Modeling In-Hospital Mortality Among Patients Undergoing Percutaneous Coronary Intervention with Acute Myocardial...

Area of Science:

• Cardiology
• Medical Informatics
• Machine Learning in Healthcare

Background:

• Acute myocardial infarction with cardiogenic shock (AMI-CS) presents significant morbidity and mortality.
• Treatment strategies for AMI-CS, including mechanical circulatory support devices like Impella and intra-aortic Balloon Pump (IABP), require nuanced understanding.
• Heterogeneity in patient response to these interventions necessitates personalized risk assessment.

Purpose of the Study:

• To develop a machine learning model to identify patient subgroups with varying mortality risks when treated with Impella versus IABP for AMI-CS.
• To pinpoint clinical features associated with the greatest change in mortality risk for specific patient profiles.
• To elucidate the heterogeneity of treatment effects in AMI-CS management.

Main Methods:

• Utilized a large cohort of 15,796 patient visits from 369 sites in the National Cardiovascular Data Registry.
• Developed a machine learning-based mortality model to analyze treatment effects of Impella versus IABP.
• Identified specific clinical subgroups and their associated mortality risk changes.

Main Results:

• The overall estimated excess mortality effect of Impella versus IABP was 10.4 ± 0.8%.
• Identified a subgroup of 282 patients where Impella use was associated with a decreased mortality risk compared to IABP.
• Patients in this subgroup were characterized by younger age, higher systolic blood pressure, higher rates of salvage percutaneous coronary intervention, higher initial creatinine, and lower hemoglobin.

Conclusions:

• Impella devices, despite an overall higher mortality risk compared to IABP in AMI-CS, demonstrate potential benefit in specific patient profiles.
• The study highlights significant heterogeneity in treatment effects, underscoring the need for individualized risk stratification in AMI-CS.
• Machine learning models can effectively identify these nuanced treatment effects, guiding clinical decision-making for mechanical circulatory support in AMI-CS.