Generative Artificial Intelligence for Chest Radiograph Interpretation in the Emergency Department.

Importance: Multimodal generative artificial intelligence (AI) methodologies have the potential to optimize emergency department care by producing draft radiology reports from input images. Objective: To evaluate the accuracy and quality of AI-generated chest radiograph interpretations in the emergency department setting. Design, Setting, and Participants: This was a retrospective diagnostic study of 500 randomly sampled emergency department encounters at a tertiary care institution including chest radiographs interpreted by both a teleradiology service and on-site attending radiologist from January 2022 to January 2023. An AI interpretation was generated for each radiograph. The 3 radiograph interpretations were each rated in duplicate by 6 emergency department physicians using a 5-point Likert scale. Main Outcomes and Measures: The primary outcome was any difference in Likert scores between radiologist, AI, and teleradiology reports, using a cumulative link mixed model. Secondary analyses compared the probability of each report type containing no clinically significant discrepancy with further stratification by finding presence, using a logistic mixed-effects model. Physician comments on discrepancies were recorded. Results: A total of 500 ED studies were included from 500 unique patients with a mean (SD) age of 53.3 (21.6) years; 282 patients (56.4%) were female. There was a significant association of report type with ratings, with post hoc tests revealing significantly greater scores for AI (mean [SE] score, 3.22 [0.34]; P < .001) and radiologist (mean [SE] score, 3.34 [0.34]; P < .001) reports compared with teleradiology (mean [SE] score, 2.74 [0.34]) reports. AI and radiologist reports were not significantly different. On secondary analysis, there was no difference in the probability of no clinically significant discrepancy between the 3 report types. Further stratification of reports by presence of cardiomegaly, pulmonary edema, pleural effusion, infiltrate, pneumothorax, and support devices also yielded no difference in the probability of containing no clinically significant discrepancy between the report types. Conclusions and Relevance: In a representative sample of emergency department chest radiographs, results suggest that the generative AI model produced reports of similar clinical accuracy and textual quality to radiologist reports while providing higher textual quality than teleradiologist reports. Implementation of the model in the clinical workflow could enable timely alerts to life-threatening pathology while aiding imaging interpretation and documentation.

Video self-instruction for police officers in cardiopulmonary resuscitation and automated external defibrillators.

INTRODUCTION: Police officers often serve as first responders during out-of-hospital cardiac arrests (OHCA). Current knowledge and attitudes about resuscitation techniques among police officers are unknown. Hypothesis/problem This study evaluated knowledge and attitudes about cardiopulmonary resuscitation (CPR) and automated external defibrillators (AEDs) among urban police officers and quantified the effect of video self-instruction (VSI) on these outcomes. METHODS: Urban police officers were enrolled in this online, prospective, educational study conducted over one month. Demographics, prior CPR-AED experience, and baseline attitudes were queried. Subjects were randomized into two groups. Each group received a slightly different multiple-choice test of knowledge and crossed to the alternate test after the intervention, a 10-minute VSI on CPR and AEDs. Knowledge and attitudes were assessed immediately before and after the intervention. The primary attitude outcome was entering "very likely" (5-point Likert) to do chest compressions (CC) and use an AED on a stranger. The primary knowledge outcomes were identification of the correct rate of CC, depth of CC, and action in an OHCA scenario. RESULTS: A total of 1616 subjects responded with complete data (63.6% of all electronic entries). Randomization produced 819 participants in group 1, and 797 in group 2. Groups 1 and 2 did not differ significantly in any background variable. After the intervention, subjects "very likely" to do CC on a stranger increased by 17.2% (95% CI, 12.5%-21.8%) in group 1 and 21.2% (95% CI, 16.4%-25.9%) in group 2. Subjects "very likely" to use an AED on a stranger increased by 20.0% (95% CI, 15.3%-24.7%) in group 1 and 25.0% (95% CI, 20.2%-29.6%) in group 2. Knowledge of correct CC rate increased by 59.0% (95% CI, 55.0%-62.8%) in group 1 and 64.8% (95% CI, 60.8%-68.3%) in group 2. Knowledge of correct CC depth increased by 44.8% (95% CI, 40.5%-48.8%) in group 1 and 54.4% (95% CI, 50.3%-58.3%) in group 2. Knowledge of correct action in an OHCA scenario increased by 27.4% (95% CI, 23.4%-31.4%) in group 1 and 27.2% (95% CI, 23.3%-31.1%) in group 2. CONCLUSION: Video self-instruction can significantly improve attitudes toward and knowledge of CPR and AEDs among police officers. Future studies can assess the impact of VSI on actual rates of CPR and AED use during real out-of-hospital cardiac arrests.