BMC Emergency Medicine最新文献

{"title":"RapidNeuroGuide: an NLP-enhanced AI platform for digital headache triage and clinical decision support.","authors":"Nail Besli, Burcu Bulut","doi":"10.1186/s12873-026-01730-5","DOIUrl":"10.1186/s12873-026-01730-5","url":null,"abstract":"<p><strong>Background: </strong>Headache is among the most frequent presenting complaints in emergency departments worldwide, yet differentiating life-threatening secondary causes from benign primary headaches remains a persistent clinical challenge.</p><p><strong>Objective: </strong>This study describes Phase I of the RapidNeuroGuide development roadmap, comprising the development and an initial literature-derived technical proof-of-concept evaluation of a three-layer AI-driven headache triage engine that integrates critical threshold detection, SNNOOP10 red-flag scoring, and expert-weighted risk assessment.</p><p><strong>Methods: </strong>A total of 240 PubMed-indexed headache articles were screened, and 121 full-text documents were acquired from legally accessible sources. The S-PubMedBERT-MS-MARCO sentence-transformer model was used to identify clinically relevant sentences, which were mapped to 23 structured clinical fields using a deterministic terminology dictionary comprising 103 expressions. The resulting 158 curated, literature-derived case representations constituted the proof-of-concept evaluation dataset.</p><p><strong>Results: </strong>Exact agreement with the predefined reference triage classification was observed in 110 of 158 case representations (69.6%), while 157 cases (99.4%) were assigned within one adjacent triage category. One case (0.6%) differed by more than one category. All 47 case representations carrying a predefined CRITICAL reference classification were assigned to the CRITICAL category within the evaluated dataset. The NLP pipeline produced usable structured parameter sets for 114 of 119 machine-readable documents (95.8%), with a mean cosine similarity of 0.878 across 570 retained sentences.</p><p><strong>Conclusions: </strong>RapidNeuroGuide demonstrated preliminary technical feasibility and measurable agreement with predefined reference triage categories in this Phase I proof-of-concept evaluation.</p>","PeriodicalId":9002,"journal":{"name":"BMC Emergency Medicine","volume":"26 1","pages":""},"PeriodicalIF":3.1,"publicationDate":"2026-08-11","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://www.ncbi.nlm.nih.gov/pmc/articles/PMC13499320/pdf/","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"148788001","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":3,"RegionCategory":"医学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Spatial heterogeneity in risk-adjusted return of spontaneous circulation after out-of-hospital cardiac arrest: the urbanization-adjusted RACA model.","authors":"Dokyeong Lee, Martin Bender, Eiko Spielmann, Ulrike Grittner, Christof Prugger, Julian Friebel","doi":"10.1186/s12873-026-01715-4","DOIUrl":"10.1186/s12873-026-01715-4","url":null,"abstract":"<p><strong>Background: </strong>Registry-based benchmarking of return of spontaneous circulation (ROSC) requires well-calibrated risk adjustment across Emergency Medical Services (EMS) systems. Existing risk-adjustment models, however, do not account for urbanization-related heterogeneity, which may contribute to systematic urbanization-stratified miscalibration, bias cross-system benchmarking, and misdirect quality improvement. We developed and validated an urbanization-adjusted ROSC after cardiac arrest (URACA) by incorporating Eurostat's Degree of Urbanisation (DEGURBA) into RACA and assessed whether it improves calibration across urbanization levels without compromising discrimination.</p><p><strong>Methods: </strong>Using 2014-2023 German Resuscitation Registry data, we identified 145,780 adult (≥18 years) out-of-hospital cardiac arrests and randomly split cases into development and validation cohorts. DEGURBA was recoded as metropolises (≥500,000 inhabitants), cities (<500,000), and non-urban areas (all others). We refitted RACA and fitted URACA using logistic regression with multiple imputation for missing values. In validation, we assessed calibration (calibration belts), discrimination (area under receiver operating characteristic curve; AUROC), prediction error (Brier score), and reclassification (continuous net reclassification improvement; cNRI).</p><p><strong>Results: </strong>Refitted RACA exhibited urbanization-stratified miscalibration, overpredicting in non-urban areas and underpredicting in metropolises and cities. URACA reduced these systematic deviations: in non-urban areas, predicted ROSC closely matched observed ROSC across the full risk range, and in metropolises and cities, the mid-range underprediction was resolved. Discrimination and overall accuracy were unchanged (AUROC 0.735 vs 0.737; Brier 0.202 for both), while reclassification improved (cNRI 9.4%).</p><p><strong>Conclusions: </strong>URACA mitigates urbanization-stratified calibration bias while preserving discrimination, enhancing the comparability of risk-adjusted ROSC benchmarking across EMS systems.</p>","PeriodicalId":9002,"journal":{"name":"BMC Emergency Medicine","volume":"26 1","pages":""},"PeriodicalIF":3.1,"publicationDate":"2026-08-10","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://www.ncbi.nlm.nih.gov/pmc/articles/PMC13466119/pdf/","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"148722988","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":3,"RegionCategory":"医学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Impact of a high-profile sudden cardiac death on emergency department utilization: a multi-campus interrupted time series and exploratory phenotyping study.","authors":"Chenxi Wang, Han Cai, Xiujuan Hu, Jiayu Liu, Guanhua Li, Li Zhang","doi":"10.1186/s12873-026-01723-4","DOIUrl":"10.1186/s12873-026-01723-4","url":null,"abstract":"<p><strong>Background: </strong>High-profile public health events can be temporally associated with significant shifts in healthcare-seeking behavior. We aimed to quantify the surge in emergency department (ED) volume and characterize the demographic shifts following a high-profile sudden cardiac death (SCD) event of a nationally prominent public figure.</p><p><strong>Methods: </strong>A retrospective study was conducted across two major hospital cohorts (N = 15,981). We employed interrupted time series (ITS) analysis to estimate the immediate level shift in daily ED visits. An XGBoost model with SHAP (SHapley Additive exPlanations) interpretability was developed as an exploratory phenotyping tool to characterize the surge based on clinical and behavioral features. Findings were confirmed using an internal validation cohort.</p><p><strong>Results: </strong>Following the SCD event, daily ED volume exhibited a significant immediate level shift (Main: +24.4%; Internal Validation: +34.7%,p < 0.001). A subtle but robust demographic shift toward younger populations was observed (Median age decrease: 2.2 years,p < 0.001). Machine learning identified Age, Nighttime Presentation, and Cardiovascular Department attendance as the features most strongly associated with the surge phenotype. Notably, the SHAP dependence analysis revealed a sharp, non-linear escalation in surge probability for individuals aged < 45 years, particularly during nocturnal hours. Despite the volume surge, no significant increase in mortality or acute myocardial infarction diagnoses was observed, indicating that the surge was predominantly associated with low-risk individuals exhibiting reassurance-seeking behavior.</p><p><strong>Conclusion: </strong>Media-driven events can be followed by a structured, large-scale surge in emergency department utilization, predominantly among younger, low-risk individuals seeking reassurance. These demographic and behavioral findings are consistent with the cyberchondria phenomenon, which may produce a dilution of clinical severity without increasing hard endpoints, potentially straining acute care resources. Targeted public health communication and adaptive triage strategies are essential to preserve emergency capacity during periods of heightened media-associated health concerns.</p>","PeriodicalId":9002,"journal":{"name":"BMC Emergency Medicine","volume":"26 1","pages":""},"PeriodicalIF":3.1,"publicationDate":"2026-08-08","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://www.ncbi.nlm.nih.gov/pmc/articles/PMC13545800/pdf/","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"148890981","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":3,"RegionCategory":"医学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"ECG-based detection of occlusion myocardial infarction: a dedicated deep neural network versus multimodal large language models and physicians - a retrospective diagnostic accuracy study.","authors":"Emin Hüseyin Akar, Kamil Kokulu, Ekrem Taha Sert","doi":"10.1186/s12873-026-01716-3","DOIUrl":"10.1186/s12873-026-01716-3","url":null,"abstract":"<p><strong>Background: </strong>Accurate electrocardiogram (ECG) interpretation for acute coronary occlusion is a critical, time-sensitive task in emergency care. The field is increasingly shifting from the ST-segment elevation myocardial infarction paradigm to the broader concept of occlusion myocardial infarction (OMI). This study compared the diagnostic accuracy of a dedicated occlusion-detection deep neural network, Queen of Hearts (QoH), two general-purpose multimodal large language models (LLMs), and emergency physicians for ECG-based OMI detection. Response consistency and confidence calibration of the LLMs were also evaluated.</p><p><strong>Methods: </strong>This retrospective diagnostic accuracy study used 36 twelve-lead ECGs from patients referred for emergent coronary angiography, including 24 with angiographically confirmed acute coronary occlusion and 12 without. Five emergency medicine specialists, five emergency medicine residents, and two LLMs, ChatGPT 5.2 and Gemini 3 Pro, interpreted all ECGs under a standardized moderate-risk acute coronary syndrome scenario. The same ECGs were submitted to QoH. Each LLM evaluated the set five times in separate sessions, whereas QoH was queried once because it provides deterministic output. In the primary head-to-head analysis, each interpreter or interpreter group contributed a single decision per ECG. Areas under the curve (AUCs) were compared using the DeLong test, binary metrics using the exact McNemar test, and reliability using Fleiss' kappa.</p><p><strong>Results: </strong>QoH showed the highest discrimination for OMI (AUC 0.96), followed by ChatGPT (0.81) and Gemini (0.62). QoH also achieved the highest sensitivity (95.8%, missing one of 24 occlusions) and accuracy (86.1%), with significantly higher sensitivity than emergency specialists (p = 0.016). Among physicians, specialists performed best (accuracy 75.0%; sensitivity 66.7%; specificity 91.7%). ChatGPT showed high specificity (91.7%) but low sensitivity (54.2%), whereas Gemini performed least well overall (accuracy 55.6%). The LLMs produced variable interpretations across repeated queries (Fleiss' kappa 0.24-0.49), and Gemini showed marked overconfidence (Brier score 0.40).</p><p><strong>Conclusions: </strong>In this OMI-focused study, QoH showed the highest discrimination and sensitivity for acute coronary occlusion. General-purpose LLMs showed clinically important limitations and do not currently support autonomous OMI diagnosis. These findings support further evaluation of task-specific models for ECG-based occlusion detection and suggest, at most, an adjunctive, human-supervised role for current general-purpose LLMs.</p>","PeriodicalId":9002,"journal":{"name":"BMC Emergency Medicine","volume":"26 1","pages":""},"PeriodicalIF":3.1,"publicationDate":"2026-08-07","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://www.ncbi.nlm.nih.gov/pmc/articles/PMC13545760/pdf/","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"148890969","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":3,"RegionCategory":"医学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Airway management in acutely poisoned patients with Glasgow Coma Scale less than 9: a retrospective cohort study.","authors":"Sabrina Schmoll, Eva Mehrl, Stefanie Geith, Florian Eyer, Christian Rabe, Tobias Zellner","doi":"10.1186/s12873-026-01714-5","DOIUrl":"10.1186/s12873-026-01714-5","url":null,"abstract":"<p><strong>Background: </strong>A Glasgow Coma Scale (GCS) ≤ 8 is frequently used to guide prehospital intubation, but its generalizability to acutely poisoned patients is uncertain. We compared outcomes of poisoned patients with GCS ≤ 8 managed with or without prehospital intubation.</p><p><strong>Methods: </strong>We retrospectively included poisoned patients with a GCS ≤ 8 found as helpless persons in public settings and admitted to a tertiary toxicology center. Patients were grouped by prehospital intubation status. Clinical characteristics, management, and outcomes were analyzed using nonparametric and categorical tests, the analyses were exploratory due to the limited sample size.</p><p><strong>Results: </strong>Among 84 patients, 10 (11.9%) were intubated prehospitally, and 74 (88.1%) were not. Median GCS was lower in intubated patients (3 vs. 7). Aspiration pneumonitis in non-intubated patients was rare (6.8%) and did not differ significantly between groups. Vasopressor therapy within 24 h was required in one of 74 (1.4%) non-intubated patients and two of 10 (20.0%) intubated patients. Antidotes were frequently administered, especially in non-intubated patients, predominantly in the prehospital setting. Two non-intubated patients (2.7%) required delayed mechanical ventilation. No deaths occurred.</p><p><strong>Conclusions: </strong>Non-intubated patients showed favorable outcomes with low complication rates and infrequent delayed intubations, while antidotes were frequently administered in both the prehospital and in-hospital settings. These results suggest that GCS alone may be insufficient to guide airway management in poisoning, supporting an individualized, clinically guided approach.</p>","PeriodicalId":9002,"journal":{"name":"BMC Emergency Medicine","volume":"26 1","pages":""},"PeriodicalIF":3.1,"publicationDate":"2026-08-06","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://www.ncbi.nlm.nih.gov/pmc/articles/PMC13445855/pdf/","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"148683383","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":3,"RegionCategory":"医学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Tourniquet use in different emergency medicine professional groups, results from a regional cross-sectional online survey.","authors":"P Hilbert-Carius, H Wrigge, K Zur Nieden, B Stichert, A Großstück, Michael Lautenschläger","doi":"10.1186/s12873-026-01674-w","DOIUrl":"10.1186/s12873-026-01674-w","url":null,"abstract":"<p><strong>Background: </strong>Traumatic haemorrhage is one of the most common causes of death in patients with injuries. Targeted pre-hospital haemostasis can reduce morbidity and mortality. The use of tourniquets is advisable for external bleeding in the extremities that cannot be treated with compression or wound packing. There are recommendations for the use of tourniquets. Little is currently known about the level of training and frequency of use of various professional groups of potential tourniquet users.</p><p><strong>Methods: </strong>A regional cross-sectional online survey was conducted on the level of training, frequency of use, and adherence to guidelines for tourniquets among various pre-hospital professional groups (emergency medical technician- EMT, advanced emergency medical technician-AEMT, Paramedical, Helicopter Emergency Medical Services-Technical Crew member-HEMS-TC, pre-hospital physician), to get a regional convenience sample. To obtain an overview of the level of training and use of tourniquets among various professional groups, we conducted a regional online survey on ground-based emergency medical services in the region of Halle/Northern Saale District in Saxony-Anhalt, as well as on helicopter emergency medical services in the Halle/Leipzig metropolitan region. The different professional groups were compared with regard to theoretical and practical training, frequency of use, and adherence to the existing guidelines.</p><p><strong>Results: </strong>178 participants (74% male, 26% female) took part and answered the questionnaire. Among respondents who reported professional group were 47% pre-hospital physicians, 32% paramedics, 9% EMT, 7% HEMS-TCs, 4% AEMT. Statistically significant differences were found in theoretical education, with the lowest rate of education in the EMT group (67%) and the highest rate in the paramedic group (96%). When comparing theoretical education of all the emergency medical service personnel (91%) against the pre-hospital physicians (73%), it was significantly higher in the emergency medical service personnel group. No differences were found in practical training or frequency of use. Physicians had a significant higher adherence to the guidelines for location for applying a tourniquet.</p><p><strong>Conclusion: </strong>In addition to theoretical education, no statistically significant differences were detected in this regional sample in terms of practical training or frequency of use. With regard to adherence to existing guidelines, physicians had a significantly higher adherence for location for applying a tourniquet. Standardized training and periodic refresher education may help ensure consistent tourniquet knowledge and application across prehospital professional groups.</p>","PeriodicalId":9002,"journal":{"name":"BMC Emergency Medicine","volume":"26 1","pages":""},"PeriodicalIF":3.1,"publicationDate":"2026-08-03","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://www.ncbi.nlm.nih.gov/pmc/articles/PMC13445843/pdf/","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"148677028","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":3,"RegionCategory":"医学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Prehospital Emergency Nurse competence and pain management in patients with acute abdominal pain: a retrospective observational study.","authors":"Rasmus Bjerén, Carl Magnusson, Johan Herlitz, Martin Flores Bjurström, Magnus Hagiwara, Denise Bäckström","doi":"10.1186/s12873-026-01707-4","DOIUrl":"10.1186/s12873-026-01707-4","url":null,"abstract":"<p><strong>Background: </strong>Pain is common in emergency medical services, and abdominal pain is among the most frequent complaints. This study examined whether documented pain assessment, analgesic treatment, reassessment, and documented end-of-encounter pain status in patients with acute abdominal pain differed by clinician competence (Prehospital Emergency Nurse [PEN] vs. non-PEN), patient sex, age, and prehospital interval.</p><p><strong>Methods: </strong>This retrospective observational study was a secondary analysis of adult patients with acute abdominal pain in a previously described cohort. Electronic patient records were reviewed in full, including Numerical Rating Scale (NRS) scores and free-text pain descriptions. Primary outcomes were documented pain assessment, analgesia administration, and any documented reassessment. Secondary outcomes were documented NRS use, analgesia among patients with moderate-severe pain, reassessment after analgesia, and mild pain at the last documented assessment. Age and prehospital interval were dichotomized at the median. Mann-Whitney U and Fisher's exact tests were used, with p < 0.01 considered statistically significant.</p><p><strong>Results: </strong>Of 840 sampled records, 816 were included. Median age was 64 years (IQR 41-79), and median prehospital interval was 38 minutes (IQR 27-60). Compared with non-PEN encounters, PEN encounters had higher rates of pain assessment (67% vs. 53%, p = 0.003), analgesia administration (50% vs. 37%, p = 0.005), and reassessment (31% vs. 17%, p < 0.001). Among patients with moderate-severe pain, analgesia was more frequent in PEN encounters (81% vs. 58%, p < 0.001). Mild pain at the last documented assessment was also more common in PEN encounters (23% vs. 8%, p < 0.001). Younger patients and cases with longer prehospital intervals had higher rates of pain assessment, NRS use, analgesia administration, and reassessment, but not mild pain at last assessment.</p><p><strong>Conclusions: </strong>The PEN encounters showed higher rates of documented pain assessment, analgesia administration, and reassessment, with a higher proportion of patients having mild pain at the last documented assessment. Differences related to age and prehospital interval were mainly limited to process measures. Further research is needed to clarify whether clinician competence and care processes translate into improved pain relief.</p>","PeriodicalId":9002,"journal":{"name":"BMC Emergency Medicine","volume":"26 1","pages":""},"PeriodicalIF":3.1,"publicationDate":"2026-07-29","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://www.ncbi.nlm.nih.gov/pmc/articles/PMC13425982/pdf/","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"148629457","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":3,"RegionCategory":"医学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Artificial intelligence for pediatric fracture detection: impact on diagnostic revisions and patient recall rates in a tertiary emergency setting.","authors":"Oliver Johannes Deffaa, Johanna Pape, Dominik Schlösser, Franz Wolfgang Hirsch, Martin Lacher, Maciej Rosolowski, Daniel Gräfe","doi":"10.1186/s12873-026-01697-3","DOIUrl":"10.1186/s12873-026-01697-3","url":null,"abstract":"<p><strong>Purpose: </strong>Artificial intelligence (AI) can support and enhance radiologists in musculoskeletal imaging, but evidence of clinical benefit is still lacking. This prospective study aimed to estimate the range of expected effect size on patient recall rates whether after implementation of an AI system for fracture detection affects in a paediatric emergency setting during out-of-hourse care.</p><p><strong>Methods: </strong>Children and adolescents (2-18 years) undergoing appendicular skeletal radiography between April and September 2025 during on-call hours at a tertiary referral hospital were eligible. On every second day, automated fracture detection by a commercial AI system (TechCare Kids, Milvue, Paris, France) was available for the treating physician. Endpoints were the rate of diagnostic revisions leading to patient recall the next day, therapeutic changes, length of stay at the emergency department, need for senior consultation, subjective diagnostic confidence, and AI accuracy.</p><p><strong>Results: </strong>Among 1515 screened patients, 667 were enrolled (median age 11.0 years, 61% male). Fractures were present in 296 cases (44.3%). AI accuracy was 95.1%. Diagnostic revisions occurred in 8.6% without AI, in 5.7% with AI support (risk ratio 0.66; 95% CI 0.37-1.19). Resulting therapeutic changes were rare in both groups (2.0% vs. 0.4%; p = 0.10). No significant differences were observed for secondary endpoints, including need for senior consultation (14.3% vs. 12.5%), subjective diagnostic confidence, or length of stay (2.3 h vs. 2.4 h).</p><p><strong>Conclusion: </strong>Given the small effect size for patient recall rates in an academic medical setting, it is questionable whether the efforts conducting a confirmatory study with adequate statistical power would be justified by the expected incremental benefit for the investigated outcomes.</p>","PeriodicalId":9002,"journal":{"name":"BMC Emergency Medicine","volume":"26 1","pages":""},"PeriodicalIF":3.1,"publicationDate":"2026-07-29","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://www.ncbi.nlm.nih.gov/pmc/articles/PMC13421735/pdf/","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"148618362","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":3,"RegionCategory":"医学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Incident location at home and emergency medical services on-scene time for older adults in Japan: a population-based cross-sectional study.","authors":"Rumiko Tsuchiya-Ito, Keiko Ueno, Chie Teramoto, Haruka Imura, Tomonori Sano, Kazuaki Tanabe","doi":"10.1186/s12873-026-01694-6","DOIUrl":"https://doi.org/10.1186/s12873-026-01694-6","url":null,"abstract":"<p><strong>Background: </strong>As population aging increases the demand for emergency medical services (EMS), reducing prehospital time has become crucial for improving patient outcomes and operational efficiency. Accessible housing environments may facilitate patient transportation and reduce on-scene time (interval between ambulance arrival and departure from the scene), but evidence is scarce. This study examined the associations between incident locations at home and EMS on-scene time among older adults in Japan.</p><p><strong>Methods: </strong>This population-based cross-sectional study analyzed 26,058 people aged ≥ 65 years transported from their homes to hospitals by the Higashi-Hiroshima Fire Department's ambulances between January 2016 and December 2024. We examined the associations between two exposure variables (floor level and in-home location) and on-scene time using generalized linear models with gamma distribution and log-link function. Adjusted mean ratios (aMRs) and 95% confidence intervals (CIs) were calculated. Then, a correspondence analysis using text data from the ambulance dispatch records was conducted to generate further hypotheses regarding the factors characterizing each in-home location.</p><p><strong>Results: </strong>The median EMS on-scene time was 16 min (interquartile range: 13-21). Higher floors were significantly associated with progressively longer on-scene times. When compared with living rooms, corridors (aMR: 0.95, 95% CI: 0.93-0.97) and gardens (aMR: 0.95, 95% CI: 0.93-0.98) had significantly shorter on-scene times. In contrast, toilets (aMR: 1.08, 95% CI: 1.04-1.12), bathrooms (aMR: 1.09, 95% CI: 1.05-1.13), and kitchens (aMR: 1.06, 95% CI: 1.03-1.10) had significantly longer on-scene times. Correspondence analysis showed that garden and other locations were positioned at the \"exogenous\" pole along the horizontal axis (representing type of injury or disease). The toilet, bathroom, and kitchen, which were associated with longer on-scene times, were positioned toward the \"loss of consciousness\" pole along the vertical axis (representing level of consciousness).</p><p><strong>Conclusions: </strong>Higher floors and confined areas (toilets and bathrooms) were associated with longer on-scene time. Future studies are invited to examine the effect of patients' consciousness levels to ensure that paramedics can rapidly reach and transport emergency patients regardless of location.</p>","PeriodicalId":9002,"journal":{"name":"BMC Emergency Medicine","volume":"26 1","pages":""},"PeriodicalIF":3.1,"publicationDate":"2026-07-24","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"149227079","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":3,"RegionCategory":"医学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Performance of GPT-5.5 thinking for Manchester Triage System-based emergency department triage: agreement with an expert reference standard and prediction of short-term outcomes.","authors":"Sureyya Tuba Fettahoglu, Adem Az, Ozgur Sogut, Tarik Akdemir, Muhammed Furkan Ozden, Salih Fettahoglu, Mucahide Samli Helvacı, Enes Sisman, Mehmet Ozcomlekci, Emir Bilici, Nurbaki Kilic, Tugay Saricicek, Ayse Feyza Celik, Zekeriya Colak, Zulkuf Eralp","doi":"10.1186/s12873-026-01704-7","DOIUrl":"https://doi.org/10.1186/s12873-026-01704-7","url":null,"abstract":"<p><strong>Objective: </strong>To evaluate the agreement between Manchester Triage System (MTS) category assignments generated by GPT-5.5 Thinking and an expert-derived reference standard in a selected cohort of emergency department (ED) patients, and to assess the model's ability to predict hospital admission and 72-hour mortality.</p><p><strong>Methods: </strong>This single-center observational study prospectively sampled consecutive ED presentations during a prespecified continuous 72-hour window, with triage-time variables and secondary outcomes obtained from hospital records. Patients classified as MTS Category 5 (Blue; nonurgent) by the expert reference standard were excluded from the analytic cohort. GPT-5.5 Thinking was prompted in a standardized single-shot format using structured clinical vignettes derived from triage data. The primary endpoint was agreement between model-assigned MTS categories and an expert reference standard established by two independent blinded emergency medicine specialist rater teams, with disagreements resolved through senior adjudication. Exact agreement, within-1-level agreement, overtriage, undertriage, and dangerous undertriage were calculated. Secondary analyses evaluated prediction of hospital admission and 72-hour mortality using diagnostic performance metrics.</p><p><strong>Results: </strong>A total of 1,841 patients were included. Overall agreement between GPT-5.5 Thinking and the expert reference standard was substantial (κ = 0.721; 95% CI 0.691-0.740). Exact agreement rate was 62.0%, and within-1-level agreement rate was 99.0%. Overtriage occurred in 35.3% of cases, undertriage in 2.7%, and dangerous undertriage in 0.2%. Exact agreement rate was significantly lower in patients aged ≥ 65 years than in younger adults (57.1% vs. 64.2%, p = 0.013). For hospital admission, accuracy rate was 79.9%, sensitivity 83.8%, positive predictive value (PPV) 45.2%, and negative predictive value (NPV) 95.9%. For 72-hour mortality, accuracy rate was 86.5%, sensitivity 72.3%, PPV 12.6%, and NPV 99.1%.</p><p><strong>Conclusion: </strong>Within this selected ED cohort, GPT-5.5 Thinking showed substantial agreement with expert MTS-based ED triage and generally preserved the ordinal hierarchy of clinical urgency. Its performance was characterized by frequent adjacent-level agreement, predominant overtriage, and very rare but nonzero dangerous undertriage. The model also showed stronger rule-out than rule-in characteristics for hospital admission and 72-hour mortality.</p>","PeriodicalId":9002,"journal":{"name":"BMC Emergency Medicine","volume":"26 1","pages":""},"PeriodicalIF":3.1,"publicationDate":"2026-07-24","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"149227076","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":3,"RegionCategory":"医学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}