Physicians' Academy for Cardiovascular Education

New accurate prediction models for heart failure hospitalization and mortality

Development and validation of multivariable models to predict mortality and hospitalization in patients with heart failure

Literature - Voors AA, Ouwerkerk W, Zannad F, et al. - Eur J Heart Failure 2017;19(5): 627-634

Main results

Final full prediction models:

Final compact prediction models (strongest variables):

Risk score

The risk score included the following cut-off points for optimal classification: NT-proBNP >4000 pg/mL, BUN >11 mmol/L, HDL <1.05 mmol/L, age >70 years, sodium <140 mmol/L, hemoglobin <12 g/dL, eGFR (CKD-EPI formula) <40 mL/min and SBP <140 mmHg.


In the validation cohort, the c-statistic for the full models were 0.73, 0.64 and 0.68 for mortality, HF hospitalization and their composite, respectively, and 0.72, 0.61 and 0.67 for the compact models.


In a large European study, new prediction models for all-cause mortality and HF hospitalization were developed and validated, which perform better compared with existing prediction scores and use information that is usually readily available in routine clinical setting. Variables in the all-cause mortality prediction models were different from those in the HF hospitalization models. Based on these findings, a simplified risk score for use in clinical practice was also developed (see link below).


Show references

Go to Risk score Find this article online at Eur J Heart Fail.

Share this page with your colleagues and friends: