Godswill-07/rivers-hiv-cascade-analysis

"HIV care cascade analysis for Rivers State, Nigeria — Excel, Python, Power BI # Rivers State HIV Care Cascade Analysis An end-to-end data analytics project tracking HIV testing, treatment, and viral suppression outcomes in Rivers State, Nigeria — built on real PEPFAR public health data (2016-2025). **View the interactive dashboard → --- ## The Story While validating this dataset, I found my "Total Tested" figure came out to **12 million people — in a state with a population of roughly 7 million.** That discrepancy led to uncovering a real data structure issue in PEPFAR's reporting: the same population is reported at multiple overlapping age-band granularities, and summing them together silently double- and triple-counts real people. This repo documents the full pipeline that found, diagnosed, and corrected that issue — along with several other data quality problems (Excel-corrupted values, precision loss, and structural inconsistencies) — before producing final analysis. ## Key Findings - **Viral suppression rate rose from 72% (2017) to 96% (2023)**, a sustained multi-year improvement, now plateaued just below the UNAIDS 95% target - **The steepest cascade drop-off occurs between testing positive and starting treatment** — a linkage gap worth targeted intervention - **Children (ages 1-14) consistently show lower suppression rates than adults** across the full study period — flagged with an explicit small-sample caveat, not overstated as fact ## Pipeline | Stage | Tool | What happened | |---|---|---| | 1. Data validation | Excel | Caught floating-point ID corruption, scientific-notation precision loss, embedded line breaks | | 2. Reshaping & cleaning | Python (pandas) | Wide→long reshape, Excel date-corruption recovery for age bands, numerator/denominator suppression rate calculation | | 3. Modeling & visualization | Power BI | Star schema (fact table + 3 dimension tables), DAX measures, 4-page interactive dashboard | ## Data Cleaning & Pipeline Architecture The raw PEPFAR MER dataset required extensive programmatic reshaping and …