Cost-sharing is a commonly used mechanism to control prescription drug costs; there is, however, a lack of information on how drug cost-sharing affects clinical outcomes and total medical costs, which is particularly concerning given that the most extreme version of cost-sharing (no coverage) is associated with poor health, and that poor drug adherence may lead to higher downstream medical costs. To address these issues, we propose to evaluate the effects of the level of drug cost-sharing on patient clinical outcomes and total direct medical costs between 2000-2005, using a quasi-experimental pre-post design with concurrent controls, within the Kaiser Permanente (KP) integrated health delivery system. In this natural experiment, nearly half of the 330,000 members age 65 years and older experienced a new pharmacy benefit cap on January 1, 2001; and 70% of these same members experiencing a new multi-der drug co-payment during the following year (1/1/2002). The level of cost-sharing will be the main predictor, with co-payment levels in effect until patients exceed their benefit cap if any, and """"""""full costs"""""""" applicable after patients exceed their cap. The main outcome measures are emergency department (ED) visits, hospitalizations, deaths, and total annual direct medical costs, i.e. the sum of all outpatient, inpatient, and pharmacy costs (including patient out-of-pocket drug costs). We will investigate these outcomes by the level of cost-sharing within the Overall Population (age 65 years and older on 1/1/2000), and within select Vulnerable Populations, i.e. patients with low socio-economic status (SES), existing chronic diseases, existing high drug use, or those who are new to the health system. We will use a proportional hazard model to test the hypothesis that higher levels of cost-sharing are related to higher rates of adverse clinical outcomes, and a two-part model to test the hypothesis that higher levels of cost-sharing are related to lower total direct medical costs. We will have the ability to detect even small changes in our outcomes, e.g. 80% power to detect a difference of one death/1,000 person-years. We will make adjustments for relevant patient and organizational factors, such as measures of SES, case-mix, physician, and medical center. These factors may influence drug use independently or modify, the association of cost-sharing level and patients' decisions to use prescription drugs. In short, there is a tremendous need to understand the safety and economic effects of cost-sharing as its use increases, in other words primum non nocere. This study will be the first to address these issues within a broad sample of Medicare beneficiaries.
