ABSTRACT
BACKGROUND: Despite a growing consensus that better adherence with evidence-based medications can save payers money, assertions of cost offsets may be incomplete if they fail to consider additional drug costs and/or are biased by healthy adherer behaviors unobserved in typical medical claims-based analyses. OBJECTIVES: The objective of this study was to determine whether controlling for healthy adherer bias (HAB) materially affected estimated medical cost offsets and additional drug spending associated with higher adherence. SUBJECTS: A total of 1273 Medicare beneficiaries with diabetes enrolled in Part D plans between 2006 and 2009. RESEARCH DESIGN: Using survey and claims data from the Medicare Current Beneficiary Survey, we measured medical and drug costs associated with good and poor adherence (proportion of days covered ≥ 80% and <80%, respectively) to oral antidiabetic drugs, ACE inhibitors/ARBs, and statins over 2 years. To test for HAB, we estimated pairs of regression models, one set containing variables typically controlled for in conventional claims analysis and a second set with survey-based variables selected to capture HAB effects. RESULTS: We found consistent evidence that controlling for HAB reduces estimated savings in medical costs from better adherence, and likewise, reduces estimates of additional adherence-related drug spending. For ACE inhibitors/ARBs we estimate that controlling for HAB reduced adherence-related medical cost offsets from $6389 to $4920 per person (P<0.05). Estimates of additional adherence-related drug costs were 26% and 14% lower in HAB-controlled models (P < 0.05). CONCLUSIONS: These results buttress the economic case for action by health care payers to improve medication adherence among insured persons with chronic disease. However, given the limitations of our research design, further research on larger samples with other disease states is clearly warranted.