Using the ARIMA Model to forecast sexual and gender-based violence cases reported to a tertiary hospital in Lusaka, Zambia

Abstract

Introduction: sexual and gender-based violence (SGBV) is a global public health problem and women are disproportionally affected. Sexual and gender-based violence has an acute and chronic impact on health. We set out to apply autoregressive integrated moving average (ARIMA) models to make predictions on the number of SGBV cases reported to the Women and Newborn Hospital of the University Teaching Hospitals in Lusaka, Zambia. Methods: we retrospectively collected monthly number of SGBV cases from April 2017 to December 2019 as a primary outcome measure. We utilized the following strategy for the analysis; we developed the Box-Jerkins methodology through identification, estimation and diagnostic checks before using the model for forecasting. Structural ARIMA models were used to model the number of SGBV cases. We considered the simplest parsimonious lowest order model based on the lowest volatility, highest log-likelihood value, most significant coefficients, and the lowest Akaike information criteria and Bayesian information criteria values. Results: the ARIMA (1,1,1) model was able to forecast the number of SGBV cases. The forecasted mean monthly was 31.7 (95% CI: 29.9-33.3). The prediction error assessed by mean absolute percentage error was 10.3%, and the root mean standard error was 6.6. Further, a comparison of the observed and predicted values showed no statistically significant difference (p=0.753). Conclusion: the ARIMA model can be a useful tool for planners and clinicians to manage survivors of sexual and gender-based violence at the Women and Newborn Hospital.