Development of Decision-Making Model for Predicting Delirium Tremens in Alcohol Use Disorder

Objectives: To identify predictive factors associated with the onset of delirium tremens (DT) and produce a ‘decision tree’ that clinicians could utilize when deciding which patient undergoing alcohol withdrawal symptoms is more likely to develop DT. Methods: Consultation answer notes, demographic, and laboratory data of 243 inpatient consultation cases regarding alcohol withdrawal symptoms were collected. Receiver operating characteristic analysis (ROC) was performed to identify predictors of DT development. Results: DT was developed in 26.75% of consulted cases. ROC revealed that delirium history was the most significant predictor of DT development (κ=0.342; χ2=29.415, p<0.001). Subjects with a delirium history had a conversion rate of 59.5%. Those without such history had a conversion rate of 19.0%. The second most significant factor was body mass index (BMI), with a cutoff point of 23.8 kg/m2 (κ=0.224; χ2=9.043, p<0.01). Among subjects without delirium history, those with BMI ≥23.8 kg/m2 had a conversion rate of 28.6% and those with BMI