• Justin K Brooten, Jaime L Speiser, Jennifer L Gabbard, David P Miller, Simon A Mahler, Adam S Turner, Rebecca L Omlor, Michelle M Mielke, and David M Cline.
• Department of Emergency Medicine, Wake Forest University School of Medicine, Winston-Salem, North Carolina, USA.
• Acad Emerg Med. 2025 Jan 16.

ObjectivesIdentifying patients in the emergency department (ED) at higher risk for in-hospital mortality can inform shared decision making and goals-of-care discussions. Electronic health record systems allow for integrated multivariable logistic regression (LR) modeling, which can provide early predictions of mortality risk in time for crucial decision making during a patient's initial care. Many commonly used LR models require blood gas analysis values, which are not frequently obtained in the ED. The goal of this study was to develop an all-cause mortality prediction model, derived from commonly collected ED data, which can assess mortality risk early in ED care.MethodsData were obtained for all patients, age 18 and older, admitted from the ED to Atrium Health Wake Forest Baptist from April 1, 2016, through March 31, 2020. Initial vital signs including heart rate, respiratory rate, systolic blood pressure, diastolic blood pressure, mean arterial pressure, pulse oximetry, weight, body mass index, comprehensive metabolic panel, and a complete blood count were electronically retrieved for all patients. The prediction model was developed using LR. The ED early mortality (EDEM) model was compared with the rapid Emergency Medicine Score (REMS) for performance analysis.ResultsA total of 45,004 patients met inclusion criteria, comprising a total of 77,117 admissions. In this cohort, 52.8% of patients were male and 47.2% were female. The model used 35 variables and yielded an area under the receiver operating characteristic curve (AUC) of 0.889 (95% CI 0.874-0.905) with a sensitivity of 0.828 (95% CI 0.791-0.860), a specificity of 0.788 (95% CI 0.783-0.794), a negative predictive value of 0.995 (95% CI 0.994-0.996), and a positive predictive value of 0.084 (95% CI 0.076-0.092). This outperformed REMS in this data set, which yielded an AUC of 0.500 (95% CI 0.455-0.545).ConclusionsThe EDEM model was predictive of in-hospital mortality and was superior to REMS.© 2025 Society for Academic Emergency Medicine.

4 keywords...

hide…