Merryheart/kncv-nigeria-tb-analysis

End-to-end TB data analysis project examining patient journey across KNCV Nigeria's program states — from screening to treatment outcome. Includes data generation, cleaning, EDA, and predictive modeling. This is a mock project using synthetic data. This project is not affiliated with or endorsed by KNCV Nigeria. # KNCV Nigeria TB Analysis End-to-end TB data analysis project examining the patient journey across KNCV Nigeria's program states — from screening to treatment outcome. Includes data generation, cleaning, exploratory data analysis, statistical testing, predictive modeling, and a Power BI dashboard. ## Project Overview **Analysis Question:** From screening to success — tracking the TB patient journey across KNCV Nigeria's program states, and understanding what predicts treatment outcome. **Data:** Synthetic dataset of 60,000 TB patient records (2020–2024) modeled on real Nigerian TB burden data and KNCV Nigeria's program structure. ## Repository Structure ``` kncv-nigeria-tb-analysis/ ├── data/ │ ├── raw/ │ │ └── kncv_nigeria_tb_data_raw.csv │ └── processed/ │ ├── kncv_nigeria_tb_data_cleaned.csv │ └── feature_importance.csv ├── notebooks/ │ ├── 01_kncv_tb_cleaning_analysis.ipynb │ ├── 02_kncv_tb_eda.ipynb │ ├── 03_kncv_tb_statistical_analysis.ipynb │ └── 04_kncv_tb_prediction.ipynb ├── scripts/ │ └── generate_kncv_tb_data.py └── images/ ├── page1.png └── page2.png ``` ## Key Findings - Lagos, Kano and Oyo account for the highest TB burden among KNCV program states - TB case notifications dropped in 2020 due to COVID-19 disruption and recovered by 2023 - DR-TB patients have significantly worse treatment outcomes than DS-TB patients - PLHIV have lower treatment success rates (47.11%) compared to HIV-negative patients - Relocation is the leading reason for loss to follow-up - No state met the WHO 90% treatment success rate target ## Prediction Model Logistic Regression model trained on 35,884 confirmed TB patients to predict treatment success. - **Accuracy:** 59.5% - **ROC-AUC:** 0.58 - **Top predictors:** Treatment regimen, PLHIV status, child under 15 ## Dashboard Built in Power BI with 2 pages: - **Page 1:** Patient Journey Overview — KPIs, case notifications trend, case finding methods, TB type distribution, geographic map - **P …