due to limitations in space of each health facility´s pharmacy for archiving records, the health facilities routinely discard FILAs after an unspecified period of time. As a result, it was not possible to manually confirm the ART pick-up dates of all eligible patients identified from the OpenMRS system as planned at study outset. As such, analysis of patient retention for this study was limited to only those patients from which we could link data from the electronic medical record, or other registries, and the FILAs that were located. Outcomes of interest were calculated by subtracting the actual ART pick-up date from the scheduled ART pick-up date and assumed, based on Mozambican protocols, that the quantity of ART provided was always the same and sufficient for a 30-day supply, regardless if you received your ART at the facility or through a CASG. One exception to this were patients who first initiated an ART regimen during the study period that consisted of AZT+3TC+NVP. These patients only received 15 pills for their first two-week induction. Following this, they received the full 30-day supply as described. While medication stock-outs do continue in Mozambique, it is beyond the capacity and scope of this work to analyze if each patient actually received a 30-day supply of ART or not. We then categorized patients based on the number of uninterrupted days without ART. Patients were categorized as “adherent to ART pick-up” if the number of uninterrupted days was 0–14 days and “non-adherent to ART pick-up” if ≥15 days, our primary outcome of interest. For purpose of analysis, patients who were classified as LTFU, transferred or who had died, were censored at the time of becoming LTFU, transfer or death. Statistical analysis was performed using IBM SPSS Statistics, version 20 (International Business Machines Corporation Statistical—Package for the Social Sciences). Baseline socio-demographic and clinical characteristics of patients who were adherent and not-adherent to ART pick-up were compared using the chi-square test. Adherence to ART pick-up was assessed using the Kaplan Meier estimate (survival analysis), with graphical visualization of the mean time to first non-adherence to ART pick-up of ≥15 days (survival curve). Patients were followed for a minimum of 24 months but depending on their initiation date it could be longer. Data were censored after 1,278 days (3.5 years) of follow-up. The variables with a significant association with non-adherence to ART pick-up (p< 0.2) in univariate analysis were individually stratified into survival curves (Kaplan Meier) by comparing the two curves using the Breslow test. Continuous variables with significant association with non-adherence to ART pick-up were categorized after analysis using a Receiver Operating Characteristic (ROC) curve to determine the cutoff point that maximizes the division between adherence and non-adherence. Finally, variables significantly associated with non-adherence to ART pick-up in univariable analysis (p<0.2) were analyzed using Cox proportional hazards models. The co-factor Marital Status lost significance in the multivariate analysis and was therefore not included in the final model. The evolution of dropout during the study period was illustrated by Kaplan Meier's estimates in which we used scheduled dates for ART pick-up and added 14 days to those dates, rather than actual dates of ART pick-up. As described above, patients who started an ART regimen during the study period consisting of AZT+3TC+NVP, received a two-week supply of ART for their initial induction period. As such, it is possible for non-adherence to ART pick-up outcomes to occur as early as 30 days post ART initiation. This study protocol was approved by the Institutional Bioethics Committee for Health of Zambézia Province (Comité Institucional de Bioética para Saúde- Zambézia, CIBS-Z), as well as the Institutional Review Boards (IRBs) of Federal University of São Paulo (Universidade Federal de São Paulo, UNIFESP) and Vanderbilt University. Additional approval was obtained from the Provincial Health Directorate of Zambézia Province (Direção Provincial de Saúde da Zambézia, DPS-Z). All data were fully de-identified prior to accessing them and all three research ethics review committees waived the need for informed consent.

N/A

• Dima Health