Artificial Intelligence ECG Diastolic Dysfunction and Survival in Cardiac Intensive Care Unit Patients

Figure 1. Flow diagram demonstrating inclusion and exclusion criteria for the final study population, including assignment of true negative, false negative, false positive, and true positive groups based on AIECG and TTE estimation of filling pressures. AIECG indicates artificial intelligence ECG; CICU, cardiac intensive care unit; LVDD, left ventricular diastolic dysfunction; and TTE, transthoracic echocardiogram.

Figure 2. CICU and in‐hospital mortality according to admission AIECG diastolic function grade, with unadjusted and adjusted odds ratio and 95% CI values from logistic regression, in the overall cohort (A) and sensitivity analysis cohort (B). AIECG indicates artificial intelligence ECG; aOR, adjusted odds ratio; CICU, cardiac intensive care unit; and uOR, unadjusted odds ratio.

Figure 3. In‐hospital mortality according to admission AIECG diastolic function grade, according to admission diagnosis in the overall cohort. AIECG indicates artificial intelligence ECG; and OR, odds ratio;

Figure 4. In‐hospital mortality according to admission AIECG diastolic function grade and medial mitral E/e' ratio by TTE (A) or severity of left ventricular systolic dysfunction based on LVEF by TTE according to ASE guidelines (B). AIECG indicates artificial intelligence ECG; ASE, American Society of Echocardiography; LVEF, left ventricular ejection fraction; LVSD, left ventricular systolic dysfunction; and TTE, transthoracic echocardiogram.

Figure 5. CICU and in‐hospital mortality according to elevated filling pressure group based on admission AIECG and TTE, in the overall cohort (A) and sensitivity analysis cohort (B). AIECG indicates artificial intelligence ECG; aOR, adjusted odds ratio; CICU, cardiac intensive care unit; DF, FN, false negative; FP, false positive; TN, true negative; TP, true positive; TTE, transthoracic echocardiogram; and uOR, unadjusted odds ratio.

Figure 6. Kaplan–Meier curves demonstrating overall 1‐year survival according to admission AIECG diastolic function grade (A) and elevated filling pressure group based on admission AIECG and TTE with medial mitral E/e' ratio as the “gold standard” (B) with unadjusted and adjusted hazard ratio and 95% CI values from Cox proportional‐hazards regression. AIECG indicates artificial intelligence ECG; CICU, cardiac intensive care unit; DF, diastolic function; FN, false negative; FP, false positive; HR, hazard ratio; TN, true negative; TP, true positive; and TTE, transthoracic echocardiogram.

Figure 7. Kaplan–Meier curves demonstrating 1‐year survival according to admission AIECG diastolic function grade in admission diagnosis subgroups of interest: ACS (top left), HF (top right), CS (bottom left), and CA (bottom right). ACS indicates acute coronary syndrome; AIECG, artificial intelligence ECG; CA, cardiac arrest; CICU, cardiac intensive care unit; CS, cardiogenic shock; DF, diastolic function; and TTE, transthoracic echocardiogram.

Figure 8. Kaplan–Meier curves demonstrating 1‐year survival in hospital survivors according to admission AIECG diastolic function grade (A) and elevated filling pressure group based on admission AIECG and TTE with medial mitral E/e’ ratio as the “gold standard” (B) with unadjusted and adjusted hazard ratio and 95% CI values from Cox proportional‐hazards regression. AIECG indicates artificial intelligence ECG; CICU, cardiac intensive care unit; DF, diastolic function; HR, hazard ratio; FN, false negative; FP, false positive; HF, heart failure; HR, hazard ratio; TN, true negative; TP, true positive; and TTE, transthoracic echocardiogram

Figure 9. Kaplan–Meier curves demonstrating one‐year survival according to admission AIECG diastolic function grade, both overall (A) and among hospital survivors (B) in the sensitivity analysis cohort, after excluding patients used to derive the AIECG algorithm. AIECG indicates artificial intelligence ECG; CICU, cardiac intensive care unit; DF, diastolic function.