Prognostic Implications of Machine Learning Algorithm-Supported Diagnostic Classification of Myocardial Injury Using...

Area of Science:

• Cardiology
• Biomarkers
• Machine Learning in Healthcare

Background:

• High-sensitivity troponin assays increase identification of myocardial injury.
• Type 1 myocardial infarction (T1MI) is less common but has established evidence.
• Understanding temporal cardiovascular event risks across myocardial injury types is crucial.

Purpose of the Study:

• To assess the time course of cardiovascular events in different myocardial injury classifications.
• To compare risks of mortality, recurrent myocardial infarction, heart failure, and arrhythmia over three years.

Main Methods:

• Utilized machine learning algorithms to classify myocardial injury (T1MI, acute/type 2 myocardial infarction [T2MI], chronic injury, no injury) from hospital encounters.
• Analyzed temporal hazard for major adverse cardiovascular events over three years.

Main Results:

• Among 176,787 episodes, T1MI was 6.9%, acute/T2MI was 6.0%, and chronic injury was 26.7%.
• All injury types showed early mortality risk; T1MI had the highest short-term recurrent myocardial infarction risk.
• Acute/T2MI and chronic injury demonstrated persistent long-term heart failure risk, unlike T1MI.

Conclusions:

• Non-T1MI myocardial injury classifications are associated with substantial and persistent late cardiac events.
• There is a need for evidence-based therapeutic strategies for non-T1MI.
• Opportunities exist to improve long-term outcomes for all myocardial injury types with existing and novel therapies.