Decades-ahead forecasting of disease trend leveling and decline: leveraging birth-cohort effects

Accurate long-term prediction of disease trends is crucial for public health planning and resource allocation. Traditional methods like age-standardized rate extrapolation and the Lee-Carter model often face limitations in predictive accuracy. The age-period-cohort model offers a promising alternative. We employed a Monte Carlo simulation to model disease rate changes from 2001 to 2040 under various scenarios influenced by age, period, and cohort effects. The predictive performance of the age-period-cohort model was compared with linear extrapolation, restricted cubic spline extrapolation of age-standardized rates, and the Lee-Carter model. Evaluation metrics included bias, variance, and mean square error. The age-period-cohort model showed superior predictive accuracy, closely aligning with true values, especially in scenarios dominated by cohort effects. In contrast, restricted cubic spline extrapolation, the Lee-Carter model, and linear extrapolation demonstrated progressively poorer performance. The age-period-cohort model effectively anticipates decades-ahead stabilization and decline of disease rates, outperforming traditional forecasting methods. It is recommended as a robust tool for guiding public health policy and resource distribution.