Impact of prescription coverage on hospital and physician costs: a case study of medicare beneficiaries with chronic obstructive pulmonary disease

Abstract

Background: It is widely believed that appropriate use of prescription medicines can reduce avoidable hospitalizations and more expensive nonpharmacologic therapies, but identifying such cost offsets in operational programs is elusive. Any possible impact would be most apparent in patients with medication-sensitive disease conditions, such as chronic obstructive pulmonary disease (COPD). Objective: The goals of this study were to develop an observational study design appropriate for estimatingpotential cost savings in the US Medicare budget as a result of extending drug coverage to persons with particular chronic diseases and to apply these study methods, in an exploratory analysis, to a sample of Medicare beneficiaries with COPD. Methods: Spending for drugs, hospitalizations, and physician services was compared for COPD patients withand without prescription coverage using data from the 1999 and 2000 US Medicare Current Beneficiary Survey. To control for channeling bias, multivariate matching on observable variables was combined with tests for missing variable bias. The matching algorithm used propensity score weighting to ensure comparability between the 2 groups on all observed characteristics at baseline. Results: Our sample comprised 462 beneficiaries with prevalent COPD in the year 2000: 384 (83.1%) hadprescription coverage the entire year and 78 (16.9%) had no coverage. After adjustment, drug coverage was associated with 61% higher spending on medications and 29% lower spending on physician services (both, P < 0.05). Hospital costs appeared slightly lower for those with drug benefits, but the difference was not statistically significant. No statistically significant effects were found for services specific to COPD. However, effect sizes were large even for nonsignificant findings. Conclusions: Although this analysis did not establish a strong causal link between drug benefits and lower costs, 11 of our 12 comparisons had signs consistent with the cost-offset hypothesis.