Deep learning model using 12-lead ECG predicted 1-, 3-, and 12-month mortality in ICU patients with QTc≥440 ms, achieving AUROCs of 0.852, 0.834, and 0.815.
Does a deep learning model based on 12-lead ECG improve prediction of short-term mortality compared to SAPS-II in ICU patients with QTc prolongation?
A deep learning model using 12-lead ECGs effectively predicts short-term mortality in ICU patients with QTc prolongation, outperforming the standard SAPS-II score.
Absolute Event Rate: 0% vs 0%
Abstract Background The prolongation of the QT interval is strongly linked to increased short-term mortality risk, especially in settings like the intensive care unit (ICU), where prolonged QT intervals are common. Purpose This study aims to develop a deep learning model using the 12-lead electrocardiogram (ECG) to predict short-term mortality risk in ICU patients with QTc interval prolongation. Methods Data were obtained from the MIMIC-IV database. Restricted cubic splines (RCS) examined the relationship between QTc interval and mortality risk in patients' most recent ECG. For ECGs with QTc ≥440 ms, data were selected where the interval from ECG to death was within 1, 3, and 12 months as positive samples. A 1:1 random sampling ensured data balance among patients without recorded death. Twenty percent of the dataset was designated as an independent test set, with five-fold cross-validation during training. Model performance was assessed using AUROC, AUPRC, sensitivity, specificity, accuracy, NPV, precision, and F1 score. The model's performance was compared to SAPS-II using NRI and IDI indices. Results RCS analysis showed a U-shaped relationship between QTc interval and mortality risk, with HR exceeding 1 when QTc surpassed 446.3 ms. AUROCs for predicting mortality at 1, 3, and 12 months were 0.852 (95% CI: 0.834–0.870), 0.834 (95% CI: 0.816–0.850), and 0.815 (95% CI: 0.799–0.830), respectively. Compared to SAPS-II, NRI values were 0.184, 0.187, and 0.211 (P 0.001), and IDI values were 0.107, 0.088, and 0.115 (P 0.001), respectively. Conclusion QTc interval≥440 ms is significantly associated with increased short-term mortality risk in ICU patients. The deep learning model based on 12-lead ECG effectively identifies potential short-term mortality risk in patients with QTc interval prolongation.
Wang et al. (Sat,) reported a other. Deep learning model using 12-lead ECG predicted 1-, 3-, and 12-month mortality in ICU patients with QTc≥440 ms, achieving AUROCs of 0.852, 0.834, and 0.815.