AIMS

To create and validate a clinical prediction rule which is easy to manage, reproducible and that allows classifying patients admitted for heart failure according to their one-year mortality risk.

METHODS

A prospective cohort study carried out with 2565 consecutive patients admitted with heart failure in 13 hospitals in Spain. The derivation cohort was made up of 1283 patients and 1282 formed the validation cohort. In the derivation cohort, we carried out a multivariate logistic model to predict one-year mortality. The performance of the derived predictive risk score was externally validated in the validation cohort, and internally validated by K-fold cross-validation. The risk score was categorized into four risk levels.

RESULTS

The mean age was 77.2years, 49.7% were female and there were 611 (23.8%) deaths in the follow-up period. The variables included in the predictive model were: age≥75, systolic blood pressure<135, New York Heart Association class III-IV, heart valve disease, dementia, prior hospitalization, haemoglobin<13, sodium<136, urea≥86, length of stay≥14 and Physical dimension of Minnesota Living with Heart Failure questionnaire. The AUC for the risk score were 0.73 and 0.70 in the derivation and validation cohorts, respectively, and 0.73 in the K-fold cross-validation. The percentage of mortality ranged from 8.08% in the low-risk to 58.20% in the high-risk groups (p<0.0001; AUC, 0.72).

CONCLUSIONS

This model based on routinely available data, for admitted patients and with a follow-up at one year is a simple and easy-to-use tool for improving management of patients with heart failure.