The information that many plans use needs to be open to physicians for review, and the formulas need considerable improvement
When the American Medical Association last summer sent 47 health plans across the country a letter citing the results of a series of Rand Corp. studies that called insurers’ cost profiles of physicians into question — and then suggested that health plans reevaluate their profiling programs — one of Rand’s lead researchers, Ateev Mehrotra, MD, MPH, feared what people would think. “It painted us as antiprofiling, and that’s not the case,” he says. “I fully understand the need for it; we just need to figure out better ways of doing it.” Physician cost profiling isn’t as reliable or accurate as it should be, Rand has found. Depending on specialty, for example, between 38 and 100 percent of physicians have cost profiles that are unreliable. The logic behind the idea of profiling, however, and the ever-increasing need to reduce health care costs are incentives enough to continue to work toward creating profiling systems that make sense, both for commercial health plans and for Medicare, he says. In addition to his work at Rand, Mehrotra is an internist on the faculty of the University of Pittsburgh School of Medicine. He earned a bachelor’s degree from Massachusetts Institute of Technology; a master’s in public health from the University of California–Berkeley; and a master of science from Harvard University. He earned his MD from the University of California–San Francisco and completed a residency in internal medicine and pediatrics, as well as a fellowship in internal medicine, at Harvard. He spoke recently with MANAGED CARE Editor John Marcille.
MANAGED CARE: The Patient Protection and Affordable Care Act calls for Medicare to institute physician cost profiling next year. How is that effort progressing?
ATEEV MEHROTRA, MD, MPH: When we are talking about physician profiling, people think of two areas — quality and cost. To date, most of the effort has been on the quality side, with the implementation of new codes and so forth so that Medicare can start encouraging physicians to do quality profiling. The way that Medicare has done it, which seems to be very intelligent so far, has been to first create incentives for the provider to do the quality profiling, rather than pay for reporting. The next step is to increase the size of those incentives so that it becomes detrimental not to do the reporting — for example, you don’t get your increase in your payment — and then move to pay for public reporting and then to pay for performance. That’s the arc we’ve seen on the hospital side, and it appears that will be the arc that Medicare is pushing toward on the physician side.
MC: Will cost profiling be ready by 2012?
MEHROTRA: Medicare has put a lot of resources in place to start moving in that direction, but we don’t have enough information right now to see whether it is going to be able to be implemented in the timeline that they are hoping for based on the legislation. There are so many nuances to be addressed. Our research and others’ research have highlighted how complex this is from a methodological perspective.
MC: Some commercial health plans have been doing physician cost profiling for a decade or more, yet your research has found quite a few problems with it. Does it work?
MEHROTRA: We have concerns about patients’ response to profiles and about the reliability and validity of profiling. We also don’t have the evidence to show that cost profiling saves money. Still, there’s a lot of hope. Health plans are desperate to find interventions for saving money, and it is possible that this could work.
MC: What are your concerns about patients’ response?
MEHROTRA: We need to ask, are patients responding to these systems as we would expect? These profiles are often associated with consumer-directed health care, and plans often use cost and quality profiles to place physicians in tiers so that one physician is labeled as high value and another is labeled as low value, with copayments that give patients a financial incentive to see the higher value physician. With the tiering of pharmaceutical plans, people did respond, but now we are not talking about switching from a branded drug to a generic drug; we are talking about switching providers, and these systems may not work as well as in pharmacy.
MC: Why not?
MEHROTRA: Consumers typically equate money with quality. If you buy a more expensive car or TV, it is usually a better car or a better TV. So if the health plan tells its members that Dr. Jones is more expensive than Dr. Smith, and that’s the only information it gives them, they might think that the more expensive doctor probably orders more tests and does more overall, so he is therefore better than the lower-cost doctor. Plans have to be very conscious of how this information is presented, as it could have the opposite effect.
MC: Are patients paying attention to the profiles?
MEHROTRA: One of the things we have to determine is whether the information we are providing to the patient is useful information or noise. Is it a true signal about the provider? So when you say that this doctor is more expensive than another doctor, is that true, or is it going to be different next year? When we measured reliability, which is a statistical measure of noise, we found that across a number of specialties, there was more measurement noise than there was signal. So we believe that at least with the current measures we reviewed, there was not enough signal to make it useful for patients.
MC: Can we say the profiles are wrong?
MEHROTRA: It’s difficult to capture whether the profiles the plans are creating for physicians are valid measures of their practice patterns. One way to look at that question is to ask, If you measured the cost profile in a different way — if you assign the cost using one rule and then you do it using another rule — how often do the profiles agree on the final answer? We found that using two different rules often gets us very different answers, and that raises concerns about validity.
MC: You’ve described the need to open the black box of physician cost profiling. What is so mysterious?
MEHROTRA: It’s hard to know exactly what’s happening out there, because every health plan seems to be doing it differently. Early on, the focus was on utilization measures. What fraction of the prescriptions that you write are generic? What is the emergency department utilization rate for your patients? That’s where plans started. Some plans are still focused there, and others have moved into episode-based cost profiles, which is where we have focused our research. Instead of just looking at the specific services, now the focus is on a larger episode cost for a condition, so that all of the costs related to diabetes or all of the costs related to an episode of pneumonia, for example, are included. More and more health plans are using commercial programs such as those developed by Ingenix and Thomson Reuters to generate these episode-based cost profiles for physicians.
MC: Could the use of commercial programs lead to more standardization?
MEHROTRA: It is hard to know. Using standard methods is advantageous for plans and providers. On the other hand, there is concern when the exact methodology is proprietary and therefore black box. I believe this is why Medicare is supporting the development of open-source programs.
MC: With all of the concerns, why are health plans still doing this and why is Medicare moving in this direction?
MEHROTRA: Payers are looking for levers to decrease health care costs without impinging on patient choice. We are hearing a lot about this new era, especially with the passage of the Affordable Care Act, where we are going to have an increased pressure to reduce costs. We need to try things. Does it have to be perfect? Many people argue that it doesn’t have to be perfect. All of us recognize that many of the costs in the U.S. health care system are driven by physicians. Physicians are the ones ordering the medications, ordering the expensive tests, or getting the patient hospitalized. So we as physicians have a lot of power, and to the degree that we can change our behavior, we can affect that. There’s a lot of money at stake.
MC: If you can find the flaws in these cost profiling systems, can health plans do the same?
MEHROTRA: We have not seen insurers doing much work to measure the reliability of the cost profiles. We hope that plans will start measuring reliability regularly so that they can better assess this issue of measurement noise. There are some complexities, but we hope that health plans build in the analytic ability that will enable them to do that.
MC: At some point will there be a standard or a best practice for cost profiling?
MEHROTRA: One of the criticisms of our own work that I am very conscious of is that we are knocking down what is being done now, but we are not offering a solution. I can tell you that there are some real concerns about the cost profiles being done, but I can’t tell you what would be a better system. It’s not enough to say the current system is not working; we hope to work on better systems for the future, because this cost problem is not going to go away, and the desire for physician cost profiles is not going to go away.
MC: What challenges have you had in investigating the current approaches?
MEHROTRA: We got the raw data from a couple of health plans in Massachusetts, and we had to create the cost profiles in a way that we understood what health plans are doing. We used the Ingenix system, and then made the choices we believe plans are making. The process really highlighted for us the challenge that it’s currently not possible to go to a health plan’s Web site and easily understand the steps they are using to create these cost profiles. That’s what we meant by opening the black box. We want health plans to disclose how this is done. Other people have also been calling for that transparency. When he was attorney general, current New York Governor Andrew Cuomo asked health plans that were doing cost profiling in that state to be transparent about the methods that they were using.
MC: Do you believe that physicians really want the information?
MEHROTRA: You can be transparent but still be complex. I don’t really know how many physicians are going to go out and try to figure this out on their own, but making the information available is important.
MC: The AMA sent out the letter about your research to health plans last summer; will physician groups continue to monitor this?
MEHROTRA: That centers on the question of whether there are going to be significant financial incentives tied to it. If they see it as a financial incentive, physician groups are going to put a lot of energy there.
MC: We started out by discussing the two types of profiling: cost and quality. What is the relationship between them?
MEHROTRA: Much of the work done to date has put cost and quality on two separate axes of overall value, with the assumption that they have little or no relationship. For example, when a health plan places providers in tiers, it will first evaluate them on quality and then on cost. So people have asked, Is there a relationship between the two? One would think that if a provider has higher quality, he could prevent complications and potentially that would lead to lower costs down the pike. Yet some of the preliminary work that we’ve done on that topic has not shown much of a relationship between the two areas. But that is still an open question that needs more work.
MC: In the profiling systems we have now, is patient severity a factor?
MEHROTRA: Current profiles try to deal with whether one physician’s patients are sicker than another’s, but we don’t know if they do it sufficiently well; I still have concerns. Another aspect of this is whether socioeconomic status is an issue. If two physicians have the same patients, but one’s patients are poorer or less educated, does it require more resources to get them the same level of care? If so, is that physician going to look high-cost even though that’s not his or her fault? On the other side, physicians who take care of a wealthier population are saying, “You haven’t seen my patients: They are more demanding, they want MRIs and other expensive tests, and therefore taking care of a wealthier population leads to increased costs.” The Affordable Care Act actually specifies that there be more work on adjustment for socioeconomic factors. We have an active project to study that question, and it’s an interesting one. You see both sides of the coin. It is the Lake Wobegon effect applied to health care. Every physician’s patient population is sicker or more expensive than average.
MC: Aren’t we supposed to be encouraging more patient decision making? Are demanding patients going to be turned away if they could hurt a physician’s cost profile?
MEHROTRA: You have to be very conscious of how providers might respond to cost profiling. You do not want to create any unintended consequences, or you at least want to minimize the unintended consequences. We obviously don’t want a system that encourages a physician to say, “Ms. Jones is really demanding or is very sick, so I’m going to refer her to the doctor down the street so it doesn’t hurt me financially.” As we create these cost profiles, we have to be conscious of how physicians will respond, and we have to think about how to create the systems so that they discourage that. Then we have to monitor for that type of behavior.
MC: You said earlier that cost profiling isn’t going away.
MEHROTRA: Increasing health care costs are going to be a bigger and bigger issue that health plan executives need to face. The need to find interventions that change that trend are therefore going to be that much more important. Even though we have a lot of concerns, physician cost profiling is not going away.
MC: Thank you.
For further reading
- Adams JL, Mehrotra A, Thomas JW, and McGlynn EA. Physician Cost Profiling — Reliability and Risk of Misclassification: Detailed Methodology and Sensitivity Analyses, Santa Monica, Calif.: Rand Corp., TR-799-DOL, 2010.
- Adams JL, Mehrotra A, Thomas JW, and McGlynn EA. Physician Cost Profiling — Reliability and Risk of Misclassification. The New England Journal of Medicine, Vol. 362, No. 11, March 18, 2010, pp. 1014–1021.
- Mehrotra A, Adams JL, Thomas JW, and McGlynn EA. The Effect of Different Attribution Rules on Individual
- Physician Cost Profiles. Annals of Internal Medicine, Vol. 152, No. 10, May 18, 2010, pp. 649–654.
- Mehrotra A, Adams JL, Thomas JW, and McGlynn EA. Cost Profiles: Should the Focus Be on Individual Physicians or Physician Groups? Health Affairs, Vol. 29, No. 8, August 2010, pp. 1532–1538.
We don’t have the evidence to show that cost profiling saves money. Still, there’s a lot of hope.
The relationship between a provider’s cost and quality is still an open question that needs more work.