Background: As an example of the process that could be used to evaluate and optimize the performance of quality measures in routine practice, we evaluated whether the Healthcare Effectiveness Data and Information Set (HEDIS) measure assessing the "persistence of β-blocker treatment after a heart attack" correlates with post-myocardial infarction (MI) outcomes and whether or not there are alternative specifications of this construct which are better predictors and/or may be more easily applied.
Research design: The study included a retrospective cohort of 8672 post-MI patients 18 years old and above. We assessed the strength of the association between the different adherence measures and the composite clinical outcome using multivariable Cox models. We compared the predictive capacity of each adherence definition model to one that did not contain adherence by computing the change in C-statistics and the continuous net reclassification improvement indices (NRIs).
Results: Adherence was associated with clinical outcome reductions, with hazard ratios ranging from 0.48 (95% CI, 0.27-0.85) to 0.81 (95% CI, 0.67-0.99). None of the adherence measures, including the HEDIS definition, significantly changed the C-statistic relative to a model that did not include adherence. However, the short-term adherence measure (having 72 d covered during the first 90 d postdischarge) showed a large change in NRI (correctly reclassifying 12% of cases and 16% of noncases; NRI: 28%; 95% CI, 22%-38%), although did not significantly differ from the change in NRI with the HEDIS measure.
Conclusions: We identified an adherence measure that showed a predictive ability as good as that of the HEDIS definition to measure β-blocker use after MI, halving the time of assessment required, and thus, allowing for the implementation of quality improvement interventions in a more timely manner.