Previsão da quantidade de consultas especializadas no HU-UFJF e em Juiz de Fora: uma abordagem através da análise de séries temporais

Abstract

This study aims to forecast the number of specialized outpatient consultations at the University Hospital of the Federal University of Juiz de Fora (HU-UFJF) and in the municipality of Juiz de Fora using time series analysis techniques. Considering the strategic role of university hospitals within Brazil’s Unified Health System (SUS) and the growing importance of demand forecasting for resource planning, three methodologies were applied: Exponential Smoothing (MAE), the Box-Jenkins univariate model (MBJ), and Artificial Neural Networks (ANN). The analyzed series showed irregular behavior influenced by epidemiological seasonality, academic cycles, and the effects of the COVID-19 pandemic, highlighting the need for robust predictive models. Model estimation, performance evaluation through RMSE, and in-sample error comparison were conducted. Results showed that Artificial Neural Networks outperformed traditional methods for both series, yielding the lowest prediction errors. Based on the best-performing model, forecasts were generated through December 2026, enabling the assessment of HU-UFJF’s participation in the municipal healthcare network and the anticipation of future demand fluctuations. The findings reinforce the value of time series modeling as a strategic tool for hospital management, supporting efficient resource allocation and improving healthcare delivery.