alex166-cmd/forecasting_ncds_qech_mw

This is a project that was done as part of requirements to earn a bachelors degree in Applied Statistics at The Catholic University of Malawi. The projects' main aim was to forecast the prevalence of non communicable diseases at Queen Elizabeth Central Hospital using data from 2010 to 2022. # Forecasting Non-Communicable Diseases (NCDs) at QECH, Malawi **Undergraduate thesis project — BSc Applied Statistics, The Catholic University of Malawi (2024)** A statistical forecasting study using time series modelling to project the future prevalence of non-communicable diseases (NCDs) at Queen Elizabeth Central Hospital (QECH), Malawi's largest referral hospital — with the goal of supporting evidence-based resource allocation and health system planning. --- ## 📌 Research Problem Non-communicable diseases (cardiovascular disease, diabetes, hypertension, cancer) are rising rapidly in sub-Saharan Africa, yet health facilities often lack data-driven forecasts to guide staffing, bed capacity, and drug procurement decisions. This project asked: **Can historical NCD admission data at QECH be used to accurately forecast future case loads, and what do those forecasts mean for hospital planning?** --- ## 📊 Data - **Source:** Queen Elizabeth Central Hospital clinical records - **Period:** 10 years of monthly NCD admission data - **Diseases modelled:** Hypertension, Diabetes Mellitus, Cardiovascular Disease - **Variables:** Monthly case counts, seasonal indicators, trend components --- ## 🔧 Methodology | Step | Method | |---|---| | Stationarity testing | Augmented Dickey-Fuller (ADF) test | | Decomposition | Classical time series decomposition (trend + seasonality + residual) | | Model selection | Auto-ARIMA (minimising AIC/BIC) | | Forecasting | ARIMA model — 24-month forecast horizon | | Model evaluation | RMSE, MAE, MAPE | --- ## 📈 Key Results - **Hypertension** showed the strongest upward trend, with forecast models projecting a **~18% increase** in monthly admissions over the next 24 months. - **Diabetes** admissions followed a seasonal pattern, with peaks observed in the first quarter of each year — consistent with post-harvest dietary changes. - **Cardiovascular disease** admissions showed the highest volatility, with ARIMA(1,1,1) selected as the bes …