Delivering Better Value in U.S. Health Care: The Role of Physician Performance Measurement
With a Keynote Address by Secretary of Health and Human Services Michael Leavitt
About This Event

If we are to improve health care, we need effective ways to measure how health care providers deliver care. Speakers at this conference will assess how existing physician performance measurement tools operate and how we can improve their accuracy, reliability, transparency, and usefulness. For example, both the Government Accountability Office and the Medicare Payment Advisory Commission have recently examined how Medicare claims data might be used to measure relative resource use by individual physicians across particular "episodes of care." Such clinically equivalent evaluation periods are generally based on the entire package of services required to treat a particular illness, regardless of treatment location or duration. Those efficiency measures could then be combined with measures of the quality of care to help develop overall profiles of their performance, along with incentives to improve it and reward better care. A number of health care purchasers, such as large insurers, employer groups, and state government bodies, are increasingly measuring physician performance through promising but limited mechanisms.

Secretary of Health and Human Services Michael Leavitt will give a keynote speech on the objectives and achievements of the Bush administration’s value-driven health care initiative, highlighting how chartered "value exchanges" could evaluate and report on health care value.

Panelists will discuss the practical side of getting "from here to there" in physician performance measurement and value-driven health care, with particular attention to the difficult and complex tradeoffs ahead concerning such issues as what should be measured; how to do it; how to aggregate and pool data from multiple sources, including Medicare claims data; how to interpret and use the findings; and what level of transparency is needed to ensure that performance measures are fair, sustainable, and accurate.

Agenda
12:30 p.m.
Registration and Luncheon
1:00
Keynote Speaker:
Secretary of Health and Human Services Michael Leavitt
2:00
Presenters:
Donald Casey, M.D., Atlantic Health
Elliot Fisher, M.D., Dartmouth Institute for Health Policy and Clinical Practice
Robert Krughoff, Center for the Study of Services and Consumers’ Checkbook
Arnold Milstein, M.D., Mercer Health & Benefits
Steven Pearson, M.D., America’s Health Insurance Plans
Earl Steinberg, M.D., Resolution Health
Moderator:
Thomas P. Miller, AEI
4:30
Adjournment
Event Summary

November 2007

 

Delivering Better Value in U.S. Health Care: The Role of Physician Performance Measurement

 

 

If we are to improve health care, we need effective ways to measure how health care providers deliver care. Speakers at this conference assessed how existing physician performance measurement tools operate and how we can improve their accuracy, reliability, transparency, and usefulness. For example, both the Government Accountability Office and the Medicare Payment Advisory Commission have recently examined how Medicare claims data might be used to measure relative resource use by individual physicians across particular "episodes of care." Such clinically equivalent evaluation periods are generally based on the entire package of services required to treat a particular illness, regardless of treatment location or duration. Those efficiency measures could then be combined with measures of the quality of care to help develop overall profiles of their performance, along with incentives to improve it and reward better care. A number of health care purchasers, such as large insurers, employer groups, and state government bodies, are increasingly measuring physician performance through promising but limited mechanisms.

At an AEI event on November 5, 2007, Secretary of Health and Human Services Michael Leavitt gave a keynote speech on the objectives and achievements of the Bush administration's value-driven health care initiative, highlighting how chartered "value exchanges" could evaluate and report on health care value. Panelists discussed the practical side of getting "from here to there" in physician performance measurement and value-driven health care, with particular attention to the difficult and complex tradeoffs ahead concerning such issues as what should be measured; how to do it; how to aggregate and pool data from multiple sources, including Medicare claims data; how to interpret and use the findings; and what level of transparency is needed to ensure that performance measures are fair, sustainable, and accurate.

Michael Leavitt
Secretary of Health and Human Services

There is currently no health care "system." What exists is a health care sector in the absence of a system. Economic systems exist everywhere--a cell phone represents a piece of an economic system. Different vendors operate on a single platform in order to optimize the value and allow for competition on price and quality. Health care is different. For example, if a patient calls their doctor to ask the cost of a colonoscopy, the doctor cannot tell them. Instead, the patient must get the colonoscopy, and then the doctor negotiates with the patient’s insurer to determine how much health care the patient has consumed.

It would be better if there were pre-made episode of care bid sheets for cost comparison; if there were standard quality comparisons; and if doctors, hospitals, and insurers were all motivated to provide the best care at the lowest cost. This would allow for competition across a cohesive system based on value. Creating this system would require electronic medical records, quality measures, standard price units in the forms of episodes of care, and incentives to provide high quality at a low cost.

Arnold Milstein
Mercer Health & Benefits

The essential question on performance rating is: how close to perfect is good enough for physician performance measures to be used by health care professionals? Physician performance measures have been used for a number of years by HMO and PPO insurance providers, but never before has this kind of information been publicly distributed. Higher quality at lower cost is necessary, and the way to get there is through transparency on quality. Currently, however, there is hesitation over whether performance measures are good enough. Some advocate proceeding with less than perfect measures.

The Institute of Medicine has stated that 30-40 percent of health care spending is wasted--that is, it does not do anything to improve anyone’s health or satisfaction. The institute also estimates that a not insignificant portion of this is due to differences in cost per unit of service that seemingly have no bearing on quality. Health care spending outgrows the GDP by around 2.5 percent annually. Given these statistics, it is imperative that the health care system be motivated to create better and lower-cost care so that every year the efficiency gain more than offsets any increase in cost over the American income.

There are a few arguments against moving forward. One is that ratings are costly. This is true, but ratings allow for better value judgments that will ultimately lower costs. There is also friction around individual physician measurement. People argue that perhaps physician groups, as opposed to individuals, should be analyzed. Research suggests that groups with individually identifiable members are more productive than groups with anonymous members, which provides a strong reason to proceed with ratings of individual physician performance. When it comes to outcomes of patients with chronic conditions who see multiple doctors, each physician who has seen a patient for more than a year should be accountable for the outcomes. It cannot be true that because a patient sees multiple physicians, none of them is responsible for the outcome.

Will there be benefits from the use of imperfect measures? Judith Hibbard, Jean Stockard, and Martin Tusler published a study in Health Affairs in 2005 entitled "Does Publicizing Hospital Performance Stimulate Quality Improvement Efforts," which was based on the distribution of what is generally considered a very imperfect measurement: hospital level obstetrical complication rates. In Wisconsin, one part of the state was given no information on the obstetrical complication rates, which is the status quo. In another region, this information was distributed to health care professionals, including physicians. In a third region, the information was widely distributed to the public. Two years later, the first region showed no overall improvement in quality of care, the second region showed moderate overall improvement, while the region where information was publicly distributed showed very significant improvement. This study provides strong evidence that the public can benefit from less than perfect quality measures.

The current round of consensus-based physician measurements cannot be tracked with claims data and thus cannot be evaluated efficiently. Consumer leaders should decide how to measure physician performance. Attribution paralysis around single patient/multiple physician scenarios needs to stop, along with the toleration of blindness to quality. And hospitals must be required to submit lab value data with claims in order to lower the expense of retrieving data from hospital records. 

Elliot Fisher
Dartmouth Institute for Health Policy and Clinical Practice

A big problem with health care today is the huge discrepancy in spending across the nation that is not accounted for by differences in illness or prices, but rather by how much care is provided. Regions with high health care expenditure tend to have more beds than doctors; more episodes of care; more doctors assigned to each patient; and patients seeing more specialists, resulting in more imaging. Ironically, regions with high health care expenditure also tend to have worse health outcomes.

The differences in regional health care expenditure are driven by local capacity and clinical culture. Only about 10 percent of the cost differential is attributable to malpractice, and none of the difference is explained by patient preferences. What exists is a payment system that ensures that doctors stay busy. In areas with more doctors, there tend to be more episodes of care, and the intensity of a particular physician practice is strongly correlated with spending.

The role of transparency is to create better value judgments on health care expenditure. The current model is that health care outcomes stem from one-on-one encounters with physicians, meaning that each time a different doctor adds to a patient file, there has been a move towards better health. This leads to an inadequate continuity of care, creating an environment with no incentive for physician interaction. There needs to be a change in this model to either attribute health outcomes to all physicians involved in the care of a particular patient or foster shared accountability between physicians for patient care. The model also needs to look at the care of populations as opposed to individual care.

Rapid development of performance measurement is needed but can occur simultaneously with the use of current measures. Waiting for consensus-based measures is not an option, as they are slow to come and ignore the organizational context of care. Measures must be taken to fix the broken payment system of medicine, establish physician group oriented outcome and cost measurement, provide incentives for high-value care, and remove the public from the mindset that more care is synonymous with better care.

Robert Krughoff
Center for the Study of Services and Consumers' Checkbook

There are certain principles to be adopted that will be important in establishing health care measures. To the extent consistent with patient privacy, as much useful data as possible should be made available for measuring and reporting on providers at all levels. Inventiveness in the development of statistical measures should be encouraged, as well as dialogue about the strength and weaknesses of current measures. Users and publishers of measures who do not meet standards or principles on issues such as disclosure of analysis and data used should be discredited.

The government should work to gather as much useful data as possible from various sources such as Medicare Advantage plans, prescription drug claim systems, and numerical lab data from hospitals. It should assemble data into useful forms, linking claims, hospital, pharmacy, and lab data. It should take all feasible steps to patient deidentify data in ways that will preserve the ability to link data across patients and episodes. It should distribute patient deidentified data at cost to companies that will adhere to set reporting principles. And it should make data that cannot be patient deidentified available to researchers and other entities that meet Institutional Review Board and other privacy board standards, even if the sole purpose of research is to produce measures of individual providers.

There is an obstacle to achieving Medicare data availability at the physician level, and that is the Center for Medicare and Medicaid Service's (CMS) assertion that it is barred from releasing data with physician identifiers by a 1979 decision by a U.S. district court in Florida. Consumers’ Checkbook brought a Freedom of Information Act case to the U.S. district court in the District of Columbia seeking information from Medicare part B claims data with physician identifiers. The purpose of the lawsuit was to nullify CMS's argument that it is barred from releasing physician identified data. On August 21, 2007, the court ordered the Department of Health and Human Services (HHS) to release the data. On October 19, the government filed a notice that it would appeal. This is ironic, given the current HHS pledge of transparency. Various groups have supported the lawsuit, and amicus briefs may be filed in support. CMS will have certain provider organizations such as the American Medical Association (AMA) filing amicus briefs for their side. The lawsuit has certain limits in that it has not asked for any patient identified data. If Consumers’ Checkbook wins, while CMS must provide data on counts of procedures by doctors, it may still be impossible to get data that would allow one to link multiple episodes of a single patient over time.

Why is it important to push for the release of raw data rather than trusting HHS for analysis? First, having more than one source for measures will lead to the development of the most valid measures. Second, the gears of the government tend to grind slowly. Third, HHS will be under political pressure from provider organizations such as the AMA to draw as few distinctions between providers as possible.

Donald Casey
Atlantic Health

Effective quality measurements should be evidence based. The goal should not be evaluation alone, but rather using evaluation to improve care. Measurements should also be actionable--that is, areas being measured need to be under physician control. Administrative data is not always actionable data, which is a major problem in using such data. Measurement should also be reliable and valid, capturing the most meaningful aspects of care. Measures should be feasible to evaluate. And measures should be timely, as past data becomes increasingly irrelevant over time.

Some cost versus quality studies focusing on individual performance show zero correlation between cost and quality at the physician level, rather than showing that higher quality costs less, which is what most expected to see. There are other studies that, rather than looking at individual physicians, look at composite measures on health systems. These studies did find that higher quality achieves lower cost and that higher quality systems had lower readmission rates and better health outcomes. It is more useful to focus on aggregate data for systems than to focus on individual physician measurement because such analysis yields useful results and implications.

Focus needs to be placed on quality improvement as opposed to accountability, and measures need to be the right measures. What is needed is for those right measures to become easy to evaluate through creating an infrastructure that will allow for efficient processing of essential information. It is claimed that measure analysis will be expensive, but the overall savings they will provide in terms of efficiency will more than offset the cost of extraction. Health information technology seems unmotivated by market forces to create interoperable data standards. Unless the federal government steps in and forces this to happen, gathering the right data will remain extremely difficult.

Earl Steinberg
Resolution Health

Performance measurement starts with evidence. Evidence comes from either clinical trials or expert opinion. This evidence then becomes incorporated into clinical practice guidelines and standards, the difference being that there is less tolerance for deviation from a standard than a guideline. Performance measures are then based on these standards and guidelines.

Creating a system under which to collect data for measurement is no trivial exercise. There are problems with attribution, programming, and expense. There are many potential sources of data, but the two that tend to be focused on are claims data and chart abstraction. Each of these has its own advantages and problems. Chart abstraction can be extremely expensive and subject to methodological error, as people must do the abstraction. Claims data tends to be a very practical but limited approach. It is interesting that some studies have used claims data and chart abstraction for the same population and found that performance at the physician level can look better through the analysis of claims data than through the analysis of chart abstraction.

Scoring and statistical issues also arise around performance measurement. Scores can be aggregated for doctors, groups, or performance by groups in specific areas of care. Questions arise, such as: Are some measures better than others? If so, how much weight should be applied to each measure? What is the minimum number of observations necessary for a physician to be measured accurately? What difference in a score represents not only statistical significance, but also clinical significance?

Steven Pearson
America's Health Insurance Plans

Three issues have been looming over the health care system for some time: escalating costs, admitted and acknowledged quality gaps, and variations in care. To deal with these problems, it was first thought that by involving health care providers such as physicians' professional organizations in the development of clinical guidelines, there would be an efficient and high-quality health care system. This has been done to a large extent, and yet there is still room for much improvement. Now, ideas are being considered such as sharing quality measurement and cost with consumers and using them as tools to drive down costs, and sharing quality measurements with health plan providers so they might retool to create a pay-for-performance system and tier physicians into performance groups.

Measurement and data are keys to these new approaches. Quality measures are being developed to analyze the structure of care, the process of care, and the outcomes of care. Some have even looked to measure patient satisfaction with care. But the major difficulty with developing measures is framing them in such a way so that they are both actionable and useful for physicians. Health plans and purchasers have been demanding information to evaluate different measures with little or no coordination on what it is exactly that determines the quality of care. This has been burdening physicians for years, and for this reason, having the CMS data available for everyone to access is going to be a good thing for patients and families.

Collaborative efforts are needed to pull together as much data and information as possible when it comes to evaluation. America’s Health Insurance Plans (AHIP) is working through grants to create an overarching quality steering committee to try to create collaboration and consistency on performance measures. AHIP is also providing funds to create an infrastructure for the identification of doctors and health care providing organizations so that data can be accurately pooled across health insurance plans. Once this is in place, it will be important for physicians to see how they differ from each other and also for consumers to see how physicians differ from one another.

AEI research assistant Walton Dumas prepared this summary.

View complete summary.
Also Visit
AEIdeas Blog The American Magazine

What's new on AEI

image The Census Bureau and Obamacare: Dumb decision? Yes. Conspiracy? No.
image A 'three-state solution' for Middle East peace
image Give the CBO long-range tools
image The coming collapse of India's communists
AEI Participants

 

Thomas P.
Miller
  • Thomas Miller is a former senior health economist for the Joint Economic Committee (JEC). He studies health care policy and regulation. A former trial attorney, journalist, and sports broadcaster, Mr. Miller is the co-author of Why ObamaCare Is Wrong For America (HarperCollins 2011) and heads AEI's "Beyond Repeal & Replace" health reform project. He has testified before Congress on issues including the uninsured, health care costs, Medicare prescription drug benefits, health insurance tax credits, genetic information, Social Security, and federal reinsurance of catastrophic events. While at the JEC, he organized a number of hearings that focused on reforms in private health care markets, such as information transparency and consumer-driven health care.
  • Phone: 202-862-5886
    Email: tmiller@aei.org
  • Assistant Info

    Name: Neil McCray
    Phone: 202-862-5826
    Email: Neil.McCray@aei.org
AEI on Facebook