World journal of emergency medicine最新文献

Background: To provide a comprehensive analysis of the landscape of artificial intelligence (AI) applications in cardiac arrest (CA).

Methods: Comprehensive searches were conducted in PubMed, the Cochrane Library, Web of Science, and EMBASE from database inception through 10 June 2025. Studies that applied AI in both in-hospital cardiac arrest (IHCA) and out-of-hospital cardiac arrest (OHCA) populations across the following domains were included: prediction of cardiac arrest occurrence, prognostication of CA outcomes, applications of large language models (LLMs), and evaluation of cardiopulmonary resuscitation (CPR) and other AI-driven interventions related to CA.

Results: The scoping review included 114 studies, encompassing data from 9,574,462 patients in total. AI was most commonly applied to the prediction of CA (overall, n=40; IHCA, n=30; OHCA, n=4; and both, n=6), CPR-related decision support during CA (n=16), and post-arrest prognosis and rehabilitation outcomes (overall, n=38; OHCA, n=21; IHCA, n=3; and both, n=14). Additional application areas included LLM-based applications (n=8), emergency call handling (n=4), wearable device-based detection (n=3), heart rhythm identification (n=2), education (n=2), and extracorporeal cardiopulmonary resuscitation (ECPR) candidate identification (n=1). Across all application scenarios, the highest area under the receiver operating characteristic curve (AUROC) value for pre-arrest CA prediction in IHCA patients was 0.998 using a multilayer perceptron (MLP) model, whereas the optimal AUROC for pre-arrest CA prediction in OHCA patients was 0.950 using extreme gradient boosting (XGBoost) or random forest (RF) models. For CPR-related decision support during CA, the highest AUROC achieved was 0.990 with a convolutional neural network (CNN) model. In prognostic prediction, the optimal AUROC for IHCA patients was 0.960 using XGBoost, while for OHCA patients it reached 0.976 using an MLP model.

Conclusion: This review shows that AI is most commonly used for the prediction of CA and CPR-related support, as well as post-arrest and rehabilitation outcomes. Future research directions include drug discovery, post-resuscitation management, neurorehabilitation, and clinical trial innovation. Further studies should prioritize multicenter clinical trials to evaluate AI models in real-world settings and validate their effectiveness across diverse patient populations. Overall, AI has significant potential to improve clinical practice, and its role in CA application is increasingly important.

Background: Acute pain is a sudden experience secondary to injuries and varies in perception among individuals. In trauma patients, it can negatively affect respiratory function, immune response, and wound healing, making it a significant public health concern. This study is to determine the prevalence and factors associated with acute pain among emergency trauma patients.

Methods: A multicenter cross-sectional study was conducted. Data were collected via interviewer-administered questionnaires and patient chart review. The data were analyzed via the statistical package for social science version 25. Bivariable and multivariable logistic regression analyses were used. Variables with a P-value <0.05 were considered statistically significant.

Results: A total of 397 patients were included in the study, for a response rate of 96.8%. The prevalence of pain during admission was 91.9% (95% confidence intervals [95% CIs]: 88.8%-94.4%). Blunt trauma (adjusted odds ratio [aOR]=2.82; 95% CI: 1.23-6.45), analgesia before admission to the emergency department (aOR=2.71; 95% CI: 1.16-6.36), documentation of pain severity in the chart (aOR=2.71; 95% CI: 1.16-6.36), analgesia provided within two hours after admission (aOR=7.60; 95% CI: 2.79-20.68), use of non-pharmacological pain management methods (aOR=3.09; 95% CI: 1.35-7.08) and availability of analgesia (aOR=3.95; 95% CI: 1.36-11.43) were associated with acute pain experience.

Conclusion: The prevalence of acute pain among emergency trauma patients was high in the study area. Analgesia should be administered prior to admission, and non-pharmacological pain management should be implemented. Moreover, training on pain assessment and management should be provided for healthcare providers in the emergency department.

Background: Sepsis is a highly heterogeneous organ dysfunction syndrome. There is limited evidence regarding phenotypes and clinical outcomes in sepsis patients with initial normal lactate levels. We sought to identify the lactate-based clinical phenotypes and outcomes of sepsis patients.

Methods: The Medical Information Mart for Intensive Care IV (MIMIC-IV) and eICU databases were used to conduct a retrospective cohort study. Adult sepsis patients were included. Lactate was measured via blood gas, and the same assay type was used across both databases. Serial lactate measurements were analyzed via a two-point classification system based on the highest values recorded during two consecutive 24-hour periods following ICU admission. The first measurement window (T1) comprised the initial 24 h post-admission, whereas the second window (T2) covered 24-48 h post-admission. The lactate difference was defined as the numerical change between the highest lactate level at T2 and the highest level at T1. The time interval between these two measurements was fixed, with T2 commencing immediately after T1, together encompassing the first 48 h post-ICU admission. A normal lactate level was defined as ≤2 mmol/L, and an elevated level was defined as >2 mmol/L. Sepsis patients were stratified into four trajectory phenotypes: (1) normal‒normal (N‒N); (2) normal-elevated (N‒E); (3) elevated-normal (E‒N); and (4) elevated-elevated (E‒E). The primary outcome was in-hospital mortality.

Results: This study enrolled 6,926 sepsis patients. The clinical phenotypes of the sepsis patients were as follows: N‒N (24.4%), N‒E (3.8%), E‒N (36.4%), and E‒E (35.3%). The in-hospital mortality rates of sepsis patients with the four phenotypes from the MIMIC-IV and eICU databases were as follows (N‒N: 18.9% vs. 17.6%, P=0.66; N‒E: 35.3% vs. 29.2%, P=0.45; E‒N: 16.6% vs. 14.2%, P=0.14; E‒E: 43.6% vs. 37.8%, P=0.01). After adjusting for age, sex, Sequential Organ Failure Assessment (SOFA) score, vasopressor therapy, and infection sites, the N‒E phenotype was associated with a higher risk of in-hospital mortality (odds ratio [OR] 1.44; 95% confidence intervals [95% CI] 1.11-1.86; P=0.006; adjusted OR 1.61; 95% CI 1.23-2.11; P<0.001). The E‒N phenotype was associated with the most favorable outcomes for in-hospital mortality in the multivariable analysis (adjusted OR 0.41; 95% CI 0.36-0.46; P<0.001). The E‒E phenotype was associated with the highest risk of in-hospital mortality in the overall cohort (adjusted OR 3.00; 95% CI2.67-3.37; P<0.001).

Conclusion: In sepsis patients with normal initial lactate levels, serial lactate measurements could be valuable for prognostic assessment.

Background: Sepsis is a prevalent and severe condition, with microcirculation disruptions playing a crucial role in its progression. Endothelial cell (EC) injury is the primary factor behind microcirculatory issues. This review is to outline the pathomechanism, organ heterogeneity, biomarkers, and therapeutic implications of endothelial dysfunction in sepsis, offering references and insights for the clinical management of sepsis.

Methods: A systematic search of Web of Science and PubMed from inception to June 10, 2025, limited to English publications, was conducted. Two reviewers independently identified studies on EC injury in patients with septic microcirculatory dysfunction. Duplicate articles based on multiple search criteria were excluded.

Results: Fifty-nine articles, including cell, animal, and clinical studies, were included. These studies reported the effects of EC injury on the microcirculation in sepsis, including changes in vascular permeability, coagulation dysfunction, vasomotor regulation, and inflammatory responses. These pathways interact and ultimately lead to septic microcirculation disorders.

Conclusion: Sepsis-induced endothelial dysfunction involves various interconnected mechanisms, which collectively compromise ECs and impede microcirculatory perfusion. Future research should enhance current understanding of endothelial injury mechanisms, develop synergistic multi-target strategies to disrupt this cycle, and facilitate the clinical application of endothelial markers for early intervention and dynamic assessment.

Background: This study aims to develop and validate a machine learning-based in-hospital mortality predictive model for acute aortic syndrome (AAS) in the emergency department (ED) and to derive a simplified version suitable for rapid clinical application.

Methods: In this multi-center retrospective cohort study, AAS patient data from three hospitals were analyzed. The modeling cohort included data from the First Affiliated Hospital of Zhengzhou University and the People's Hospital of Xinjiang Uygur Autonomous Region, with Peking University Third Hospital data serving as the external test set. Four machine learning algorithms-logistic regression (LR), multilayer perceptron (MLP), Gaussian naive Bayes (GNB), and random forest (RF)-were used to develop predictive models based on 34 early-accessible clinical variables. A simplified model was then derived based on five key variables (Stanford type, pericardial effusion, asymmetric peripheral arterial pulsation, decreased bowel sounds, and dyspnea) via Least Absolute Shrinkage and Selection Operator (LASSO) regression to improve ED applicability.

Results: A total of 929 patients were included in the modeling cohort, and 210 were included in the external test set. Four machine learning models based on 34 clinical variables were developed, achieving internal and external validation AUCs of 0.85-0.90 and 0.73-0.85, respectively. The simplified model incorporating five key variables demonstrated internal and external validation AUCs of 0.71-0.86 and 0.75-0.78, respectively. Both models showed robust calibration and predictive stability across datasets.

Conclusion: Both kinds of models were built based on machine learning tools, and proved to have certain prediction performance and extrapolation.