Temporal Trends of Malaria Incidence in Southwestern Nigeria: An ARIMA Analysis of Hospital-Based Surveillance Data

Abstract

Understanding temporal patterns in reported malaria cases is essential for surveillance, healthcare planning, and resource allocation. This study examined trends in malaria case counts and developed a forecasting model using hospital-based surveillance data from Ota, Ogun State, Nigeria. Monthly malaria records from January 2016 to December 2019 were obtained retrospectively from General Hospital, Ota. Descriptive statistics summarized temporal patterns, while Pearson correlation analysis examined the relationship between male and female case counts. Time-series modelling followed the Box–Jenkins autoregressive integrated moving average (ARIMA) approach. Candidate models were compared using the Bayesian Information Criterion, and residual autocorrelation was assessed using the Ljung–Box test. Reported malaria cases fluctuated throughout the study period but exhibited an overall declining trend. The estimated trend equation, Malaria Cases = 161,610.958 − 80.025 × Year, indicated a progressive reduction in recorded cases. Male and female case counts followed similar temporal patterns and were strongly positively correlated (r = .991, p < .01). Among the candidate models, ARIMA(1,1,1) provided the most suitable fit according to the model selection criterion. The Ljung–Box test detected no statistically significant residual autocorrelation, supporting the adequacy of the fitted model for the observed series. Forecasts suggested a continued decline in reported malaria cases over the forecast period. These findings document a reduction in hospital-recorded cases between 2016 and 2019 and indicate the potential value of ARIMA modelling for surveillance planning. The model may support short-term resource allocation, although predictive accuracy requires evaluation against subsequently observed data. Continued monitoring and assessment of environmental predictors could strengthen future forecasting. Interpretation should remain specific to hospital-reported cases because population incidence cannot be established from case counts alone.