Data mining-based clinical profiles of substance use-related emergency department utilizers

Substance-use is a prevalent presentation to the emergency department (ED); however, the clinical characterization of patients who are treated and discharged without admission for further treatment is under-investigated. The study aims to define and characterize the clinical profiles of this patient population.

Patients' presentations were examined by clinical data mining (chart review) of ED records of substance use-related events of individuals discharged without admission for further treatment. Records (N = 199) from three major hospitals in New York City from March and June 2017 were randomly sampled with primary diagnosis of alcohol, opioid-related and other psychoactive substance-use presentations. Qualitative thematic coding of clinical presentation with inter-rater reliability was performed. Quantitative distinctive validity tested independence through Pearson's chi-squared and analysis of variance using Fisher's F-test.

Six distinct clinical profiles were identified, including, High Utilizers (chronically intoxicated with comorbid health conditions) (36.7%), Single Episode (20.1%), Service Request (14.1%), Altered Mental Status (13.6%), Overdose (9.0%), and Withdrawal (7.5%). The profiles differed (p < 0.05) in age, housing status, payor, mode of arrival, referral source, index visit time, prescribed treatment, triage acuity level, psychiatric history, and medical history. Differences (p < 0.05) between groups across clinical profiles in age and pain level at triage were observed.

The identified clinical profiles represent the broad spectrum and complex nature of substance use-related ED utilization, highlighting critical factors of psychosocial and mental-health comorbidities. These findings provide a preliminary foundation to support person-centered interventions to decrease substance use-related ED utilization and to increase engagement/linkage of patients to addiction treatment.