Mortality risk stratification using artificial intelligence-augmented electrocardiogram in cardiac intensive care...

Area of Science:

• Cardiology
• Artificial Intelligence
• Medical Diagnostics

Background:

• Left ventricular systolic dysfunction (LVSD) is a critical condition impacting cardiac intensive care unit (CICU) patient outcomes.
• Transthoracic echocardiography (TTE) is the standard for assessing LVSD, but advanced algorithms may offer complementary insights.

Purpose of the Study:

• To evaluate if an artificial intelligence-augmented electrocardiogram (AI-ECG) algorithm can stratify mortality risk in CICU patients.
• To determine if AI-ECG risk stratification is independent of LVSD detection by TTE.

Main Methods:

• Analysis of 11,266 CICU patients from Mayo Clinic (2007-2018) with AI-ECG.
• Extraction of left ventricular ejection fraction (LVEF) data from TTE for patients with available data.
• Multivariable logistic regression used to analyze hospital mortality, adjusting for LVEF.

Main Results:

• Higher AI-ECG probability of LVSD correlated with increased hospital mortality (aOR 1.05 per 0.1 higher, P=0.003), independent of LVEF.
• Lower LVEF (by TTE) was inversely related to hospital mortality (aOR 0.96 per 5% higher, P=0.02).
• A stepwise increase in hospital mortality was observed across true negative, false positive, false negative, and true positive AI-ECG predictions compared to TTE.

Conclusions:

• AI-ECG prediction of LVSD is significantly associated with hospital mortality in CICU patients.
• The AI-ECG algorithm provides valuable risk stratification beyond echocardiographic LVEF.
• Electrocardiographic patterns recognized by AI-ECG hold significant prognostic value for underlying myocardial disease.