Background: Cardiac arrest (CA) is one of the leading causes of death among patients in the intensive care unit (ICU). Although many CA prediction models with high sensitivity have been developed to anticipate CA, their practical application has been challenging due to a lack of generalization and validation. Additionally, the heterogeneity among patients in different ICU subtypes has not been adequately addressed. Objective: This study aims to propose a clinically interpretable ensemble approach for the timely and accurate prediction of CA within 24 hours, regardless of patient heterogeneity, including variations across different populations and ICU subtypes. Additionally, we conducted patient-independent evaluations to emphasize the model’s generalization performance and analyzed interpretable results that can be readily adopted by clinicians in real-time. Methods: Patients were retrospectively analyzed using data from the Medical Information Mart for Intensive Care-IV (MIMIC-IV) and the eICU-Collaborative Research Database (eICU-CRD). To address the problem of underperformance, we constructed our framework using feature sets based on vital signs, multiresolution statistical analysis, and the Gini index, with a 12-hour window to capture the unique characteristics of CA. We extracted 3 types of features from each database to compare the performance of CA prediction between high-risk patient groups from MIMIC-IV and patients without CA from eICU-CRD. After feature extraction, we developed a tabular network (TabNet) model using feature screening with cost-sensitive learning. To assess real-time CA prediction performance, we used 10-fold leave-one-patient-out cross-validation and a cross–data set method. We evaluated MIMIC-IV and eICU-CRD across different cohort populations and subtypes of ICU within each database. Finally, external validation using the eICU-CRD and MIMIC-IV databases was conducted to assess the model’s generalization ability. The decision mask of the proposed method was used to capture the interpretability of the model. Results: The proposed method outperformed conventional approaches across different cohort populations in both MIMIC-IV and eICU-CRD. Additionally, it achieved higher accuracy than baseline models for various ICU subtypes within both databases. The interpretable prediction results can enhance clinicians’ understanding of CA prediction by serving as a statistical comparison between non-CA and CA groups. Next, we tested the eICU-CRD and MIMIC-IV data sets using models trained on MIMIC-IV and eICU-CRD, respectively, to evaluate generalization ability. The results demonstrated superior performance compared with baseline models. Conclusions: Our novel framework for learning unique features provides stable predictive power across different ICU environments. Most of the interpretable global information reveals statistical differences between CA and non-CA groups, demonstrating its utility as an indicator for clinical decisions. Consequently, the proposed CA prediction system is a clinically validated algorithm that enables clinicians to intervene early based on CA prediction information and can be applied to clinical trials in digital health.

Background and Objectives: Few comparative studies have evaluated wearable single-lead electrocardiogram (ECG) devices and standard multi-lead ECG devices during exercise testing. This study aimed to validate the accuracy of a wearable single-lead ECG monitor for recording heart rate (HR) metrics during graded exercise tests (GXTs). Methods: A cohort of 50 patients at a tertiary hospital underwent GXT while simultaneously being equipped with wearable single- and conventional multi-lead ECGs. The concordance between these modalities was quantified using the intraclass correlation coefficient and Bland–Altman plot analysis. Results: The minimum and average HR readings between the devices were generally consistent. Parameters such as ventricular ectopic beats and supraventricular ectopic beats showed strong agreement. However, the agreement for the Total QRS and Maximum RR was not sufficient. HR measurements across different stages of the exercise test showed sufficient agreement. Although not statistically significant, the standard multi-lead ECG devices exhibited higher noise levels compared to the wearable single-lead ECG devices. Conclusions: Wearable single-lead ECG devices can reliably monitor HR and detect abnormal beats across a spectrum of exercise intensities, offering a viable alternative to traditional multi-lead systems.

Cancer remains a leading cause of mortality worldwide. Chemotherapy is effective but can cause febrile neutropenia, deplete neutrophils, and increase infection risk. Although timely fever detection is critical to prevent complications, periodic temperature monitoring has inherent limitations. Wearable devices (WDs) for real-time body temperature (RBT) monitoring are emerging as tools for early fever detection, though their clinical feasibility and usability remain unverified. This study assessed the clinical feasibility of a novel WD by comparing its temperature concordance and fever detection capability against a standard reference in patients with cancer. The RBT system’s clinical maturity, including its capacity for early fever detection and monitoring treatment-related changes, was evaluated. Furthermore, satisfaction among clinicians and patients was analyzed in a hospital setting. In this prospective study, patients with hematologic malignancies at high infection risk admitted to the Chonnam National University in Hwasun Hospital, South Korea, were enrolled. Data from 47 patients were analyzed to compare MT100D as WD (SEERS Technology Co., Ltd., South Korea) readings with axillary thermometers. Temperature concordance was evaluated using the intraclass correlation coefficient (ICC), and diagnostic performance was compared with the reference thermometer. The early fever detection performance of the MT100D was analyzed to evaluate the clinical effectiveness of the early response system. Usability was evaluated through a five-point Likert scale survey completed by clinical nurses. The MT100D demonstrated good agreement with the reference thermometer (mean difference: −0.19 ± 0.35 °C; ICC: 0.78). The RBT system achieved sensitivity, specificity, and area under the receiver operating characteristic curve values of 81.50%, 96.32%, and 96.60%, respectively. The early fever detection rate was 77.1%, with the RBT system detecting fever an average of 1.13 ± 1.28 h earlier than the reference. Usability and satisfaction assessments showed high patient satisfaction (mean 4.69 ± 0.5; range 4.57–4.80) and moderate clinician satisfaction (mean 3.63 ± 0.75; range 2.89–4.11). Among patients with cancer, including those at risk of febrile neutropenia, the MT100D demonstrated strong concordance with standard thermometry, enabling early fever detection. The RBT system shows promise for the early identification of neutropenic risk during chemotherapy.

BACKGROUND Cardiac arrest (CA) is one of the leading causes of death among patients in the intensive care unit (ICU). Although many CA prediction models with high sensitivity have been developed to anticipate CA, their practical application has been challenging due to a lack of generalization and validation. Additionally, the heterogeneity among patients in different ICU subtypes has not been adequately addressed. OBJECTIVE This study aims to propose a clinically interpretable ensemble approach for the timely and accurate prediction of CA within 24 hours, regardless of patient heterogeneity, including variations across different populations and ICU subtypes. Additionally, we conducted patient-independent evaluations to emphasize the model’s generalization performance and analyzed interpretable results that can be readily adopted by clinicians in real-time. METHODS Patients were retrospectively analyzed using data from the Medical Information Mart for Intensive Care-IV (MIMIC-IV) and the eICU-Collaborative Research Database (eICU-CRD). To address the problem of underperformance, we constructed our framework using feature sets based on vital signs, multiresolution statistical analysis, and the Gini index, with a 12-hour window to capture the unique characteristics of CA. We extracted 3 types of features from each database to compare the performance of CA prediction between high-risk patient groups from MIMIC-IV and patients without CA from eICU-CRD. After feature extraction, we developed a tabular network (TabNet) model using feature screening with cost-sensitive learning. To assess real-time CA prediction performance, we used 10-fold leave-one-patient-out cross-validation and a cross–data set method. We evaluated MIMIC-IV and eICU-CRD across different cohort populations and subtypes of ICU within each database. Finally, external validation using the eICU-CRD and MIMIC-IV databases was conducted to assess the model’s generalization ability. The decision mask of the proposed method was used to capture the interpretability of the model. RESULTS The proposed method outperformed conventional approaches across different cohort populations in both MIMIC-IV and eICU-CRD. Additionally, it achieved higher accuracy than baseline models for various ICU subtypes within both databases. The interpretable prediction results can enhance clinicians’ understanding of CA prediction by serving as a statistical comparison between non-CA and CA groups. Next, we tested the eICU-CRD and MIMIC-IV data sets using models trained on MIMIC-IV and eICU-CRD, respectively, to evaluate generalization ability. The results demonstrated superior performance compared with baseline models. CONCLUSIONS Our novel framework for learning unique features provides stable predictive power across different ICU environments. Most of the interpretable global information reveals statistical differences between CA and non-CA groups, demonstrating its utility as an indicator for clinical decisions. Consequently, the proposed CA prediction system is a clinically validated algorithm that enables clinicians to intervene early based on CA prediction information and can be applied to clinical trials in digital health.