Background: Emergency Department (ED) crowding is often attributed to a slow hospitalization process, leading to reduced quality of care. Predicting early disposition with cardiac-presenting patients is challenging: most are ultimately discharged, yet those with a cardiac etiology frequently require hospital admission. Existing scores rely on single-time-point data and often underperform when patient risk evolves during the visit.
Objective: To develop and validate a real-time deep-learning model that fuses serial 12-lead electrocardiogram (ECG) waveforms with sequential vitals and routinely available clinical data to predict hospital admission early in ED encounters.
Methods: We conducted a retrospective cohort study using the MIMIC-IV, MIMIC-IV-ED, and MIMIC-IV-ECG databases. Adults presenting with chest pain, dyspnea, syncope, or presyncope and at least one ECG within their ED stay were included. Two evaluation cohorts were defined: all stays with ≥1 ECG (N=30,421) and a subset with ≥2 ECGs during the encounter (N=11,273). To predict hospital admission, we first established two baseline models: a tabular model (random forest) trained on structured clinical variables including demographics, triage acuity, past medical history, medications, and laboratory results, and an ECG-only model that learned directly from raw 12-lead waveforms. We then developed a multimodal deep-learning model that combined ECGs with sequential vital signs as well as the same static tabular features. All models were restricted to data available during the stay up to the time of the last ECG. Performance was assessed with stratified 5-fold cross-validation using identical splits across models.
Results: The multimodal model achieved an Area Under Receiver Operating Characteristic (AUROC) of 0.911 when trained on all eligible stays. The model predicted disposition after the final ECG was taken, which was a median of 0.3 hours after triage and 4.6 hours before ED departure. Baseline models performed worse: the ECG-only model had an AUROC of 0.852, and the tabular random forest had an AUROC of 0.886. In the subset requiring at least two ECGs within the stay, ECG-only reached an AUROC of 0.859, and random forest, with the longer interval to chart tabular data, reached a higher AUROC of 0.911. The multimodal model had AUROC 0.924, and outperformed baselines in each cohort (paired DeLong P<.001).
Conclusions: Serial ECGs, when integrated with evolving vitals and routine clinical features, enable accurate, early prediction of ED disposition in cardiac-presenting patients. This open-source, reproducible framework highlights the potential of multimodal deep learning to streamline ED flow, prioritize higher-risk cases, and detect evolving, time-critical pathology.
Clinicaltrial:
Background: A 12-week digital health program for nonalcoholic fatty liver disease (NAFLD) previously showed feasibility in engagement, program retention, and clinical outcomes. This study investigates whether improvements in cardiometabolic risk factors achieved during a 12-week active program were sustained over a subsequent 6-month follow-up period.
Objective: The primary objective of this analysis was to evaluate whether the clinical improvements achieved after a 12-week program were maintained over the subsequent 6-month period, which did not include coaching or new intervention materials. In addition, the study aimed to assess participants' retention and engagement with the maintenance program.
Methods: In a 9-month, single-arm study using the Sidekick app (Sidekick Health), individuals with NAFLD and BMI >30 or metabolic syndrome or type 2 diabetes were included. The initial 12 weeks focused on providing education about diet, physical activity, stress management, and sleep, followed by 6 months without coaching or new intervention materials. The measured outcomes encompassed demographics, body composition, liver fat assessed using magnetic resonance imaging-proton density fat fraction (MRI-PDFF), and blood markers.
Results: Of the 34 participants who completed the first 12 weeks, 28 (82%) completed the 9-month study measurements. The median age was 63.0 years (IQR 53.5-71.0) and 57.1% (16/28) were women. At 9 months, compared to baseline, the mean weight loss was 4.0 kg (SD 5.0; P<.001). Liver fat decreased by 2.5% (SD 4.5; P<.001), with an 18.4% relative reduction. Systolic blood pressure decreased by 8.3 mm Hg (SD 13.4, P<.001) and diastolic by 2.5 mm Hg (SD 6.0; P=.02). Waist circumference decreased by 4.7 cm (SD 7.1; P<.001) and median glycated hemoglobin A1c (HbA1c) decreased by 19.5 mmol/mol (P<.001).
Conclusions: Sustained improvements in liver fat and metabolic markers suggest that Sidekick Health's digital program is a promising strategy for managing NAFLD without requiring continuous coaching.
Background: Cardiac rehabilitation (CR) is essential for recovery from cardiovascular disease. However, patients often encounter challenges in navigating the transition from acute hospital care to CR. Mobile health (mHealth) technologies may support this critical phase; however, evidence regarding their clinical practice remains limited. The HERO app (developed by REDOX GmbH) was developed to address the needs of patients with cardiovascular disease for orientation, emotional support, and motivation during this transition.
Objective: This study aims (1) to explore how mHealth technologies tailored for patients with cardiovascular disease can support their needs regarding orientation, emotional balance, and motivation during the transition from the acute hospital to CR and (2) to evaluate the user experience and acceptance of the HERO app as targeted pathway support.
Methods: A mixed methods study was conducted with patients with cardiovascular disease using study diaries, questionnaires, and semistructured interviews. Participants were purposively recruited in acute hospitals and rehabilitation settings. Quantitative data were analyzed descriptively, and qualitative data were analyzed using content analysis after Mayring.
Results: Eight participants used the app for an average of 14 (range 4-23) days. The app was perceived as a helpful short-term resource. It supported patients in understanding their condition, planning for CR, and regaining motivation. Participants highlighted the value of combining objective information with peer experiences. Suggestions for improvement included more personalized self-management guidance and a precise onboarding process to increase accessibility and usability.
Conclusions: Based on the findings, we propose 4 pillars of mHealth support for cardiac care transitions, including timely access, actionable guidance, peer support, and short-term usability. These pillars could inform the design of patient-centered mHealth tools for care transitions.
Background: Patient education and self-management support are critical for atrial fibrillation (AF) management. Conversational artificial intelligence (AI) has the potential to provide interactive and personalized support, but has not been evaluated in patients with AF.
Objective: This study aimed to evaluate the feasibility of a conversational AI intervention to support patients with AF postdischarge.
Methods: This was a single-blinded, 4:1-parallel-randomized controlled trial with process evaluation of feasibility and engagement. The primary outcome was the change in Atrial Fibrillation Effect on Quality-of-Life (AFEQT) questionnaire total score between groups. Patients with AF (18 y and older) were recruited postdischarge from Westmead Hospital cardiology services and randomized to receive either the intervention or usual care. The 6-month intervention consisted of fully automated conversational AI phone calls (with speech recognition and natural language processing) that regularly assessed patient health and symptoms and provided self-management support and education. These phone calls were supplemented with an online survey (sent via text message or email) containing replicated call content when participants could not be reached after 3 call attempts. If participant responses were concerning (eg, poor overall health, low medication confidence, and high symptom burden), they would be followed up with an ad hoc phone call and directed to clinical care if required. A semipersonalized education website was also available as part of the intervention, and participants were encouraged weekly (nudges delivered via text messages or emails) to visit it.
Results: A total of 103 patients (mean age, 63.7 y, SD 11.2 y; n=72, 70% male) were randomized (82 to the intervention); the target sample size was 385. The difference in the AFEQT total score was nonsignificant (adjusted mean difference 2.08, 95% CI -7.79 to 11.96; P=.46). An exploratory prepost comparison revealed an improvement in total AFEQT score in the intervention group only (baseline: 69.9, 95% CI 64.4 to 75.5; 6 months: 79.9, 95% CI 74.9 to 84.8; P=.01). Participants completed 4 of 7 outreaches on average, and 88.4% (304/344) of completed outreaches were reported as useful.
Conclusions: This proof-of-concept study demonstrates the feasibility of conversational AI in supporting patients with chronic conditions postdischarge. Intervention participants had improvement in their atrial fibrillation quality of life, though the forced shortening of the evaluation was unable to demonstrate a significant difference between groups.
Background: Atrial fibrillation (AF) is a prevalent chronic condition with increasing incidence worldwide. AF increases the risks of stroke, heart failure, and myocardial infarction and imposes a substantial burden on the health care system. Cardiac rehabilitation programs, while effective, often have low patient adherence. Recent evidence suggests that cardiac telerehabilitation, where patients are given home monitoring devices, could enhance adherence and outcomes. The program "Future Patient-Telerehabilitation of Patients with AF" (FP-AF) was created to assess the effects and potential benefits of cardiac telerehabilitation on patients with AF.
Objective: The objective of this study is to explore the experiences of patients participating in the FP-AF program.
Methods: This qualitative sub-study is part of the multicenter, randomized controlled FP-AF trial, which included 208 patients. Semi-structured interviews were conducted on 14 patients, randomly selected from participants in the intervention arm of the FP-AF program. The patient interviews, guided by self-determination theory, focused on patients' experiences with the FP-AF program, including the use of telerehabilitation technologies and a web-based portal called the "HeartPortal." Interview responses were analyzed using NVivo software (version 14.0; QSR International), with thematic coding based on interview guides and methodological guidance elaborated by Brinkmann & Kvale. The study adhered to ethical guidelines, with informed consent obtained from all participants.
Results: Based on the interviews, the following themes were identified: the home monitoring devices are viewed positively by the patients; the HeartPortal is a useful digital toolbox; patients develop new coping strategies for living with AF; the measured values are useful for the patients; the community of practice is beneficial; and the FP-AF program creates a sense of security.
Conclusions: Participation in the FP-AF program enhanced patients' sense of security, empowerment, and knowledge about AF. This improvement was due largely to a combination of patients' use of the HeartPortal and the educational sessions at health care centers. Telerehabilitation for patients with AF may be a useful way of researching this group of patients with a focus on rehabilitation and may be an effective means of offering rehabilitation to this group in the future.
Background: Remote patient management (RPM) using smartphone-enabled health monitoring devices (SHMDs) can be an effective, value-added part of cardiovascular care. However, cardiac patients' adherence to RPM is variable. Personas are fictional representations of users with common behaviors, needs, and motivation and can thereby help guide tailoring of interventions to be meaningful and possibly more effective. Personas can be used to understand the needs of the patient group and guide tailoring toward more personalized and effective eHealth intervention.
Objective: The aim of this study was to develop data-driven personas for patients with myocardial infarction (MI) based on both quantitative and qualitative results.
Methods: This study used a mixed methods design involving (1) database analysis of patients with MI (N=261) SHMD usage data (blood pressure [BP], weight, step count) over the course of a one-year care track and (2) semistructured interviews with patients with MI (N=16) currently using SHMDs. Overall, 12-month adherence rates were calculated based on the number of weeks patients performed the prescribed home measurements with the SHMDs.
Results: A cluster analysis was conducted on the self-monitoring data resulting in four distinctive usage patterns labeled as stiff starting (low adherent in first 6 weeks: 13%, 34/261 of users), temporary persisting (decreasing adherence: 24%, 62/261), loyally persisting (continuously adherent: 26%, 68/261), and negligent quitting (nonadherent: 37%, 97/261). Health outcomes (BP, step count, and weight) were analyzed based on these patterns. More adherent usage patterns show better controlled BP when compared to less adherent usage patterns, suggesting that adherence is associated with health outcomes. Patient experiences regarding adherence or nonadherence to the RPM relating to the four distinctive usage patterns were uncovered by means of semistructured interviews, providing insight into adherence factors most relevant for each of the clusters. Thus, 4 distinct personas were developed by data collection (database analysis and semistructured interviews), persona segmentation, and persona creation, named Tamara, Sam, Peter, and Kim.
Conclusions: This study identified 4 personas regarding adherence experiences and usage patterns of patients within an RPM care track. Adherent usage patterns were characterized by improved BP and step count. These personas can guide future tailoring of eHealth interventions to maximize patient adherence.
京公网安备 11010802042870号
京ICP备2023020795号-1
扫码关注我们
求助内容：
应助结果提醒方式：