Predicting Health Spending by Risk Level and Severity in a Colombian Health Promotion Entity

OBJECTIVES: This study has two main purposes. Firstly, explain how the health risk stratification model of the population of a Health Promotion Entity in Colombia is developed according to multimorbidity and comprehensive spending. Secondly, show how the predictive model is applied to health risk management.

METHODS:: The database encompasses information for 6,280,262 users distributed across the four quarters of 2022 (22,571,675 records). It incorporated consumption metrics, disease-related indicators and demographic factors. Two-part models were employed to predict healthcare expenditure. The coefficients derived were utilized to compute the relative weights of each subgroup, facilitating the establishment of a case-mix

RESULTS: Models incorporating variables related to multimorbidity provided a more comprehensive explanation of healthcare expenditure. In the initial phase of the two-part models, a logit model was employed, while positive costs were modelled using log-linear OLS regression. This approach yielded an adjusted R2 of 54%. Leveraging the derived weights, we devised a case mix system tailored for capitation purposes.

CONCLUSIONS: Risk and severity groups are well-suited for risk adjustment in the Colombian healthcare system, offering advantages over alternative systems. It effectively manages health risks among the insured population and is valid for capitated adjustments using the case-mix system developed in this study. However, for broader implementation in other contexts, validation with clinical and economic data from diverse populations and healthcare systems is essential.