2005 HSR&D National Meeting Abstract
1014 — Using the Addiction Severity Index to Predict Health Service Use in a VA Substance Abuse Clinic
Wryobeck JM (Ann Arbor VA/University of Michigan)
Chermack ST (Ann Arbor VA/University of Michigan)
Closser MH (Ann Arbor VA/University of Michigan)
Blow FC (Ann Arbor VA/University of Michigan)
The purpose of this study was to examine predictors of substance abuse services and a broad array of ancillary services (partial hospital, mental health counseling, psychiatric, medical, urgent care) for veterans admitted for outpatient substance abuse treatment. This study addressed limitations of prior work by examining a broad array of health services beyond substance abuse treatment, and demonstrating the incremental validity of combining the Addiction Severity Index (ASI) with demographic and clinical diagnostic information.
Subjects were 260 veterans, who were administered an ASI thru the Substance Abuse Clinic (SAC) of a major VA Medical Center. Data included demographics (e.g., age, sex, race), clinical diagnoses (mental health and medical) assessment data (e.g., ASI composite scores), as well as measures of six month health service utilization (e.g., outpatient substance abuse, mental health counseling, psychiatric medication management, partial hospitalization, medical outpatient services, and urgent care visits, overall service use).
Multivariate analyses (Tobit regression models) were conducted for each services outcome to examine the impact of adding ASI data to patient data (demographics, psychiatric and medical diagnoses). Compared to models with only patient data (as well as models with only ASI scores), full models containing both ASI scores and clinical showed marked increases in R2 for most services outcomes (substance abuse, mental health, partial hospital, medical, urgent care), although increases were minimal for psychiatric appointments and overall service use. ASI scales (e.g., psychiatric, medical) and medical diagnoses generally predicted service use (e.g., partial hospital, medical) in an expected manner, and a comorbid diagnosis of depression stood out as a predictor of overall services use.
Adding psychometric data (ASI scores) to demographics and clinical data improves the ability to predict an array of services utilization, and the results support the predictive validity of ASI scores for medical and intense partial hospital services.
This study demonstrates the utility of combining patient (demographics, clinical diagnoses) with ASI data to predict subsequent services use. Such an approach could be useful to the VA in conducting needs assessments and examining the effectiveness of care processes.