Centro de Documentacion de Fundación MAPFRE - Improving healthcare cost prediction for chronic disease through covariate clustering and subgroup analysis methods

Notas: Sumario: Predicting healthcare costs for chronic diseases is difficult because these costs depend not only on medical factors but also on patients' own perceptions and behaviors. To better capture this complexity, this paper introduces a new statistical framework that combines covariate clustering with finite mixture regression models. This approach groups highly related variables and identifies patient subgroups, improving both model interpretability and prediction accuracy in high-dimensional, noisy, and correlated data settings. The method uses a penalized clustering structure and a dedicated EMADMM algorithm to handle the challenging optimization problem. Through simulations and real-world data from diabetes patients, the framework shows strong stability and effectiveness: it improves predictions by sharing information across related variables and reveals meaningful behavioral patterns in patients' self-perception data.Registros relacionados: En: Astin bulletin. - Belgium : ASTIN and AFIR Sections of the International Actuarial Association = ISSN 0515-0361. - 12/05/2025 Volume 55 Issue 2 - may 2025 , p. 375 - 394Materia / lugar / evento: Matemática del seguro Seguro de salud Enfermedades crónicas Costes económicos Predicciones estadísticas Modelos estadísticos Modelos de simulación Otros autores: Huang, Yifan
Cao, Yang
International Actuarial Association
Outras classificações: 6

‌ Ver detalhe do número