Nathan R. Hoot

Background and Objectives: Decision-making capacity is essential for informed consent, yet its assessment in emergency departments (EDs) is often subjective and inconsistently documented. Older adults are particularly vulnerable to impaired capacity during acute illness. We aimed to develop a scalable, electronic health record (EHR)-based approach to support early identification of older ED patients at risk of decisional vulnerability using the Medical Information Mart for Intensive Care (MIMIC)-IV database. Methods: We conducted a retrospective cohort study of 51,195 ED patients aged ≥65 years. Clinicians manually reviewed 2000 patients using a conservative consensus protocol to establish a consensus-derived proxy for decisional vulnerability. Such an approach yielded a definitive reference subset (capacity vs. no capacity) and an “uncertain” category when consensus was not achieved. We developed a weak-supervision expectation–maximization (EM) label model that combined multiple noisy labeling functions derived from triage vital signs, acuity measures, arrival mode, and large language model (LLM)-classified chief complaints to estimate the probabilistic risk of impaired capacity. Model discrimination and calibration were assessed on an independent holdout subset of definitive reference labels using receiver operating characteristic area under the curve (ROC-AUC), precision–recall area under the curve (PR-AUC), calibration plots, and Brier score. To support clinically conservative use, operating thresholds were a priori constrained to limit automated flagging to ≤15% of patients, with the remaining ones deferred for clinician review. Results: On the definitive holdout set, the weak-supervision model achieved an ROC-AUC of approximately 0.855 and a PR-AUC of approximately 0.837. Calibration assessment demonstrated residual miscalibration in the generative posterior, which improved after a lightweight discriminative refinement step (logistic regression trained on EM-derived probabilistic labels), reducing the Brier score to approximately 0.224 on holdout evaluation. Under the prespecified operational constraint (≤15% auto-flagged), the model functioned as a conservative, selective alerting strategy, achieving high specificity and positive predictive value while identifying only a minority of patients with decisional vulnerability. Conclusions: This study demonstrates a methodological proof of concept for using weak supervision to model a retrospectively defined proxy for decisional vulnerability from routinely collected ED EHR data. The framework is intended to support conservative, triage-oriented prioritization. Further prospective validation, external testing, and workflow governance are needed before clinical implementation.

Objective: Emergency department (ED) sign-outs provide continuity of care for patients at shift changes. However, these periods have been associated with communication breakdowns and medical errors. We explored information factors (stage-of-the-case and the explicitness of the sign-out narrative) to determine their impact on clinicians’ diagnostic accuracy and confidence. Methods: Clinicians from 2 academic ED sites reviewed patient vignettes sign-out narratives (N=69). Using a within-subjects experiment, the explicitness of the sign-out information (i.e., broad statement about labs versus stating each test) and the patient’s stage in the diagnostic process (e.g., labs pending, treatment occurring, etc.) at the time of sign-out varied. Results: Explicitness (i.e., expansion in description) did not significantly impact diagnostic accuracy (χ²=0.364, df=1, P=0.55) or confidence in the diagnosis (t=0.514, df=273, P=0.61). However, diagnostic accuracy (33%) and confidence in the diagnosis (48.9%) of middle-stage stimuli cases were lower compared with late-stage cases for diagnostic accuracy (60%) (χ²=19.945, df=1, Pt=5.404, df=273, P2=0.491, df=1, P=0.48), and clinician role (χ²=0.024, df=2, P=0.99). Conclusions: The stage of the patient’s case affects both clinicians’ diagnostic accuracy and confidence in their diagnosis, whereas the explicitness of the information described does not. Patients in the earlier stages of care are most impacted. At sign-out, acknowledging the stage of care the patient is in may provide additional support to clinical decision-making.

Left without being seen (LWBS) is a quality care metric associated with patient-centered outcomes. The risks affecting LWBS are complex and interventions targeting certain risks have diverse effects. We aimed to use different artificial intelligence and machine learning (AI/ML) algorithms to identify the risks affecting LWBS, implement triple interventions specifically targeted at such risks, and compare daily LWBS rate changes before and after the intervention. This is a retrospective observational study. Single urban Emergency Department (ED) daily throughput data from March 1, 2019, to February 28, 2023, were used for AI/ML model prediction. Model performance including accuracy, recall, precision, F1 score, and area under the receiver operating characteristics (AUC) were reported. The top risks affecting the LWBS were identified using the important function of the AI/ML feature. Triple interventions were implemented. The average daily LWBS rate was compared before (March 1, 2019, to February 28, 2023) and after (June 1, 2023, to May 31, 2024). A total of 1919 daily throughput metrics were analyzed, including 1461 daily metrics before the intervention, 92 daily metrics during the wash period, and 366 daily metrics after the intervention. Using data before the intervention, the Extreme Gradient Boosting (XGBoost) and Random Forest AI/ML algorithms predicted LWBS with a similar favorable performance. The 3 common factors influencing the increased daily LWBS rate were triage-to-bed (wait time), boarding time, and door-to-triage in the ED. Rapid triage, direct bedding, and boarding reduction (triple intervention) were implemented on March 1, 2023. We found 4.82% of daily LWBS before the triple intervention compared to 1.93% of daily LWBS after the triple intervention (P < .001). AI/ML approaches can identify common factors that are highly related to LWBS with favorable performance. Triple interventions targeting these factors can reduce the daily LWBS rate by approximately 60%, indicating the efficiency of the ED operational management.

Introduction: Medical errors are frequently under-reported, yet their appropriate analysis, coupled with remediation, is essential for continuous quality improvement. The emergency department (ED) is recognized as a complex and chaotic environment prone to errors. In this paper, we describe the design and implementation of a web-based ED-specific incident reporting system using an iterative process.Methods: A web-based, password-protected tool was developed by members of a quality assurance committee for ED providers to report incidents that they believe could impact patient safety.Results: The utilization of this system in one residency program with two academic sites resulted in an increase from 81 reported incidents in 2009, the first year of use, to 561 reported incidents in 2012. This is an increase in rate of reported events from 0.07% of all ED visits to 0.44% of all ED visits. In 2012, faculty reported 60% of all incidents, while residents and midlevel providers reported 24% and 16% respectively. The most commonly reported incidents were delays in care and management concerns. Conclusion: Error reporting frequency can be dramatically improved by using a web-based, user-friendly, voluntary, and non-punitive reporting system.