CARVE OUTS AND COST SHIFTING

Harvard University (Medical School), Boston MA

Investigators

Linked publications & trials

Abstract

Description: The proposed research will examine market forces that lead plans to carve-out some element of care. Carve-outs can be found in cancer, mental health, substance abuse, and some radiological services. The study will rely on three different data sources. The first is information from plans in the AAHP annual industry survey, the second is data from 10 employers, while the third is a natural experiment with a carve-out in Puerto Rico. Using the contracting, financial, and clinical parts of the AAHP data, they will examine decisions to carve out MH/SA, pharmaceutical, cardiac, cancer, and diabetes care. The list of factors that will influence the choice are itemized on page 223. Some apply to the plan and some to the market area. "Independent" variables will also include plan type, type of payment (capitation), as well as the number of providers of the appropriate type. The second data set is for 10 employers who have a mental health care carve-out with UBH, but who allow choices of other plans for non-MH/SA services. They will collect additional information on primary care provider payment arrangements. The sample will consist of patients with a diagnosis of depression. The dependent measures will be MH/SA, medical, and Rx spending. The major independent variables are PCP payment structure, demographics, plan design, and employer. They will also use data from Medstat on plans that changed their MH/SA arrangement during this period, with some non-changing plans as controls. The dependent measure will be overall Rx spending, as well as for specific drugs used in treatment of depression. This will be done as a pre-post design with a contemporaneous control group. The last study will be of a natural experiment in a MH/SA carve out in Puerto Rico. During the mid nineties, a carve out was introduced in two thirds of the state. Data come from three rounds of a community survey conducted in the state on a sample of about 2700 adults; there was an average of about 300 MH/SA users in each round. The analysis is a pre-post comparison with a contemporaneous control group. The dependent measures are visits to the general medical sector and to specialty mental health. They will use difference-in-difference approaches, as well as negative binomial regression models.