Local and regional variations in healthcare have been well documented and raise important issues about quality of care in the system.¹ A payer's coverage policies are intended to define whether a particular service or technology will be offered and, if so, set forth the conditions of use. There is growing interest among payers, providers, and policy makers about the use of scientific and economic evidence in healthcare coverage decision making to encourage more appropriate use and reduce variation.^2,3

Managed care organizations have a strong interest in understanding practice variations and encouraging appropriate use to improve outcomes and reduce costs. They have an additional interest in regional variation within the Medicare program because the calculation of the Medicare Advantage premium is tied to fee-for-service adjusted average per capita costs. Regional variation strongly influences the amount of a Medicare Advantage premium. Because Medicare Advantage plans must follow the coverage policy in the county where a beneficiary resides, plans offering services over a large geographic region or nationally face operational challenges when Medicare coverage policies differ. Moreover, recently enacted Medicare provisions are moving the program toward new regional models, raising additional concerns about the implications of policy variations and practice patterns on participating health plans.⁴

Although Tunis⁵ has recently described efforts to enhance Medicare's national coverage process, most Medicare coverage decisions are decentralized. Medicare relies on a patchwork of nearly 50 local contracting organizations (called carriers for Part B and fiscal intermediaries for Part A) that develop thousands of local medical review policies (LMRPs) applicable only within their own jurisdictions. The 1965 Medicare statute authorized local contractors to process claims as a buffer between providers and government.⁶ Contractors subsequently acquired the coverage policy making function through administrative action in the 1990s.

The Medicare Payment Advisory Commission and the US General Accounting Office have recommended elimination of Medicare's local coverage process,^7,8 and the 2003 Medicare Prescription Drug Improvement and Modernization Act directs the Centers for Medicare & Medicaid Services to work to achieve greater coverage policy consistency. However, little empirical work has been done to examine the extent of, reasons for, or consequences of variation in local coverage.

Our research team has been involved in a multiyear study of Medicare's local coverage processes. The variations in the size, resources, and stability of the diverse contracting organizations have been reported elsewhere.⁹ Significant variations in contractor productivity (number of policies each contractor posted), use of evidence (based on evidence cites in LMRPs), and effective dates of policies among local contractors have also been found.¹⁰ These findings document variation across a wide range of variables.

There has been no systematic analysis of variation in policy content, to our knowledge. For example, while we know when local policies covering deep brain stimulation (DBS) were issued, how those policies differ in specific detail one from another has not been analyzed. What can the analysis tell us about variation that would inform payers, providers, and policy makers? To answer these questions, we developed 6 case studies that reveal patterns of policy variation with important implications for Medicare.

CASE STUDIES

Selection Methods

Medicare carriers are required to post LMRPs on the Centers for Medicare & Medicaid Services Web site with prescribed data fields.¹¹ We downloaded all LMRPs on May 31, 2001. We observed that all LMRPs are not the same. We identified 3 types of LMRPs: "new technology" (NT), "technology extension"(TE), and "utilization management"(UM) of widely diffused interventions, defined as follows: NT policies provide guidance for, and limitations on, the use of new clinical interventions; TE policies focus on new uses of procedures or technologies already covered for other uses; and UM policies focus on widely diffused procedures to avoid misuse or overuse. Two physician consultants coded all policies in our database. More than 85% of LMRPs focus on widely diffused technologies (UM); the rest (< 15%) evaluate new technologies (NT) or extensions of technologies (TE). A complete discussion of the coding methods has been published previously.¹⁰

For our case studies, we established the following criteria: (1) We chose only policies from among 48 carrier files (n = 5213) because carriers are more active policy makers than fiscal intermediaries and their regions are more specifically drawn. A complete discussion of the complex structure of Medicare contracting organizations has been previously published.⁹ (2) We selected only cases in which our physician consultants agreed on the coding by type. (3) To enable meaningful comparisons, we selected only case studies with at least 20 posted policies.

Using these criteria, we selected 6 case studies from a pool of 80 cases that met our criteria. We selected 2 case studies from each of our 3 policy types, 1 diagnostic and 1 treatment or procedure in each type. The 6 case studies represent a detailed analysis of 195 separate coverage policies. Although additional case studies might confirm or challenge our findings, the clear patterns we found suggest that our results will be replicated with further research. The case descriptions are drawn from posted policies on the Centers for Medicare & Medicaid Services Web site.¹¹

Deep Brain Stimulation (NT, Treatment). Deep brain stimulation is a neurosurgical procedure that uses electrical stimulation of subcortical structures (the thalamus or the basal ganglia) to control tremors.