What you're going to hear about this session are the details on indicators--a more popular topic these days than when the IOM first took on the process of developing a set of access indicators.

I'm going to start by putting the indicators issue into a policy context. I'll touch on why it is important to have indicators, what we are doing with them, and how indicators fit with health care reform and the general changes in the health care system.

I'll finish by trying to give you a sense of where I think the data systems and the policy on data are going to allow us in the future to do a better job of developing new indicators and measuring the indicators more directly.

Over the next few days, you are going to hear about different ways of measuring access to primary care.

What I want to do is step back a little bit and ask, "Why do we care about this issue at all?" And I think the answer is that we see this very rapid movement toward managed care and integrated delivery systems. What was once implicit in the definition of primary care is now becoming more explicit. And that's the notion that primary care is the starting point for responsibility and accountability in the health care system.

An HMO capitated fee is essentially a budget for taking care of a patient. The employer says to a plan, "We've negotiated this fee, and you are now responsible for this patient's quality and access. And I want to get good value for this money".

We've taken care of the cost part with the negotiated premium. And so more employers are now asking HMO's to prove to them that they are getting good value for the premium. Many of you have probably heard of a development called HEDIS. It is an effort on the part of the HMO industry to devise a set of indicators that can be used across the board so there is no need to come up with different indicators for every employer.

In turn, the HMO's are saying to the primary care physicians, "I'm going to profile you and decide whether you can provide the kind of performance I'm looking for in order to take care of these patients within the budgeted capitated fee". This is a trend we're beginning to see. And this is where indicators come in.

At the community level we're just beginning to see this trend. We are going to be saying to public health, "We want you to provide some performance indicators, too. We want you to show us--through outcome measure and interventions--that the public health dollar we're spending is really worth it".

We initially started down that path with the 1990 objectives. There are many more year 2000 objectives. The trick will be to come up with a limited set of efforts and outcomes that we can hold health departments responsible for.

The federal perspective on this is where the market doesn't work to distribute physicians. The federal government must figure out how to deal with the production process (in terms of education) and the distribution process. As a result, we get programs like the National Health Service Corps. (States are faced with a similar set of issues.)

From the federal perspective, we really need to determine the level of effort required here given everything else going on in the public sector.

I have been working recently with the Public Health Foundation and the Public Hospital Institute to get a better sense of how much primary care and what type is delivery by those institutions. We've never really quantified the proportion of care delivered in the public sector vs. the private sector, nor have we determined the nature of that care in terms of its comprehensiveness.

The major policy concerns that I believe we are really dealing with that lead to the need for indicators include the following.

The first is the need to allocate resources for building primary care capacity. From our standpoint, this has to do with the National Health Service Corps, which is growing, and community health centers. There is a concern about how the community health centers, in general, will be absorbed into the delivery system given all the changes.

In the Health Security Act, there was a fair amount of money allocated to helping health centers and other kinds of essential community providers form their own networks or integrate successfully into other networks. That money isn't there now, and we're going to have to figure out what to do. There will be all sorts of issues about who really needs this money and to what extent the private sector can deal with the capital needs. A part of these capital needs is information systems.

The second policy concern, I think, is the need to negotiate contracts and monitor managed care plans. I've talked to state Medicaid agencies, and even some counties that are trying to develop plans for indigent populations not covered under Medicaid. They are trying to figure out how to write these contracts so they can assure that there, indeed, will be access for the vulnerable populations--those traditionally have used the system in a more resource-intensive manner, and have required a whole set of enabling services in order to be successfully cared for.

I think the work we've done thus far with indicators can be built upon to begin to look at what it is we want to include in this process of making plans accountable. Remember, in the Health Security Act, they were called accountable health plans.

The third broad policy issue, I believe, is the integration of public health and primary care.

As part of the second national primary care conference, HRSA commissioned 42 case studies to look at health care reform at the grass roots level, including state-initiated efforts and market-driven reform.

Some of these case studies involved health departments trying to decide whether they are in or out of the business of personal health care services; if they are out of the business of personal health care services, what do their population-based services have to do with the personal health service system; and can these population-based services be developed in a mutually supportive way to accomplish the overall goals of health care reform.

I think exactly how we're going to hold health departments responsible for population-based activities, and whether or not they are going to play a role in holding plans responsible, is a big issue yet to come. Again, I think the work on indicators will inform that process.

Another movement that's coming down the road on the heels of 1115 waivers, is an effort on the part of the States to consolidate federal funding streams in categorical and block grant programs, and instead of the reporting requirements we currently impose on the States, substituting benchmarked outcomes for performance.

This fits in well with Vice-President Gore's performance review, so the White House is very interested in it. I think we will need to address the issue of how these benchmarks will provide us with the kind of information we need to determine whether the money is being put to a good use.

Snaking its way through these issues is another set of issues that I will discuss in a later session, and they are concerned with enabling services. We need to understand and come to deal with the insurance barriers, and all sorts of other nonfinancial, structural, and cultural barriers that stand in the way of people taking full advantage of health services. I see this as an emerging issue. You've probably heard it discussed at the state level and among various interest groups. I think you'll see it throughout all of the sorts of issues we've just discussed.

I'm going to stop here and segue to Janet Mitchell who has something more specific to say about indicators.

DR. JANET MITCHELL:

I would like to mention another reason for developing access indicators besides the ones already mentioned. And that is, we need to be able to obtain an objective picture of how well we're doing as a nation in assuring access to care for all Americans.

Much of what we know about access has to do with what we read in the paper or hear on television. You hear about the poor person who is turned away from the emergency room of a private hospital, or the woman without health insurance who waited too long to see a doctor about the lump in her breast.

What we often don't know is just how often these access problems occur; whether they occur more often for some subgroups of the population than for others; and whether the problems are getting worse or better. For policy purposes, it seems to me to be absolutely vital to be able to monitor trends in access over time. And that's even more important as the states increasingly begin to experiment with state health reform.

Now, in order to track changes in access, we obviously need to develop indicators to measure access. Economists traditionally have used economic indicators to monitor the health of the economy.

So measures like unemployment, inflation, and new housing starts have been used as indicators to tell us whether the economy is heading into a recession or is starting into a boom.

No one of those economic indicators alone can provide this information, but taken together, a series of indicators tells economists whether the economy is expanding or contracting.

Well, the same is true for monitoring access. We can't use just one indicator; we need multiple indicators to determine if access problems exist and whether those problems are improving or worsening over time.

As part of an earlier Robert Wood Johnson Foundation project, we developed a list of access indicators that could be used in much the same way that economic indicators are used. The result was the Access Chart Book, published by the Foundation.

What I would like to do is to talk about why we chose the indicators that are in the Chart Book.

To begin with, we believed it was very important to be able to measure access for persons of all ages. So our indicators covered the health care needs of all parts of the life cycle, including family planning, prenatal care, child health, adult health, and dental health. For each one of these health care areas, we chose two different types of indicators: utilization indicators and outcome indicators.

Utilization indicators measure contacts with the health care system. These include physician visits, immunization rates, mammography screenings, and Pap tests.

Outcome indicators measure the health status consequences of contacts with the health care system or the consequences of not having the contacts. And that includes measures like infant deaths, avoidable hospitalizations, the percent of breast or cervical cancers that are diagnosed at a late stage, and so forth.

Now, changes in either utilization indicators or outcome indicators provide information about changes in access to care. But when both types of indicators are changing, and changing consistently, then we have a lot more confidence that access to care is either improving or getting worse, depending on how the indicators are changing. Therefore, we felt it was very important for each of the areas we looked at to have both utilization and outcome indicators.

Let me give you an example of a couple of indicators from the adult health area. One of our utilization indicators was the percent of women age 50 and over who had received a mammogram in the last year. The outcome indicator that went along with this utilization measure was the percent of women with breast cancer who had their breast cancer diagnosed in a late state after it had already metastasized.

Mammography is a demonstratively effective screening tool for detecting breast cancer. And when breast cancer is detected at an early stage, when the cancer is localized, the chances of survival are dramatically improved. If it's detected in a late stage, the changes for survival are very low.

With this particular example, if we saw either a decline in mammography screening or an increase in the percent of breast cancers diagnosed late, we would be concerned about access barriers to routine cancer screening for women. If we saw both mammography screenings declining and late stage breast cancer increasing, then we could be reasonably sure that there were, in fact, some significant barriers to access for cancer screening.

With these particular indicators, we had good news. Mammography utilization was increasing across different income levels and the percent of women with breast cancer diagnosed at a late stage was declining over time.

Despite these overall improvements in access, there were substantial disparities between poor and non-poor women--disparities that have not changed over time. Even though things are getting better, they are not getting any better, any faster, for poor women.

Let me give you just a few examples of how we paired utilization indicators with outcome indicators, for some of the other areas of the Chart Book.

• In the family planning area, we looked at family planning visits and the percent of births that were unintended at the time of conception.
• In the areas of both child health and adult health, we looked at physician office visits and avoidable hospitalizations (hospitalizations that could have been prevented had an individual received effective, timely, ambulatory care).
• In the area of prenatal care, we looked at the number of women who received first trimester prenatal care and neonatal mortality.
• In the area of adult dental health we looked at dental visits in the past year, and the number of adults who had lost all of their teeth.

We also looked at physician participation in Medicaid because pregnant women who are Medicaid-eligible will get services early if they have good access to private physicians.

In order to really determine where access problems exist, it is very important to examine these indicators for subgroups of the population, rather than just for the population as a whole.

If you looked at these indicators for an entire state, or even for an entire county or city, it could mask access problems that exist only for a small group within that state or city or county.

In subgroups like the poor, the uninsured, racial and ethnic minorities, the disabled, individuals living in physician shortage areas, and new emigrants, there are many barriers to access.

But financial barriers, we felt, were one of the most important. And they are certainly the barriers most amenable to federal and state policies. So for this reason, our chart book emphasized utilization and outcome indicators for the poor wherever the data permitted.

Many analyses that you'll see in Health USA, or similar publications, focus on differences by race and ethnicity. These differences are important, but they reflect many factors besides the ability to pay.

While we did measure utilization and outcome indicators for Whites, Blacks and Hispanics--since those are the three ethnic groups for which data is most widely available--we reported our indicators for those groups only when we could not get comparable data for the poor and non-poor.

There is very little data of a time-series nature available that would let you specifically look at individuals with various kinds of insurance coverage. That's why we tended to emphasize poverty status. Many of the poor are uninsured and will have problems accessing care.

I would like to give you a very brief overview of the kinds of data we used to construct our access indicators. We found we were able to draw on a wide variety of data from many different sources.

• We relied heavily on household surveys. These surveys provided data on physician visits, usual sources of care, immunization rates, and cancer screenings, such as Pap tests and mammography.
• We used data from vital statistics, in particular, mortality rates, low birth weights, and prenatal care rates.
• We used hospital discharge abstracts or hospital claims. These are the data that are used to construct the rates for avoidable hospitalizations, also known as ambulatory care sensitive admissions.
• We used data from cancer registries. This is how we were able to construct our measures of the percent of cancers diagnosed at a late stage.
• We used the morbidity and mortality reports that the Centers for Disease Control produces. These gave us information on rates of congenital syphilis and rates of vaccine preventable diseases, like measles and pertussis.
• We used the area resource file to construct our measures of physician supply (e.g., the numbers of OB/GYN's in low income counties and whether that number is growing or falling over time).
• We used some survey data on physicians so we could look at trends in physician participation in Medicaid over time.

What I want to do now is really take a look at the frustrations with the existing data.

When I first got involved with indicators, I was amazed that there was all this data available.

But as I began to think about the equity issues of comparing different population groups and about actually being able to measure these phenomena, I was increasingly frustrated.

Let me begin by discussing the elements of what I call the "ideal data system". The ideal data system is upside down. Washington State's data plan comes very, very close to this ideal.

My perspective in going through this exercise was how am I going to better measure the indicators. I was probably the only one thinking about that because when other people look at enrollment and encounter data they are usually thinking about administrative simplification.

In the ideal system, where there's universal coverage, you have everybody enrolling in a plan. That gives you the opportunity to collect information about enrollees (including such things as race) once and accurately.

We presumably could collect information about people's language ability, which would allow us to retrospectively examine whether non-English speaking enrollees had problems within the health care system.

It also would allow us to explore who is enrolled and who is not. Even in the most optimistic days of the Health Security Act, we realized there were going to be people who wouldn't be able to enroll for some reason or other--they were homeless or mentally ill.

The plan was that all providers would be required to fill out some type of common encounter form. Whether this encounter form would have looked like the HCFA 1500 and what it would have contained were still open to debate.

This encounter form would really have allowed us to directly determine whether people were getting mammograms and Pap smears instead of relying on the health interview survey.

Now, the ultimate future is the computerized medical record. But we seem to be fairly far away from that.

The problem with an encounter form is that some HMO's are already moving toward the computerized medical record and see the encounter form as a detour on that path.

There is also a lot of skepticism out there among HMO's who have tried the encounter form and have had trouble obtaining consistent diagnosis data. Nobody really understands how to code "reason for visit".

What we really need to think about with any data set is what is needed in order to track access. It's very difficult to write questions that get at whether a bad result on a Pap smear was actually followed up. And it's just as difficult to rely on people's recall.

Moving on now to provider data. Competing health plans will have to submit RFP's showing that they, indeed, provide enough coverage in an area, serve the number of people they said they would serve, and have the required array of providers. This is the same kind of issue that's currently going on with bidding on managed care Medicaid contracts.

Ultimately you would get information on the structure of the plan with regard to providers.

One of the frustrating things about monitoring access now is that we don't have good variables to measure how the health care system is organized. We've relied on these really crude physician to population ratios and physician surveys indicating whether or not they accept Medicaid.

If we knew more about the actual structure of the delivery system, we could then compare the level of access in different kinds of delivery systems.

As far as population-based health status, we would be able to get information on the enrollment form that would allow us to identify sort of census tracts, and subsequently do more sophisticated work using the vital statistics data.

So, we would be able to begin to look back and forth between the health status of community and the health status of individuals in that community by linking those kinds of data sets.

Another development is that the social services systems are starting to get themselves organized, as well. And they're talking about integrating data systems and scheduling systems.

Ultimately, that system needs to be hooked up with the health care system so those social factors that influence health can be brought to bear on the actual delivery of health services.

And finally, there are the patient surveys. It was the intent under the Health Security Act to spend hundreds of millions of dollars on satisfaction surveys because it was perceived as a good way of measuring quality. I think there is a place for surveys even in a very comprehensive data system. For the people who don't use services, you want to know why. There is room for these measures of satisfaction. But they are not a substitute for other indicators.

As we talk about improving our data systems, it's like looking at a map and trying to plan your route without knowing your destination. So I want to discuss some of the goals I think are important to be able to more directly measure access nationally.

Another frustration I faced with existing data sets was their limited capacity to provide small area information. This is an issue you all face, so we are forced to do some creative things.

That's why I want to quickly run down the list of data available today that you have to learn to use to your best advantage.

• Hospital discharge data. Not every state has a hospital discharge data system.
• Surveys. As bad as surveys are for measuring things directly, they're better than nothing, especially if we can begin to get the national surveys geocoded, and introduce some synthetic estimates for populations. Then we can say, "Given your demographic profile, in X place, this is what you would expect the responses to be on the health interview survey". It would allow you to focus more on the populations at risk, rather than doing general surveys. Because when you do general surveys, you don't capture enough of the at-risk people to analyze the problems and determine what you could do about them.
• Chart audits. For example, what the PRO does in Maryland, in terms of oversight of the managed Medicaid program. They have a whole set of quality screens. They actually go into provider's offices and look for evidence of outreach activities and health education enabling services. It's very intrusive for providers, but it may be one way of guaranteeing some of these access indicators.
• Provider surveys. States have had a terrible time getting routine, consistent information about where physicians are located and whether they accept Medicaid.
• Special studies. In my breakout session, I'll spend some time talking about the work we did on enabling services.
• Standard setting. We're currently working with the Latino Coalition in California to help them develop an instrument that HMO's will use to determine if they are providing culturally competent services, and identifying the kinds of access problems Hispanics and East Asians are having within the HMO setting.

I think we're beginning to get limited management information systems in specific places, for specific populations, and for specific types of services. There is a whole set of capital investment infrastructure issues that the federal government and the states need to begin talking about. If you want to speak with me about your ideas about where federal capital would be most beneficial and needed in terms of your efforts to move in these directions, I would be interested. Maybe we can have some interaction on this issue during my breakout session.