Panel Discussion: Knowing What You Are Paying For: Benefit Design in a Changing Market

Chair: John Hsu, M.D., M.B.A, MSCE, Physician Scientist and Internist, Division of Research (DOR) for Kaiser Permanente Northern California.

Panelists: John Bertko, F.S.A., M.A.A.A., Vice President and Chief Actuary, Humana Inc.

Michael Chernew, Ph.D., Professor in the Department of Health Care Policy, Harvard Medical School.

Jay Krueger, Vice President of Benefit Analytics, Washington Mutual.

Introduction: The health insurance market is changing rapidly and there are concerns about rising costs, quality, and value, and the financial burden and distribution of costs among different groups. The question is: Who should pay?

Recently, there has been a rise in market-based solutions. There is innovation from payers, plans, and policy makers, and each is critical in determining which plans are on the market. Among the main questions for this panel are: (1) Is there really a fundamental shift in the market toward individual consumers bearing more risk? (2) If so, what information do stakeholders need and want? The panel asks these questions keeping in mind that benefit design decisions that are made every year and the benefit design cycle repeats every year. In theory, these are not one-time decisions but decisions that are made repeatedly. The panel looks at a range of different benefit designs including traditional cost sharing (deductibles and copayments), coverage limitations, formularies, networks, uncovered services, and carve outs. The panel also looks at health savings accounts and the possibility of using nonfinancial incentives to influence behavior. Many of these strategies can be targeted at a plan level or over large populations or subpopulations. It is also possible to apply the mechanisms to individuals, for example smokers; to a particular type of care; to particular indications; or based on net clinical value.

We begin with brief comments from each panelist, followed by an open discussion and audience participation.

John Bertko: I'm the chief actuary for Humana, a super regional health plan. We have been in the consumer-directed health plan (CDHP) business for five and a half years. On that basis, I'm ready to declare partial victory on the consumer-directed health care part of it. Our experience has shown that we can beat a trend in the overall insurance market by two, three, or four percentage points. This is not a complete victory, but it is a victory, nonetheless.

If you ask me why it works, I'm going to say it's like many new treatments: it works for some people and it doesn't work for others. What's the mechanism? We're not quite sure but, if you follow all the steps, it seems to work most of the time.

The second thing is that Humana is now “number two” in Medicare Part D. We are in all but four states and we have over two million members signed up for Medicare Part D and nearly a million members in Medicare Advantage. Seniors have done a remarkable job of choosing the right plans for themselves. In spite of all the confusion, many people chose low-cost plans when they need insurance. Other seniors chose higher cost plans and we're still in the initial stages of understanding the preferences of seniors.

After Part D was passed, but before anyone except the Office of the Actuary (OACT) and the Congressional Budget Office had estimates, the member premium, which is about 25 percent of the whole Part D premium, was estimated by OACT to be about $37. Last year, after the bids were in, the unweighted average of 1,400 or so bidders was $32. One of the reasons I say that seniors did a superb job of choosing is that when Secretary Leavitt [Michael Leavitt, Secretary, U.S. Department of Health and Human Services] revisited those numbers in February and then again more recently, the revised weighted average premium—that is, the average premium among people choosing United, Humana, Wellpoint, or Coventry (the first four with the most members)—was $24. So, there has been about 25–30 percent savings by seniors, their advocates, and their families. This is good news and it shows that consumerism has worked out well for seniors and for the Treasury.

The last thing—to tie together the comments for CDHPs and for Part D—is that this is not a benefit driven issue. The benefits are a necessary component, but having big deductibles and cost sharing isn't enough; much more information is needed.

Aetna, for example, announced it was going to expand a program describing actual allowed charges after discounts for physicians beyond the Cincinnati market to six or seven markets. Our company is on record, in Wisconsin and particularly in Milwaukee, regarding giving hospital prices for the most common episodes for which people might go into a hospital. You also need information on quality. These are important components of making CDHPs work.

Jay Krueger: I represent Washington Mutual (WaMu), a national bank located in Seattle. I'm responsible for contracting with our benefit vendors, measuring their performance, and tracking the finances. We are in year three of a multiyear strategy, and we've seen some remarkable successes in a short time.

About 110,000 individuals are covered by WaMu, including about 55,000 employees and 5,000 retirees. Our participants give birth to six babies a day. Our average demographic is female, 36 years old, making $36,000 a year. She watches her dollar, and her ability to find the best value are being realized in our program.

We have a clear vision and mission at WaMu, with the objective of driving consumerism, because it takes shared responsibility to maximize value. Year one of our multiyear strategy consisted of outsourcing benefits administration, year two consisted of consolidating our health plans, and year three will see the evolution of our plan design. We outsourced our benefits administration to the group that's now known as Accelerate HRO (Human Resources Outsourcing), a joint venture between EDS and Towers Perrin. We chose Accelerate because they had a unique value proposition based on what they were doing with WebMD. Specifically, they were marrying administration to the concept of delivering the right information at the right time to facilitate consumerism. Wouldn't it be great if you could log on to a website, see information on quality and cost, see who's available and when, and then make a health care decision?

In year two, we consolidated 25 national plans down to three: Kaiser, United Healthcare, and Health Net. With Kaiser and Health Net, we have developed the concept of health advocacy, where you have a health care concierge to help you understand your needs and help plug you into the right channel. Year three has been about the right plan design.

We believe in the concept of integrated disability and absence management. Fifty-five percent of our health care costs are driven by five percent of our population. Almost 90 percent of those individuals go on some level of disability. So, we're building a concept called First Day Absence Management, where as soon as someone is absent from work for longer than three days, we reach out to them and help them get involved in their care to minimize the transition to short-term and long-term disability. We are planning to launch this in January 2007. Then, in 2008, we're going to focus more on the HSA, HRA, and consumer-directed accounts.

WaMu Health is our internal portal hosted by Accelerate HRO. It's our one stop shop for information. We're using our purchasing power to force people in the direction of information integration, because we need to be able to facilitate better service and better outcomes. We're providing single sign-on through WaMu Health to all our vendors. The portal represents the data model for integration of information throughout the continuum of care.

The concierge—whether it's a 24-hour nurse line, a nurse chat, or any other form of outreach—focuses on developing a relationship with employees and driving them to the right health management program, for example, disease management, employee assistance, disability management, or others.

We also have two data warehouses in development. One is with Thompson Med Stat, where we want to understand the interactions of our employees as they go through the health care system. It will be logging web pages and activities and then assessing whether those activities produce better outcomes. We are also looking to Kaiser and United Healthcare to develop an operational data warehouse in order to facilitate the clinical aspects of care.

Last year, we were able to return 25 million dollars to WaMu and have a zero health care trend, so we believe we're making progress.

Michael Chernew: Health care costs are high and getting higher. Additionally, there are a lot of problems in the health care system in terms of quality. I'm going to talk about how we might deal with these issues through benefit design.

In today's marketplace there are several important trends. One of them is to control costs by shifting costs to enrollees, as exemplified by CDHPs, health savings accounts (HSAs), and other types of high deductible or coinsurance programs. Another is to strive to improve quality. Two popular ways to do that are disease management programs, where people with particular conditions are enrolled in programs to try to get them to manage their diseases better, and pay for performance, where financial incentives are used to get physicians to do what you think they should do.

The problem is that often these trends work at cross purposes. We looked at the copayments people were paying for pharmaceuticals inside and outside disease management programs and found that, at the same time the disease management programs were asking people to take better care of themselves, they were being charged more for doing exactly that. The disease management companies were allowed to use every method to get people to manage their chronic diseases better except money.

You see this as well in the Medicare Health Support Programs, where nonfinancial incentives are used but the programs are limited in their ability to make financial incentives congruent with nonfinancial initiatives. The problem is that price responsiveness is not closely tied to medical appropriateness. For example, among people who take statins for primary or secondary prevention of coronary artery disease, those with higher copayments stop their statins earlier. The primary and secondary prevention groups respond similarly to price even though the latter derive greater benefits from statins. Economists believe that the people who receive the smallest benefit should be the ones who drop off first and the people with the highest benefit should be the ones who stay on their medication, but this is not true in health care, where higher cost sharing reduces both appropriate and inappropriate use.

We need to use cost sharing in a clinically sensitive way. This is called value-based insurance design, or VBID. VBID could be used to coordinate cost sharing with disease management and pay for performance. For example, you might reduce the copayment for services from which people will benefit, which you identify through disease management and pay-for-performance programs. The amount people pay out-of-pocket when they go to the physician would vary based on the clinical situation, patient traits, the particular medication, and so on. My colleague, Allison Rosen, wrote a paper about how Medicare could save money if they gave away angiotensin-converting enzyme (ACE) inhibitors or angiotensin receptor blockers (ARBs) for free to patients with diabetes.

Some firms have done this. Pitney Bowes lowered copayments for diabetes and asthma medications and reported favorable outcomes. The City of Asheville had a comprehensive diabetes management plan where they lowered copayments for a whole range of services for these patients. ActiveHealth Management, in New York, identifies patients who would benefit from particular services as part of disease management, and they adjust the copayments for those people. The University of Michigan has a plan called M Healthy, focusing on diabetes, where they basically give medications away to diabetics—not just medication for blood sugar control, but for hypertension, cardiovascular disease, and other things. There seems to be remarkably little resistance from companies that have implemented this approach.

The solution in health benefit design will have to be to allow different components to work together. By using financial and nonfinancial incentives synergistically, we can make consumers behave as if they were better informed.

I remain a skeptic about whether information alone can encourage people to do the right thing. When you consider the number of people who have had heart attacks and stop their statins, the problem is not lack of information. There's something more fundamental about decision making, and insurance can help by subsidizing only the care that provides clinical value.

Finally, there are people who spend their time trying to figure out how to get individuals with chronic diseases to manage their diseases better. These programs have a range of names. I call them disease management or pay-for-performance, and there are other types of programs that would fit under these rubrics. You can use more nuanced benefit design to support these activities so the message to people is, “you should do this and we're not going to charge you to discourage you from doing it.” In years to come we'll see a much more sophisticated benefit design to support the things we want to support while discouraging the things that we want to discourage.

John Hsu: We've heard that clinically these value-based methods should be the direction we're heading in and we should be aligning our incentives. We've also heard that some of these programs can be very successful. Is this the future? Or is this a niche product?

John Bertko: I would say it's definitely the future, but administrative costs will be higher and the confusion factor is also something to consider. Nonetheless, it's going to happen. Copayments for magnetic resonance imaging scans (MRIs) vary by the location of service. The most popular drug copayment we have is a four-tier copayment. Some companies are offering a seven-tier copayment and you have to implement it with considerable communication. The question is whether researchers will lead the way and be helpful as opposed to being in the way.

Jay Krueger: I hope that these modified plan designs—the whole notion of consumerism—and delivering the right information at the right time really move the dial. However, we cannot forget about the holistic nature of medicine. Information is not just an information technology (IT) system delivering the Amazon green, red, and black, but it's also delivering the information around mental health and counseling options and about getting patients plugged in to the right resource to ensure that every component of their treatment is being addressed. It's a three-legged stool-plan design; enablement by information systems; and shared responsibility.

John Hsu (to the audience): Feel free to step up to the microphone, and please state your name and your position.

Anton Cooper: Anton Cooper, Washington State Healthcare Authority. As I recall, Pitney Bowes looked at hypertension drugs as well and did not choose to subsidize them because it did not improve compliance. Presuming that's the case, how narrowly do you have to analyze your population to see what measures would work for them, and how many resources can purchasers realistically devote to doing that for individual population groups?

Michael Chernew: Many firms have close working relationships with a range of employee benefit consulting firms and actuarial firms that do a very good job at this type of thing. Otherwise, you could buy an off-the-shelf product from a company. There are some logistical issues one would need to work out, but you could do that more broadly or more narrowly. To start with, you might just tie incentives into existing disease management programs. In general, once you've identified the set of services that subgroups of your population should receive, don't increase the copayments and add another barrier for people. Often those who set copayments are different from those who decide what type of care people need.

Jay Krueger: I have five analysts on my team, a contract with MedStat, and a handful of consultants I work with to manage the performance of all the health management plans—disease management being one of them. The point is that the data are so important to making better business decisions and impacting outcomes that the investment amounts to a rounding error when compared with the 300 million dollars a year we spend on health care.

John Bertko: The level and class of the intervention have to be tailored. It's not possible to have a nurse visit every person that shows any signs of a particular disability. Also, Jay's got 100,000 people he's looking out for. That's a good base, because some of these approaches need big populations and we've got 10 million people who are insured by Humana. Once you get away from high incidence diseases, we need data that we can point to. On the subject of disease management firms, we buy those mostly from the outside. There's a lot of hyperbole in this area and one of my jobs is to make sure things are proven.

Marcia Gold: Marcia Gold, Mathematica. I want to ask you to talk about the purchaser community …. It sounds nice that we can give people information and create things that'll help them make decisions. But the question is, how much are you hearing from the people who pay for the health benefits that they want to invest in this area?

John Bertko: Purchasers want to pay less, but they are also looking for better value. My personal philosophy is that we should always give choices to individual consumers. Some at the low end of the pay scale live paycheck to paycheck and they need options that are copayment plans that don't have much risk. The solution we have works across the board and we get better value. It means the people in the copayment plans are actually choosing a little bit better and saving some money.

Jay Krueger: Consumerism is not a panacea. A great example is 401(k) plans. Employers are giving a benefit to employees and expecting them to invest it responsibly. But the fraction of people who are engaged and who actively manage their portfolios in a way that is going to help them meet their long-term obligations is very small. If we believe our employees will act similarly with health care as they do with 401(k) plans, what do you do? There has to be a balanced approach. What's great about consumerism is that it's a language business executives understand because they speak it every day with their own customers.

The idea is to instill the basic fundamentals in your employees, giving them the tools and encouragement to take on more responsibility. At WaMu, we pay on average, 75 percent of employees' health care costs. People in the lowest salary tiers pay 6 percent of their costs and those in the highest salary tiers pay close to 40 percent. We use a communication campaign that reminds people to care and to use the resources that are available, and that tries to engage them in creative ways through incentives. Over time we're starting to see people's behavior change and we're measuring that through MedStat. We look longitudinally across people's health care experiences, ultimately tying it to outcomes, which are health risk scores and cost to the corporation.

Michael Chernew: The employee benefit consultants will make employers take this matter seriously. It's not that employers are out to do this; it's that people in the benefit consultant community find ways to do the analyses. Once those types of firms can find those types of data and people like John and Jay develop products, you can then buy at Humana where it's all automated. It requires work, because there's a range of things they have to do to make it simple for the individual, including integration at the point of service.

John Hsu: Consultants are a big part of this. Jay mentioned the point that you need a lot of data and you really need to follow your own population and a criticism of consultants is that it's a black box. What other data do you need? What other information do you need? Are consultants really enough?

Jay Krueger: At Washington Mutual, we wouldn't pay consultants unless they could empirically demonstrate that they are adding value. We use that same methodology regardless of who we're dealing with throughout the delivery continuum. We have innovative contracts that speak to service levels as well as performance guarantees, that are data driven, and that require vendors to integrate with one another and demonstrate their value. What we care about is outcomes, but data are the key.

John Bertko: Consultants would do a better job if they had better tools and better data. One of the things we need widely disseminated is the Medicare fee-for-service database.

Unidentified Audience Member 1: I have a question for Jay. You said you saved 25 million for the company and I'd like you to be a little more definitive. What does that really mean? And I would like to get some sense of where you think you got the most bang for the buck or was it just the whole package?

Jay Krueger: We self-fund 85 percent of our lives. We go through the actuarial exercise of projecting what our premium should be and then we use that to determine the cost to our employees and what the employer pays. Money is then put into an account from which we pay claims and quarterly we calculate our incurred but not reported (IBNR) expense. The 25 million dollars comes from release of reserves due to us coming in significantly under our health care budget. A lot of our success comes from self-funding and taking on risk. Additional factors are network discounts and, probably more important, actual change in behavior and consumption patterns.

Wally Gerard: Wally Gerard, Brigham and Women's. Many of us want more data. You mentioned the Medicare fee-for-service data and many of us have the concern that as the marketplace becomes more private, certainly with Part D, we can't call anything a success until we look at data. The Centers for Medicare and Medicaid Services (CMS) has said they are not going to release plan-specific data. What does that do for our ability to assess how effective Part D is?

John Bertko: I can only give you my interpretation. The data from hospitals and physicians on the Part A/B side are still available. Medicare is collecting prescription drug events data. Presumably the aggregated Part D data are going to be of the same quality or better than the Parts A and B data.

Paul Ginsburg: This question is about two issues you didn't address. One is benefit structure design to encourage use of certain providers, hospitals, and physicians over others. The other is about tailoring copayments to the patients or their diseases. Is there much going on as far as varying the copayment based on the notion that some services are seen as more elective or as having less certain effectiveness than others, which we're more confident we want to encourage?

Michael Chernew: The companies I mentioned would fit into that latter category of lowering copayments by services, not patients, although ActiveHealth can more efficiently target patients and maybe Humana has a product that does that as well. The University of Michigan did target patients. With regards to using copayments to steer people to a type of service or site of care, the challenge is that the more you lower the copayment for valued services, the less incentive there is to find the cheapest service. How you integrate those together affects the complexity of design, and you want to do it in a way that's not too complicated. The building blocks are there, but there is not a lot of that going on yet.

John Bertko: There's been a modest turnaround in terms of employer attitudes. In the early to mid '90s, managed care and tight networks were in vogue and they achieved some zero percent trend years. Then, as the economy heated up, everybody wanted the ability to go to any doctor. We haven't reversed that completely, but acceptance of tiered networks and limits on which provider to go to seems to be beginning to emerge. I would look for more tiering and more restrictions on the network.

Jay Krueger: Tiered networks are going to be in, at least from Washington Mutual's perspective as well as other purchasers that I interact with. We're flying individuals to centers of excellence for transplants. One of our primary providers, United Healthcare, has just come out with quality measures that are now on physician and facility look-up tools. We're starting to make those strides and going to tiered networks.

John Hsu: So the question is really about varying cost sharing by provider settings, services, and drugs. And the question is how well do we understand the value? Cost is fairly straightforward, but assessing quality is much harder. For instance, if you're talking about selective serotonin reuptake inhibitors (SSRIs) or biologics versus other types of drugs, the data aren't always there. What is the feeling from the panelists?

Michael Chernew: There are a lot of clinical areas where we don't know enough to do a good job at determining value and the default seems to be to charge people more in those areas. However, most of the money is spent on diseases that are reasonably well studied. We do know enough to do things better than we seem to be doing now. I'll put it another way. You would never want to impose the same copayment for everything. We don't know exactly how the copayment should vary, but we certainly can do better than the current system.

John Bertko: We don't know enough yet, at least on the industry side. We are experimenting. We have a product for prescription drugs called Rx Impact, where instead of grouping drugs by unit costs—generic versus brand versus nonpreferred brand—we group them on other dimensions. Essentially, it boils down to important, less important, and lifestyle oriented. In terms of market appeal, we've had some sales, but most employees are basically in the four-tier standard copayments because they understand that better.

Jay Krueger: You hit on a very important thing: employers' comfort level. Employers understand brand to generic, but, we know that's not always the answer any more. The question is how do you communicate in a simple way that people can understand and then have a marketplace that works to complement this information so that people can make the best decision. We're trying to work through that, specifically around prescription drugs.

Unidentified Audience Member 2: Michael, all the examples you gave involved copayments and most of the discussion has been about copayments. But the consumer-driven movement, which we heard at the beginning is working, has mainly been about deductibles. Are deductibles going the wrong way? Are they too blunt an instrument or can you do the same things with deductibles that you're talking about doing with copayments?

Michael Chernew: In consumer-driven health plans there are exemptions where you get first dollar coverage for certain types of preventive services. I would just expand the definition of preventive services to include important services for the management of chronic diseases. I'd like to see exemptions for certain types of services for certain types of patients in particular situations, and I think we will see that.

Unidentified Audience Member 3: I've got two questions for John and Jay. Could they talk about behavioral health services and whether consumers are being exposed to high deductibles for those services or if it's going to be modified and look something more like what Mike is talking about? Also, is there a greater interest in providing consumers with cost transparency or price data for different providers or different hospitals? Or, is that something you see internally and then try to steer your consumers to the more cost-effective hospitals or cheaper hospitals based on what you know?

John Bertko: In terms of mental health services, we try to have integrated approaches like prescription drugs and managed mental health care with lower copayments but higher direction and management. It keeps copayments from being too high, but constrains the network for those providers quite a bit. Our goal as a company is to make costs transparent to consumers in as simple a way as we can.

Jay Krueger: We're delivering claims data information—actual costs for services and facilities—to our employees so they can automatically load it into their personal health record for viewing. The second part is being able to use the cost data in benefit selection during the next open enrollment, where employees can use a cost comparative tool to see what impact different benefit designs will have on them.

Kenton Johnson: Kenton Johnson, Blue Cross Blue Shield of Tennessee. Jay, my question for you has to do with data integration. You're describing the way Washington Mutual manages total health care costs across the spectrum and that involves medical claims, pharmacy claims, disability, health risk assessment data, and disease management data. How do you get all the data integrated? Could you discuss how you get the pharmacy benefits manager (PBM) to work with the health insurer, how you get MedStat to work with Kaiser, and so on.

Jay Krueger: Driving integration is the most challenging part of my job. We do it through sheer purchasing power. We say here's what you're going to do, here's the timeframe, and here are the outcomes we want. We only talk to those individuals who can deliver. That's why the purchasers are going to play a big role in changing the landscape of health care.

How does it tactically happen? In many different ways. We have nurses and clinicians, who are literally looking at two or three screens because United Healthcare hasn't been able to integrate its IT systems yet. United has done great things with its system, using it as the repository and producer of data. I've already made comments about using MedStat, which is more of an analytical tool set rather than a practical or operational one, to integrate data.

Kenton Johnson: My question for Michael has to do with value-based benefit designs. I am interested in hypertension because there's a good degree of unmanaged hypertension that results in high-cost conditions later on. It seems like an ideal condition for giving people free drugs, but there are cheaper and more expensive ways to treat hypertension and ideally insurers would want to treat it in the cheapest way possible. Have you in your analysis looked not only at the idea of providing free medication for people with hypertension (or other diseases), but also encouraging them to make certain choices?

Michael Chernew: You don't see this as much in hypertension, although I know of a company that reduced copayments for ACE inhibitors and ARBs. But they had the concern that if they made everything free there would be no incentive to treat efficiently within the class. To address this, they reduced the copayments for generics to zero, and they reduced the copayments for other preferred brands by half, so there was still a differential. In the Rx Impact type approach, you get a certain amount of money based on the efficient or valued way of doing it and that's your allowance. If you choose a more expensive course, you pay the difference. It's a combination of reference pricing and value-based purchasing benefit design.

Whatever you do, it's important to think through the clinical ramifications. It requires a little nuance in how you balance subsidizing certain things that are important, making sure you're not subsidizing everything.

Amy Taylor: Amy Taylor, AHRQ. John, you said that consumer-driven plans have been a success. What do you mean by a success? I presume you mean that Humana has designed plans and consumers have bought them. But Tom Rice today at lunch talked about the concern that people with higher incomes and who are healthier will buy the consumer-driven health plans, and that could hurt the market for people who are poorer and in less health. Do you have any data on who's buying these plans?

John Bertko: We sell across whole employer groups. The Smart Suite is a total replacement plan, so it covers everyone with a choice element in it. The early movers tend to be healthier people and higher income folks. We now have 40 percent of our employees in these plans and our population looks like Jay's. It's highly female, they make about $35,000 a year, and many choose to take on some risk in terms of their payment at the point of service, but they save a lot of money on their monthly checks. By success I mean that we have beaten the trend over the last 5 years by about three or four percentage points.

Kevin Bosick: Kevin Bosick, UC San Francisco. Michael, have you looked at any type of benefit designed for bigger ticket items like medical devices and the example I'd like you to consider is from my own field, hip and knee replacement surgery. The implants vary in cost from about $4,000 to about $14,000 with no data to suggest there's any difference in patient outcomes or quality. Something I've considered with United and other payers is creating a benefit design around differential copayments for devices that have no proven efficacy beyond say, a gold standard.

Michael Chernew: I agree. One problem with the high-deductible health plans is that, for things like joint replacements, they don't do a good job of providing incentives. The problem in a lot of clinical situations is that you need to put the cost sharing where the incentive is. If someone needs a hip or a knee, you provide them full coverage up to the cheapest amount, like the allowance idea, and then you charge them above that where the decision matters.

On value-based design, you said there was no difference across devices. I would go further and say that even if one option is a little bit better, if it's also much more expensive there should be a price differential. I'm comfortable with the idea that you pay more for better quality.

With consumer-directed health plans, people have changed their behavior and they're consuming less because we're charging them more. But you have to make sure they're cutting back on the things that you want them to cut back on. The crucial question concerns the health consequences, or how to use cost-sharing in ways that preserve clinical value.

John Hsu: I'm going to finish up with one final question which is really a request to elaborate. What type of tools and data do we need?

John Bertko: I have folks who work for me who analyze data, and we're part of an effort by a large consulting firm to aggregate data. It's a difficult and expensive proposition and so I will say that there is such a rich data set in Medicare fee-for-service that it's probably what we ought to start doing tomorrow. Getting data on people under age 65 is equally important, and there needs to be a mechanism for doing that. I'm a fan of private–public partnerships and there needs to be something done here that's more aggressive than what we're doing today. Private companies need health services researchers and the government to lead us on this.

Michael Chernew: I edit a journal and we have a lot of articles that come across the transom and data quality is an issue. But more important, it's about study design and control groups. It's not just that you need a lot of people. You need to have a comparison group or some sort of quasi-experimental design. It's important to have either the right control group or the right study design in order to control for other trends. I think this is probably what happened in the disease management world. There was so much enthusiasm and so much skepticism because the study designs weren't good. It's not primarily an issue of statistical precision, it's an issue of bias and it can lead you to make worse decisions. The advantage for companies to send their data to organizations that collect data from other companies is that the data are standardized so people can learn how to use them.

Companies also need data from before and after an intervention for both the treatment and control groups, in the same databases. Credible study designs, which require integrated data and control groups, are crucial.

Jay Krueger: As a purchasing coalition we have 100,000 lives. One of the great things about being in the Pacific Northwest is that we've got Boeing, Starbucks, and Microsoft, a lot of large organizations with employees in the area who share the same vision. Being able to establish those coalitions to drive change in the marketplace to get the data we need is key. An example is the Puget Sound Health Alliance, which is Ron Sims's, our King County Executive, attempt to bring a purchasing coalition together to impact change. Also, being near the University of Washington and being able to work with some of the faculty and researchers has been beneficial. My last comment is that Washington Mutual's employees own the data. We owe it to them to get them the data they need to make the proper decisions for their health care.

John Hsu: Thank you very much. Please join me in thanking our panelists.

Prepared for publication by Kristin Rosengren, Director of Communications at AcademyHealth.