sickle-in-africa/registry

Introduction This repository contains the analysis code for the SickleInAfrica consortium Phase I registry datasets from three African countries: Ghana, Nigeria, and Tanzania. The study focuses on understanding key health outcomes by employing hierarchical modeling to assess age at diagnosis, age at hydroxyurea initiation, hydroxyurea use, and other adherence measures. The hierarchical approach captures variations at hospital, and country levels, providing robust estimates that account for both within-country and between-country differences. The analysis includes scripts for data preprocessing, imputation, and model implementations using various R packages. The codebase ensures a reproducible workflow, supporting consistent analysis throughout the study, including harmonization of the datasets. The code is mainly in the R programming language (version 4.4.1). This analysis was performed in the following steps: Data Cleaning Initial data preparation involved correcting errors, consolidating of site individual datasets, and ensuring the accuracy of the datasets. Data Standardization Variables were standardized to the SickleInAfrica core data elements to ensure comparability across different datasets, facilitating more accurate analysis. Data Filtering Relevant subsets of the data were selected based on specific criteria, and the data was checked for consistency. Data Exploratory Analysis Preliminary analysis was conducted to understand data distributions. Data Imputation Missing data were addressed using multilevel multiple imputation, we used the R MICE package (v3.16.0) for this. Data Long Survivor Detection (Upper Extreme Outliers) Identified individuals with unusually high ages, considered as long survivors, were detected and examined as potential outliers. Univariate Modelling Individual variables were analyzed independently using Generalized Linear Models (GLM) and linear models (lm) in R to assess their impact and significance in the context of the …