MCHB/EPI Miami Conference — December 7 - 9, 2005

Linking Data and Policy to Programs — Transcript

DANIEL WELSH: Hi everyone. I see some friendly faces from the past. I guess what I'm going to do is--this going to be a little bit more nuts and bolts and I'm going to try and tell a story and as some stories, they don't come out with happy endings.

I'm examining risk factors, I'm looking at services and I'm looking at outcomes during the perinatal period. I'm playing around with it because for the last couple of years I've been doing nothing but matching and linking data. Doing nuts and bolts of developing a Web site on children. It all of a sudden dawned on me at some point, oh, maybe you should start showing some findings and producing some information for policy makers.

So you become my guinea pig because when I go back I'll start doing some reporting to the program managers because we've been in a sort of continuous two year re-organization and I've just gotten sort of bureaucratic-ed out and now during lunch time I now get to do research.

So lets move and talk about the population. We're are looking at the 2000 year age cohort for the District of Columbia and what we did was linked newborn screening data with the birth file. Then we added the Medicaid data and we added social services and in our jurisdiction it is called ACIS. After that I was able to obtain data for WIC, Healthy Start, Early Intervention, and viola, somewhere I got the infant death file too. So we've got, in terms of administrative data sets, a pretty good picture of data, and now whether it is good data or not is another issue.

Administrative data, keep in mind, is used for program performance and activities and as a side issue for research. In the District, we have certain areas that are under reported. For example, substance abuse, alcohol and smoking I know are badly under reported. Keep in mind that the District is a very urban area and if you are thinking of us as a state, we are the most urban state and in reality we are a city. People receive services that aren't covered in the administrative data sets, so keep that in mind. We weren't able to use hospital discharge data, which I would have loved to because they wouldn't provide the identifiers. And finally, I'm not talking about here the quality of services, more about the use of services.

So let's look at some data. I'm going to ignore, I think most of this group probably knows about birth file risk factors and birth file data and in your handout in the back, the last sheet are all the risk factors we used. But here what I wanted to do was focus on some of the data elements that pop up from link data. Obviously, the first one is living in poverty. By merging Medicaid and Social Services data you can create that variable and get to play with it. Another example is no newborn testing. By linking your newborn screening program with your birth file you find out the kids that weren't tested and then you can create, you know, in my mind is a variable that we hope is followed up by somebody. You pick up from the Medicaid file in our jurisdiction, information on foster children. The transfer to Children's Hospital is coming from a newborn testing file because our Children's Hospital does some genetic testing so we can pick up some of that data that they had a transfer from a birthing hospital to Children's, obviously indicating a serious condition.

What I did was take approximately 40 data elements and I recoded them into a dichotomist variable 01. Zero being neutral, one being having a factor that is defined and then adding the factors for each child to give a sort of a crude concept of what is going on. None of the children--there wasn't a single birth in the District without a risk factor and it went up to 16 risk factors. What I did was to classify them into three groups because frankly I'm interested in some of the program impact in terms of what we are doing. So my interest is really the high- and medium-end group. The low-risk group I'm hoping is doing great and we can leave them alone and not have program intervention.

Okay lets look at the groups. From this you can see that the three groups that race changes as the risk level changes. Similarly with education, you are going with the high-risk group nearly half of the group hasn't graduated. Moms haven't graduated from high school. Age is a mixed bag but generally the higher the risk you end up with larger teen populations and also you end up with larger older women populations. And finally not married, we have in the high-risk group almost 100 percent are getting to that level of not being married. So you know clearly we have some family structural and stability problems among our population.

Now lets go back and look at some of those individual risk factors in relationship to the three groups. So for example you can see something like very low birth weight, you can see the change from one group to the other. What we are looking at here is comparing differences among groups. Generally, the high-risk population has more; is poorer, has more complications, more medical risk factors and if you look at things like multiple address you can see--and multiple addresses is a created or linked variable by linking all these files together you can look at how many addresses they had by registering with these different programs. Obviously someone who has a lot of address changes is showing some issues in terms of stability again within the community.

So you get that and so I'm feeling pretty good now in terms that I've got groups that look different and maybe they act different. So we are going to switch gears. Don't lose those risk factors. You've got to keep that in your mind because I'm coming back to this. We're going to look now at the services that were provided during the perinatal period and essentially up to as three year olds. Fifty percent of our population or at least for this age cohort is in Medicaid. Over 40 percent is in WIC. A third of the population is receiving social services and then much smaller populations, targeted populations, are served by Healthy Start, Early Intervention and SSI. Let's look at services a little bit more from a different perspective and look at the participation among multiple services.

Essentially what you have here is that 50 percent of the population is receiving two to three services which I found very encouraging because we've spent a lot of time on coordination and that famous collaboration stuff. Amazingly the no program group is 28 percent and I, you know, you'll see a little bit more data in a moment on how to interpret that. Actually the amazing one is there is a small group that has five services. Now picture this: you have to have been in Social Service, Medicaid and WIC and then you had to have two of the following three services: Healthy Start, Early Intervention, SSI. That is a whole lot of services and just to me fascinating that we've got those kinds of mixes going on and lack of things both positive and negatively.

Now I'm going to bring it back and we are going to look at those risk groups. These all get a little bit more complicated so just hang with me. This was the table that I was looking at lunchtime. Tuna fish sandwich and I looked at it and said this is great. I've got dark blue on the first column of each of the groups. That is lack of assistance or participation in programs. That goes down as the risk level goes up. I think that is great.

I've got the far column, that sort of greenish, light blue. That goes up as the risk level increases. In other words, that high-risk group, they've got 8 to 16 risk factors in their make-up. So I want to see them receiving Medicaid, Social Services and I'm really thrilled that they are also involved in direct services like WIC, Healthy Start, Early Intervention.

The yellow group that is the third column involves only direct services. I don't want to see at the high-risk--really I don't at the medium- and high-risk group, I don't want to see just receiving direct services alone because that means we are missing out somewhere in terms of coordinating services with other programs. But overall I'm really happy with this and the one that would bother me is the seven percent no assistance in the high-risk category. Clearly, that would be an area that I would want to recommend to the program people to focus on that population in the future.

Now, we are going to move to outcomes. Okay. What's going on? I can look in my administrative data sets I can look at essentially two outcomes: infant death and disability and then everybody else. Here I'm presenting in terms of risk groups that you've seen the nice pattern of data and the way things are working out that made me very happy. Now I'm less happy because I've got infant mortality in the low-risk population. I've got the largest amount of infant mortality and disabilities in my medium-risk group. I don't have a clue and that's why I'm sharing this with you. Sometimes you share some data that you can't figure out.

Okay, so now we've got--we're dealing with risks, we're dealing with outcomes. Lets add back in services. This is a complicated table which means also that I didn't figure out a good way to report the information and hopefully over time I'll come up with some better ways. I've pulled out the disability group and I've pulled out the infant death group from my three risk groups and they are on the far side. They show the services they are receiving. The three left columns show the services receiving now of the children with the various risk factors, but not having any disabilities yet and not having--because the criteria for infant mortality we're not looking at mortality over time here, we are just looking at infant deaths because of the nature of the files. What obviously concerns me is that infant death file or infant death column, you've got 70 percent of the population has received no assistance, no services. That kind of shook me up and at the same time you've got a whole lot--the infant death is a relatively small group and I kept here the actual numbers for the District of Columbia and the populations we were working with because the assumption here is that I can work with program people and good research at some point will tell them in some way or other that we can get better at what we are doing.

Generally, the rates in the District are improving. There is some argument that the reason they are improving is that poor people are being pushed out of the District as we are gentrifying the city. But clearly we've got work to do and clearly or unclearly there is work to do in terms of--within the District on research on what is going on.

So in summary, I'm excited about what link data this SSDI project that has been going on at a national level of each of the states to do these kinds of linking data. I think adding new risk factors to the traditional ones is a good thing and a lot of the risk factors that we are going to be adding or at least that I'm looking at relate to program performance by the moms through their kids. For example, if you're in Medicaid and you are not active in Medicaid that becomes a negative risk factor.

With link data it is as important to look at the populations that are receiving services as the populations that aren't receiving services. Mostly we've never been able to do that. Usually with this sort of link data process you get that kind of information starting to pop out. Clearly, I need to do more research on relationships on risk factors, services and outcomes. And then how to target or coordinate programs like Healthy Start and Early Intervention on where they should spend their bucks and what populations they should address.

In this address I didn't talk about geographic differences. The District has wards, which are like counties because it is not relevant to a national population. It is very relevant to the local population but what happens is that you'll have a program like Healthy Start that targets a specific area to provide services and what I'm seeing is that geographic area you've got a mix. You can pretty easily eliminate the low-risk population but the medium- and high-risk groups are much more difficult to separate and you've got a lot of them. Also there may be relationships with the program succeeding. In other words, Healthy Start intervening and reduces risk factors and that would be great. So that leads me to a good place to stop and continue for the next time that MCH EPI is in another good jurisdiction like this.