Possible Alternatives to Diagnosis-Based Denominators for Addiction Treatment Quality Measures

Consumers of healthcare quality measures are often unaware of how measured performance may be driven by diagnosing practices rather than the provision of high quality care. Reliance on quality metrics that depend on documented diagnoses can therefore subvert comparisons between clinicians, facilities or systems. In this study, three versions of an alcohol use disorder (AUD) treatment quality measure were calculated: method 1--the usual denominator including all diagnosed patients; method 2--a "population-based" denominator including the entire facility census; and method 3-an epidemiologically-derived denominator comprising the expected prevalence of AUD based on case-mix characteristics and geographic region. Performance rankings under the three specifications were calculated. Changes in percentile rank of up to 30-45% were observed between methods. Therefore, much of the observed between-facility differences on diagnosis-based quality measures may reflect variation in the propensity to diagnose rather than real differences in performance. Stakeholders must decide which of the validity threats produced by these different methods is least worrisome.