Recently, insurers have become more aggressive about implementing a cost-saving strategy they have always employed behind the scenes: provider profiling. Now, profiling is being used to steer patients away from physicians whom insurance companies deem to charge more than average for their services or provide low-quality care. In extreme cases, health plans have used provider profiling to discharge physicians from their rosters. To a health plan, provider profiling appears to be a reasonable strategy for controlling costs. A closer look, however, shows the flaws in this approach.
THE COST OF CARE
Health plans obtain information about physicians' treatment patterns from informatics companies. These agencies collect claims data (ie, CPT codes and ICD-9 diagnoses) from insurers across the country and use proprietary "grouper" software to sort the information into episode treatment groups (ETGs). Informatics agencies then calculate how much it costs physicians to complete a discrete episode of care for a particular condition (see Defining an Episode of Care) and use this information to identify an average cost for each ETG.
When profiling physicians, insurance companies compare an individual doctor's cost per ETG with the average value supplied by the informatics company. This information is used to calculate doctors' efficiency scores. Doctors are also compared with their peers. Because informatics companies do not distinguish between comprehensive ophthalmologists and subspecialists, they incorrectly compare physicians with greatly differing practices. Subspecialists are more likely to have higher costs of care due to the complex nature of the patients referred to their practices. The situation becomes egregious if patients begin their care with a comprehensive ophthalmologist at the beginning of the year and are later referred to a subspecialist for surgery. The insurer assigns all the costs incurred by the comprehensive provider to the subspecialist, thereby further increasing the specialist's average cost of care.
Ingenix (Eden Prairie, MN), a wholly owned subsidiary of United Health Group (Minneapolis, MN) currently controls 90% of the informatics market. To date, Ingenix's database includes information solely from commercial insurance providers and only about patients who are younger than
65 years. Data on older patients are lacking, because the Centers for Medicare & Medicaid Services (CMS) have been wary of sharing their statistics with informatics agencies.
RELEVANCE TO OPHTHALMOLOGY
An analysis of the role of informatics in rating physicians' efficiency raises several questions. Is the average cost of care synonymous with its appropriateness or its effectiveness? Are insurers using risk adjustment to identify patients with comorbid conditions or more severe disease that will require more expensive care? Does the process always assign costs to the correct provider?
For ophthalmologists, the answer to these questions is definitely no. First, no valid evidence supports equating the average cost of care with appropriate treatment. In fact, a recent review of glaucoma cases (identified by claims data) showed that physicians did not provide the level of care recommended by the AAO's Preferred Practice Patterns.1
Second, the informatics companies do not have a fully developed methodology for assessing the efficiency of ophthalmic care. Until a few months ago, Ingenix's proprietary grouper technology was a carefully guarded secret. The company has since released part of its methodology in an effort to be more transparent. A review of the software's mapping architecture for ophthalmology reveals that, except for diabetic retinopathy, it does not use levels of severity to differentiate between more and less complex cases. To the software, all cases of primary open-angle glaucoma (ICD-9 365.11) are the same. In addition, the methodologies for ophthalmic analyses do not use comorbid conditions to identify patients who are likely to require more expensive care. As Lee and Blumberg pointed out in the March 2008 Archives of Ophthalmology, accurate diagnosis and staging of severity are critical to the understanding of the quality and measurements of outcomes of care.2
Third, an unpublished review of limited data provided by insurers to ophthalmologists who had been profiled in Massachusetts revealed many instances in which ophthalmologists were assigned costs for unrelated procedures and diagnoses such mammograms, hematologic malignancies, systemic diabetes, and manifestations of rheumatoid or thyroid disease. Analysts for the informatics companies acknowledge that data are occasionally misassigned across all medical specialties, but they assert that these errors do not materially affect individual providers' efficiency scores. The latter point is difficult to confirm, however, because physicians are not allowed to scrutinize in detail or to correct errors in their personal data. Even more disturbing are mistakes propagated in Ingenix's training materials. For example, the Ingenix module incorporates codes for glaucoma testing and diagnosis into an ETG for cataract surgery.
The methods by which health plans assess the quality of care appear to be even more haphazard than those used to calculate costs. In fact, how insurers use their claims data to determine the quality of care is unclear. Some health plans use their own "black box" information, whereas others obtain data from contractors such as Mercer Human Resources, Inc., and Resolution Health, Inc. (RHI). Many times, these companies tell a profiled physician that they had too few episodes for which quality was known to base their analysis on these data. In these cases, the physician's profiling score is determined by cost alone.
Recently, some commercial insurers began using measurements similar to those employed by the CMS' Physician's Quality Reporting Initiative (PQRI) to assess the quality of care. Whereas the PQRI program allows physicians to use a specific code to indicate that they have completed a quality measure, commercial plans are attempting to use claims data to determine if physicians have satisfied quality measures. For example, RHI cited a return to the OR within 30 days of cataract surgery as an appropriate quality measure. A similar PQRI measure proposed by the AAO excludes multiple patients with high-risk characteristics such as lens subluxation or pseudoexfoliation. RHI's adaptation of the quality measure, however, does not allow for excluding high-risk patients from analysis. Under this system, surgeons who want to care for higher-risk patients could be unfairly labeled as providing poor-quality care.
EFFECT ON POLICY
Informatics companies share their analyses of claims data only with insurers. Doctors typically do not know or cannot decipher how health care plans apply the information provided by informatics agencies. To complicate matters further, these data are used differently in various states. For example, in Massachusetts, the commission charged with choosing health plans for state employees indicated that it would only contract with insurance companies that implemented a tiered plan for their insured members. Under this system, the participating health plans provide patients with a list of physicians and their associated tiers. Patients who choose physicians from a higher tier have higher copays than those who pick a doctor from a lower tier. It is virtually impossible for any specialist (ophthalmologist, rheumatologist, orthopedic surgeon, etc.) to be assigned to the lowest tier, which is reserved for low-cost primary care physicians.
Unfortunately, the information that patients receive about the tiering process is confusing and often misleading; it implies that physicians in the higher tiers with higher copays provide a lower quality of care. Many specialists, however, were placed in higher tiers because their profiling scores were based solely on cost. The rationale behind such decisions is not always apparent to patients.
In addition, many patients mistakenly believe that higher copays represent a larger overall payment to the physician. They may not realize the insurance company actually covers proportionally less of the full payment and is using the higher copay to shift some of the cost onto patients.
Since Massachusetts instituted a tiered system, the head of the Group Insurance Commission has recommended the policy's cost-saving benefits to other states. The Massachusetts Medical Society, along with five physicians, has filed suit against the Group Insurance Commission and two participating health plans, with the statement that "The [tiered program] lacks a valid and accurate methodology to properly evaluate physicians and therefore harms physicians and misleads patients."3
In other states, health plans have used analyses of claims data to remove specialists from their plans (see Appealing Changes in Policy), thereby leaving patients without adequate choices for specialized care. When insurers create networks of lower-cost physicians and promote those networks to customers as providing a higher quality of care at lower cost, they often omit specialty surgeons from their panels. This limitation of choices then forces patients to pay higher out-of-network prices for specialized care.
In reality, physicians who routinely treat difficult cases will always spend more per episode than those who offer more general services. One can argue that the treatment of more complex cases by a specialist or subspecialist is often the most cost-effective way of providing high-quality care. The profiling/tiering system, however, encourages patients to stay with the more general practitioner, and patients may avoid subspecialty care due to the higher out-of-pocket cost.
CONCLUSION
The methodologies that insurance companies currently use to analyze claims data are significantly flawed: they do not differentiate between patients with mild and severe disease or account for the higher costs inherent in treating patients with comorbidities. To be effective, a cost analysis program must be based on a sufficient volume of episodes, must compare subspecialists with each other rather than with comprehensive ophthalmologists, and should use only carefully extracted data to compare surgeons with medical ophthalmologists. Until informatics and insurance companies institute these controls, physicians and patients will continue to be casualties of provider profiling.
Cynthia Mattox, MD, is the vice chair of ophthalmology at the Tufts University School of Medicine and is the director of the Glaucoma and Cataract Service at the New England Eye Center, both located in Boston. She acknowledged no financial interest in the products or companies mentioned herein. Dr. Mattox may be reached at (617) 636-8108; cmattox@tuftsmedicalcenter.org.
