Health care process quality measures usually are designed by expert panels attempting to synthesize nuanced clinical evidence and subsequently operationalized using administrative data. Many quality measures are then adopted without directly validating their presumed links with outcomes. Later efforts to validate process measures often yield negative results, leaving policy makers without a defensible means of measuring quality. This article presents an alternative strategy for developing and validating process quality measures. The development of an alcohol use disorder (AUD) treatment quality measure is used as an example.
An expert panel generated a range of candidate process quality measures of AUD treatment derivable from administrative data that were then tested to determine which had the strongest associations with facility- and patient-level outcomes. Outcome and process data were from 2701 US Veterans Health Administration patients starting a new episode of care at 54 VA facilities.
Several of the candidate process-of-care quality measures predicted facility- and patient-level outcomes. Having at least 3 visits during the first month of specialty AUD treatment was correlated with improvement on the Addiction Severity Index Alcohol composite at the facility level, r = 0.41 (95% Confidence Interval 0.16–0.61), and at the patient level, r = 0.07 (CI: 0.03–0.11).
These “prevalidated” quality measures can now be judged for the extent they map onto the extant clinical literature and other design requirements. The development and validation strategy we describe should aid in efficiently producing quality measures in other areas of health care.