Citations

Backround: Atrial fibrillation (AF) is a common complication following veno-venous extracorporeal membrane oxygenation (VV-ECMO), leading to further complications from anticoagulation or stroke. The Artificial Intelligence AF Prediction Dashboard developed at Mayo Clinic generates two predictive outputs: (1) a continuous probability score and (2) a binary threshold-based signal to predict the likelihood of AF at an individual level after 12-lead electrocardiogram (ECG) is obtained. Using both continuous risk (%) and binary (yes/no) likelihood, we present a novel method to understand AF risk after VV-ECMO cannulation based on pre-ECMO 12-lead ECG. Methods: We reviewed a total of 147 patients on ECMO between 2018 and 2024, from which 85 VV-ECMO patients with complete data were included. AI Dashboard outputs were quantified as binary (1/0) or percent likelihood for AF. R was utilized for statistical analysis. Results: 85 patients met our inclusion criteria from which AF after cannulation occurred in 31 (36%). 17 patients had a history of AF prior to VV-ECMO initiation. The mean age of our cohort was 49 years with a mean BMI of 29.3. Baseline AI Dashboard continuous assessment identified 39 patients (46%) compared to the binary assessment identifying 31 patients (36%). Logistic regression in the binary assessment demonstrated an odds ratio of 1.08 (95% CI 1.04 – 1.12) with a p value of <0.001. The binary AF Prediction yielded a 58% sensitivity, 76% specificity, 58% PPV, and 76% NPV, with an accuracy of 67%. The continuous AF Probability score achieved 72% balanced accuracy with sensitivity 74%, specificity 70%, PPV 59% and NPV 83%, indicating fair overall performance and suggesting good utility for ruling out AF in low-risk patients. Although both outputs provided moderate predictive accuracy, the continuous score reported before ECMO cannulation demonstrated stronger overall performance with fewer false negatives and higher rule-out value. Conclusions: We demonstrate the novel use of an AI predictive model in the VV-ECMO population for the prediction of AF risk with both continuous and binary risk cutoffs from baseline ECG. This data highlights the need for further development in ECMO-specific AI models to help mitigate risk after the onset of atrial arrhythmias. As AI-driven tools become increasingly integrated into care, their performance must be continuously validated and recalibrated for specific populations to ensure their safe and effective use.

Problem Statement: The prognostic value of troponin elevation in septic ICU patients without known history of cardiac disease remains unclear, particularly after accounting for overall illness severity. Background: Minor troponin rises are frequent during sepsis, yet clinicians still lack a clear cut-point that signals danger once cardiac disease is ruled out. We asked whether a modest increase in troponin-T predicts in-hospital death beyond global illness severity. Methods: We studied 1 221 adult first ICU stays labelled “sepsis” in the eICU Collaborative Research Database and removed any case with a cardiac admission diagnosis. Implausible or missing laboratory values were imputed by multiple imputation. Hospital mortality was modelled with logistic regression that included troponin (a continuous and a binary model with threshold > 0.04 ng/mL) together with age, sex, lactate, albumin, BUN, creatinine, intubation status, vasopressor use, dialysis requirement and APACHE-IV predicted mortality. A natural-spline model explored non-linear behavior. Findings were checked in a 1:1 propensity-matched cohort. Results: Overall mortality was 11.7%. Troponin exceeded 0.04 ng/mL in 38% of patients. After adjustment for APACHE-IV probability and other covariates, a troponin level > 0.04 ng/mL was associated with almost twice the odds of death (OR ≈ 1.9), whereas the continuous model suggested only a modest trend (OR ≈ 1.1 per ng/mL). The spline curve rose steeply between 0 and 0.10 ng/mL and flattened thereafter, pinpointing an inflection near 0.05 ng/mL. Matching on all measured confounders confirmed that high-troponin patients had about three-fold greater mortality (OR ≈ 2.8, 1.55-5.05; p<0.001), highlighting that even modest troponin elevation is an independent risk marker for death in septic ICU patients. A bedside heat-map combining troponin, albumin and need for intubation illustrated a simple gradient: risk climbs from 2 % in low-troponin, well-albuminised, non-intubated patients to nearly 7 % when troponin exceeds 0.10 ng/mL in intubated, hypo-albuminemic patients. Conclusions: Even modest troponin elevations (> 0.04 ng/mL) independently predict mortality in sepsis after rigorous adjustment and matching on APACHE-IV physiology, lactate, and organ-support variables. This biologic threshold supports incorporating troponin into bedside triage and future septic cardiomyopathy trials.