A new population-based risk stratification tool was developed and validated for predicting mortality, hospital admissions, and health care costs

Abstract

Objectives: The aim of this study was to develop a new population-based risk stratification tool (Chronic Related Score [CReSc]) for predicting 5-year mortality and other outcomes.

Study design and setting: The score included 31 conditions selected from a list of 65 candidates whose weights were assigned according to the Cox model coefficients. The model was built from a sample of 5.4 million National Health Service (NHS) beneficiaries from the Italian Lombardy Region and applied to the remaining 2.7 million NHS beneficiaries. Predictive performance was assessed by discrimination and calibration. CReSc ability in predicting secondary endpoints (i.e., hospital admissions and health care costs) was investigated. Finally, the relationship between CReSc and income was considered.

Results: Among individuals aged 50-85 years, CReSc performance showed (1) an area under the receiver operating characteristic curve of 0.730, (2) an improved reclassification from 44% to 52% with respect to other scores, and (3) a remarkable calibration. A trend toward increasing rates of all the considered endpoints as CReSc increases was observed. Compared with individuals on low-intermediate income, NHS beneficiaries on high income showed better CReSc profile.

Conclusion: We developed a risk stratification tool able to predict mortality, costs, and hospital admissions. The application of CReSc may generate clinically and operationally important effects.

Keywords: Comorbidity; Health care utilization database; Population-based study; Prognostic score; Record linkage; Risk stratification.