Precision Forecasting for Mpox Outbreaks: A Rigorous Evaluation of Compartmental and Phenomenological Models Across Error Structures

The 2022 Mpox outbreak emphasized the need for effective forecasting tools to support public health responses. This study evaluates four epidemiological models, SEIR, Richards, Gompertz, and GLM, for short-term prediction of the 2022 Mpox outbreak in the United States, using CDC data from 12 May to 15 December 2022. We compare model perfor- mance across 1- to 4-week horizons with the Normal, Poisson, and Negative Binomial error structures, using metrics such as MAE, MSE, WIS, and 95% PI Coverage. The Richards model with a Normal error structure excelled, achieving the lowest MAE (227.81–283.88) and WIS (163.67–214.24), improving 29.2–36.3% in MAE over SEIR. The GLM showed competitive accuracy, while Gompertz led in uncertainty quantification (95% PI Coverage: 39.26–40.75%). The Normal error structure consistently outperformed others. These findings highlight the Richards model’s potential for early warning and inform targeted interventions, enhancing public health strategies for Mpox and similar diseases.