• A Gradient-Boosted Decision-Tree Algorithm for the Prediction of Short-Term Mortality in Acute Heart Failure Patients

    Background: Acute heart failure (AHF) is associated with significant morbidity and mortality. Effective patient risk stratification is essential to guiding hospitalization decisions and clinical management. Clinical decision support systems can be used to improve mortality predictions in emergency care settings.

    Methods : A gradient boosted decision tree machine learning algorithm (MLA) was developed on retrospective patient data. An algorithm was developed to predict 7-day mortality using age, sex, vital signs and laboratory values. Algorithm performance was compared to the Emergency Heart Failure Mortality Risk Grade (EHMRG), a validated risk score for prediction of 7-day mortality in the emergency department (ED). Model performance was assessed by area under the receiver operating characteristic curve (AUROC), sensitivity and specificity.

    Results : The MLA was trained and tested on 236,275 total ED encounters, 1,881 of whom were positive for AHF. The MLA and EHMRG demonstrated an AUROC of 0.84 and 0.78, respectively, for prediction of 7-day mortality. The MLA outperformed the EHMRG on all assessed metrics (Table 1).

    Conclusions: An MLA can successfully predict 7-day mortality in AHF patients with higher performance than a conventional mortality prediction model. In ED settings, this risk stratification tool may assist clinicians in making critical decisions about patient disposition by providing early and accurate insights into an individual patient’s risk profile.

    Author bio

    Cardiovascular Revascularization Medicine, Volume 28, Supplement, July 2021, Page S19


    Read the full article on Science Direct: https://www.clinicalkey.com/#!/content/playContent/1-s2.0-S1553838921003456?returnurl=null&referrer=null

This site uses cookies. By continuing to browse the site you are agreeing to our use of cookies. Review our Privacy Policy for more details