Synapse
⌘+K
Synapse
PulseExploreJournal ClubResearchersJournals
Instagram
HomeJournal ClubExplore
July 10, 2026Kardiologia PolskaOpen Access

Predicting 3-year mortality after myocardial infarction in a structured care program: A comparative analysis of Cox regression and machine learning models

View Full Paper
Ask AI
Bookmark
Share

Why the study?

Accurate risk stratification after MI is essential, but whether machine learning offers an advantage over conventional statistical models in real-world clinical settings remains uncertain.

Do machine learning models improve the prediction of 3-year all-cause mortality compared to Cox regression in patients after myocardial infarction?

Population

965 patients enrolled in a structured post-MI care program between 2018 and 2022

Comparison

Cox proportional hazards regression vs three machine learning models

Design

Retrospective cohort study with temporal validation

Follow-up

3-year

Key result

Machine learning models did not outperform Cox regression in predicting 3-year all-cause mortality after myocardial infarction (AUC 0.712-0.738 vs 0.742; P=0.59).

Authors

CSCyntia Szymańska-ŁyczkowskaRDR DankowskiJFJulita Fedorowicz

Discussion

Loading...

Member takes

Overview

ML models offer no advantage over Cox regression for post-MI mortality prediction in moderate cohorts; leaves open utility with larger datasets or richer predictors.

Key Points

  • The study aims to compare Cox regression and machine learning models for predicting 3-year mortality after myocardial infarction.
  • Retrospective cohort study of 965 patients in a structured post-MI care program with 3-year follow-up.
  • Development cohort (n=740) and validation cohort (n=225) created to analyze model performance.
  • Cox regression and three ML models (Random Forest, XGBoost, Random Survival Forest) were developed using clinical variables.
  • 13% of patients died during the follow-up period (n=126).
  • Cox model showed an AUC of 0.742, while Random Survival Forest was 0.738, Random Forest 0.727, and XGBoost 0.712 (P=0.59).
  • Random Survival Forest achieved the lowest Brier score of 0.105, but no ML model outperformed Cox regression.

Study Design

Type

Cohort (n=965)

Structured PICO

Do machine learning models improve the prediction of 3-year all-cause mortality compared to Cox regression in patients after myocardial infarction?

P
Population
965 patients enrolled in a structured post-myocardial infarction care program between 2018 and 2022, followed for 3 years.
E
Exposure
Machine learning models (Random Forest, XGBoost, and Random Survival Forest) using baseline clinical variables
C
Comparator
Cox proportional hazards regression model
O
Outcome
3-year all-cause mortalityhard clinical

Main Result

p-value: p=0.59

Machine learning models did not outperform traditional Cox proportional hazards regression for predicting 3-year mortality in a moderate-sized post-MI cohort.

Limitations

  • Moderate-sized real-world cohort with limited predictors
  • Requires external validation in independent cohorts before clinical implementation

Cite This Study

Szymańska-Łyczkowska et al. (2026) conducted a cohort in myocardial infarction (n=965). Machine learning models vs. Cox proportional hazards regression was evaluated on 3-year all-cause mortality (p=0.59). Machine learning models did not outperform Cox regression in predicting 3-year all-cause mortality after myocardial infarction (AUC 0.712-0.738 vs 0.742; P=0.59).

synapsesocial.com/papers/6a508ea56eeac72a437a1731https://doi.org/10.33963/v.phj.113620
View Full Paper
Ask AI
Bookmark
Share