Descriptive statistics (e.g. means, standard deviations/variances, medians with inter-quartile ranges, and frequency distributions) will be used to describe participant demographic characteristics. The chi-squared statistic test will be used for categorical variables whereas T-test (parametric) or Wilcoxon rank sum test (non-parametric) will be applied for continuous variables. Transformations of outcome variables will be explored and performed if appropriate. Multivariate logistic regression models will be performed to examine associations between covariates. Since data is collected on one subject over time, the data will be treated as longitudinal data. Generalized linear mixed effect models will be fitted to determine associations of interest with time and intervention effects included as fixed effects and allowing for subject-specific random effects. Alternatively, generalized estimating equations can be used to account for any correlations of repeated data measures, when assessing determinants of PrEP adherence at time points provided we have balanced data with ignorable missing patterns. Time-to-event analysis will be conducted using Kaplan-Meier plots. To determine possible associations with covariates, a Cox-proportional hazard will be used. If the proportional hazard assumption does not hold, an accelerated failure time model will be used. Missing data due to loss to follow-up is inadvertent and inevitable. The extent of missing data will be assessed in the primary analysis. The primary analysis will be conducted under the assumption of data Missing at Random (MAR). As such, longitudinal likelihood-based data analysis, utilizing all the observed pre-deviation data from each participant will be employed. Thereafter, multiple imputation (MI) method will be employed under the primary MAR assumption, and Rubin’s rule will be followed to combine results. Sensitivity to deviation from the MAR assumption will be investigated. Intention to treat (ITT) assessment with complete case approaches where necessary will be implemented in cases of non-adherence to randomization assignment. Depending on the extent of non-adherence, the “as treated (AT)” approaches may be employed to analyse participants according to the intervention received regardless to their randomized allocation. A marker of level of adherence can also be included in the models. Patterns of PrEP use will be identified using latent class models or clustering techniques and the association of factors with these latent classes will be examined using multinomial logistic regression modelling. For all statistical investigations, tests for significance will be two-tailed. Analyses will be conducted using STATA 13.0 or R version 4.

Qualitative data analyses of participant and staff IDIs, CAB FGDs, CMEs, and staff SEPs will be conducted using a constant comparison approach guided by the study IMB framework. Data and results will be triangulated using matrices to refine findings [61]. Transcripts will be analysed by the qualitative research team, with results presented to study leadership for review and to guide additional analyses. Socio-behavioural factors and contextual (urban v. rural) influences influencing young women’s action plans for PrEP uptake and adherence will be used to create risk and adherence profiles to inform eventual intervention tailoring and implementation.

Note: The content above has been extracted from a research article, so it may not display correctly.



Q&A
Please log in to submit your questions online.
Your question will be posted on the Bio-101 website. We will send your questions to the authors of this protocol and Bio-protocol community members who are experienced with this method. you will be informed using the email address associated with your Bio-protocol account.



We use cookies on this site to enhance your user experience. By using our website, you are agreeing to allow the storage of cookies on your computer.