Predictors of Higher-Intensity Care in Epinephrine-Treated Emergency Department Encounters for Anaphylaxis and Allergic Reactions: A Retrospective Cohort Study With Internal Validation
Year: 2026
Type: preprint
Study design: Observational Study
Abstract: Background Clinicians treating anaphylaxis in the emergency department (ED) have little objective guidance on which patients will need more than a single dose of epinephrine. We examined whether features recorded at the initial ED assessment are associated with higher-intensity care and tested the stability of these associations across approaches to missing data, repeat encounters, and outcome definition. Methods Adult ED encounters that received epinephrine for anaphylaxis or an allergic reaction were drawn from MIMIC-IV-ED. The primary outcome was a composite of recurrent epinephrine, intensive care unit (ICU) admission, mechanical ventilation, or vasopressor use. A prespecified seven-variable logistic model (age, sex, ambulance arrival, initial oxygen saturation, asthma, hypertension, food trigger) was fitted with multiple imputation. Sensitivity analyses used complete cases, a saturation-free model on all encounters, a first-encounter cohort, and patient-clustered standard errors. Internal validation nested the imputation step inside bootstrap resampling and repeated 10-fold cross-validation on all encounters. ICU admission, recurrent epinephrine, and life support (ventilation or vasopressor) were modeled separately. Results Among 424 encounters (399 patients), 185 (43.6%) met the primary outcome; none died in hospital. Asthma (adjusted odds ratio 3.65 (1.45–9.23)), hypertension (3.45 (1.64–7.29)), older age (1.16 (1.02–1.33) per decade), and a food trigger (2.50 (1.22–5.12)) were associated with the outcome. The food estimate weakened in complete cases (1.71 (0.77–3.80)) because encounters lacking an initial saturation were disproportionately food-triggered and severe; the saturation-free model on all 424 encounters reproduced the association (2.43 (1.19–4.95)). Age, asthma, and hypertension estimates were similar in the first-encounter cohort and with clustered errors. Discrimination was modest (nested cross-validated AUC 0.682; bootstrap-corrected calibration slope 0.90). Asthma, hypertension, and food trigger were more strongly related to ICU admission (AUC 0.883) than to recurrent epinephrine. The lowest tertile of cross-validated risk still had an event rate of 30.3%. Conclusions Asthma, hypertension, older age, and a food-related trigger identify epinephrine-treated encounters at higher risk of escalation, particularly ICU-level care. None of the examined risk strata had an event rate low enough to support routine immediate discharge. Timing of ICU-level events relative to the first dose could not be verified; external validation is required.
Source: Research Square
Institutions: Sağlık Bilimleri Üniversitesi, Ağrı İbrahim Çeçen University
Language: English
Subfield: Immunology and Allergy
Field: Medicine
Domain: Health Sciences
Keywords: anaphylaxis, allergic reactions, epinephrine, emergency department, asthma, hypertension, food allergy, intensive care unit admission, mechanical ventilation, vasopressors, oxygen saturation, internal validation, missing data, MIMIC-IV-ED, risk prediction model, discrimination and calibration, bootstrapping, multiple imputation
Open Access status: green
OpenAlex ID: W7236728936
