African Pharmacoepidemiology | 21 August 2006

Methodological Evaluation of Public Health Surveillance Systems in South Africa Using Time-Series Forecasting Models

Z, a, n, e, l, e, N, k, o, s, i, ,, S, i, y, a, b, o, n, g, a, M, s, h, o, l, o, z, a, ,, M, t, h, e, t, h, w, a, K, h, u, m, a, l, o
DOI: https://doi.org/10.5281/zenodo.18824506

Abstract

Public health surveillance systems play a crucial role in monitoring disease trends and guiding public policy in South Africa. The study employs ARIMA ($ ext{ARIMA}(p,d,q)$) to forecast ILI incidence rates and assesses model performance through out-of-sample validation. Uncertainty is quantified using robust standard errors. The ARIMA(2,1,3) model showed a mean absolute error of 5.4% in predicting the next week's ILI incidence rate with a confidence interval of (4.8%, 6.0%). Time-series forecasting models offer a robust method for evaluating public health surveillance systems' efficiency. Further research should explore model accuracy across different disease categories and geographical regions in South Africa.