Efforts to steer patients to lower-cost physicians may be based on misleading rankings, study finds

The first major assessment of physician cost profiling found that about one-fourth of the 13,788 physicians studied would be misclassified under the system of cost ranking commonly used by insurance plans, according to findings published in the March 18 edition of the New England Journal of Medicine.

Cost profiling increasingly is used to decide which physicians will be included in health plan networks, and patients often are provided financial incentives to use physicians that have lower cost profile scores.

"Our findings raise questions about the utility of cost profiling tools for high-stakes activities such as tiered health plans and the likelihood that wide use of these strategies will reduce health care spending," said John L. Adams, the study's lead author and a senior statistician at RAND, a nonprofit research organization. "Consumers, physicians and those who pay for health care are all at risk of being misled by the results from these tools."

Studying 28 physician specialties in detail, researchers found that only about 40 percent of physicians had cost profile scores that were at least 70 percent reliable -- a common threshold for reliability -- and fewer than 10 percent of physicians had cost profiles that were at least 90 percent reliable.

Among physicians in a hypothetical two-tiered insurance plan, nearly 40 percent of internists and nearly two-thirds of vascular surgeons labeled as lower cost were not lower cost, according to the RAND study. Physicians in surgical specialties, in particular, appear to have low reliability cost profile scores, while dermatologists' cost profile scores were the most reliable.

Health purchasers have focused cost-saving efforts on physicians because they write the orders that can drive increased heath care spending. Health plans are limiting the number of physicians who receive in-network contracts, offering patients lower co-payments to see preferred doctors and paying bonuses to doctors who keep spending down.

Researchers from RAND Health analyzed information from insurance claims for 2004 and 2005 from four health plans in Massachusetts that provide coverage to about 80 percent of the non-elderly with private insurance. They used commercially available software to examine the costs of treating episodes of common illnesses such as diabetes and heart attack, assigning each episode of care to a physician and creating a cost profile for each physician based on all similar episodes of care.

Using statistical tools, researchers evaluated the reliability of physician cost scores by considering factors such as the number and types of patients physicians treated. The results show that the reliability of cost-profiling scores were unacceptably low for physicians in most of the specialty groups.

Researchers also examined how reliability scores might change across several different scenarios, such as requiring at least 30 episodes of treatment in order to create a profile and different methods for assigning episodes to physicians. While some scenarios modestly increased reliability, the results still fell short, according to the study.

"These ranking systems may be useful for some purposes, but they are not reliable enough at this point to make decisions about encouraging patients to see certain providers or excluding some doctors from insurance networks," Adams said. "Much work remains to be done to improve these systems before they are used for high-stakes activities."

He said the current systems may be useful for efforts such as warning physicians that their treatment methods appear to cost more than those used by their peers and urging them to reexamine their practice styles.

While cost profiling shows promise as a strategy to reduce health costs, it cannot be successful until more-robust tools are developed to use claims data and other information to create reliable cost profiles for physicians.

Funding for the study was provided by the U.S. Department of Labor. Other authors of the study are Elizabeth A. McGlynn of RAND, Dr. Ateev Mehrotra of RAND and the University of Pittsburgh School of Medicine, and J. William Thomas of the University of Southern Maine.