Özen, Mert; Sari, Abdullah; Sarı, Zehra Özlem Peker

Introduction: Emergency medical services (EMS) transport a het erogeneous patient population, ranging from patients requiring critical care to those discharged after limited emergency department (ED) evaluation. Simple scores based on early available variables may help describe these different EMS-arrival phenotypes more consistently. This study aimed to derive and internally validate two practical scores for critical care need (CCN) and potentially inappropriate ambulance use (PIAU) among EMS arrivals to a tertiary ED. Material and Method: This single-center retrospective study included patients transported by the national 112 EMS system to a tertiary academic ED between Jan uary 1, 2018 and December 31, 2018. The CCN outcome was defined as intensive care unit-level admission, transfer for intensive care requirement or ED death. PIAU was defined as green ED triage, ED discharge, no computed tomography or magnetic resonance imaging, no non- radiology clinical consultation and no hospital admission. Separate multivariable logistic regression models were used to derive the CCN and PIAU scores. Beta coefficients were converted into simplified integer points. Discrimination was assessed using receiver operating characteristic curve (AUC) and internal validation was performed using stratified 10-fold cross-validation. Results: A total of 9,500 EMS arrivals were included. Critical care need occurred in 1,018 patients (10.7%), while PIAU was identified in 852 patients (9.0%). The CCN Score showed good discrimination for critical care need, with an AUC of 0.825 and a 10-fold cross-validation AUC of 0.828. At t he optimal cut-off of ≥4 points, sensitivity was 78.6%, specificity was 72.8% and negative predictive value was 96.6%. The PIAU Score also showed good discrimination, with an AUC of 0.853 and a 10-fold cross-validation AUC of 0.856. At the optimal cut-off of ≥11 points, sensitivity was 80.3%, specificity was 81.1% and negative predictive value was 97.6%. Conclusion: The CCN and PIAU scores showed good discrimination for identifying two clinically relevant ends of the EMS-arrival spectrum. These simple point-based tools may support future research on early risk stratification, ED resource planning and EMS system feedback, but they require exte rnal validation before clinical implementation.

Çeliker, Fatma Beyazal; Hürsoy, Nur; Kaba, Esat; Er, Hüseyin; Kayaoglu, Ahmet Taha; Kolluk, Hafsa; Çubukçu, Yusuf; Budak, Dilara

Introduction: This study aims to evaluate the success of the ChatGPT model in detecting emergency pathologies in pediatric emergency ultrasound reports and to compare the effectiveness of zero-shot and few-shot prompting techniques. Material and Method: A total of 196 pediatric emergency ultrasound reports were included in the study. Firstly, the ultrasound reports were anonymized. A board-certified radiologist labeled the reports as positive or negative for emergency pathology. ChatGPT-3.5 and 4 models were tested with different prompting techniques (zero-shot and few-shot) and at different times (March and June). Performance of each model was assessed using accuracy, sensitivity, specificity, and F1-score. The McNemar test was used to compare the binary classes. Results: The most frequently identified emergency pathologies were testicular torsion, appendicitis, and intussusception. ChatGPT-4 (March, zero-shot) achieved the highest success rate (accuracy 94%, sensitivity 91%, specificity 95%). In binary classification, a stat istically significant difference was only found between ChatGPT-3.5 zero-shot (June) and ChatGPT-4 zero-shot (June) responses (p<0.05). The accuracy rates of all models ranged from 88% to 94%. Few-shot prompting did not show a significant performance advantage over the zero-shot technique. Conclusion: ChatGPT models are a successful method for detecting the presence of emergency pathology when compared to the reference standard of radiologist interpretation. Our study demonstrates the potential for large language models to be used as triage tools in clinical workflows.