Skip Navigation Archive: U.S. Department of Health and Human Services www.hhs.gov
Archive: Agency for Healthcare Research Quality www.ahrq.gov
Archive print banner

This information is for reference purposes only. It was current when produced and may now be outdated. Archive material is no longer maintained, and some links may not work. Persons with disabilities having difficulty accessing this information should contact us at: https://info.ahrq.gov. Let us know the nature of the problem, the Web address of what you want, and your contact information.

Please go to www.ahrq.gov for current information.

Session 3: Using Data To Tell The Safety Net Story (continued)

Cindy DiBiasi: Let's talk also about the need for accurate interpretation because it would seem that interpretation is really what this is all about.

Andrew Bazemore: Absolutely and of course we started with some pilot efforts to do this mapping and we needed to take it back to the clinic and see could they actually use this? Would they find this information helpful? So we conducted a series of informal focus groups and key informant interviews with the leaders of our clinic network and almost uniformly to a man found positive responses. Just about every leader or clinician had a new question that they could drum up for us to ask and answer using mapping.

Throughout this series, they really felt like the most important uses of the maps early on would be for strategic planning and resource allocations as well as possible expansion of their network.

As you mentioned, the most important piece is to pair what we do in maps and what we see in the data, as John mentioned, was knowledge of the existing community. So we used their information and found that they said these maps were very easy to interpret, particularly since they knew the roads, the clinics, the individual patients in the communities they served very well. They found that what this allowed them to do was take otherwise useless data or data they weren't using, bring it together and bring it to life and really get to know their communities much better.

Over the course of that day, we generated a long list of questions, which we continue to work on and I should also note that the static mass that you are seeing in our presentation here today were really deemed least useful by the clinics. They said having a dynamic mapping source, having something that they can interact with, take layers on and off of, such as pull population data from or add another water or sewage line to, really allowed them to answer their questions in real time, which really reinforced our notion that a Web-based interface, particularly broadly used, would be the most useful way of mapping the safety net.

Cindy DiBiasi: Andrew, we will be back with some questions for you. But let's now talk to Pete Bailey from the South Carolina Budget and Control Board. Pete is the author of the chapter entitled Integrated State Data Systems. South Carolina does have an integrated data system. Pete, what types of data are included in that system?

Pete Bailey: Before I answer that question, I would sort of like to mention a quotation that I think is very useful. Someone once said that good health is what happens when everything else is OK.

So you will see in our integrated data system that we are covering more than health; that we are trying to cover the whole human services side so that you will see that we have private sector-type data in terms of hospitalizations, emergency room visits, outpatient surgery and home health visits. These are the actual bills. We have the Medicaid eligibility and claims, the state employee/teachers health plan eligibility and claims and vital records. A good portion of the free clinic database in South Carolina. Then, as related to state agencies, we covered the social services side. That is like (unclear), food stamps, abuse and neglect and foster children. The mental health and alcohol/drug side, that is like hospitalizations, all mental health center visits. We cover the public education side. For example, school readiness, achievement scores, and exit exams. Other health agencies like the health department, disabilities and special needs and voc rehab. Then we cover crash files from public safety and the criminal justice system in terms of juvenile justice and arrest data like from South Carolina law enforcement. So you can see that we have tried to cover a full range of human services from health, physical and mental to social services to education to criminal justice to the private sector health data.

Cindy DiBiasi: How do you link individual records from the different systems?

Pete Bailey: I think that is a very important question because the date that I just mentioned to you, believe it or not, we have the ability to link and track people across this whole system. That means we do get identifiers, but we use these identifiers to create a unique tracking number so that with the identifiers, the tracking number that we would create for you, and by the way, this is a randomized number. It is not a code that you can break. That tracking number would stay with you whether you came through the emergency room or the social services department. What we do is use that tracking number to link the statistics from the different integrated systems so that at no time do personal identifiers end up with the statistical data.

Cindy DiBiasi: So what added value does an integrated system provide over individual administrative data sets?

Pete Bailey: I think the most interesting example to cover that is children with special healthcare needs. When we pulled this system together and we began working with children with special healthcare needs in the Department of Health and Environmental Control, they basically knew about children's rehabilitative services, which they ran, and Baby Net. That covered around 13,000 people. Once we began working with them and they gave us all the ICD9 codes of children with special healthcare needs, then we were able to sweep through like the Medicaid data system, all the hospitalizations, and all the emergency rooms. Once we swept through, you can see that we ended up in the 300,000's in terms of children with special healthcare needs. The great thing about this is once you have this cohort or this sub-group of population, you can do this for a large number of areas. I am just using children with special healthcare needs as an example. But once you are able to do that, then you can link that cohort like children with special healthcare needs over to social services and say what is their abuse and neglect rate of children with special healthcare needs? Or you could link it to juvenile justice systems to see if by chance they are ending up more in the juvenile justice system because of misunderstanding or link it to school. By the way, we link children with special healthcare needs over to the school data and found that they were not doing near as well as children without special healthcare needs.

Then the fact that Medicaid is such a large piece of this cohort. In the Medicaid system you know everything about what happens to an individual. So you can get down to physician rates. You can get down to hospitalizations, whole mental side so that there are enormous advantages to being able to pull a large cohort together and then jaunt to different areas like social services or law enforcement.

Cindy DiBiasi: If you would, give us an example, Pete, of the analysis from the integrated data system and how those results were used.

Pete Bailey: The most interesting example I think that I could use is how we ended up linking the Medicaid data system to the educational data side. We began working on the educational side. They wanted a school report card. To do a school report card they wanted to use a poverty indicator. The only poverty indicator they had from the educational side was students on free and reduced lunch. They felt that high school kids would not take free and reduced lunch because of embarrassment and so they were not pleased with that as an indicator for poverty. So we suggested why not, let's link and unduplicate free and reduced lunch and Medicaid because our Medicaid included children's health insurance program, SCHIP. Well, when we did that it turned out to be 54% of the children in public schools in South Carolina. But by being able to do that, for the first time then we could see how the Medicaid children do in school or how (unclear) in general. So the chart that you are seeing is the duplicated free and reduced lunch and Medicaid. You can see you have four categories, below basic to basic to proficient and advanced. You can see these enormous numbers in terms of the below basic and the basic for the poor. The ramifications of this has to be that that means the poor are not going to be able to go to college. They are not going to end up in the professional jobs. So this sort of opened up an enormous door. Our governor at the time, virtually all you could hear out of him was education, education, and education. That is across the country. But once were able to link health with education, he began saying health and education, not just education.

Since then we have several projects that are really interesting. We have a school district that is looking at four high schools that are pretty differing high schools. We are pulling the kids that are Medicaid kids so that they can look in their high school, and one high school was two-thirds Medicaid, they can look at their doctor visits. They can look at the emergency room visits, their hospitalizations, even pharmaceuticals. The things that we see, for example, in this particular high school, their number one visit to physicians was for pregnancies. The other thing that we saw so much of is mental illness and mental conditions that we were somewhat caught about. So this type of linkage has tremendous potential.

Cindy DiBiasi: The same in the safety net population. An integrated system such as your could really be helpful in addressing problems.

Pete Bailey: The first time that I actually, I didn't know a lot about it. I went to a conference on safety net and they kept talking about it and I thought, who is the safety net population? Someone said, well, it is basically the Medicaid, the (unclear), the food stamps and the uninsured. I thought, well gosh, we can just swing through our populations and identify that population. So you will see that we did that and it was like shocking because see, what you are seeing here is that South Carolina, and this certainly doesn't cover all the uninsured. It was those that have hit the system through emergency room, but you see that 24% of our population is safety net. Look at the non-white, which is heavily African-American, look at the 1-14, 74% in the safety net. So just being able to identify this cohort. When I mean identify that means with this tracking number now that we can put our arms around them, we can look at them in mental health. We can look at them in hospitalizations. We can look at them in vital records. That is a fantastic thing. So what it means is that you can take this safety net population, look at their health problems, look at their problems around birth, look at abuse and neglect, look at law enforcement issue. Just a wonderful thing.

Cindy DiBiasi: Was your integrated system useful in helping to locate and enroll children in SCHIP?

Pete Bailey: It was tremendously helpful there because we were able to run through the food stamps, tenant, which most were on Medicaid and the children that were using the emergency room that at that point in time were uninsured, link them over to the Medicaid eligibility to see if they still were not covered by SCHIP or Medicaid. But by being able to do that, then we used our DIS system to actually map, down to the Census block level and we then created density maps so that they had outreach workers going out and finding children that were uninsured. It was very useful to them. The other thing that we did is that we took the free and reduced lunch files from the educational side, link them over to the Medicaid eligibility side, to see how many of them were not on Medicaid, which gave us by school and school district the volume, potential volume of uninsured. Of course then the outreach people really concentrated on those schools. Then the great thing about it is that as you initiated programs, we could use the emergency room data and the children that were coming that were uninsured to test and see how well our programs, what we were trying to go out and do outreach and find them, how well they were working.

The really interesting thing about this effort is that between 1998 and 2001, now this is strictly looking at emergency room data. Children that had come in there without insurance. We cut that number by 38%. For some counties it was up as high as almost 70%. I think we broke our bank in terms of the Medicaid system, but it did show that you can go out and find these children and you can get them enrolled.

Cindy DiBiasi: It must also, being able to identify and track this many people in the safety net population, must help in dealing with disparities.

Pete Bailey: It really does because in almost any area you look, for example, let's say under the Medicaid, under hospitalizations or mental health or whatever area you start to move in, that you are looking for disparities. Where you see disparities, you don't know how much of it is due to for example, age, sex and poverty. Typically, you can control for age. Typically you can control for sex, but poverty is very difficult. But by having the safety net population, when you look to mental health, that means you know that subset of population that is poor. So you can control for poverty. So if you see disparities, like we did with education, when we were able to look at the poor and see the massive disparities, we thought well, we have known that there are disparities in the educational system in the percent of children below basic. We thought that the disparity between races was due to poverty. Once we had the poverty data, we could look at poverty and then race. It is not just these two disparities. We don't know what it is due to. So being able to do that linkage really does tremendously help you understand disparities.

Cindy DiBiasi: Now it is one thing to figure out how to deal with the data, but how does a state that might be interested in developing an integrated system, work to the political and organizational issues that must surely arise in this?

Pete Bailey: One thing, it is very important, and I am not sure, as long as I have been working in the health system, that we fully understand this. That is that every state is different organizationally and that each state has to work through its political structure and personalities. They are very important. I can simply tell you certain things that we think have been really important to us. We are a neutral organization. That is, we do not offer any programmatic services. So there is no way that we are competitive with any agency that we are working with in terms of programs. So I think if you can possibly put your integrated data system in a neutral organization and the other thing about that neutral organization is absolutely they should have no power at all over data. As much as you have seen in terms of data systems that we have, we have no control over any of those data systems. What that means is that we have not upset the balance of power in South Carolina and that is really important to do. So you certainly have to be able to try this. We have had so many years that we have been able to do a good job, both in terms of agencies in the private sector. As I mentioned, you have got to be able to preserve that existing power structure and respect partner's roles. Then of course you have got to have the appropriate controls of data.

I think that maybe it is appropriate to say that if we can do it in South Carolina, a real liberal state; they ought to be able to do it everywhere. What is most important here is not thinking about your organizational structure. Not thinking about use, but thinking about the people we believe we can help. Then concentrating on accessing the data, not improving your particular organizational structure.

Cindy DiBiasi: Good advice in so many areas. Thank you, Pete. We will come back to you during the question and answer. But in a moment, we are going to open up the discussion for questions from our listening audience. But first let's go to Christine Shannon, who is the administrator in the Office of Health Planning and Medicaid for the New Hampshire Department of Health and Human Services. Christine, how would you use the data and tools that we talked about today?

Christine Shannon: Well, we have seen here today that there are different types of data that we can use to tell the safety net story. We can use this data alone or we can certainly can fit them together to answer questions and to solve problems. And answering questions and solving problems are critical for us in making our planning decisions, where do we target our resources? Do we do an overall campaign? Target the whole state or rather do we look at certain areas? Policy decisions, what the impact may have on clients if you decide to do something about services for clients, still maintain access. And also informing decision makers, not only within our state, but sometimes when we apply for grants, or community health centers are looking to expand their services, we can also inform decision makers there.

Cindy DiBiasi: Could you give us some specifics on how you might use it in your area?

Christine Shannon: The GIS mapping? Let me give you a few examples of how we have used GIS mapping and how we plan to use it. Basically, I think what GIS mapping says for me is there are so many things we don't have pictures of. We are finding that GIS is a tremendously useful tool for assessing provider network adequacy. We have looked at our pharmacy participation and we are in the process of looking at dentists and physician participation in Medicaid to see how we could focus our Medicaid recruitment effort.

We have also looked at CHC coverage adequacy to determine whether or not where we might target resources for future expansion. We are also right now looking at Medicaid participation to give some answers to questions that our CHC's have on needs in their communities. How many Medicaid participants we have got in the area where I deliver my services.

The next three slides that you are going to see illustrate these examples that I have just given you. The slide on Medicaid pharmacy providers, what you are seeing here is the northernmost county in New Hampshire, the Canadian border. Now from where I sit, you could have given me a list of participating pharmacies and the towns they were in. That would have been nice. But now when I see this spatial relationship here, it really gives me a good picture of access and tells me that I might make some decisions or we might make decisions that could impact access to Medicaid participants in being able to get their pharmaceuticals. Basically what I think this slide shows you is that the GIS mapping gives clarity that the tables and lists just can't possibly give you.

The next picture here is a slide we did a couple of years ago. We did this slide to show a couple of things. One was we wanted to portray to policymakers just what our community health center coverage looked like in New Hampshire. Then we also wanted to see where we might have areas that were lacking access to community health center services.

The last thing that I mentioned, trying to answer some questions for our community health centers is we just started taking a look at this. We are looking at where the percentage of the population in New Hampshire that participates in Medicaid resides. Right now we just started doing this. We have a picture here that shows you where participants reside in terms of counties in New Hampshire. We are going to try and break that down a little further.

Cindy DiBiasi: What about the emergency department data? How have you been able to use that? I am particularly interested in the ED algorithm that John Billings shared with us today.

Christine Shannon: Well, the ED algorithm is a good example of why participating in sessions like this are most useful. First of all, we used the administrative data to take a look at the emergency department use. We use our Medicaid claims and we also used hospital discharge data. We set out initially looking at our outpatient usage for Medicaid clients and we zeroed in on emergency departments. We examined our Medicaid-funded emergency visits to see how Medicaid clients compared to non-Medicaid clients in terms of emergency department use. One of the things that John Billings and others have raised here today is you look at a piece of this information and it raises more questions. It has actually been kind of fun to start exploring the questions that this has raised.

We looked at the time of day that emergency department visits occur and sort of blew some conceptions that I had that everybody goes at night, which we found wasn't true. Over 60% of both Medicaid and non-Medicaid emergency department use was happening during the hours of 6 am and 6 pm.

We also wanted to look at the reasons people go to the emergency room. We found that for non-Medicaid as well as non-Medicaid the reasons are pretty much the same. However, our Medicaid clients are using emergency departments at two to three times the rate of those with other types of payment. We also found that our Medicaid clients are using the emergency room more often for common illnesses versus injury.

Cindy DiBiasi: So once you get that date, then how does that inform decisions from policymakers and decision makers?

Christine Shannon: Well, what we are looking at is now we could use this information in assessing our Medicaid physician provider network. As I mentioned before, we are also going to use it for our dental network. We are going to try to identify gaps in services and also steps in access. What we are going to do is we are going to take the easy data, the administrative data that I just talked about, and then do some GIS mapping. I have been told by my staff that today's Geo-Access software is going to permit us to assess physician networks in terms of availability and by distance and capacity. I think that is very important as we look at; we are a state that no longer has a Medicaid managed care product. We are going to be looking at how to maintain access for our Medicaid clients to physician services. This is going to be a first step for us to take a look at our physician network adequacy.

Cindy DiBiasi: Interesting. Christine, we will be back with you in a moment.

In a moment, we are going to open up the discussion for questions from our listening audience. There are two ways you can send in your questions. We encourage you to ask your question by phone. If you are already listening via the phone, press "*1" to indicate that you have a question. If you are listening through your computer and want to call in with questions, just dial 1-888-469-5316 and then press "*1". While asking your question on the air, please do not use a speakerphone to ask your question. If you are listening to the audio through your computer, please turn down your computer volume after speaking with the operator. There is a significant time delay between the Web and telephone audio.

If you want to send a question via the Internet, simply click on the button marked "Q&A" on the event window on your computer screen. Then type in your question and click the "Send" button. One important thing, if you prefer not to use your name when you communicate with us, that is fine. But we would like to know what state you are from and the name of your department or organization. So please provide those details regardless of the way in which you transmit your question.

As you are formulating your questions or queuing up on the phone lines, I would like to say a few words about our sponsors. The mission of AHRQ is to support and conduct health services research designed to improve the outcomes and quality of healthcare, reduce its cost, address patient safety and medical errors and broaden access to effective services.

AHRQ's User Liaison Program serves as a bridge between researchers and state and local policymakers. ULP not only brings research-based information to policymakers so that you are better informed, but we also take your questions back to AHRQ researchers so they are aware of priorities at the state and local levels. Hundreds of state and local officials participate in ULP workshops each year. The audio conference is being co-sponsored by the Center for Health Services Financing and Managed Care and the Department of Health, Resources and Services Administration or HRSA. HRSA is the Department of Health and Human Services access agency. It assures the availability of quality healthcare to low-income, uninsured, isolated, vulnerable and special needs populations.

Its mission is to improve and expand access to quality healthcare for all Americans. I would like to take a quick moment to thank Rhoda Abrams, the director of HRSA's Center for Health Services Financing and Managed Care. She has been instrumental in helping to develop and produce these safety net products.

We would appreciate any feedback you have on this Web-assisted audio conference and at the end of today's broadcast, a brief evaluation form will appear on your screen. There are easy to follow instructions on how to fill it out. Please be sure to take the time to complete this form. For those of you who have been listening by telephone only and not using your computer, we ask that you stay on the line. The operator will ask you to respond to the same evaluation questions using your telephone keypad.

Your comments on this audio conference will provide us with a valuable tool in planning future events that better suit your needs. Also if you please, E-mail your comments to the AHRQ User Liaison Program at info@ahrq.gov. Now let's go to some questions from the audience. This one is for John Billings from Jessica McCann. She wants to know, "Do local health departments collect the data on hospital admissions for ambulatory care-sensitive conditions? I have never seen these type of data before.

John Billings: Well, there are about 35 or 36 states that have statewide data systems with hospital discharge data in them. I forgot, what state was she from?

Cindy DiBiasi: She was from, it doesn't say, actually.

John Billings: Oh. Well, I don't know whether her state does or doesn't. Then there are a handful of states where the local hospital association has that sort of data. They have it to help hospitals understand the patterns of utilization and market share. So you can often get it from the state hospital association. But it is typically not a local department of health. It is typically a statewide department of health or data authority. In each state, of course, it is a different place.

Cindy DiBiasi: OK. On the phone from Nevada we have Barry who is on the line. Hello? Hello, is Barry there? Seems like we may have lost him. I will try to get him back in a little bit. Here is a question for Pete Bailey from Dee Hunsaker. "Can you present data from the integrated system at local levels such as the county level or census tract level?"

Pete Bailey: Yes, all of the data that we get, of course we have got county codes, zip codes. But also with the addresses we have the ability to run through address match software to go down to anywhere from longitude-latitude all the way up to block groups, census tracts. We are still way behind in that but in some ways we have made a lot of progress on the; of course, it requires good address match. For example, I know that when we have just done working on Medicaid and I think we ended up with around a 75% match. So a lot of addresses are just not good. Our address match system is really good because we work it through the E911 system. But we can go down, and the interesting thing which we have not done this and I hope more states will move into this arena, is once you have done this type of (unclear) coding, you can go down to a house in senate districts or even congressional districts, and once we start doing our data at that level I think we can begin to put some of our elected officials on the spot.

John Billings: It may also be useful sometimes if you can geo-code addresses to get below the zip code level. Many densely populated or urban areas, zip code, which seems like a small area is actually quite a large area. That can be big differences from one part of the zip code to another. So listening to Pete suggests that you can also do that at the zip code level where you have got that sort of situation.

Cindy DiBiasi: Robin?

Robin Weinick: I just wanted to echo what John was saying that depending on what kind of area you are in, a zip code may not even be a continuous area. They are designed for postal service delivery routes rather than for the convenience of researchers and policymakers and planners. So it is really worth paying attention to that. Pete?

Pete Bailey: I think that your best area, as those of you who know census well, know that bought groups are as far as you can go down in terms of socio-economic data. So by looking at block groups, you can tell how homogeneous your area is in terms of poverty or education. But in the way you can get over it, I don't know if people have done this. We have done a lot of this is the way you can get around small numbers is to take your block groups and group them into several groupings, like three groupings. So your low socioeconomic, your middle and your high. Then when you put your data together, you have got larger numbers and you can really see the effects of poverty or education.

Cindy DiBiasi: We are going to go back to the phone and try to get Barry back from Nevada. Are you there? Hello?

Barry: Hello, I'm here.

Cindy DiBiasi: Hi. Do you have a question?

Barry: Yes. When they started this reference to using administrative records, specifically diagnostic codes tied to payment and how those were usually valid because they were tied to possible truth fraud issues. What I wanted to point out is that those are, I figured them as often quite invalid specifically because they are tied to payment. My background is in mental health and there is any number of times where I have seen a person admitted for one diagnosis as opposed to another because you could only get paid for the admitting diagnosis.

Then a more recent one that folks may be aware of is Zyban, the antidepressant that works for smoking cessation. The insurance plan that we have here at least, if it is prescribed for major depression, it is covered by an insurance plan, but if it is prescribed for smoking cessation it is not. As a result, I rather suspect that if you were to look at administrative records, you would think that we had had a major epidemic of major depression amongst people trying to quit smoking. This is one example of the sort of things that can happen.

John Billings: I think that is a good illustration of caution is in order whenever you do this sort of thing. First of all, I think it is fair to say that since DRGs have come into play where people really being are being paid based on the diagnostic code, there have been lots of good studies suggesting that the accuracy of the coding, apart from gaming, which is real, has gotten much better. Gaming is what the gentleman is suggesting where you are trying to get a higher payment level or a different payment level or you are trying to disguise something. That is a real problem.

I think in the mental health area it is a particular problem. We have always struggled with trying to understand those codes where there is a lot of slippage here, there and everywhere. So for those sort of things, extra caution seems to be required.

Cindy DiBiasi: OK. A question from Kathleen DeMinor. I'd ask also everybody who is writing in with E-mails if they could please let us know what state they are from. That may be helpful to the panelists.

Her question is, "If you don't have the resources to do the in-depth examination of emergency room medical records using the four-category algorithm developed by John Billings, is there any way to make use of the data simply by identifying certain typical diagnoses that should be primary care treatable such as otitis media?

John Billings: Sure, you could do that. What you should know is that this ED algorithm is incredibly easy to use. AHRQ has helped and HRSA has helped make it available at no cost to anybody. It is on a Website that we can let you know about where you can get it. But it really doesn't take an enormous amount of effort to run the algorithm. You can run it in Access, which is a pretty common program. You can run it SPSS and you can run it in SAS. So it is not terribly difficult. Having said all that, there is nothing wrong with saying, all right, in our community what we are really interested in is otitis media. What you are going to find, interestingly, is otitis media is pretty high up there on the utilization level, but the winners on emergency room use in most communities are things like sore throats, fever of unknown origin, abdominal pain, common cold, that sort of thing. So you have a lot to choose from. What I would suggest is doing a quick frequency distribution of the top 50 diagnoses and take a look at them. It is usually pretty eye opening because they are usually not serious emergencies; that is not to say there are not important emergencies that go on in emergency rooms, but there is a lot of primary care going on in that setting. You will see that by a simple frequency distribution of the ICD-9 code.

Robin Weinick: I wanted to second what John said. If you have to look at the date to find information on otitis media, you are going to be looking at the same data using the algorithm, which is very, very easy to use and it is available free of charge on AHRQ's Website at www.ahrq.gov/data/safetynet. We will give you that address at the end of the broadcast as well.

Proceed to Next Section

The information on this page is archived and provided for reference purposes only.

AHRQ Advancing Excellence in Health Care