Synapse
⌘+K
Synapse
PulseExploreClubsResearchersJournals
Instagram
HomeClubsExplore
November 30, 2022Frontiers in Cardiovascular MedicineOpen Access

Machine learning outperforms traditional models in predicting HFmrEF mortality and rehospitalization with a 0.83 C-index.

View Full Paper
Ask AI
Bookmark
Share

Why the study?

Heart failure with mildly reduced ejection fraction is recognized as a unique phenotype, but risk stratification models for mortality and HF re-hospitalization are lacking.

Do machine learning models improve risk prediction for mortality and re-hospitalization in patients with HFmrEF compared to traditional models?

Population

Patients with HFmrEF (45-49%) enrolled in the TOPCAT trial

Comparison

Eight ML-based risk prediction models

Design

Post-hoc prognostic model development and validation study

Follow-up

Maximum of 6 years

Key result

Machine learning models outperformed traditional models in predicting mortality and heart failure re-hospitalization in patients with mildly reduced ejection fraction, with LASSO Cox regression achieving a C-index of 0.83 for 6-year mortality.

Authors

HZHengli ZhaoPLPeixin LiGZGuoheng Zhong

Discussion

Loading...

Member takes

Overview

ML models may aid HFmrEF risk stratification; leaves open prospective validation before clinical adoption.

Study Design

Type

Observational (n=519)

Multicenter

Yes

Structured PICO

Do machine learning models improve risk prediction for mortality and re-hospitalization in patients with HFmrEF compared to traditional models?

P
Population
Patients with heart failure with mildly reduced ejection fraction (HFmrEF, LVEF 45-49%) enrolled in the TOPCAT trial
I
Intervention
Machine learning (ML)-derived risk prediction models (8 models constructed including 72 candidate variables)
C
Comparator
Traditional risk prediction models
O
Outcome
Risk of mortality and HF re-hospitalization at 1- and 6-year follow-upshard clinical

Main Result

Effect estimate: C-index 0.83 (95% CI 0.68-0.94)

Machine learning models, particularly LASSO Cox regression and random forest, outperform traditional models in predicting mortality and re-hospitalization in patients with HFmrEF, highlighting the prognostic importance of KCCQ scores.

Limitations

  • Missing values of biomarkers such as circulating natriuretic peptides and high-sensitivity troponin
  • Patients were not treated with SGLT-2 antagonists due to the time period of the TOPCAT study
  • Excluded adult patients with symptoms of HF and documented LVEF <45%
  • Study population was predominantly white males, which may limit generalizability to the broader population

Cite This Study

Zhao et al. (2022) conducted an observational in Heart failure with mildly reduced ejection fraction (HFmrEF) (n=519). Machine learning models (LASSO Cox regression and Random Forest) vs. Traditional statistical models was evaluated on 6-year all-cause mortality prediction (LASSO Cox regression) (C-index 0.83, 95% CI 0.68-0.94). Machine learning models outperformed traditional models in predicting mortality and heart failure re-hospitalization in patients with mildly reduced ejection fraction, with LASSO Cox regression achieving a C-index of 0.83 for 6-year mortality.

synapsesocial.com/papers/6a1581809b87f33fc69f9dbdhttps://doi.org/10.3389/fcvm.2022.1042139
View Full Paper
Ask AI
Bookmark
Share