Improving 1-year mortality prediction in ACS patients using machine learning

Area of Science:

• Cardiology
• Clinical Risk Stratification
• Biomarkers

Background:

• The Global Registry of Acute Coronary Events (GRACE) score is a standard tool for assessing risk in acute coronary syndromes (ACS).
• A need exists for improved prediction models for 1-year all-cause mortality in ACS patients.

Purpose of the Study:

• To develop and internally validate a novel risk stratification model for predicting 1-year all-cause mortality in ACS patients.
• To compare the performance of the new model against the established GRACE 2.0 Score.

Main Methods:

• Analysis of 2,168 ACS patients from the Swiss SPUM-ACS Cohort (2009-2012).
• Evaluation of numerous linear models using combinations of 8 out of 56 variables.
• Determination of 1-year all-cause mortality in 95.8% of patients, with a mortality rate of 4.3%.

Main Results:

• The novel SPUM-ACS Score, incorporating age, plasma glucose, NT-proBNP, LVEF, Killip class, PAD, malignancy, and CPR, outperformed the GRACE 2.0 Score.
• The SPUM-ACS Score demonstrated superior predictive accuracy compared to the GRACE 2.0 Score's 5-fold cross-validated AUC of 0.81 (95% CI 0.78-0.84).
• Key predictors identified include age, trimethylamine N-oxide, creatinine, PAD/malignancy history, LVEF, and hemoglobin.

Conclusions:

• The SPUM-ACS Score effectively highlights the importance of age, heart failure markers, and comorbidities in predicting mortality after ACS.
• External validation and refinement in larger cohorts are necessary before clinical application of the SPUM-ACS Score.