Enrique Hernandez-Laredo, Centro Universitario UAEM Tianguistenco, Universidad Autónoma del Estado de México, Tianguistenco, Estado de México, México
Brenda Y. Reza-Curiel, Centro Universitario UAEM Tianguistenco, Universidad Autónoma del Estado de México, Tianguistenco, Estado de México, México
Ángel G. Estévez-Pedraza, Centro Universitario UAEM Tianguistenco, Universidad Autónoma del Estado de México, Tianguistenco; Facultad de Medicina, Universidad Autónoma del Estado de México, Toluca. Estado de México, México
Background: Diabetes mellitus (DM) is a global public health problem whose prevalence has quadrupled in the last 30 years. Objective: To analyze the time series of DM mortality in the period 1990–2022 in the most populated municipalities of the State of Mexico: Chimalhuacán, Ecatepec, Naucalpan, Nezahualcóyotl, Tlalnepantla, and Toluca. Materials and methods: Time series from the dynamic cube of deaths from DM mortality in the National Health Information System were used. The time series were characterized using autoregressive integrated moving average (ARIMA) models, seasonal (SARIMA), and with exogenous regressors (SARIMAX). Stationarity was evaluated using augmented Dickey-Fuller (ADF), Phillips-Perron (PP), and Kwiatkowski- Phillips-Schmidt-Shin (KPSS) tests; seasonality was evaluated using the seasonal-Dickey-Fuller test. Performance was evaluated using R², RMSE, and Ljung-Box, ARCH, and Jarque-Bera tests. Results: The models showed acceptable fit, with high R² values (> 0.80), without serial autocorrelation or heteroscedasticity in most municipalities; seasonal patterns and a structural break associated with the COVID-19 pandemic were identified. Conclusions: These models are useful for explaining the evolution of DM mortality, although their predictive capacity depends on the strengthening and availability of health information.
Keywords: Diabetes mellitus. Mortality. Time series analysis. ARIMA models. Mexico. Epidemiology.