Identifying levels of alcohol use disorder severity in electronic health records

Alcohol use disorder (AUD) is conceptualized as a dimensional phenomenon in the DSM-5, but electronic health records (EHRs) rely on binary AUD definitions according to the ICD-10. The present study classifies AUD severity levels using EHR data and tests whether increasing AUD severity levels are linked with increased comorbidity.

Billing data from two German statutory health insurance companies in Hamburg included n = 21,954 adults diagnosed with alcohol-specific conditions between 2017 and 2021. Based on ICD-10 alcohol-specific diagnoses, patients were classified into five AUD severity levels: 1 (F10.0, T51.0 or T51.9); 2 (F10.1); 3 (F10.2); 4 (F10.3/4); 5 (K70 + or one of the following diagnoses: K70.0-4, K70.9, K85.2, K85.20, K86.0, 10.5-9, E24.4, G31.2, G62.1, G72.1, I42.6, K29.2). Generalized estimating equation regression models for count data (Poisson distribution) were used to assess associations with the Elixhauser Comorbidity Score (ECS).

Across the study period, the annual prevalence of any AUD diagnosis varied between 2.7% and 2.9%. A dose-response relationship was observed between AUD severity and ECS, indicating that individuals with higher AUD severity experience more comorbid conditions, particularly cardiovascular and liver diseases.

The proposal to define AUD severity levels based on ICD-10 diagnoses allows for a more nuanced analysis of AUD in EHR data.