Improving Public Health Supply Chains

Time Series Techniques for Medication Demand Forecasting

Abstract

The National Pharmaceutical Assistance Policy (PNAF) in Brazil aims to ensure universal access to essential medications through primary care. To achieve this goal and reduce healthcare access inequalities, efficient health system supply chains are crucial. This study evaluates time series forecasting methods, specifically exponential smoothing (ETS) and autoregressive integrated moving average (ARIMA) models, to predict the demand for captopril, a widely used antihypertensive drug, in São Paulo’s Basic Health Units. Data on medicine consumption and demand from January 2018 to March 2023 were collected and analyzed to address current inefficiencies in demand prediction, compared through the Mean Absolute Percentage Error (MAPE). Results indicate that the ETS model achieved the best performance in captopril demand forecasting, with a MAPE of 2.26%, significantly improving on the 77.97% MAPE of the existing methodology. Holt-Winters seasonal models and ARIMA also demonstrated robust predictive capabilities, with MAPEs of 3.81% and 3.47%, respectively. This research highlights the potential of data-driven forecasting techniques, such as the ETS model, to optimize resource allocation, ensure medication availability, and improve service quality, providing a framework for future applications in similar contexts.