Ninth Annual Maternal and Child Health Epidemiology Conference / December 10-12, 2003
Changes in Level of Prenatal Care Between Pregnancies
in Relation to Preterm Birth
BAO-PING ZHU: If you would turn to page 96, please. I don't have handouts, so this will be helpful. Data make me happy. I miss a time and since I--there is a lot of time for me, I have a little bit of time to talk about different things. I miss a time that I could actually get my hand dirty playing with data, but now I'm the State Epidemiologist; I don't have time to do anything anymore. Anyway, this study was conducted before it went to Missouri, just in case you might wonder why I use Michigan data. As background information, low birth weight preterm delivery, the second leading cause of infant death, is the leading cause of death for African American infants, is the leading cause of neonatal death, is the leading cause of pregnancy-associated hospitalization among women. However, the etiology is still largely unknown. There is a lot of debate about the role of prenatal care as to whether there's any causal relationship with the infant outcomes. However, most of the studies are based on cross-sectional data, and the association may be due to confounding critics say.
This graph shows the trends in the preterm delivery in the United States between 1981 and 2001, which is the latest year of data, because as you can see that--if I can get the mouse to work--the overall trend is actually increasing. In '81, it's below 10; it's about 9.5. In 2001, it's close to 12. So the overall trend is increasing in the U.S. population, but that trend is largely driven by the increase among the white women. Among black women, it increased from '81 to about 1990 and then decreased a little bit. So the racial disparity is also decreasing a little bit in recent years. The objective of this study is to study the preterm delivery. PTB stands for preterm birth in relation to changes in prenatal care using a longitudinal design. The reason we did use the longitudinal design was that the most other studies have used the cross-section of design, and people would criticize about that.
And the data for this study, we used data from the maternity-linked Michigan birth records from '89 to 2000, and the linkage method was deterministic linkage. We used the social security number, first name, middle initial, and last names, and date of birth and other information to link the records. We identified during this period 422,590 Michigan women who delivered over 565,000 singleton live births whose birth orders were second or higher. The reason for that is that we need to study at least two consecutive births in order to study the changes in prenatal care between two pregnancies. We measured prenatal care using the color chart index, which categorized prenatal care levels into adequate plus, adequate intermediate, inadequate or no care, and unknown. We eliminated the women who had adequate plus because of concerns about preexisting medical conditions. We also eliminated unknowns for obvious reasons.
This is just a recap of how Kotelchuck classify the Kotelchuck. On the horizontal axis is a number of visits as percent of ACOG recommendation. So if it's less than 50 percent, it's always inadequate. So in this area, it's always inadequate. If it's less than 50 percent as a percent of ACOG recommendations--ACOG recommendations are based on the gestational age. It's always classified as inadequate or if the prenatal care started after the fifth month is always inadequate. And then based on the different combinations, you get intermediate, adequate, and adequate plus care. Now, I was struggling as to how to explain the statistical methods that I used in this analysis, and then I remembered from my statistics class that I was taught that if you use notations to confuse people, they might think you know something. So what I did was come out with a little notation. Basically, you start with the previous pregnancy. You had three categories: adequate care, intermediate care, and inadequate care.
And this is the risk for preterm delivery. So it's R1--one stands for this first level. Dot--it means we have not categorized yet. The same goes for this. R2 is for intermediate level, and inadequate level is three dot. And then remember, this is a longitudinally linked data set. These women will follow it up until they had theoretically--you know, we're not knocking on the door and say, "Are you pregnant yet?" So we linked the records, so using the retrospective method, we followed up them until they had the subsequent pregnancy. So these women would have three different kinds of possibilities for the subsequent pregnancy. They either state--the first pregnancy is adequate, and the second pregnancy is also adequate. And they also could change from adequate care for the first pregnancy to intermediate care in the second pregnancy. They also had adequate care for the first pregnancy and inadequate care for the third pregnancy. So the second thing here represents the second level of second pregnancy. So there are two methods that I used to--two measures, I guess. One is RD, the risk difference, which basically takes this one, the risk here minus risk there. The reason we did it, the RD, there were two reasons. One is that it looks like the association was additive, not really multiplicative, if you know what that means. So we tried to use RD, which basically is this risk minus the previous risk for the same cohort of women.
So the same goes here. And the problem with risk difference analysis is that it's not easy to do multivariate analysis, and these days if you don't do multivariate analysis, they think that you're not prepared as a scientist. So we kind of had to use the risk ratio to study which--this risk ratio is basically this risk. I use this one as reference tool. Basically, those women who stayed adequate--so in those words, had the best outcome--all those women who had intermediate care for the first pregnancy but had improved to adequate care during the second pregnancy, all the inadequate care for the first pregnancy and adequate care for the second pregnancy, and these will become the reference group. This is the best scenario, will become the reference group. And again, the ratio of this risk divided by that. And because these would be different women, different cohorts of women, we needed to control for the confounding variables. RD because they were the same women, we did not need to control for confounding (inaudible). I hope this is helpful for you to understand this.
For the analysis, we stratified the data by parity, basically in the first-second births, and second-third birth, and third-fourth birth. For the second set of analysis basically which is the risk ratio analysis, we actually use logistic regression to control for the number of variables: mother's age, mother's education, Medicaid status, smoking during pregnancy, (inaudible) acknowledge because in Michigan, prior to 2003, they did not collect information about marital status. So we used this as a surrogate measure for marital status. We also controlled for outcome of previous pregnancy, basically, a live birth or fetal death. Okay, now some results. This graph shows the percent of preterm delivery by change in the prenatal care. As you can see, I think there are--well, let me show you, those women who stayed adequate for the first pregnancy, the second pregnancy is also adequate care. There is very little difference in the risk, the percent of preterm delivery. There is also not very much difference in the adequate for the first one to intermediate in the second one.
However, if the first one is adequate pregnant care but the second one is inadequate care, the difference is really huge. The same goes for the other categories. You see the differences are really here. These are the first case--in this case is the adequate, the first pregnancy, inadequate was second pregnancy, and here is the intermediate for the first pregnancy and the inadequate for the second pregnancy. The third case is the first pregnancy was inadequate care, and second pregnancy is inadequate care. There's a large decrease here, okay? Okay. And also, this category also had a pretty big difference. The first one is inadequate care, and the second one improved to intermediate care. The rest of them, the risks were not very different. That's the first and second births, and this was the second-to-third birth pairs. Remember, these are the birth pairs, and it is with the same women follows through time. The same goes here as well. That was a visual way to show the difference in really the risks.
This graph shows the risk differences; it basically shows the similar picture. As you can see, if the first pregnancy is adequate care and second pregnancy is adequate care, there's a slight decrease but not statistical significant. The light blue bars are not statistically significant. The red bars are statistical significant differences. So if they state adequate care from the first pregnancy to the second pregnancy, there's a slight non-significant decrease. Same goes here. But if they had adequate care for the first pregnancy, however, the second pregnancy is inadequate care, there's an increased risk. This is an absolute increase in the risk, 5.2 percent, is highly statistically significant. And here, if the first one is intermediate care, the second one is adequate care, there's not much difference. There's a slight decrease but not statistically different. And then the same goes here. If the intermediate care for the first pregnancy and the intermediate care for the second pregnancy, basically there's no difference. However, if the first is intermediate care and the second is inadequate care, then the increase in the risk.
Here it shows just the opposite of the picture. If the first is inadequate care, the second one improved to adequate care, there's a large decrease in the risk. And the same goes here, and the same goes--and this one is if the first one is inadequate care and the second one is also inadequate care, there's an increase in the risk. Makes a lot of sense. It's difficult to explain, but it makes a lot of sense. Basically, the message is if they improve the prenatal care from the first pregnancy to the second pregnancy, the risk for preterm delivery decreases. If the care level got worse, then they're increasing the risk. So this is for the first and second pregnancy, and this is for the second and third pregnancy. Basically, a very similar picture, very similar picture. And then the second set of analyses, I used the logistic regression. Remember, the other way to do it is to look at the ratio. And these two bars are using the adequate for the first one and adequate for the second one as a reference group, and if a woman changed from adequate care to inadequate care between two pregnancies--this is from the first and second pregnancies--the risk for prenatal care is five--I mean the relative odds ratio--this adjusts for the other variables--is five point three. And this one has a smaller but also statistically significant odds ratio.
And this group is using the intermediate care to adequate care as reference group, and if a woman changed from intermediate care to inadequate care, there's a large increase in the risk: 4.1. And these two bars shows using the improved category from inadequate care to adequate care as a reference group. If a woman stays the same, there's a much higher risk: 4.4 times as high in terms of odds ratio, and there's also an increase risk or a significant risk there. And then the second or third birth pairs, the pictures are very similar. Again, the logistic regression analysis using the risk ratio also demonstrates that if the care between the first pregnancy and the second pregnancy improved, then the risk for preterm delivery decreases, whereas if the care level between the first and second pregnancy got worse, then the risk would increase. There are a number of limitations in this study. The first one is the effect of the prenatal care or are these observations, these findings, due to the effect of prenatal care or the effect of changes in life? In other words, we don't know why a woman had adequate care for the first pregnancy and inadequate care for the second pregnancy. It could be the effect of prenatal care. It could also be changes in life.
For example, loss of insurance, divorce, or some other stressful factors in the life. And with the vital statistics data, there's no way to answer this question. It may be the subject for further studies. But there's a real possibility that it could be due to confounding, although we controlled for the other confounding variables. The second limitation is that because the linkage, the vital statistics data we assembled covered a very long range from '89 to 2000, so there could be interstate migration. However, just a couple of months ago, I saw in the "USA Today" that between I believe 1980 and 2000, only about 15 percent of people have moved out of state. Remember, all these variables, all the birth records were assembled for the entire state of Michigan. So we estimate that that migration should be less than 15 percent in 20 years' period, then from '89 to 2000, it should be less than 15 percent. So in other words, we have about 85 percent of the women that did not move out of Michigan. The third limitation is the inability to link all possible records. As you can see, there could be records that were not linked, and the effect of that is unknown. And the fourth one is the common problem with inaccuracy of gestational age estimation. As we all well know that the gestation estimation is not always accurate. Okay, keep those things in mind.
Here are the conclusions. The change of prenatal care between consecutive pregnancies was associated with risk of preterm delivery. If the prenatal care got from better to worse, then the risk would increase. If the prenatal care goes from worse to better, then the risk would decrease. It looks like the observation was independent of the other reproductive risk factors, and so that advocates for adequate prenatal care for all pregnancies, not just the first pregnancy, not just the previous pregnancy, but for all pregnancies. Torture the numbers, and they'll confess to anything.