AMCHP 2005 ANNUAL CONFERENCE
DELIVERING RESULTS, IMPROVING PREGNANCY & BIRTH
February 19-23, 2005
RUTH SHROCK: Thank you, Daniel. And good morning to all of you. Before I start, I want to give a thank you to the principal author of this paper, Wyda Yee. Wyda is an epidemiologist on my staff, and unfortunately he was not able to be here to give his paper because of our travel restrictions. So any questions that you have that are very detailed about methodology, I'll ask you to refer them to him, and his information is in the back.
I also want to thank Cory Hamilton, who is our previous WIC director ‑‑ I think we're just about to hire a new WIC director ‑‑ who gave us permission to use Ohio WIC data for this study. And then, finally, to Dr. David Shore, who is our division chief.
And I would be remiss if I didn't thank MCHB and HRSA for requiring in our SSDI grants that we link WIC and birth certificate data, because that really is the true impetus for us beginning to look at that, is that requirement. We just went over to them and said we are really supposed to do this. And after some years of negotiation, they have let us do that. So I thank you for making that a requirement. And also to Michael Kogen (ph) because Michael has been in conversation with Dr. Yee to talk about some methodological issues.
Now, the purpose of the presentation is to explore relationships between participation in WIC prenatally and birth outcomes, and we wanted to look at this by neighborhood income level. And I will say that the reason why we decided to look at neighborhood incomes is because Dr. Yee is a sociologist by training, and he really likes to look at the effect of neighborhood and area on health. And since WIC didn't have a burning question that they wanted answered with this, he decided to go down the road with looking at outcomes by neighborhood.
What I'm going to do through this presentation is talk about, start with the methodology that we used, and then to present our findings by birth outcomes, and then I'll give you a look at some of the differences in the WIC participants versus non‑WIC participants by some of their risk factors, the demographic factors, and some of their behaviors. Anything that we can find on the birth certificate.
Dr. Yee used a probabilistic linking method to link the 2001 and 2002 WIC prenatal files with the 2002 Ohio geocoded birth data. That link enabled us to incorporate all the maternal and infant variables from the birth into the analysis that you're going to see. And then these linked WIC and birth records were further linked to the U.S. census by census block group. And the census block group is a smallest geographic unit in which we can get
per capita income variables.
Dr. Yee then constructed three types of neighborhoods. And this is just an arbitrary construction. And he based it on per capita income. And by doing this, this allowed us to have some kind of a proxy for income because if we're going to compare WIC participants with other ‑‑ their outcomes with the outcomes of the entire population, we needed to be able to figure out some rough way of assigning an income level to them since the eligibility for WIC is 185 percent of poverty or below. And in Ohio , since women on Medicaid and CHP also can be on the WIC program during their pregnancy, we have a few women who are up to a 200 percent, roughly, of poverty that come into the WIC program just during the prenatal period.
Now, to look at the these income levels, the red represents the low‑income neighborhood, and Dr. Yee said, well, anything ‑‑ if an average income of a block group is less than 25 percent, the 25th percentile, that will be the low‑income neighborhood, and that cutoff in the census was just under $15,000 per capita. And we had just over 2000 block groups that fell into that category.
Conversely, the high income neighborhoods were defined as those neighborhoods where the income per capita was over the 75th percentile, and that mean income per capita was just over $23,000, and here again, just over 2000 block groups fell in that category. And everything in between was labeled medium income, just to help you figure out how we're assigning people.
We mapped this out for Franklin County in Ohio , which is the county where Columbus , Ohio is, where health department is. And as you see, the red would picture the lowest income block groups. And this is very consistent with what we know. So that gave us some feeling that we're probably going in the right direction.
Now, what did we get when we did the match? Of the 147,832 total resident births in Ohio , we were able to geocode 90 ‑‑ just a little over 92.1 percent, and we assigned those geocoded cases to block groups. If you can't geocode them, though, you can't assign them to a block group. So we had 7.9 percent missing a block group. These 7.9 percent cases were excluded from any other analysis.
And then when we looked at WIC, we had, in 2001 and 2002, 56,683 pregnant WIC participants who could have delivered in 2002, and we linked them with the 2002 birth files, and we were able to link about 90.4 percent of those pregnant women in WIC with the birth file.
A couple of factors could have influenced that link. First of all, if a child was adopted or if there was a paternity case or other court case, we would not have access to those files. Those are sealed in Ohio , as imagine they probably are other places.
If the mother did not carry the baby to birth, to a live birth, we would not have, because we did not match fetal deaths, we only matched live births. And, also, some women who started out in WIC could have moved and did not deliver in Ohio .
And the other thing you should remember is that we did not exclude anyone because of late entry into WIC. This is anyone who had any dose of WIC, which will be important, when we look at our results, to know.
So you can see that almost 35 percent of our total birth cohort had participated in WIC at some time during their pregnancy.
Then this slide shows you the characteristics of women by all births, to your right, in terms of what neighborhood they were assigned to and the characteristics by neighborhood of the women who were participating in WIC during their pregnancy.
So looking at all births, about 22 percent would be living in those low‑income neighborhoods, the neighborhoods where the per capita income was less than 25 percent, the 25th percentile of income. Conversely, 38 percent of the WIC participants were living in those low‑income neighborhoods. In terms of the high end, over 26 percent of all births were living in the high‑income neighborhoods, greater than 75th percentile, but about 10 percent of the WIC births were living in those sorts of neighborhoods.
The other thing you'll see here is that for all births, about 7.9 percent of the births were missing a block group. They couldn't be geocoded, so we had to exclude them. But roughly that same percent, 7.2, were also missing among the WIC women.
In order to make sure we weren't looking at some kind of a bias, or at least to figure out what our bias might be, we went in and examined what the characteristics of those women, infants, were who were in these missing block groups, and they were high‑risk group. And they were more likely to be in WIC, they were more likely to have very low birth weight. All the poor outcomes. Everything you could consider a risk was characteristic of this group. But you do need to know that they were not in our analysis.
So I'm going to start with the headline news, which is the birth outcomes. These are birth outcomes that we looked at. There's one missing off the list, and I was in my room last night figuring out why is this one missing, but I do have the tables for it, and that's just plain old low birth weight, but I'll talk to you about it.
I'll show you, for each one of these ‑‑ these are univaried analyses, and I'll talk a little bit about ‑‑ we have started to do some multivaried analysis along these to control for really quite a large number of factors and factors that are interrelated and kind of tell you in general where we are.
But if you look at the ‑‑ the dark‑colored bar is WIC and women who participated in WIC, and the lighter‑colored bar is a woman who did not participate in WIC during her pregnancy. And if we look at very low birth weight by income, you'll see that there is a statistically significant difference between those who participated in WIC and those who didn't, and WIC appeared to be protective. And I do say appeared because I don't think this is really the true case. And there was no significant difference between WIC and non‑WIC participation in those other neighborhood levels.
I also will want to say that when we did the multivaried analysis, in the medium‑income neighborhoods WIC became protective when we controlled for the other risk factors. But there wasn't a difference in the high‑income neighborhoods except for ‑‑ what Dr. Yee is doing now is beginning to look at ‑‑ he's beginning to tease out these factors by age group, by education levels, and he is finding that among the women in the high‑income neighborhoods, if they're in the age 20 to 34 years, WIC did appear to be protective for very low birth weight.
UNIDENTIFIED SPEAKER: (Inaudible).
RUTH SCHROCK: Statistically significant, yes. If I'm saying this it means ‑‑ it means statistically significant.
And you have these little bars that ‑‑ I know it's hard because some of the darkness makes it hard to see the bars, and I know that's true on your handouts too.
In terms of moderate birth weight, the pattern you see here for moderate birth weight is the same pattern that we have for low birth weight. And we see a protective effect ‑‑ this is statistically significant ‑‑ in low‑income ‑‑ of WIC for low‑income moms.
We're not seeing that same effect in the medium income or the high income. In fact, WIC is significantly worse than the non‑WIC women in the high‑income neighborhoods. However, when we do control for all the risk factors and all the confounders, we do see that ‑‑ what we see in the medium‑income neighborhoods is that we reverse this, that WIC becomes protective when we control for race and a number of others, age, income ‑‑ not income but education and that sort of thing.
For preterm low birth weight you see the protective effect only in the low‑income in the univaried analysis. We see some protection, though, in the high income when we do multivaried analysis, but only if the mother were in the age group 20 to 35 or if she were unmarried.
And I think this is going to be interesting for us as we begin to go down that road, beginning to tease out some of these different age groups. So if you look at it overall, it's a wash, but if you begin to look at some of the subgroups, you do see some protective effects even though for a whole there is no effect of WIC that we can see here.
Now, in terms of small for gestational age, WIC was not protective ‑‑ there was no effect, no difference statistically in the low‑income. WIC was not protective statistically significantly in the medium income or the high income. And this did not change in multivaried analysis. So WIC was not protective, given this kind of a study, against small for gestational age in full‑term infants. We're talking about infants who have reached 37 weeks gestation that are still less than 2500 grams.
And I think when you look at some of the characteristics in the risk factors of these moms, I think we can kind of have some hypotheses why this is.
The other index that Dr. Yee has used with some of our perinatal periods of risk work is something called the Hoffman's Index, which gives you a look at how many children are below the 10th percentile birth weight for any given gestational age. And while this shows some statistically significant difference in the unvaried analysis, when we did multivaried analysis there remained no significantly statistic differences. Now, this may be different if he looks at it, as I said, by some subgroups, which he has not done yet.
The next thing that we looked at, I just want to show you the picture of how these folks are different, WIC participants versus non‑WIC participants in the different income neighborhoods. When we showed this set of slides to our state WIC staff, they were not particularly impressed with the outcomes. They said, well, they thought that met their expectations. But what they were impressed with, when we go down some of these additional slides, was what is going on in the high‑income neighborhoods where they don't seem to target much. And they did not realize that there were as many differences or that there were maybe as many women on WIC who did come from a higher‑income neighborhood.
This is one that I don't think is truth. I know it's not truth. What I think is happening is we're ‑‑ since we allowed to be in this analysis anybody who had any dose of WIC, we really wanted to see the true effect of WIC on ‑‑ with women having multiple births. We would have to have some kind of a cutoff, because probably some of the women had their baby before they even had a chance to get into WIC, especially when someone women enter WIC late.
And when we look at our pregnancy and nutrition surveillance data that we send to CDC, this is not the picture we see. So we know we have to do this again, and I think our next speaker will show you some of those things too.
In terms of black race in WIC versus the non‑WIC, WIC is picking up a disproportionate percent of women who are black. And in Ohio , about 11 to 12 percent of our births are black. And you can see that there's a very high proportion of black births in the low‑income neighborhoods and that there is a decreasing percentage as income increases, which I think is another ‑‑ it's what we expect, but we can see it here and begin to document what some of the risk factors look like in terms of neighborhoods.
UNIDENTIFIED SPEAKER: (Inaudible).
RUTH SCHROCK: These I believe are statistically significant on their own. But you are right that I can't say for sure, but I would need Wyda here to help me. And that is a question, if I were you, I'd e‑mail him and ask him to confirm that with you.
Moms Hispanic ethnicity shows the same thing. We're serving a large percentage of mothers who are Latino or Hispanic. In Ohio , our total births are around 2 to 3 percent Hispanic. We're not a state right now that has a huge number of Hispanic statewide, but we do have large numbers depending on what county or locality you go into, and it's growing in Ohio .
WIC is serving the young women. When we looked at what percent of WIC participants were below age 19 versus mothers who were not in WIC, there's dramatic differences, and I think these are very real differences.
I got the 5‑minute sign so I will go a little bit ‑‑ I think you have the general sense, though, that we have ‑‑ you can look for yourself and see what these are. We're not serving the older women, partially because the older women may not being income eligible, because you can be living in a low‑income neighborhood and still not be income eligible, but you can be living in a high‑income neighborhood, especially if you're a pregnant teen, because you're eligible on your own behalf, and you would be eligible. So we're serving unmarried, we're serving, in general, women who have education less than 12 years.
I think the thing that concerns me is this analysis on parity looked at mothers who had four or more pregnancies, and we're serving these young women, and when you look at the percentage of 16 percent of both in the low‑income, both WIC participants and non‑WIC participants, 16 percent of them were on their fourth or more pregnancy. I think we have ‑‑ I think we're missing some golden opportunities here.
Weight gained during pregnancy, I think we're missing some very golden opportunities, because the women in WIC were not doing as well as the women who are not in WIC of terms of either not gaining enough or gaining too much. Of course, this is not a BMI because we're only getting what's on the birth certificate, and I think a session yesterday talked about some of those problems. But we know we have something to deal with, and these women are in our grasp. They are in WIC.
Then let me just go through these. Most of these that were ‑‑ previous preterm birth, there was no difference statistically in the low‑income neighborhoods whether the women participated in WIC or not. And in the medium and high‑income neighborhoods there were statistical differences against WIC. WIC consistently had more women who had pregnancy anemia.
Pregnancy induced hypertension, WIC did better in the low‑income ‑‑ in the medium‑income neighborhoods. No difference in the others.
Diabetes, no difference. Complications of labor and delivery, it was really ‑‑ WIC did better in the medium‑income, and there was no real difference in the low or high.
C‑section. The WIC moms were less likely across the board to have C‑sections, and that was statistically significant.
Congenital anomalies. No difference.
Delivered at a Level III hospital. I think one has to look at where women live, and a lot of the women who, even though in low‑income, they were statistically more likely to deliver in a Level III hospital, I think that we may need to look more about this is being where people live, this is their catchment area. The clinics that they go to may be there and they accept different sorts of pay sources.
In the high income it could be the worried well, it could be that they may be (inaudible) or advertisers or their insurance sends them there. I think there's some different things we have to look at before we can make up our minds about what this says.
No differences in this and this.
And behaviorals, real quickly, we are dealing with a group of smokers, and we've got them in our hands. And this is, of course, what we get on the birth certificate, however reliable we call this. But the women in WIC are smoking. This did not wash out in multivaried analysis at all.
We see no statistical significantly difference in alcohol. Here again, it's what we get off the birth certificate, and this is not going to be on the birth certificate in the revised one that we're going to start using next year.
Maternal prenatal care using the coddle check. In the low‑income neighborhoods there was a statistically significant difference between the WIC and the non‑WIC participants. The WIC participants were more likely not to have inadequate prenatal care. And those other neighborhoods, they were ‑‑ WIC participants were more likely. We're wondering if in higher income neighborhoods some of these may be teen pregnancies and they're not wanting to go for care, they're denying, denying, and they're not getting in late. I think those are some theories that we can look at.
As I said, this study raises more questions than we ever started off with. We've got 10,000 questions that we want to look at. Not a whole lot of ‑‑ I mean, there's differences in the medium and high income in terms of entering prenatal care at first trimester.
And I'm just going to ‑‑ we've just got a general statement here about the logistic regression, and Dr. Yee has whole tables that he could send you if you're interested. But just in low‑income neighborhoods, non‑WIC mothers were 20 percent more likely to have low birth weight babies than WIC mothers, without control of those other variables. In medium and high‑income neighborhoods, non‑WIC mothers were significantly less likely to have low birth weight babies than WIC mothers without control.
Wyda looked at these sort of variables for his logistic model, and when he did that, just looking at low birth weight ‑‑ because we could do this for everything ‑‑ in low‑income neighborhoods, after controlling for all the listed variables, non‑WIC mothers are 25 percent more likely to have a low birth weight baby than WIC mothers, which is a 5 percent increase in comparison with the no control model.
And in medium‑income neighborhoods, we switched the direction, as I said before, and now non‑WIC mothers are more likely to have low birth weight than WIC mothers. And in high income there was no difference except, as I said before, if we start drilling down on some of those subpopulations, the specific age groups and specific education levels.
So based on this, with its warts and all, we felt that there's a strong association with better birth outcomes in low and medium‑income neighborhoods if mothers participated in WIC, and that the strongest association is seen in women who participate in WIC during pregnancy in the low‑income neighborhoods.
But it raises additional questions. We need to look at dose effect, investigation, systemic biases, and we would love to link this with PRAMS, which we have PRAMS right in my area, and we can do that, and I think we will be doing that. And if we can negotiate with Medicaid ‑‑ Medicaid ‑‑ it's hard to link names with Medicaid. Medicaid agency doesn't want to give us our files easily, but we can, I think, work through and get into the files from them. So that's what we'll be doing next.
Thank you.