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

Adriaan A. Voors (Lead / Corresponding author), Wouter Ouwerkerk, Faiez Zannad, Dirk J. van Veldhuisen, Nilesh J. Samani, Piotr Ponikowski, Leong Ng, Marco Metra, Jozine M. ter Maaten, Chim Lang, Hans L. Hillege, Pim van der Harst, Gerasimos S. Filippatos, Kenneth Dickstein, John G. F. Cleland, Stefan D. Anker, Koos Zwinderman

Research output: Contribution to journalArticlepeer-review

180 Citations (Scopus)
176 Downloads (Pure)


Introduction: From a prospective multi-center multi-country clinical trial, we developed and validated risk models to predict prospective all-cause mortality and HF-hospitalizations in patients with heart failure (HF).

Methods: BIOSTAT-CHF is a research program designed to develop and externally validate risk-models to predict all-cause mortality and HF-hospitalizations. The index
cohort consisted of 2,516 patients with HF from 69 centres in 11 European countries. The external validation cohort consisted of 1,728 comparable patients from 6 centres in Scotland, UK

Results: Patients from the index cohort had a mean age of 69 years, 27% were female, 83% were in NYHA class II-III and the mean left ventricular ejection fraction was 31%. The full prediction models for mortality, HF-hospitalization and the combined outcome, yielded c-statistic values of 0.73, 0.69, and 0.71 respectively. Predictors of mortality and HF-hospitalization were remarkably different. The 5 strongest predictors of mortality were a greater age, higher BUN and NT-proBNP, lower hemoglobin and failure to prescribe a beta-blocker. The 5 strongest predictors of HF-hospitalization were greater age, previous HF-hospitalization, presence of edema, lower SBP and lower eGFR. Patients from the validation cohort were 74 years, 34% were women, 85% were in NYHA II-III and mean LVEF was 41%; c-statistic values for the full and compact model were comparable to the index cohort.

Conclusion: A small number of variables, which are usually readily available in the routine clinical setting, provide useful prognostic information for patients with heart
failure. Predictors of mortality were remarkably different from predictors of HF-hospitalization.
Original languageEnglish
Pages (from-to)627-634
Number of pages8
JournalEuropean Journal of Heart Failure
Issue number5
Early online date1 Mar 2017
Publication statusPublished - 9 May 2017


  • heart failure
  • prediction model
  • mortality
  • Heart failure hospitalization


Dive into the research topics of 'Development and validation of multivariable models to predict mortality and hospitalization in patients with heart failure'. Together they form a unique fingerprint.

Cite this