AMCHP 2005 ANNUAL CONFERENCE
DELIVERING RESULTS, IMPROVING PREGNANCY & BIRTH
February 19-23, 2005
STEPHEN SAUNDERS: It's a pleasure to be here today. And what I'm going to talk about is kind of a similar study. We've had a few more years of doing this. We've tried to work out some of the biases and see if we could really get down, drill down and see what kind of impact both WIC and case management have on birth outcomes. That's what I'm going to talk about. So here we go.
These are the folks that actually did all the work, and as you can see, the first portion is very (inaudible) you may recognize Bill Sappenfield in the back of the audience. Thanks for coming, Bill. And these were certainly our main brain power behind this initiative, so I'm going to thank them.
Illinois , just quickly, big state. Fifth largest state, 12 million people. Two‑thirds of the folks live in Chicago , but we are yet a very rural state. About 70% is rural. We have health departments in every county almost, except one or two. We have about 180,000 live births a year. It gives you the sense. And Medicaid covers about 44% of all the births in Illinois .
So I want to briefly describe these two programs that we have in Illinois . One is family case management, which is basically a care coordination program, partly financed by Medicaid, looking at outreach, case finding, assessment, (inaudible) Medicare plan, referral, follow‑up and advocacy. It's a statewide program. We contract with health departments, FQHCs and community‑based organizations. It's a $44 million state budget that is then matched by Medicaid, and so we get half of that back, or 22 million back.
We serve 281,000 infants annually. You can see it's a large program, serves pregnant women and infants and the target is all Medicaid. Pregnant women and infants and a few that are not on Medicaid but yet are low net income. You all know what the WIC program is because every state has a WIC program. Physician education, breast feeding education, (inaudible) nutrition and foods and access to healthcare. Illinois we have a big WIC program. Once again it's a statewide program. About $34 million for admin that goes out and plus as you know a big part of the WIC budget is food and we serve about half a million women, infants and children, as you know WIC goes up to age 5 in the course of a year. We have about 290,000 at any one month on the WIC program. So it's obviously a large program as well. But we're going to focus with this analysis about pregnant women.
Okay. Just quickly, the theoretical model for family case management. This is kind of a busy slide, but basically obviously a Medicaid pregnant woman is diagnosed with pregnancy. She may enter either prenatal care or family case management or WIC. I should make an aside we made a conscious effort in Illinois to combine these two programs. That is, the federal WIC program, with the state fund case management program being one program as delivered by health departments and so when the family goes in, they don't know it's WIC or case management. They just know it's MCH ‑‑ in fact they may call it WIC, but as you can see by the numbers in fact more than half the money comes from state and case management, which allows us to do a more intensive service as opposed to strictly a WIC service. Because WIC is limited by the USDA budget.
They develop a care plan. They get referred to all these kinds of services, domestic violence, education, so on. You can read the slide yourself; and of course the ultimate goal here in terms of pregnancy, our hope is that we will get more women connected with prenatal care and all the support and social services and by working with them we'll ultimately reduce low birth weight and very low birth weight. That's certainly our working hypothesis.
Background now. We have over the years, in fact for the last five or six years, been matching our vital records data, our WIC and case management data, and our Medicaid claims paid data to look at the outcomes of these programs. And so we've had the luxury that perhaps Ohio hasn't had to have been able to link these three major data sets for the last five or six years.
And when we linked them and analyzed the outcomes we saw these dramatic results, like I'll show you in a moment, that showed that when folks were on WIC and case management they had dramatically lower rates of very low birth weight. Dramatically lower rates of low birth rate and dramatically reduced Medicaid costs. We were tickled pink to see that, but more of our sophisticated epidemiological friends said, uhh, you got some biases operating here, issues operating here; let's see if we can do better.
I want to talk about, first give you the background of the earlier studies that were not controlled for these biases; then I'm going to focus the rest of the talk on where we've tried to correct for these biases, as I said with a lot of help from the CDC.
So, in fact, we started this process, we started matching data back in 1997. So what is that? Six years of data. We'll be matching 2003 here in the next couple of weeks. This is looking at only Medicaid women now. This is only looking at Medicaid recipients. We have about 80,000, 75 to 80,000 Medicaid pregnant women in the state of Illinois in a given year.
This is looking at very low birth weight. The first line is the rate of very low birth weight for WIC case management recipients. You see it runs about 1.3%, and the next slide is the very low birth rate in Medicaid, pregnant women who were not in WIC and case management. And you can see it runs between three, almost 4 percent. About a two‑fold difference. Gee, we were tickled pink with that, but like I said some folks said that was just too, if it's too good to be true, maybe it isn't true.
And if you look at low birth weight, you see the same kind of pattern. Just showing you four years here, but trust me it continues on into 2002. If you're on WIC and case, if you're a Medicaid pregnant woman and on WIC and case management, low birth rate of 6.9, 7.1 on the top line there. If you're a Medicaid pregnant woman not in WIC or case management, your low birth weight rate runs from 15 to 20%.
UNIDENTIFIED SPEAKER: (Inaudible).
STEPHEN SAUNDERS: We said once again that's fantastic but we don't believe it.
If you look at IM, it's not here. I have a slide on IM.
UNIDENTIFIED SPEAKER: The last one was ‑‑
STEPHEN SAUNDERS: I didn't read it, you're right I stand corrected. This is looking at infant mortality. We also did it for low birth weight. Here this is infant mortality, this is per thousand obviously. I stand corrected. You can see a difference almost a threefold difference in infant mortality in these babies.
And so that was fantastic. We loved it. But, of course, why isn't the rate going down that much, you could ask?
And finally you look at costs. This is Medicaid costs in the first year of life. As I told you, this file is linked to Medicaid claims paid data. We looked at all Medicaid costs, medical related costs in the first year of life, hospital, pharmacy, outpatient, everything. And you can see that the infants whose mother was in ‑‑ the Medicaid infants whose mother was in WIC and case management, prenatally, their average cost varies from 4,600 to 5,500 over those six years. And the infants whose mothers were, Medicaid infants whose mothers not in the program had about twice the cost.
So that made our Medicaid friends happy, but once again we're concerned that this is encouraging but we need to do a little further research.
So the research question then is, given all this prior experience and all these great data, is it really true? Does WIC and case management really reduce early preterm labor and small for gestational age babies? That's our question.
Let me go over a couple of biases that I think were probably operating on those prior slides. One is selection bias. It's always a problem in any non‑randomized study. And clearly with evaluating WIC or WIC and case management, we don't have the luxury of doing a randomized control study because that would be unethical. WIC has been out there too long. We can't really random myself women into the program and out of the program.
So you have to get real creative how you can address the selection bias. And obviously the theory is that folks that enroll voluntarily, or that we can find by outreach and get them into these programs, may be different by some unknown factors than the women that you can't find because they're harder to find, they're more underground or whatever, or they don't want to be in WIC. So that's what we call selection bias. And so in this analysis we tried to control for that.
I think you alluded to this bias, which is program entry, what I like to call prematurity bias. Simply put, it's that if you enroll in the WIC or case management program let's say in the ninth month of pregnancy, then you haven't, you have no chance of delivering a premature baby. All the ones that delivered prematurely are already delivered, right.
Prematurity bias. It's a major issue with these programs. I think you alluded to it. As we all know unfortunately our goal is to get folks into the WIC or case management in the first trimester or even in the second trimester. Some come in the third trimester, right? Well, those didn't have the opportunity to have prematurity, right? Because they're too far along. That would bias this study.
So how can we control for some of these biases? Here's how we did it. Women with no prenatal care or late prenatal care were not in this study at all. All women in this study that I'm going to show you both the controls and the intervention have started prenatal care before the fifth month of pregnancy. So basically we've controlled for any impact, any effect of prenatal care so this analysis is looking at the impact of WIC and case management over and above the impact of medical prenatal care. And also we do a logistic regression to look at the issue that you mentioned which are like race and education and age and so on and we'll use logistic regressions to address for that selection bias. In terms of controlling for program entry price, women who enrolled after possible preterm delivery were excluded as nonprogram participants. So they are not in the intervention group.
They're in the control group. Isn't that right, Bill? They're in the control group. So if anything, well, they're in the control group. Let me leave it at that. In other words, have you to have entered the WIC program case management early enough so you could have the opportunity to have a premature baby. We also looked for controlling for medical risk factors, because it could be, for example, that folks that get into WIC and case management are lower medical risk, that women that have pre‑existing medical conditions for some unknown reason self‑select out of these programs. So we try to control for that by logistic regression and in fact we were pleased to find that when you actually look at the birth records and try to assess medical risk factors that the folks in WIC and case management had higher rates of preterm birth had higher rates of pregnancy and hypertension and so on.
Controlling for the black box. That we didn't do. This should say the black box of case management and the black box of WIC. What I mean is if they were in, they were in. We didn't try to judge how well the intervention was delivered by the local public health nurse. And you can appreciate when you have 102 counties and you know three or 4,000 people doing this program, some do it better than others, right some are having a more intensive effect than others so on. We didn't control, this is a bias we could not control for this issue of what is the black box, what is inside the box? It's you're either in the program or you're not. We didn't try to judge the quality of your intervention in the program.
Some basic data. We did this over three years, 2000, 2001, 2002. In those three years there were 71, 75 and 75,000 Medicaid women in Illinois , enrolled in that year in Medicaid program that Medicaid paid for their delivery, and let's look at 2002. You can see that in 200266% of those Medicaid women were in both WIC and case management which is our program desired model. 10% were in WIC only. Sometimes you can't get it in both, and that's mostly, that happens, and 8% were in case management only and 15% you couldn't find. They were in neither program. So that 15% is the control group. Are you with me? That's 15% of Medicaid women.
Okay I should mention we have a relatively sophisticated MIS system it's called cornerstone. It supports all this work, and all this work is recorded locally in this distributed management information system that's in every health department, every WIC provider, all their work is done on this computer, so that allowed us to assess that aspect. And this data file is what is linked to vital records and Medicaid claims paid.
Okay. We linked the birth record to Medicaid. The cornerstone. Concocted the files, and we only looked at singleton births. We excluded multiple births, which a lot of other studies have done, because the obvious issues with multiple birth. We looked at completed months of pregnancy. They had to enter the programs prior to the gestational age of the outcomes. I'll show you what that means in a second. As I mentioned before, this is important, they had to enter prenatal care before the fifth month of pregnancy. We have basically controlled for prenatal care. They had to be in the program at least a month so that we could hope they had some program effect as opposed to just entering that week. Whatever effect that might be. We don't know what's in the black box but they were there in the black box for one month.
We did the logistic regression, we looked at outcomes of early preterm birth, and I'll show you in a minute we define early preterm as fifth or sixth month delivery, which is pretty early. Late preterm, which would be defined as the seventh month of delivery of pregnancy and SGA term, which was eight month at SGA, and we did all the logistic regression for all the usual stuff, age, race, education, parody, and so and so on.
All right. What did we find? Bottom line: Looking at the early preterm births or the proportion of early preterm births, actually if you look at the bottom point, the bottom dot point, just for us nonneonatologists the average weight of a baby born in the fifth or sixth month is between 496 grams and 1637 grams. So we're talking about small babies here. Mostly very low birth weight babies.
The odds ratio, when we did all that mumbo jumbo with the logistic regression, all the stuff I said where we tried to control for those various selection bias, program entry bias and so on, was .76. And the confidence interval is .69 to .83. What it says is folks that participated in these programs, the combined program, were 24% less likely to deliver a very low birth weight baby in this weight range or more accurately said deliver a baby at five to six months of gestational age.
So that made us feel good. Now if you go back to those prior slides I showed you, it was almost a 50% effect, right? All right. Those first slides. Where we didn't control for anything. We saw sort of a raw effect of a 50% reduction. We said that was not possible.
I'm happy to believe 24% reduction. If we could achieve a 24% reduction in very low birth weight over and above whatever contribution prenatal care has, I would be happy. Because as I told you, we adjusted for age, education, race, ethnicity, maternal, marital status, smoking, alcohol, parody, medical risks and parody again. I guess we did it twice. I don't know.
And there were 203,00 folks in this analysis. So it's a big study. So we feel fairly confident that this is fairly true.
Looking at ‑‑ well this is the same slide, looking at various combinations. We just talked about that.
Looking at late preterm birth, that is, preterm birth in the seventh month of pregnancy. If you look at the bottom part of this slide, 50th percentile for late preterm birth is 19, 18, 27, 67 grams. So these are some of your moderately low birth weight babies. We see sort of a similar effect, somewhere between 70 and 80%. So that means or .7 to .8, means about 25 to 20 percent reduction. Not quite as impressive as the early preterm or more extreme preterm. But then if you look at it efficacy wise, this group does not have the same morbidity as the younger group or mortality or Medicaid cost.
And finally, looking at SGA, just like you found, we found no impact on term SGA, whether in the program or not. The odds ratio was .97. But if you look at the confidence interval, it crosses one. No effect. That was a little bit of a surprise to us because, hypothetically, you would think that certainly WIC being in the Christian program, might have some impact on term SGA, because obviously SGA one of the risk factors for SGA is adequate nutrition adequate weight gain so on doesn't pan out. That's what you found in Ohio as well. I'm not sure what that means but that's what it says.
Okay. Conclusion: We think ‑‑ we're very impressed with these results, that 24% reduction in extreme prematurity that I showed you. We think these are mostly conservative estimates. We excluded women without prenatal care. We moved WIC case management late into the comparison group so we could avoid the prematurity issue bias. So in the comparison group there were some women that were also WIC and case management, but they entered late.
So that would, if anything would make our estimates more conservative. We adjusted for a wide array of demographic and health status variables, k regression, and I think we've addressed four of the five sources bias that have been reported in the previous study. The one bias we probably didn't well adjust for is sort of the black box issue, what is this intervention, and probably also a dose response. We didn't try and assess did they have three WIC visits versus four WIC visits versus five case management. We didn't go that far. You had it or you didn't. That's part of the part we didn't check.
And limitations, obviously with any of these kinds ‑‑ as you know there's been a lots of studies in the literature looking at case management. They're not randomized. We have to accept that. And I already said we did not account for variations in service delivery, which is the black box issue, so to speak.
Conclusions: We think that this WIC and case management, which I believe is a good way to deliver WIC because it's more comprehensive than WIC because we've got more resources than nurses can spend more time with them has a protective effect for both early and late preterm labor, although as I showed you on the data, the effect was a little bit better or stronger for more extreme prematurity, which I think, if anything, is good.
24% reduction. I already said that. Neither program seems to impact SGA term. I don't know why. And as I pointed out before, these program impacts are in addition to whatever impact, and I do believe prenatal care is affected, by the way. Whatever impact prenatal care had was held constant in this evaluation.
We feel that, there's been a number of studies that showed this general thing, but we think this one is more rigorous in terms of control for biases. At least we've tried to control for all these biases that we could, short of a randomized controlled trial. And obviously the public health (inaudible) is that this would be good for public health, which I don't think is a surprise to anybody in this room. What I didn't put up here of course is this would also save Medicaid money, which is a good thing, because why spend money when you don't have to spend it.
And I think that's the last one. So that's what I had to say. And I guess we're going to take questions; is that right? Thank you.