The extracted data from the ePMS was matched and merged with the patient data obtained from the PCBs using the unique ART numbers allocated (unique patient identifiers) at each ART clinic. A Microsoft Excel (Microsoft Corporation, Washington, DC, USA) spreadsheet was created from the merged data. The extracted anonymized data was saved into a password-protected excel file to prevent any unauthorized access or alterations of the data. The merged complete excel spreadsheet was imported into the SPSS statistical software (IBM SPSS version 28, IBM Corp. USA) for analysis. Descriptive statistics were carried out to describe the demographic and clinical characteristics of the adolescent participants included in the study, at baseline and during the following 36-month period. Bivariate analysis was executed utilizing the Chi-square test to determine the association, and the significance thereof, between retention in care and the demographic and clinical variables (age, sex, duration on ART, age at ART initiation, HIV disclosure status, WHO stage at ART initiation, current ART regimen, ART regimen at initiation, TB screening results, viral suppression status and healthcare facility level). Comparisons were performed between retention in care and demographic and clinical parameters at 6, 12, 18, 24 and 36 months post initiation on ART. Fisher’s exact test was used as an alternative to the Chi-square test in instances of sparse data (< 5 in any cell).
The Cox regression (Cox Proportional Hazards model) analysis was performed to adjust for potential confounders and interactions, to determine predictors for retention in care. The Cox proportional hazard model utilized the backward stepwise analysis, with the initial model inclusive of all candidate variables. The least significant variable was subsequently removed at each iteration until none of the nonsignificant variables remained. Variables were removed from the model at a set significance level of p < 0.05. The Complete Case Analysis (CCA) was used as less than 5% of the cases had missing data on all variables in the main analysis. Survival analysis was assessed with “Patient retention status at the end of the period” as the outcome of interest. A comparative survival analysis for the age and sex of the study participants using Kaplan-Meier survival curves was conducted [24 ]. Using Cox regression, factors influencing retention in care at months 6, 12, 18, 24 and 36 were established. Both the unadjusted and adjusted hazard ratios and their p-values were computed.
The Cox regression (Cox Proportional Hazards model) analysis was performed to adjust for potential confounders and interactions, to determine predictors for retention in care. The Cox proportional hazard model utilized the backward stepwise analysis, with the initial model inclusive of all candidate variables. The least significant variable was subsequently removed at each iteration until none of the nonsignificant variables remained. Variables were removed from the model at a set significance level of p < 0.05. The Complete Case Analysis (CCA) was used as less than 5% of the cases had missing data on all variables in the main analysis. Survival analysis was assessed with “Patient retention status at the end of the period” as the outcome of interest. A comparative survival analysis for the age and sex of the study participants using Kaplan-Meier survival curves was conducted [24 ]. Using Cox regression, factors influencing retention in care at months 6, 12, 18, 24 and 36 were established. Both the unadjusted and adjusted hazard ratios and their p-values were computed.
Free full text: Click here
