OneMando42/kenya-hiv-cascade

HIV treatment cascade analysis across Kenyan counties using R # Kenya HIV Treatment Cascade Analysis **Author:** Michael Mando | Epidemiologist & Biostatistician **Tools:** R (tidyverse, ggplot2, survival, broom) **Data:** Simulated patient cohort based on Kenya HIV Estimates (NACC/NASCOP 2023) --- ## Overview This project analyses the **HIV treatment cascade** across six counties in Kenya — examining gaps between people living with HIV (PLHIV), ART coverage, and viral suppression. The analysis is modelled on the type of routine program data encountered in real-world health information systems such as KenyaEMR and DHIS2. The 95-95-95 UNAIDS targets require that: - 95% of PLHIV know their status - 95% of those diagnosed are on ART - 95% of those on ART achieve viral suppression This analysis investigates how a facility cohort performs against these benchmarks and what patient-level factors predict viral suppression. --- ## Research Questions 1. What proportion of patients in the cohort are on ART and virally suppressed? 2. How does ART coverage and viral suppression vary by county? 3. What patient-level factors (age, sex, CD4 at entry, year of enrolment) predict viral suppression? 4. Does late presentation (CD4 < 200) affect time to ART initiation? --- ## Methods | Method | Purpose | |---|---| | Descriptive statistics | Cascade summary overall and by county | | Logistic regression | Predictors of viral suppression (OR, 95% CI) | | Kaplan-Meier survival analysis | Time to ART initiation by CD4 status | | Data visualisation | ggplot2 charts for all outputs | --- ## Key Findings ### Cascade Performance - **86%** of patients in the cohort are on ART - **~78%** of ART patients received a viral load test - **~90%** of tested patients achieved viral suppression - Overall cascade gap is largest at **ART initiation**, not viral suppression ### County Variation - ART coverage ranges from ~82% to ~90% across counties - Homa Bay and Kisumu show the largest gaps — consistent with national program data - Most counties fall be …