The limited predictive value of initial serum ethanol concentration for assessing risk of complicated alcohol withdrawal: a retrospective cohort analysis.

INTRODUCTION: Alcohol intoxicated patients frequently present to emergency departments and often require assessment of withdrawal risk. While Predictor of Alcohol Withdrawal Severity Score effectively predicts severe withdrawal outcomes, its reliance on patient history can be a limitation in the emergency department. Initial serum ethanol concentrations in these patients are sometimes used to make disposition decisions despite a lack of evidence to support their association with complicated alcohol withdrawal risk. METHODS: A retrospective cohort analysis was performed on 61 279 Veterans Health Administration emergency department visits from 2010 to 2019 which resulted in hospital admission after an initial serum ethanol concentration of >50 mg/dl was measured. Patients who developed withdrawal during their Emergency Department (ED) stay were excluded. Variables included serum ethanol concentration, demographics, and prior withdrawal history. Outcomes were International Classification of Diseases (ICD)-coded alcohol withdrawal and delirium tremens (DTs). Statistical analyses comprised Analysis of Variance (ANOVA), risk ratios, and logistic regression adjusting for prior withdrawal history. RESULTS: Among the cohort, 15% experienced alcohol withdrawal, and 1% developed DTs. While initial serum ethanol levels were higher in those with withdrawal (213 vs 203 mg/dl), the correlation was modest, and weaker than serum aspartate aminotransferase/alanine aminotransferase. Prior withdrawal history was a stronger predictor (Relative Risk [RR] 3.72 for withdrawal; RR 5.52 for DTs). After logistic regression adjusting for prior withdrawal history, serum ethanol concentration ≥400 mg/dl was not a useful predictor of withdrawal or DT risk (Odds Ratio [OR] ~ 1.00). CONCLUSIONS: Initial serum ethanol concentration offers a measurable but limited predictive value for complicated alcohol withdrawal risk, and no utility when a reliable prior history is available.