The use of administrative data as a substitute for individual screening scores in observational studies related to problematic alcohol or drug use

Abstract Administrative data provide a rich resource for improving our understanding of individuals with substance use disorders. The validation of administrative proxies for moderate or high risk alcohol or drug (AOD) use could enhance the ability to carry out rigorous observational research (for example, for use in the construction of comparison groups). This study used receiver operating characteristic (ROC) curve techniques to assess how well AOD-related administrative indicators predicted self-reported AOD use obtained from AUDIT/DAST screening scores. An administrative AOD indicator, derived from a combination of medical encounter and billing data, arrest records, and publicly funded AOD-related services data, demonstrated discrimination in the acceptable range (AUC: 0.72–0.78) for identifying self-reported AOD use consistent with potential need for either (1) any AOD-related intervention, or (2) intensive AOD-related intervention or treatment. These findings held up in two distinct samples: a statewide Medicaid-only sample and a single-site mixed-payer sample that included the uninsured. Our findings suggest that indicators of AOD-related problems derived from administrative data can be useful for identifying moderate or high risk AOD use in a research context. The findings further suggest that proxies for substance use disorders, such as those evaluated here, can enhance future observational studies intended to improve health care for this population.