JENNIFER CULHANE: Well, good morning and thank you very much for inviting me here today and for taking time out of your busy schedules to listen to what I have to say. I actually don’t know what works, so I’ll just sort of say that up front. I do have some ideas and in the next 40 minutes or so or 45 minutes or so, I’d like to present sort of a broad overview of the work that myself and my colleagues are conducting in Philadelphia .
We have lots of federal sponsors for our research including the Centers of Disease Control, the National Institute of Child Health and Development, the Maternal Child Health Bureau, and recently we received a Center of Excellence Award from the Pennsylvania Department of Health from the Non Formula Fund Tobacco Settlement Dollars, so there are tobacco money hard at work in Philadelphia anyway. I’d also like to just say that there’s many, many co investigators on the projects that I’m going to talk about broadly. They range from statistical demographers to perinatal epidemiologists to immunologists, to theoretical and spatial mathematicians, and so these people are all extremely important to this research.
By way of a brief overview, and I think everyone in the room probably knows these things, but we do have a real problem of preterm birth in this country. And the incidence in the U.S. is higher than any other developed nation. We really have no decrease in preterm births over the last 40 years. The etiology of half of these events is still unknown and it’s clearly the leading cause of infant mortality and morbidity in the United States . And what we really, what I’m very interested in is trying to understand this persistent and large racial and ethnic disparity in the incidence of preterm birth.
The etiology of preterm birth is really roughly that about 50 percent of the events occur from spontaneous preterm labor, an additional 30 percent from premature rupture of the membranes, and about 20 percent are medically indicated for either maternal or fetal reasons. There’s some interesting data that suggests that the proportion of preterm births accounted for by indicated preterm births, medically indicated preterm births, are rising and rising the most sharply among African American women. And so I think it is impossible to tell accurately from the current birth certificate data if a preterm birth is actually spontaneous or indicated, and that’s something we really need to understand, especially as it relates to program and policy, but this is sort of roughly the picture.
I think it’s important to understand that the length of gestation in and of itself is not a very interesting variable in my opinion. If you’re born at 36 weeks of gestation with no sequelae, is length of gestation an issue? And my answer would really be no. What we’re really interested in is living or dying of the mother or the fetus, of long-term handicapped condition, and severe neonatal morbidities that translate often into very high cost to an already overstressed healthcare system. So the outcome of gestational length is really a proxy for all of these real problems. And the point is that about the 10 percent of preterm births that occur in this country really account for 75 percent of the perinatal mortality and at least 50 percent of the long-term neurologic handicaps. So it is an issue in that we’re really talking about mortality and morbidity.
What we can see from this paper is that we’ve done a really good job at increasing survival among the extremely small infant, less than 800 grams. And you can see that from about 70 to 95 there’s been a real increase in the survival or the survivorship of the extremely small baby. But you can also see that the prevalence of disability among the extremely small survivors hasn’t improved. So we’re doing a lot for survival, not much for disability and that culminates in a lot more infants surviving with disabilities, and that is a real public health problem and is in need of attention for, specifically, for the families that deal with these infants.
What, this is just a bar graph showing that the rate of preterm birth in this country has really not gone down. If anything in the last few years, we’re seeing increase in preterm births in the United States . And we can see that there is still a very large racial and ethnic gap as represented by the yellow triangles at the top, which are total non-Hispanic black preterm births. And the pink squares at the bottom, which, are total non-Hispanic white preterm births. Now there has been quote unquote a narrowing in the gap, however it’s really driven by an increasing white preterm rate, as opposed to a falling rate in the African American population.
If you were to go to the literature, I would say that these are the current explanations that are posited to explain the overall high rate of preterm birth in this country and the ration and ethnic gap. There’s been a lot of attention on the role of inadequate use of prenatal care as a driving factor. There has been lots of emphasis placed on maternal health behavior, potentially differential rates of adverse health behaviors among minority groups contributing to differential rates of preterm births.
That there is, that teen births, that higher rates of teen births in different race ethnic groups drive the difference in the preterm birth rate. That poverty or education somehow is translated into this physiological risk. That maternal stress is a player. And there’s really a lot of emphasis currently being placed on the role of genetic polymorphisms and gene environment interactions. I’m going to spend a few minutes showing you why I think that each of these, at least the top four, are inadequate explanations for either the overall high rate, and specifically for the racial and ethnic disparity.
First off, what we see in this country every year, it’s published in, I think the American Journal of Obstetrics and Gynecology or it might be Pediatrics, believe it or not, where there’s a review of the vital records for the year. And what we see every year is that the percent of women entering prenatal care in the first trimester goes up and up and up. And what we really see is that the fastest rate of increasing use of prenatal care in the first trimester is among non-Hispanic black women. So if this was actually driving a chunk of the racial and ethnic disparity, we should start to see, some of this reflected in the other numbers.
But, I find this sort of--here’s the proof--it’s sort of in the pudding. What you can see is that for women that enter care, for white women that enter care in the third trimester of pregnancy, right? Their preterm birthrate is significantly lower than the non-Hispanic black women who, enters care in the first trimester of pregnancy. And what you can also see, really, is that roughly across these three groups anyway, and these are extremely small numbers over here, the none or the unknown, what you can see is that by in large the trimester of entry into prenatal care, it doesn’t have a big effect on the preterm birthrate for each race ethnic group. Right? But really we see this very major difference between white and black women with women entering care in the third trimester, white women entering care in the third trimester, still having significantly lower preterm birthrates than black women that enter care in the first trimester or pregnancy.
So there’s been a lot of papers written on whether prenatal care, early or adequate prenatal care will somehow reduce the adverse reproductive outcomes such as low birth rate and preterm birth. And these papers have generally looked at the provision of prenatal care in populations, which have had none, providing more prenatal care where there was little or providing advanced prenatal care where only routine care was available.
And what have we seen? In my opinion we’ve seen a substantial decrease in maternal mortality. We’ve seen a substantial decrease in particularly late stillbirths. We have seen no, or at best marginal effect on preterm birthrates and really no marginal effect on very low, on low birth weight and even very low birth weight. So here’s just a graphic depiction with what’s gone on with maternal mortality in this country, and I think this can be largely attributed to advances, medical advances that have been implemented within the prenatal care system.
This has been a lot about managing diabetic women. It’s been a lot about managing blood pressure, et cetera. And so we’ve had this huge decrease in maternal mortality in this country. And you can see that we’ve seen a significant decrease in late stillbirths, which also might be attributable or it’s hypothesized that it’s attributed to fetal monitoring in the late gestational age ranges. Oh, interestingly, we’ve seen an increasing rate of early stillbirths, most likely infectious in etiology. And of course, just to keep with the race ethnic theme, there is a race ethnic disparity in stillbirth rates, both actually in early and later stillbirths.
So what’s going on with prenatal care? I would suggest that intervening after conception may just be too late to alter the course of that pregnancy. And that, I think another piece of evidence that supports that idea is that a previous preterm birth is the best predictor of a subsequent preterm birth, which means that in the interpregnancy period, some of the risk is retained in the woman. And so I think that this notion that intervening after conception occurs is going to reduce the likelihood of an adverse outcome in that pregnancy, specifically in the areas of prematurity and low birth weight is questionable.
Okay, well what about health behaviors. The premiere health behavior that has any association with birth weight and prematurity is smoking. Other behaviors, substance abuse, sexual practices, et cetera, have very little association with adverse outcomes and that’s something I think people are kind of surprised about. And what we see in this country anyway, is that non-Hispanic white women smoke at a much higher rate than non-Hispanic black women in pregnancy. So if smoking, which confers about a 1.5 to a two fold increased risk of low birth weight and to a lesser degree, preterm birth, if this was really driving this disparity, we would see and inversion in the trends. And we would see that non-Hispanic white women would have more preterm birth than non-Hispanic black women.
So what about birth to teenagers? What we’ve seen is that birth to women age 15 to 19 there is a race ethnic gap such that black women do have more births in that maternal age range. However, the fastest rate of decline has been in the non-Hispanic black population in the last 20 years and there’s an uptick in preterm births in teen births to non-Hispanic white women. But what’s really much more interesting and this is from our Arlene Geronimus’s Weathering hypothesis.
I took her paper and I redid it on Philadelphia data and I took first-born, so the women in the data set, their only primiparous women on their first births, I looked at neonatal mortality rates in about six years of Philadelphia linked birth and death files. And then I looked at maternal age and maternal race. And what I saw, and this is supportive of Arlene Geronimus’s work, is that for these three race ethnic groups, Hispanic, White and Asian, the teens actually did look like they had a disadvantage in terms of outcome, this is neonatal mortality, compared to women age 20 to 29.
Neonatal mortality is hugely driven by preterm birth. It’s deaths that occur in the first 28 days of life, and at least 80 percent is a direct result to early gestation. And what we see here though, is that for the non-Hispanic black women, the teens actually were protected in terms of neonatal mortality compared to women age 20 to 29. Now this is exactly what Dr. Geronimus found using national data. So then I took the whole national data set and I looked at very low birth weight by maternal age and race, and what we can see is that for the non-Hispanic White women, these dark bars, there is what we call a U shaped curve.
So the teens seem to have a disadvantage compared to women in the sort of 20s and early 30s. But for non-Hispanic black women, save the extremely young women, there is really more of a linear curve here. And so really what we see is that teen births to the African American, the higher rate of teen births among African American women is not driving the racial and ethnic disparity. If anything, it’s masking a bigger disparity.
So what about education and income? We know that there is a definitely a social class gradient in health. I don’t know how many people read the New York Times yesterday about the three different scenarios for the heart attack victims, the rich white guy, the middle class black guy and the poor immigrant white woman and they had totally different experiences with the health care system. Their health improved at very different rates and that was sort of an interesting depiction of the social class gradient in health.
But Ken Schoendorf and others took the national data, again the linked birth and death certificates, and they only analyzed infants that were born to two college-educated parents, as a proxy for class, because we don’t have income measures on our vital records. And then controlled for a lot of different things, age and parity and prenatal care use and marital status, and still within this "proxy group" for higher income, we still saw that black infants were about two times as likely to die in the first year of life compared to their white counterparts, and that this difference was almost exclusively attributable to excess preterm birth rates.
So, go back to the national data and we see that, yes, there is clearly an education gradient within each race ethnic group. So it does measure something. But what’s disturbing if you ask me is that non-Hispanic black women with greater than 13 years of education, still have significantly higher preterm birthrates than non-Hispanic White women that have less than a high school education. So, yes there’s an education gradient, but there’s something else going on here across race ethnic groups that it is not simply explained by, in my opinion, use of prenatal care services, by differential health behaviors, by the proportion of teen births, or low maternal education or SES.
Now, just by way of a little dig at the genetics crowd and I say this in places and I get, I have a lot of controversial responses, but I would suggest the genetic difference or polymorphisms probably have a lot to do with individual level risk for a disease outcome. If you have a certain genetic constellation, then you may have a greater risk for a certain disease state. However, I don’t think it explains group differences in health, because we know that the within group variability, lets just take Africans versus Caucasians and what we know is that the within group variability far outweighs the across group variability.
So it’s hard to believe that it’s driving that the genetic differences are driving a whole host of health outcomes. And I’d say you can look to this literature, mostly called the acculturation literature to support the idea that social structural factors are clearly at play in driving reproductive outcomes. And what we know is that if we look at Mexico born women that live in the United States , other people measure acculturation by language preference or duration of tenure in the United States . But in this paper, these papers, they looked at Mexico born women, and what they saw is that compared to Mexican American women, Mexican women born here, in one paper, and longer term immigrants, the Mexico born women used very little prenatal care on any measure of socioeconomic status they were poor.
They had very low levels of maternal education. They reported very low levels of prenatal stress. They had really almost negligible adverse health behaviors in terms of diet, exercise, smoking, alcohol use. They had positive attitudes towards their pregnancy and they reported high levels of social support. As they "acculturate," which means live here, things happen. They use a lot more prenatal care. Right? They actually become richer. They get more education. Their prenatal stress goes up, negative health behavior--this was mainly in the area of diet changes--they have shrinking positive attitudes towards their pregnancy and shrinking reports of social integration.
And what happens is that the preterm birth rate in this group looks just like White women in this country. And out here it looks really a lot like non-Hispanic Black women. And this happens real quickly. And so I would suggest that it’s not changes in the genetic composition of this population that happens in a generation or in seven years, or when you switch your language preference from Spanish to English, but rather what is sort of the fabric of your existence in complicated ways.
So on the medical side of things, these are a lot of interventions that have been tried to reduce preterm birth. And whenever they’re done in randomized clinical trials, as opposed to self-selection or voluntary participation, and that’s a very important difference, we see that all of these things from increasing prenatal care use, to enhancing services in prenatal care, to trying to do these risk scoring systems, nutrition counseling, caloric supplementation, protein supplementation-- which has been shown to be dangerous--iron supplementation, almost all tokolytics you get about two days, drug, alcohol, tobacco programs, bed rest--another one that’s actually dangerous--hydration, home uterine activity monitoring, the provision of social support in randomized clinical trials, my read of this literature is that none of them have reduced the preterm birth rate in vary tightly controlled studies.
So really most strategies aimed at preventing preterm birth have been shown--have not been shown effective in very defined and randomized trials and it’s not surprising at all really, that the prematurity rate has not declined, given that nothing in terms of social factors have actually changed.
One thing, because I have to mention this, is that this is a recent study that was done by the National Institute of Child Health and Development using what they call 17-P. It’s progesterone in pregnancy. And what they did see is that these were intramuscular injections in a very high-risk group of women where 55 percent of the women in the control group had a preterm birth. In the women that were getting this injection weekly, 36 percent had a preterm birthrate, so they see about a 40 percent reduction in the preterm birth rate and other neonatal outcomes also looked better, intraventricular hemorrhage, necrotizing entercolitis and oxygen--the use of oxygen. Now, what’s real interesting is that almost nobody agreed to participate in this study, because this hurts. And this background rate is huge. So, I don’t know exactly what happened, but I wouldn’t change clinical practice based on this study yet. However, it’s fastly seeping into clinical care.
So in conclusion to that part of the talk, I think that behavioral and health care utilization or genetic differences do not adequately explain the overall high rate or preterm birth or the persistent racial disparity that we see in this population.
Okay, so what are we doing in Philadelphia ? We are trying to actually integrate what we call three promising lines of research into one big research agenda. And we want to look at the role of what we call Individual Level Maternal Stress Exposure. This is cognitive appraisal stress--I feel--as well as life event stress--this happened to me--as well as sort of material hardship type stuff. We’re looking clearly at the role of infection and I’ll tell you why in a minute. I don’t think you can talk about prematurity without having infection as a major player in your theory. And then we’re also looking at the role of what we call contextual level maternal stress, which is really we’re looking hard at neighborhoods and as a way to sort of understand social structural explanations.
This is really many, many studies but largely the parent study, if you will, is a big cross sectional clinical prevalence study of over 5,000 women that are entered into the study at the time of their first prenatal care visit where we do an extensive face-to-face interview and evaluate the woman for a condition called bacterial vaginosis and I’m about to get into that. Then we follow them up for their birth outcomes and on subcohort we do lots of endocrine and immune measurements all through pregnancy and in the post partum period. Women that do not speak English or Spanish have to be excluded. And then we do a lot of other analysis using the whole city-linked birth and death files.
But I don’t want to spend tons of time going into all the little discreet methods. But this is just a map that shows the--it’s old now, so there’s a lot more dots. This is the distribution of participants in our study and this is Philadelphia . And what we were hoping for is big geographic distribution and we’re doing okay. This is almost exclusively white. You can see that we have what we call hyper segregation in Philadelphia . I don’t know if that’s the case in Chicago , but we have what we call very hyper segregated neighborhoods, so that it’s very hard to study women from different race ethnic groups that reside in the same neighborhood because they basically don’t. And it makes it very hard to do these things called hierarchical linear models, which I’ll talk about in a minute.
Okay, what we ask the women is a lot about the usual suspects, the health behaviors, her health, housing, financial circumstances, social support, we use lots of validated scales for stress and depression and anxiety, this sort of stuff. We get their address. We look for bacterial vaginosis and we link to the birth certificates.
So our first simple question is does maternal psychosocial stress, and this is cognitive appraisal stress, it’s the Cohen Perceived Stress Scale, I feel, kind of questions. Do women that have lots of cognitive appraisals, stress in pregnancy, have shortened length of gestation? And if you go to the literature, there's a lot of studies that actually have looked at just this simple question and about 75 percent of these studies do show relationship between prenatal stress and shortened length of gestation. About 60 percent only in terms of low birth weights, and the effect sizes range from about 1.2 to about 2.5. That’s not that different at all from smoking.
And although we’re not going to get into it very deeply here, I just want to say that there is really a plausible physiological mechanism, that that’s not just so fluffy that nobody can understand it. And there’s lots of groups right now working hard on trying to pick a part the mechanism between stress and actually prematurity. And very briefly, and really simplified, what we see is that the mother who has a stressful exposure from the HPA axis in her brain actually gets activated and this results in over production of cortisol. And in the pregnant state, cortisol actually up regulates the production of corticotrophin releasing hormone in the placenta.
Okay, so this is the placenta. So maternal cortisol up regulates the production CRH. CRH--this CRH is bioactive--it’s released both to the maternal and fetal compartments. In the mother it up regulates the production of prostaglandins, which has a lot to do with myometrial contractility and it also actually activates the fetal adrenal glands in such a way that this ultimately comes down to production of estrogen. Okay? And the change from a progesterone dominated environment, to an estrogen dominated environment is one of the things that may in fact, initiate labor, and this is Challis and others have written a lot about this.
And Roger Smith has shown that women, that the amount of CRH in the plasma, which by the way is not measurable in the non-pregnant state. It’s stays in your brain. In the pregnant state, the amount of CRH in the plasma is actually associated with the length of gestation, even early in pregnancy. So the amount of CRH seems to predict the length of gestation. Maternal stress can up regulate the production of CRH.
So really what we see is that there is an epidemiological association between maternal stress and risk of preterm birth. There is a positive association between maternal psychosocial stress and stress mediators, and that’s done in laboratory paradigms. You can actually stress someone and measure the rise in these things. And placental CRH does, in fact, predict the length of gestation. So it is a fruitful avenue.
Okay, the second line of research that we’re trying to integrate into this stuff is the role of urogenital tract infections. And what we know is that in the very early gestational age ranges, the amount of placentas that have evidence of histologic choreo immunitis, which is an infection, and this is sub-clinical. So the person isn’t sick, the woman isn’t sick, is pushing about 100 percent in the very low gestational age ranges. So the role of infection seems to diminish as term births--as the preterm birth moves towards term. Okay, but in the early gestational age windows, one has to look at the role of infection. So really, in the etiology of spontaneous preterm birth, we think that infection is a major player in here between say 20 and 30 weeks of gestation.
There’s a lot of other pathologies in the later preterm births and there’s even some preterm births that have no pathology at all. So there are babies that can be born at 36 weeks without any problems. And so those are probably just earlier births with no pathology, but infection down here is a major player.
So what about race ethnic differences, in urogenital tract infections? This is from the VIP study. And what you can see is that for every urogenital tract infection studied, African American women have a much higher rate than their White counterparts. Bacterial vaginosis is the most common urogenital tract infection, and I say infection hesitantly because it’s really not something that you acquire. It’s just an imbalance in the vaginal flora. But if you look at every study published, and some of these studies have huge numbers of women, Black women have much more bacterial vaginosis compared to, and I could only get into non-Black counterparts.
So in prematurity we also know that an infection can be chronic, which means that it can take weeks if not months for preterm birth to happen because of an infection, because we documented these infections in women literally weeks or months prior to a preterm birth. And women who have large quantities of vaginal bacteria, which we call bacterial vaginosis, and these are gram-negative anaerobes and there’s a lot of different species that make up BV, they’re much more likely to have these same gram-negative anaerobes in their uterus. So it looks like these bugs can actually ascend into the upper genital urinary tract and they may just be harbored there and then after the membranes seal in a pregnancy, in some women, cause a problem.
However, we know that treating BV in pregnancy has quite mixed results, with a very, very big study done by the Maternal Fetal Medicine Unit showing that there was no reduction in the preterm birth rate treating with Metronidasol in pregnancy. There are other studies that show a reduction in the preterm birthrate among women that were treated and there are still other studies that show harm, so that in the group treated with antibiotics, their preterm birthrate has actually increased.
So what we wanted to know is does stress, and we haven’t gotten to the contextual level yet, at the individual level, actually modulates susceptibility to the acquisition of infections. And then we wanted to know does stress actually increase the risk of a preterm birth in the presence of an infection? And then we wanted to know are there specific endocrine and immune profiles that we can actually measure in women that are stressed and are these profiles actually associated with the likelihood of having BV, the likelihood of having a preterm birth, or the likelihood of having a low birth weight birth, which I think is less likely.
So what we saw is just very simply that the probability of having BV as it relates to the self reported proceed stress is a line. That the more stress the higher the probability of having bacterial vaginosis and what’s real interesting is it went straight up from the BV negative women to the BV intermediate women. This is sort of mixed vaginal flora healthy and non-healthy, to the full-blown bacterial vaginosis. So we saw that there was an association between perception of stress and the likelihood of bacterial vaginosis in pregnancy. And when we adjusted for everything else under the sun, we still saw that there was a, these women were still about two times as likely to have a preterm birth compared to their non-stressed counterparts.
So what is the mechanism? Behavioral? Highly unlikely because we looked at douching practices and sexual practices and a lot of other things, smoking has been associated with BV. We looked at all that and we couldn’t explain it. So is there a physiological mechanism? Is it endocrine? Does CRH or cortisol have something to do with it? What about immune? Is there a suppression associated with stress? Is there a disregulation in an inflammatory response in something called the TH1-TH2 response? So is there an endocrine immune interaction?
We’re not going to talk about this but we are just finishing a study where we measured all the endocrine and immune perimeters that might in fact, interact and cause a host to be more pro-inflammatory or immuno suppressed. we don’t really know, at three time points in pregnancy and two times in the post partum period. And our early preliminary data that we just presented at the Society for Maternal Fetal Medicine, in fact, shows that women that are stressed, this is CRH, remember?
This is a direct result of more maternal cortisol. Women that have more CRH when their cells are stimulated with something that looks like a bacteria, produce a lot of a pro-inflammatory cytokine called TNF Alpha. So in other words, it looks like stress when you show cells an infection, make a lot of pro-inflammatory cytokines, which is likely to under pen the pathology of infection. It’s really the host response and not the anaerobes themselves.
Lastly in the infection category, I wanted to mention this periodontal disease because there’s lots of research activity going on in the world right now with a huge trial in North Carolina and Alabama called the Motor trial, where they’re actually taking women that have periodontal disease in pregnancy and treating them and looking for reduction in preterm birth because their preliminary data published, I forget what JATA is, but it’s, okay, actually shows that women, that these are adjusted odds ratios that women that have preterm birth at less than 32 weeks, they have an eight fold likelihood of having a preterm birth at less than 32 weeks if they have periodontal disease compared to their non-periodontal disease counterparts. So the effects seen earlier, which is very in keeping with the infection stuff that we know about in terms of the distribution of preterm birth. So periodontal disease, interestingly, are also gram-negative anaerobes.
The third promising line of research is the role of neighborhood or social context. And really what we wanted to do is try to find a way to understand where women live. Not just what they perceive, in terms of their cognitive appraisal stress or their own life event stress, but does neighborhood or context actually matter? So we’ve taken administrative and public health data that has address level information and we geocoded these data, which means we’ve just assigned it a point, an X and Y coordinate where we can see its location on a map, and we’ve aggregated these phenomenon to various levels, census boundaries, everyone’s heard of block groups and census tracts. We’ve also made what we call raster identities, which are continuous surfaces in space.
And we’ve also programmed our own surface called, what we call kernel densities, but I’m not going to talk about that stuff today. It’s too complicated. What we, these are the data sets that we’ve been working with. For instance, shelter episodes. In Philadelphia anyone who requests an emergency shelter stay, had to go to a centralized location and they have to have a last Philadelphia address to actually get into the shelter system. We have the address of every lead test in the city. We know data on school truancy and we know test scores at the first and third grade.
We of course, have all the census data. We know where utility cutoffs happen. We know where every case of gonorrhea, syphilis and Chlamydia occurred, in terms of an X and Y coordinate, in the city. And our goal was just to investigate the relative importance of community context, as well as individual characteristics and their interactions on the risk of bacterial vaginosis in pregnancy, preterm birth, and low birth weight.
Now, the role of context sort of faded out of public health. At the turn of the century, not this past one, there was lots of interest in the role of neighborhoods and public health. And then we really entered a phase where we had what we call a Wariness of Ecological data or the Ecological Fallacy, taking a measurement at a group and ascribing it to an individual. We had tons of advances in statistical methods that kept a lot of people busy doing models on individual level data. We had really what I consider a methodological conceptual and political individualism, right?
Dominated this place since the 1950s, and we didn’t, the epidemiologists and demographers were really separated and I see many places now where they’re coming together. But since about the 1990s there’s been a lot written in the literature for what they call the new public health. And this is calls to look at the upstream social conditions that may actually underpin risk of disease. Link & Phalen wrote what I consider a sentinel paper called Social Conditions as the Fundamental Cause of Disease. And so there is call to simultaneously examine sort of the distal and the proximal causes. And so neighborhoods have come back into the literature.
And so we really want to know, does where you live affect your health, independent of your individual behaviors, your individual level of income, et cetera. And so what we did is we just started taking, looking at crude associations in Philadelphia . For instance this is, we have for every building footprint in the city, we have the tax delinquency. So for any building in the city, we know its market value and we know the proportion of its market value that it owes in tax delinquency. And so at the Block Group level, we looked for properties that owed 20 percent or more of their market value in tax delinquency as a sort of proxy for neighborhoods heading towards abandonment.
And what we, and they we just looked at the preterm births by these tax arrearages and we see that there’s clearly an association. The more tax delinquency, the more preterm birth, unadjusted for any individual anything. And then we did the same thing for homelessness, for women that had entered the shelter system, and again we can see--isn’t that amazing how there’s sort of like a real line right in there? You see that as homelessness goes up, the amount of prematurity goes up and these are at the Block Group level.
So then we took data and we took data from ‘90 to ‘95, this is tract level data, and this is data on this side from ‘96 to 2000, okay, so two different swatches of birth certificate data. And then we looked at, they were all geocoded, and then we looked at the, by the census tract this time, the percent of births to women with less than a high school education. Now, these are not the same women? Okay, but what we can see is that their census tracts, that the amount of births to less than high school educated women stays very stable within a census tract.
So if you are in a tract with high births to less than high school educated women in ’90 to ’95, then the tract has basically the same rate of births to less than high school women in ’96 to 2000, very stable. Same with births to single women and same with smoking, so there are tracts that produce it seems like over and over again high rates of antenatal smokers. What is a little disturbing is that when we looked at the preterm birth outcome, the line wasn’t nearly as consistent, right? The correlation was not nearly as strong in fact it wasn’t significant. It’s highly significant in the other pictures I just showed you.
So if there is a neighborhood effect on preterm birth, it’s small and it’s hard to find. And the data absolutely bears that out if you look at the other published reports. So what we want to know is the differential neighborhood conditions potentially contribute to the race ethnic disparity in reproductive outcomes. So we started to look and see if these exposures varied by race ethnicity. This is no longer in press. It’s out in the American Journal of Public Health. And what this slide shows you is that for--we took two years of birth certificate data and they’re unduplicated, so it’s just, a woman is in the records once. And we merged it into five years of shelter intake data.
So we looked at a five-year period prevalence rate of request for emergency shelter. And we looked at this by maternal race, maternal parity and maternal education. So this says that Black women with a high school education or less, where their birth, their index birth, represented their fourth or greater birth, about 42 percent requested an emergency shelter stay in a five-year period.
For the same demographic, except race, for White women, so in terms of education and parity, it’s about five percent. So there’s a huge difference by race ethnicity and you can see that the Latina women, which in Philadelphia are almost exclusively Puerto Rican, they look a lot more like White women. So, in fact, homelessness, a measure of severe housing instability is much more prevalent among African American women in Philadelphia .
So then we took the address on the birth certificates and we got crime data from the police department, it’s about 10 years of data, where, again, geocoded records, where we know where every, gun related death occurred in the city. And so we took the address on the birth certificate and we drew a concentric circle around that point in space at a bandwidth of 500 feet and we looked to see the proportion or the percent of records by maternal race that fell within a circle, so that they were close to a gun related death. And what you can see is about 35 percent of the Black infants had an address on their birth certificate that was within 500 feet of a gun related death, and it’s about five percent of the White women.
So then independent of the individuals from our study, we actually took the women that were in our study, and we looked at their Block Groups. We looked to see where they came from. And we looked at shelter use, tax delinquency, and assaults. And what you can see is that, and in this study, I should mention right here, that all the women were enrolled from Medicaid clinics. Everyone is poor in terms of their individual level income. So this is not a measurement of individual level income. So everyone in this study was on Medicaid or uninsured. And what you can see is that the Black poor women were far more likely to live in disadvantaged neighborhoods compared to their White and Hispanic poor counterparts.
So then we did some complicated statistics and we tried to see if we actually compare the preterm birth weight for Black and White women in the city, and this is just a crude--this is a logistic regression. It’s just a simple odds ratio. And the green bar then over here is that same logistic regression model just controlling for all the individual level characteristics we can get from birth certificates. So, the usual thing smoking, prenatal care use, and you can see that we retake about, it’s about 3.3, Black women are about 3.3 times as likely to have a preterm birth in just a crude odds ratio compared to White women. When we adjust for the individual level stuff, it’s around 2.3. So we get rid of some of the difference.
In this model right here, we actually introduced in something called random effects model, a bunch or neighborhood characteristics. And so what you can see is that, in this model right here is unadjusted for the individual level information, just looking at neighborhood effects alone, we achieve around a two point, we get another, you know, reduction to about a 2.5 times as likely and then we add the individuals and we get down to two. In this model where we only compare Black and White women that live in the same neighborhood, so in effect we hold neighborhood affect constant and we adjust for individual level characteristics.
The Black/White odds ratio gets to be about 1.8. So this, there’s still, even looking at neighborhood constant individual level attributes all adjusted for, Black women in Philadelphia are still about l.8 times as likely to have a preterm birth compared to their White counterparts. In a crude odds ratio adjusted for nothing, it’s about 3.3. Let me just move on.
Okay, so what can we do? That’s, we’ll arrive at that point. Well, I think that there’s a few things that we can do actually. We can look at multiple interventions simultaneously because we think that risk factors co-occur in populations. Okay? And most studies that have tried to intervene, to reduce preterm birth or low birth weight have focused on a risk factor for lets say, smoking. They fail to address also, depression, housing instability, diet, et cetera. So I think we could try studies that look at multiple interventions simultaneously. I think we have a real problem with the timing of interventions and I think after conception is largely too late. It’s largely too late for antibiotic use. It’s probably--it’s very hard to intervene behaviorally.
On women that smoke in the antenatal period, you know, there’s one study that paid women a big chunk of money to quit smoking during the antenatal period, and they did until about five hours after delivery. So, you know, and they, and I’m talking hundreds of dollars, it was a lot of money. And they just went straight incentives. And they had a fairly good cessation rate compared to counseling and education. And I think that we need to understand that community and social interventions are really important and risk factors that are at the community level are absolutely disproportionately visited upon minority populations. And the race ethnic and social class gradient may not in fact improve with just biomedical approaches.
So in Philadelphia we’re just undertaking a very large randomized clinical trial where women in the inter-pregnancy period who have just had a preterm birth at less than 34 weeks of gestation, are randomized to an observation group or a treatment group, and these are the things that we’re targeting. We’re aggressively treating all urogenital tract infections at multiple time points, one month, six months, 12 months, 18 months, and 24 months post partum. We’re screening and treating with as best therapy as you can get for periodontal disease. We’re working on abnormal BMI, both heavy and thin because interestingly, in the preterm birth world, women with pre-pregnancy BMI’s of less than 19 are the ones that are really at risk. It’s not obesity. It’s absolutely thinness.
Depression, there’s, you know, mixed data on the association between maternal depression and risk of adverse outcomes. There’s some very interesting physiological mechanisms. Smoking--this is a bear. It’s very, very hard. I can tell you right now women that smoke during their pregnancy or quit after they learned they were pregnant, we try to start that before they’re discharged from the hospital, and 70 percent of the women eligible will not even participate. They do not want to be in the smoking intervention. And this is my favorite because we are actually doing something about housing instability. We actually have put a roof on someone’s house. And these are grants, they’re not loans. We bought somebody a stove last week.
We’re doing, we have a fund to actually try to improve their housing and we have a health literacy intervention. We’re following these women for two years, because it’s all the money that we have. We know that 50 percent will have another birth in that period. And the last data collection time point is at 20 weeks of gestation, and the obvious major outcome variable is reduced preterm birth rates in the intervention group. So finally, I think really that we’ve focused tons on these proximal causes in public health lately, in the last 50 years. We’ve done, we have a huge emphasis on health behaviors, healthcare use.
You know, we’re starting to move into stress physiology and this is a dike. Basically, I think that these pressures that have to do with social inequality push on these proximal causes, and as you plug up one without reducing this pressure, they’ll manifest through another hole in the dike and that this is what has to do with face ethnic differences in health outcomes. Yes, these proximal causes are in the mix, but without attention to these upstream factors, we probably will not see big reductions in race ethnic disparities in cardio vascular disease, in diabetes, in reproductive outcomes, you pick it.
And then, I would like to thank the March of Dimes for all their work on remediation of preterm birth rates in this country.
Thank you.