BACKGROUND

To predict mortality in patients with acute heart failure (AHF), we created and validated an internal clinical risk score, the KICKOFF score, which takes physical and social aspects, in addition to clinical aspects, into account. In this study, we validated the prediction model externally in a different geographic area.

METHODS AND RESULTS

There were 2 prospective multicenter cohorts (1,117 patients in Osaka Prefecture [KICKOFF registry]; 737 patients in Kochi Prefecture [Kochi YOSACOI study]) that had complete datasets for calculation of the KICKOFF score, which was developed by machine learning incorporating physical and social factors. The outcome measure was all-cause death over a 2-year period. Patients were separated into 3 groups: low risk (scores 0-6), moderate risk (scores 7-11), and high risk (scores 12-19). Kaplan-Meier curves clearly showed the score's propensity to predict all-cause death, which rose independently in higher-risk groups (P<0.001) in both cohorts. After 2 years, the cumulative incidence of all-cause death was similar in the KICKOFF registry and Kochi YOSACOI study for the low-risk (4.4% vs. 5.3%, respectively), moderate-risk (25.3% vs. 22.3%, respectively), and high-risk (68.1% vs. 58.5%, respectively) groups.

CONCLUSIONS

The unique prediction score may be used in different geographic areas in Japan. The score may help doctors estimate the risk of AHF mortality, and provide information for decisions regarding heart failure treatment.