 I'm delighted to introduce Professor Diane Lauderdale, who is a professor in the Department of Health Studies here at the university. Is that on too? Thank you. I don't know. Professor Lauderdale holds degrees in religion, divinity, and library science from the University of Chicago and from Harvard. She currently directs the Master of Science program for clinical professionals in the Department of Health Studies. Today's topic, part of our Disparities series, will be births to Arab-American women before and after 9-11, evidence of stress effects. I've been asked to remind everyone that during the question and answer session, you're supposed to hold this to your mouth as I forgot to do during my introductory remarks. But I will remind you as you go. So it's a pleasure to welcome Diane. Thank you all for finding this unusual room and thank the hospital for being kind enough to let us use it today. Diane, thank you. Thank you, Mark. I did want to make one little addition to my introduction. I do actually have a doctoral degree in epidemiology, too. It's not usually the one that gets left off. I like to have the library science and divinity acknowledged, too, though, because they're so odd. So when Dr. Siegler called me a few months ago about presenting in this Disparities series, I gave him several alternatives. And he was most interested in hearing this piece of work, which I'm happy to present, but it is work that I completed a few years ago. And so if we have time at the end, I can talk about the reaction to its publication. But he was most interested because being a study of Arab-Americans, it was distinctive in the Disparities literature. So the purpose of this presentation, and can you hear me at the back? Yes, great, is to examine whether risk of low birth weight or preterm births increased for Arab-American women who are pregnant during a time of heightened ethnic discrimination, which is the weeks immediately following 9-11. A little background. So there has been, for the last 15 years or so, a lot of interest in how discrimination experiences affect health. Needless to say, this interest has grown from observations about disparities in health for African-Americans. And a number of outcomes have been examined, mental health, self-rated health, hypertension. And there was particular interest, I would say, in the first half of the decade we just left in birth outcomes. And this project grew out of that interest. So again, that literature was focused on African-Americans. Birth outcomes were a really attractive laboratory to try to study whether discrimination affected health for several reasons. Poor birth outcomes are common. At least 10% of births are preterm in the population. And that's common for a disease outcome. There are big disparities by race. Birth offers an interesting critical period for experiences, because clearly there is the nine months of gestation. And one can examine experiences that happen during that time at different points in that time or earlier in life. And so there are some interesting issues there which are clearer to try to articulate than if you're looking at something like mental health or hypertension. And finally, it was of interest because there was a potential biological mechanism, which I'll talk just a little bit about, but this truly is not my area of expertise. So the biological mechanism is corticotropin releasing hormone. So this is a hormone that's released in the placenta during pregnancy. And there have been several studies, not really a boatload of literature. But there have been several studies that have found that it is responsive to stress experiences, that the level goes up and the trajectory of increase during pregnancy is accelerated in women who report more stress of various kinds during pregnancy. And that, furthermore, independently of those studies, other studies have found that higher levels of CRH and that increased trajectory of increase is associated with an increased risk of preterm labor. That's from several studies and an increased risk of fetal growth retardation from far fewer studies. But of course it's just a risk. It's not that it's destiny. So like many biological mechanisms proposed in social epidemiology, it's intriguing. I would say it's not really rock solid. So here are the previous studies, though these are about experiences of racism or race-related discrimination. And they're all about African-American populations. And again, this is something else I just went to look. Is there been anything new in the last three to four years? And there don't really seem to have been any major studies that have changed the story from the early 2000s. So these are the stories one sees at the pieces of research one sees cited most recently. Some are case-control, some are cohorts. So in a study by Collins and others, and he's in Chicago, women with very low birth weight babies were more likely to report exposure to racism during pregnancy. So that's a case-control design. And you could wonder whether, after having had this very low preterm baby, whether that changed how women thought about what had happened to them during the pregnancy. So there's a little question about that. But then there are some prospective studies, the Rosenberg et al, that women prospectively reporting unfair treatment on the job during pregnancy were more likely to have preterm infants. However, this was just one of, I think, nine different kinds of discrimination hassle experiences that they were queried about. And it was the only one with a significant finding. So that's kind of weak evidence. In another prospective study, a big study out of North Carolina, women who prospectively reported racial discrimination were more likely to have preterm infants. And they sort of had a more summary instrument. Now these are very small associations that kind of just hit statistical significance, all of these. And in another study by Collins, women with very low birth weight, again, were more likely to report retrospectively chronic discrimination at work or finding a job. But in this study, which asked both about chronic and during pregnancy discrimination, there were no associations during pregnancy. So the literature is kind of muddy. There's some evidence. It's controversial. Not a clear cut story. But there are some problems with these studies. The biggest problem is measuring discrimination on the basis of self-reports. The women perceiving something as discrimination and then choosing to reveal it to an interviewer, there are different factors that would cause those to happen. So the same experience objectively, one woman might see as race-related and another woman might not see it as race-related. And it's altogether possible that the predictors of seeing something as race-related or not themselves are psychosocial factors which may be related to birth outcomes. It's very hard to unpack this because it's only a small proportion of the experiences that are just totally unambiguous. And even those, some women, they'll brush them away. They won't even remember them. And others, they'll really hold onto them and report them differently than others. So the self, and the people studying discrimination and health are totally aware of this. There's very articulate work that's been done by Nancy Krieger and others sort of unpacking what a problematic measure this is for self-reports. And in particular, studies that have tried to be a little more sophisticated have had separate instruments looking at those psychosocial factors that might predict the identification of something as a discrimination experience given the same experience. And so there have been these complicated interactions trying to look at coping style, internalization, and how that affects the report of discrimination on health effects, so complicated. So that's one problem with the self-reports. Another is, again, the confounding problem that all kinds of social factors that you can't measure or measure well are very likely correlated with the discrimination experiences. So certainly, certain kinds of women in certain kinds of situations are more likely to be the victim of discrimination either during the pregnancy or throughout their lives. And those kinds of structural, contextual, personal, individual factors may themselves have to do with birth outcomes, so that's complicated too. There is furthermore this little issue of trying to establish whether their experiences during pregnancy versus throughout life, and of course they're correlated. So how can you unpack that? But they speak to different potential mechanisms. And then another limitation in this literature is it has entirely focused on African-Americans. And when you're looking at the discrimination experiences, it's kind of nonsensical. Although studies do it, ask white women about their experiences of race-related discrimination. And who knows what you're getting there? I mean those seem like an odd group of women who would identify things as because they're white, things happen to them. And so the studies necessarily are comparing within the African-American community women who are perceiving things as discrimination versus not, and so that's sort of a limitation too in the study design. So what I did with this study was I sort of took a whole different look at the idea of race-related stress and discrimination and birth outcomes. So I focused on this group which had a sudden experience of a dramatic increase in ethnic discrimination. And I'll go back to remind you what was happening in the weeks just after 9-11. It's a before-and-after comparison. And the value of a before-and-after comparison is that the lifelong experiences of discrimination are likely similar for women who were pregnant right after 9-11 and women who were pregnant a year earlier, which is who I'm comparing them to. I'm avoiding self-reports altogether. So it's a natural experiment or if you please, an instrumental variable type approach in that there are some women who, while they were pregnant, earlier in their pregnancy suddenly were hit with this very stressful time period, and other women who were not pregnant and had nothing to do with other things about their lives. And so I'm not looking at comparing individual women who did or did not feel victimized by anti-Arab sentiment, but rather looking at a period effect at a population level. So in several ways it's a different population, a different kind of study design, a different measure of discrimination. It's very different from the other studies that were done. So to move back 10 years and remind ourselves of what this was like, and it does have the reaction. So the last year, again, anti-Arab, anti-Muslim sentiment has been much in the news. And it's actually had a very different treatment, I would say, than what it had 10 years ago. So 10 years ago there was really a great deal of instant national from President Bush and others reaction to try to combat the immediate anti-Arab sentiment and anti-Muslim sentiment following 9-11, which is a little different from what we've seen this past year. So just sort of a sample, Zogby, who is an Arab American in a big pollster, perhaps not the greatest pollster in the US, but nonetheless he did a poll a few weeks after 9-11 and found nationally that 20% of Arab Americans, and throughout, if I A here, is always Arab American, not African American, which is not what you're used to seeing. So 20% of Arab Americans reported personal experiences of discrimination within the four weeks after 9-11. Nationally, in the first nine weeks, there were over 700 reports of violent incidents, including several murders. Not all accurately targeted, there were a number of South Asians rather than Arabs who were victims of these. And the nine weeks is important because basically nine weeks was how long it lasted. And there was, but as I said, there was a really strong reaction on the part of the media, politicians to try to tamp this down. And one example is this AP news story from a month later, which is really playing at the heart strings. Abdul Ali Ahmed's five-year-old son is still waiting for his father to bring him candy and ice cream. Adel Karas was planning a surprise barbecue for his wife to celebrate her certification as an anesthesiologist, et cetera, two more names. All were shot and killed after the September 11 terrorist attacks, and what the FBI and local authorities are investigating as possible hate crimes. Family and friends say they came to the United States in search of the American dream. There was a great deal of publicity for these incidents and the strong sympathetic reaction, too. My data are going to come from California, and I'll describe why in a minute, but let me focus in for a minute on what was going on in California. So hate crimes directed against Arabs and Muslims increased five-fold from 2000 to 2001. And that's just included. There was no increase for the vast majority of 2001. So it's all in that few weeks. So from about 100 to 500. This from someone in the public affairs group in Los Angeles, people are being attacked. Even when the numbers are small, the ratio is great. I've noticed a great fear of everyday life. People are afraid to go to work and school. There's been a tremendous backlash. People are nervous and unhappy, and it will take quite a while before the feeling goes away. However, the incidents did go away pretty much by mid-December, and they had returned to that sort of low level of about 100 a year, pre-9-11. So there's always been some anti-Muslim, anti-Arab sentiment in the country, but it was a low level before. And this from the communications director of the American Arab Anti-Discrimination Committee, and I have to say clearly indicating why this guy is communications director because it's a great sentence. My impression is that we are rapidly returning to what one would have to unfortunately call a normal amount of hate crimes. So that's the background to remind yourselves what October and November, it was a frightening time for people who could be perceived wrongly or rightly as being Arabs or Muslims. Now, why would this affect birth outcomes? And the most relevant literature, because the salient feature of that period was this violence would be the studies of violence and birth outcomes. There have been a number of neighborhood studies which examined crime level as a contextual variable and as an important source of neighborhood stress, and they have generally found weak but real effects, consistent effects. And here are some of the studies. There was one in Chile, so a different population than most of the U.S. ones that found more pregnancy complications in neighborhoods with social and political violence. High crime census tracks in the U.S., more likely to deliver low birth weight babies. Again, another Collins study. And then Jeff Mornoff, who did this work as part of a sociology dissertation here and then went to the University of Michigan, neighborhood violent crime predicts birth weight in Chicago. Now, you can imagine, of course, these also are contextual effects, population level data. It's not about which mothers were themselves the victims of crime, but rather this sort of fear, ambient crime effect. These studies, too, though, of course, you can think, no matter how hard you try to measure the other things that might vary by neighborhood or might be individually correlated with neighborhood violent crime, it's going to be nearly impossible to truly eliminate the possibility that you're seeing confounding that things correlated with a violent crime. But it seems plausible because, again, of the stress responses. So that's the end of the background. I'm now going to turn to my methods. So the outcomes, I have two outcomes I'm going to examine, low birth weight and preterm birth. Low birth weight is a very common outcome in studies of birth outcomes, but it's one that is often criticized. And it's criticized because it's a heterogeneous outcome. It's sort of a manifestation of a couple of different underlying problems. One of them is preterm birth, which greatly increases the risk, although there are preterm births that are above low birth weight, which is 2,500 grams or about five pounds is its official definition. And certainly there are preterm births where the baby weighs more than that. And then about one third of them are full term babies that are just little babies. They're at the left end, for you over there, of the birth weight distribution. And the normal birth weight distribution does go down that far. So there are two different things going on. There's the growth restriction. There's the extreme tail, also, of a normal distribution. And then there's the preterm birth. But it's a standard metric, and it's a standard metric because it's really well measured. Every baby is really accurately measured for weight, and it's recorded on the birth certificate. So it's a little unclear exactly what it represents, but it's great, complete, accurate data. And even though it's heterogeneous, nonetheless, as a whole, it's been shown to be predictive of a number of poor outcomes. There are, of course, many, many babies who weigh less than 2,500 grams at birth who do great through life and never have any problem. But it does increase risks of developmental delays, infant mortality, and perhaps even chronic disease risk later in life. So it's sort of a fuzzy outcome phenotype. Preterm birth is a better outcome to know about. It's more clearly a risk factor for poor, subsequent events. It's defined officially as less than 37 weeks gestational age. The problem with preterm birth is not that it isn't the outcome one wants, but that the data are bad. So you need to, in birth certificate data, which I'm going to be using, it either is coming from a clinical assessment or it's coming from the date of last menstrual period. And in most birth certificates you can tell which that is. The date of last menstrual period, which is the piece of information that is more generally available, is really a messy piece of information. This is just not something that women have, you know, written on their hand and remembered. And you get crazy gestational ages when you use it. So there are women who appeared if you just subtract that out to have been pregnant for four years. And so there are lots that are impossible. There are also gestational ages that are truly impossible given the birth weight. So, you know, you get your 48-week baby that weighs 4,000 grams. I mean, there's nonsense. And you have to figure out how are you going to deal with that nonsense. So it's incomplete and it's often wrong on the birth certificates, but it's really what our story, our underlying mechanism, speaks to. So the specific study questions I'm going to be looking at was the risk of low birth weight or preterm birth elevated for Arab-American women giving birth during the six months following September 2001, beginning October 1st and going through the end of March. So it's six months, the last quarter of 2001, the first quarter of 2002. I omitted the three weeks following September 11th, the rest of September, because basically most of the women giving birth then were already not preterm. So I wanted women who are a little at risk for having preterm birth. And so it needed to be a little later. I wish that I could kind of look month by month and sort of tease out the gestational ages and what months they were pregnant, but there just aren't quite enough women to do that. So I've got this aggregate six months, which I'm going to compare to the corresponding six months a year earlier. So I'm doing this for Arab-American women and then I'm also interested in whether the pattern of, if there's a difference, whether that pattern is different for other groups of women, either all foreign born women or other women in other race ethnicity groups, because honestly it was possible that everybody was freaked out right after September 11th, and there was a broader effect. The data are all women giving birth to singleton infants in California in 2000 to 2002. Why California, you may ask. And there are several reasons. One of them is that California is just a really big, diverse state. It's hard for reasons, I'm going to, again, make clear, I needed birth certificates that actually had names, identifiable information on it, and very few states, even for researchers who've gone through lots of hurdles, will release those. But California, at least at this point, would. And they had a totally functional review process for requests for identifiable vital statistics data. And I knew that because I've previously done work with death certificates from California. So it's the biggest state for births and it's a diverse state. It also has the largest Arab-American population. And it's a diverse Arab-American population. So people are surprised to hear that usually. They think Michigan is the state with the largest Arab-American population. Michigan has the largest percentage of births that are to Arab-Americans, but it's not actually as big as California. Michigan does, however, actually collect ancestry information, which it's the only state that will let you get at Arab ancestry. So one of the follow-ups to this study was a group at the University of Michigan who sort of did a similar study in Michigan. But California, so the data were available and there are really a lot of births. And that's why it was appealing. And I'd worked with them before. A sort of a side comment. One of the joys of working throughout this period when I was doing the death certificate and birth certificate worth with California was that the point person that one dealt with to go through their IRB, the state IRB, and get the data was a terrifically helpful woman called Jan Christensen. And owing to how the email algorithm was formed, you get these emails from J. Christ in California, which catches your attention. So the data derived from birth certificates and the pieces of information that I used on the birth certificate were the date of birth, obviously, the mother's given name and maiden name, the infant's given name and surname, the infant's birth weight, the estimated gestational age, the mother's race in these categories, white, black, American Indian, Pacific Islander groups are collected, but I collapsed them for this analysis. And other, which is in all the models, but I'm never going to show because who knows who that is, Hispanic ethnicity, which I pulled out of the white and black groups. They became white, non-Hispanic and black, non-Hispanic. I only looked at singletons. And then the mother's age, education, inferred marital status, and parity. Parity just first birth or not first birth. Inferred marital status is a little interesting. In identifiable records, they will not give you the actual marital status, which they know. And they instead just give you the inferred one based on the similarity of the infant's and the mother's surname, I believe is all that's going into it, which makes it a bit of a mess because there are different ethnic groups with different conventions about changing the name at marriage. So it's who knows what it is really. And then maternal birthplace, but they don't actually have the Arab countries because like many of the vital statistics records in the US, they have divided the world into the United States, Canada, Mexico, and the rest of the world. So we just have the rest of the world. But we have a little place of birth information. We know whether people were the mothers were born in the US or the rest of the world. So the key challenge in this project was identifying the Arab-American women. So Arab is not a race or ethnicity category on vital statistics records in the US. There are two, the Office of Management and Budget has issued two guidelines for how to collect race, ethnicity, data. The earlier one from the late 1970s explicitly said that whites included people from the Middle East. So technically Arab-Americans should have been identifying themselves as white given that. Now, prior to the 1997 revision from Office of Management and the Budget, whose results you're familiar with, because that is what led to the 2000 census looking different with the option to check more than one race category and breaking apart the Arab and the Asian and Pacific Islanders. So there were changes then. And there was a lot of lobbying to include a new category for Middle East or Arab. But precisely because the lobbying was of two minds with some people saying Arab-Americans should be identified as a separate category and other people saying Middle East should be identified as a separate category, a decision was made that there was really no common ground. We weren't ready to put that in stone in the vital statistics so nothing was changed. And so there's no way to use vital statistics stated to identify Arab-Americans. So what I worked with here was an algorithm that I had worked on that combines quantitative information about how predictive surnames and given names are of origins in Arabic-speaking countries. Now, this grew out of previous work I'd done with a demographer at the Social Security Administration to develop lists of distinctive names for specific Asian-American groups. And those have been widely used. After doing that, I was asked to help him and a researcher at RAND to develop Arab-name similar identifiable algorithms. And I felt, frankly, very queasy about it because it was 2002. And I thought people will think we're doing these for security reasons, particularly because all three of us were Jewish. And so to be honest, the genesis of this project I felt both of them had been so helpful to me in my previous work that when they asked me, I needed to do it. But the genesis of this project was actually trying to mull in my mind something that was a nice public health project, a sympathetic public health project that would be a story that I could tell people for why my name was on this Arab-name algorithm project. And it was totally unexpected that it turned out to have interesting findings. So the Arab-name algorithm, and these are our co-authors, and they did pull in an Arab-American demographer to help work on it. As it happened, I actually had studied Arabic in college. I had a couple years of Arabic, which was somewhat helpful in understanding what was going on and evaluating it. So the name algorithm was empirical and probabilistic. It was derived from this file that you can only get at if you're at the Social Security Administration, which is the file of all the Social Security number applications. Not that they have an Arab-American ethnicity code, but what they have is for the foreign born, they have the country of birth. And we used a country of birth in one of the 22 Arab countries as a surrogate for being an Arab-American. So that was the standard we used to pick names that were distinctively Arab or not. Now, of course, there are many Arab-Americans, particularly younger ones, who were not born in those Arab countries. And so we cannot expect a name to be perfectly predictive of our surrogate here, because there are definitely Arab-Americans who are not born in those countries who will have those names. So there's a fuzziness here. So what we did, for every name, we took the total number of occurrences of the name, the number of occurrences with an Arab birthplace, and just made a ratio of those. So the score was the percent of occurrences with an Arab birthplace. The name list included names that had at least five occurrences because my colleague at Social Security Administration felt for confidentiality he could not release a name that was really just held by one family because we were telling people where they were born than in the file. And a score of at least 1%. In fact, we used a score that was higher, but that was the original name list, which some other investigators have used. I think FAR, you used this. Yes, OK. So the names in the Social Security files, names are unedited. Whatever someone writes down is the name on there. And so they're kind of a mess, particularly for a language where names are often more than one word. So the names could appear in the list as one word, as a hyphenated two words, or as two words, the exact same name. And then there are differences in transliteration. There are a lot of different ways that Arab names are transliterated. They, to some extent, correlate with which countries people come from, but not completely. So here is an example of five different ways that El Khalidi shows up in the file with the total number of occurrences under in that middle number column with the number 68, 12, 40, et cetera. And then the number of occurrences with an Arab birthplace. And I put this up just to show you that even though in English these are completely different names, they end up, frankly, being reasonably similarly predictive. They've all got scores ranging from 36% are Arab country origins to 65% are Arab country origins. So that's reassuring that the name, no matter how it shows up in the file, is going to be similarly predictive. And then I've got one other example down here, which misaligned. But what we have is Hossein and Hossein. And Hossein has 4,307 occurrences of which only 1% are from an Arab country. And Hossein has 63% from an Arab country. And who knows why that is? Does anyone here know the answer to this? Yes? Because Hossein is mostly a Shiite name. And most of the Arab countries are Sunnis. Right. It's what Persians. It's the Persian version. Very good. And so these are people from Iran, which is not an Arab country. And it works well. So this was actually also a validation to see this, that this was performing reasonably. People kept the spellings they should, and that's great. So here's just another example of then how we combined the given name and the sur name. And so of course, the given name is probably more informative for people who were foreign born. Once you're in the US, you have a choice of using an Arab or a non-Arab given name. But many people do, certainly in the first generation, continue to use Arab names for their children. So here's an example from Palace Walk to just show you how we added together it. So we have Khadija and Aisha. And then the last name shows up both as Al-Jawad, as two separate, and as one name. And we added together the percent who were from an Arab country with the given name and the percent with the sur name. So if either one was quite distinctive, someone was classified as Arab, and if both were a little predictive, then the sum would categorize them. This was not as sort of out of our hat as it sounds, because the validation determined that in a file of everyone who filed IRS applications, which for complicated reasons, members of the team had access to develop the names, if we picked this 20%, we ended up with the same number of people as identified themselves as having Arab ancestry and being employed in the 1990 census. So it gave roughly the right amount of people, not clear that they were the same people, but that's how we came up with the 20%. And then you see here that Nagib is highly predictive, 81%, Mahfuz, 31%. Interesting to note, Khadija is pretty predictive, 32%, whereas Aisha is only 3%. And the reason for that, I believe, is because Aisha has kind of become a more disseminated given name. It's moderately popular in Muslim African-American populations. And because they're as numbers, that's such a bigger population than the people from the Arab countries, that even though it's quite distinctively Arab in those countries, that gets washed out when you get it in the US context, where there's so many more blacks than Arab Americans. OK, so the name algorithm, as I said, what we did is we combined score threshold of 20. We also looked at the infants given name. And that was a marker in my thinking for, and I should say, so there was this team I showed you that worked on the development of the name algorithm. But this is an oddity in epidemiology. This is a paper that I'm the sole author of, so I'm going to go to I now from we. I thought that an infant having a distinctively Arabic name was going to be kind of a marker of stronger ethnic identity or more recent immigration on the part of the mother. And so I also use that in the analysis as sort of an extra variable. So name identification, which is most familiar, I think, to people from the census's work on Spanish surnames, which have been pretty widely used for decades. It's not perfectly sensitive. It doesn't identify everybody. It's not perfectly specific. It identifies some people it shouldn't identify. But it does identify a group, which is much more enriched for the group you're looking at than a group that doesn't have those names. And in this particular study, because our comparisons are not so much Arab to non-Arab, but rather Arabs to Arabs a year earlier and other groups to themselves a year earlier, it kind of doesn't matter so much, because we're identifying women the same way. And they're clearly including a lot more Arab Americans in the name identified than not. So although there are weaknesses to name identification, I think they're muted in this particular study. And to the extent that we've got other people washed in there with the Arab Americans, that should actually mute any difference. If there's a real difference for Arab Americans and not for others, it's a negative bias, which is good. It's conservative that we've got some other people mixed in. So it's not going to create a spurious association where there isn't one, I hope. The analysis is really simple on this. So I did do a baseline logistic regression analysis just to look for the 18 months, the first 18 months, of the three years of data I got from California, which did take a long time to get, I want to say, despite their having a great system. It did take like a couple of years. So from January 2000 to the middle of 2001, I looked to see just kind of as background how much did race ethnicity affect low birth weight. And I looked both unadjusted and adjusted for the main confounders we've got in the birth certificate data, maternal education age, marital status, which is a mess as I described in parody. And then, as an epidemiologist, I was in a wonderful position, which is that none of those factors actually changed for mothers within a race or ethnicity group over the course of a year. So I didn't actually have to adjust for any of them when comparing the 2001-2002 births to the 2000-2001 births. And that means that I could very simply present relative risks and not be using logistic regression and either presenting odds ratios or monkeying with it to get probabilities. So it's always a beautiful thing to be able to just present relative risks because everybody understands what that is. It's just the ratio of the two proportions. So then I look at the relative risk of low birth weight within each race, ethnicity, and nativity group for the six-month period after 9-11, compared to one year earlier. And then to look at preterm birth, I'm using a method I'll describe in a minute, which Ellen Wilcox developed, which is just uses the birth weight to examine preterm birth because the preterm birth data are so bad on the birth certificate. This was a very clever method he developed, which is available. There's a little piece from a website at the bottom there. It's sort of a plug-and-chug utility that's available at the National Institute of Environmental Health Sciences. And what it does is it assumes that the birth weights you see are the sum of two distinct distributions. One is a totally normal distribution for full-term infants. And then the other is a lower, not normal necessarily, it isn't, a distribution of lower weights that are for preterm infants. And that when you look at the birth weight distribution altogether, you can assume this symmetry based on the heavy end of the normal distribution. You can sort of mirror it and see how much extra observation you've got in the low end. Does that you follow? So as I said, there are some preterm births, near preterm births that are not low birth weight at all. But he argues effectively that really the ones we care about are the ones that are both preterm and low birth weight, because those are the serious problems, potentially. So this method, you put in all those birth weights and you get back a percentage, like 2%, 4%, whatever is this extra left tail weight. So here is the total data for all three years, just to give an idea of the size of the populations and also some comparisons of these big risk factors for low birth weight preterm birth. So there are about a half million births in California years, so we've got one and a half million over the three years, something which many years ago there was publicity about how whites had become a minority in California. And you see that here, that only a third of them are non-Hispanic whites. There is a larger proportion who are Hispanic. The black is not that big in California. 90,000 Asian and Pacific Islander about twice as big. And the people, the women I've identified by name, are just 15,000 here, which is not enormous, but it's not nothing. I highlighted the white and the Arab-American to show you that on the proportion that are first birth and the inferred unmarried, which I think is not performing as well for Arab-Americans as the whites, they look very similar to whites. Whereas for foreign born, they look just like the Asian-Pacific Islanders. This is all highly plausible. The age distribution, this is the mother's age distribution. And in pretty colors of blue and purple, we have less than 20, 20 to 29, 30 to 34 and 35 plus. And you see here that the Arab-Americans, again, look very similar to the white population with a modal category there of 20 to 29, just a few younger, and actually quite a few in the oldest category. So that's who they look most like. And education distribution, they also look, this is the top bar and the bottom bar, they also look very similar to the white population. So not quite as well, let's see. It's actually not that different from the Asian-Pacific Islander, although there's more some college in those. But anyway, basically, so this is all pretty plausible. So here is the logistic regression model, which is just showing us, just to get some idea of the baseline risk over the first year and a half before we get to 9-11. And what we see here, and white, non-Hispanic, is the referent, the odds ratio of 1, that there are things that we expect to see. So the first column is unadjusted, nothing's in the model but the race ethnicity groups, which are all mutually exclusive. We see that Black, not Hispanic, has an odds ratio of 2 and 1 half, so much elevated risk of low birth weight. There's a slightly elevated risk for Asian-Pacific Islander, a non-significantly elevated one for Native American, a very mildly elevated one for Hispanic. But because it's such a big population, it achieves statistical significance, and the Arab-Americans are not distinguishable from the white population. When we put in age, maternal age, education, parity, marital status, basically it doesn't change much. So those factors aren't really major confounders or explainers of those race ethnicity distinctions. Now we're going to look at the relative risk. This is the main result. So these are the relative risk of pre-term birth, of low birth weight in the post-9-11 period compared to the pre-9-11 period, and it's just a ratio. And you can see for all women, it was exactly the same. The relative risk, the ratio was one. So the proportion who were low birth weight, exactly the same post-9-11 as before. It's exactly one for all foreign-born women, white not Hispanic, black not Hispanic. It's very slightly but not significantly elevated for Asian Pacific Islander 1.03. No difference for Hispanic. And the only group that shows any difference is this modestly elevated relative risk of 1.34, which is statistically significant, not massively so, but at 0.022, it is significant. Then I further divided to try to sort of get within this group at the more ethnic versus less ethnic by categorizing the children's first name as being on our Arab name list or not. And one question that's been asked, great question, is did women stop giving their babies Arab names after 9-11? And the proportion with Arab names is just the same in the two-time periods, so that doesn't seem to have been affected. So the elevation and risk for the Arab Americans was really entirely explained by the risk among the women who gave their babies ethnically distinctive given names, so it's 2.25 for them and really not much elevated risk at all for the not ethnically distinctive, which is consistent with the hypothesis, because these presumably are women who identify more strongly as Arab Americans. It may also overrevers that women who dress so as to be identifiable or had other features that sort of made them more distinctively Arab, because I mean there's a range of clearly of women with Arab ethnicity who are obviously so or not to others. And now that sort of clever parsing of the birth weights into the excess left tail versus others, and sorry, so again the hyphen came out as this funny S, but what you see is that the percent of births in the residual distribution for white non-Hispanic is identical in the two-time periods, 2.2%, whereas the percent changes for the Arabic named mothers from 1.8%, so originally lower than the whites, up to 2.8%, which is quite a bit higher. Not only is the 1.8 to 2.8 a statistically significant difference, but obviously the change is different from the whites because they have no change, so no matter which way you look at it. So that's consistent with the low birth weight and with preterm birth being at least part of the story. So to summarize here, the Arab American women giving birth in California in the six months following September 2001 experienced a moderately increased risk of low birth weight and preterm birth compared to a year earlier. Other ethnic and racial groups or foreign-born women in general have no comparable increased risk and the risk of low birth weight for Arab American women is much greater among those who appear to have a stronger ethnic identification as suggested by their infants having ethnically distinctive given names. There are some major limitations. The women who are identified by name might not identify themselves as Arab Americans. These are two overlapping but distinct populations. Arab Americans with less distinctive names could actually have a different risk of low birth weight because there are other things different about them too. They might sort of tend to come from different countries that have overlapping names with some other countries or there are less, you know, U.S. born women would be less likely to have the distinctive given names meaning that their surname had to be super distinctive. So that's a little bit of an issue. Because of the given name issue, the foreign-born women might be overrepresented because they're more likely to have a distinctive given name. In fact, though, this turned out to be, it doesn't seem to be much of a problem because my name group, as you recall, 82% were born in the rest of the world. And in the 2000 census, I did a sort of a separate analysis on the census microdata sample and the women who listed an Arab country as their primary ancestry, 80% of them were foreign-born in California of those who had a child under five years of age. So it looks like it's not actually over-representing that much because mostly it is a foreign-born population for the mothers. And then there could be something that's different about California than elsewhere. It's a very, it's not a physically concentrated Arab-American population like Michigan and it has lots of different groups. There was an old, large immigration from Lebanon that is now multi-generation in California and then there are more recent groups like in the rest of the country, but that earlier one was sort of, that was a key destination. And the sample size is too small to examine month by month changes because that would be interesting to know whether second trimester, third trimester made a difference and it's just not big enough. There are alternate explanations. So one that was suggested was that maybe because security measures were heightened that women were expelled. And so some women had to leave the U.S. because they were on visas and the visas tightened up after 9-11. But in fact the number of births is about the same in the two-time period so I don't think that's it. And you'd have to assume that women who are gonna have excellent birth outcomes differentially left and that seems weird. There could be other conditions affecting Arab-Americans and no one else in California either decreasing their risk in my post-period or increasing it in the pre-period. Maybe there's something screwy about the referent period. People have asked about Ramadan. It was in November in both study periods so it's in both study periods. It doesn't move that fast so it's the same time a year almost, year by year. So what I think is strong about this paper is that the compositional demographic, health, residential factors would not change in one year within the race and ethnicity group. So a lot of things are being held constant. Things we know about and things we don't know about and cannot measure. And I've avoided the psychological and motivational issues in the recognition and reporting of experiences of discrimination here. But there are a lot of possible explanations for what's going on. So I mean it could be economic. So there was a separate study done actually by people at UIC economists that showed that earnings did decrease for Arab-Americans in the post-911 period. So there could be some financial stress behind this. There is, as I've sort of put out, there is what seemed most likely to me that the violence and the fear of violence caused a stress reaction. But it could also be that there were health behaviors change that increased risk of preterm birth, that people took up smoking to control stress or drank more or their nutrition fell apart or sort of all kinds of things. They may have just simply as that one quotation at the beginning said been afraid of going out and consequently not sought health care as assiduously as one is supposed to during pregnancy. That of course assumes that prenatal care actually affects birth outcomes and the jury is out on that. And then finally, I mean the other explanation is the stress one coupled with the violence of fear of violence. And I, I mean it seems like a plausible one but I absolutely can't say that that's what's going on here. The way that I think this contributes to the literature on discrimination outcomes is that it's limited to an acute experience during pregnancy because the lifetime stress should be the same for the women giving birth in the two different years. Unmeasured confounding is unlikely because of the design. It's a different population. And you know, it's discrimination could mean and surely does mean a completely different thing for African-Americans than for any other groups who may suffer from discrimination temporarily or a longer term in the US. And so it's quite interesting to sort of find something here as well. You can't compare the magnitude of the effects though because you know, this is not an individual comparison like all the African-American studies. It's this group population time period comparison. The people who contributed are my colleagues at SSA and RAND who helped with the name algorithms and also sort of commented on drafts of this as I went along. And the wonderful Center for Health Statistics in the California Department of Health Services for access to the data, which really, I mean it's kind of remarkable among states that they are willing to share these data, no matter how many hoops one does have to jump through. So I'd like to stop here, take questions and if there's time I can tell you a bit about the reaction. There's some funny reactions to this paper. Do you have questions? Yes? My name is Mahmoud Ismail and I'm an Arab-American, one of those people you, and thank you for studying them. The pre-term labor is, it's multifactorial. And the Arab-American community is actually ethnicity. They call them same Arab-American, but some of them come from Sudan, some come from Morocco, some come from Egypt, some come from Jordan, Syria, and stuff like that. So the most probably, we don't know, even in the Arab world, if the pre-term labor between these people in different countries is the same. And it's also most probably, is depending on socioeconomics, wherever you go. In this case, and you alluded to, I think it's stress after 9-11 and the fear factor, most probably caused some of these pre-term laborers, but it's very difficult really to prove or disprove that. Thank you again. Thank you. So that's a great point. It is a heterogeneous population from 22 countries, but presumably the distribution didn't change in one year in California of the relative Tunisian to Egyptian ratio. Yes? There's at least one other study, and maybe you can think of more or tell us about more that looked at the difference between, or the effect of some sort of national event on birth effects. And I'm thinking of a study that was done in Sweden after, or Norway or some, one of those Scandinavian countries when the king was shot, and it was a big and kill, it was assassinated. Do you recall seeing that? Okay, so the question was that there are other studies though of, no, there were more than that, that look at national disasters and how that affects birth outcomes, and they have heterogeneous findings. I don't know that one, but there's a professor in New York who has done an interesting series of studies about a Central American earthquake and finds that the problem with disasters though, the disasters that increase stress actually material affect life in other ways as well. So an assassination is not like that. That really doesn't affect life unless the country falls apart. But something like an earthquake, the food chain could really be disrupted, livelihoods could be affected. So they're not quite comparable. But I mean, it is a study design that has been used certainly. Yes, back there. Thank you very much for a very nice talk and very well designed, although it's very challenging to make this study. Now, two questions. The first one, do you have any plans on doing any subgroup analysis, looking at, let's say women who are born in the United States, but of Arab ethnicity, or women who were immigrants, recent immigrants to the United States who maybe were having fears of having to go back to their countries because of visa or that's one. The second thing is, do you plan to find or identify the kids of those women who had at least low birth weight? Would you follow them like over 20 years, see if there's any difference in the outcome for those offspring as well? So those are both interesting questions. The first question was looking at the foreign born versus US born women and differential effects. And I did that and the problem is only 18% of the women which is already a small group where US born. And so I couldn't really say anything about them. So that's why I took that infant name as a different way of kind of trying to look at a dose response of Arabness in the population, if you will. The follow-up, no, I mean, those are good questions, but those study designs just basically of looking at long-term outcomes of preterm birth, there are many people who do them. You need a very large population to do it. I have no way to follow these particular infants and really promised never to link these data to any other data, making that hard to do. So the research reaction was that a really fabulous MD, PhD student at the University of Michigan, whose family is Egyptian, just really liked this study and wanted to repeat it in Michigan and was able to get access to the ancestry data and found nothing. And so this is why, you know, is there something different about California? Is it the fact that the great majority of Arabs in Michigan live in just a couple of neighborhoods so they're in very tight enclaves or perhaps not sort of as vulnerable to the outside world as in California? Or is that a fluke or is this a fluke? And there is no answer to that. The direct reactions, I mean I presented this once at the National Epidemiology meeting and as the talk ended, someone in the audience sort of quickly left and then after the questions and then I was talking to people. And she came back in and came up to me about two minutes later and she said that she had run out to call her daughter because she was so excited about this study and this was now about five years ago. I was presenting it and then she said, but I spent the whole time staring at you trying to decide what country you were from, what kind of Arab you were. And she decided I was Lebanese, which I think is a really, if I'm gonna be an Arab, I think that was a good choice for what I look like. And I said I'm Semitic, but that was interesting. And then I got called by the Arab desk at the BBC who wanted to interview me, but they wanted to interview me in Arabic. And I said I don't know that much Arabic, I can read a little, but I can't be interviewed in it. And he said, but there must have been an Arab who was working with you on this. No, there wasn't. And it really, it went on for like five minutes, he was incredulous and that was the end of that interview. And I didn't really wanna be, I did not want this to get international coverage in a sense because I felt a little bad about publicizing all the anti-Arab incidents in the US in those few weeks. It seemed a little embarrassing and I figured if I got interviews, I'd get asked a lot about that and I didn't really wanna be asked or talk about that. I was being a demographer. This did not get much coverage in the US. There was a fantastically detailed, accurate column about it in the Washington Post by their survey editor who I think just subscribed to Demography, which is, I published it in Demography because I needed a lot of pages of front matter which clinical journals wouldn't give you to sort of describe all this stuff about the names and the violence and all. And it was a great column but only got picked up by a few newspapers and that was that. However, I did discover when I put together a social epidemiology course a couple of years ago that the paper is in a lot of social epidemiology syllabi around places. So people in the field of discrimination do know about it, but it didn't really, other things have gotten more press, yeah. This is more a comment than a question but I'm struck that regarding your methodology, what you kind of see as a limitation, which is that your name algorithm might be inaccurate, actually maybe isn't a limitation and maybe as a plus in this context since people who commit hate crimes aren't always accurately identifying Arab Americans themselves. Right, so there's a Venn diagram there somewhere of people who self-identify, people who other people think are Arab, people who are identified by name and there's just various amounts of overlap and it's not really clear what's most relevant. Thank you for saying it's a strength. I still think it's, I mean, a limitation, actually those other two pieces of the Venn diagram would be the ones one most wanted who other people thought were Arab and who thinks that they are. Although the self-identification, self-identification of ethnicity is much squishier than people think and the Census Bureau has done beautiful studies showing that people that Hispanic and Native American are incredibly volatile self-identifications depending on the context and historical factors and individual factors, I mean even year to year from the same individual and I imagine around this incident there are some people who might have ceased to identify themselves as Arab American so not having the self-identification perhaps does solve a problem there. Yeah, yes. This is that it really begs for follow-up for you might be into something that it could become a better formula that only relation to the Arabs after the 11 because it is of course most likely to do that. These women were under stress and of course most likely to do that your statistical birth are true but true to truth does not necessarily make the cause of relation unless there is a partially a third one that both relate to and let's temporarily call that a stress and if stress causes that that the women have birth less that means that all kind of stress could do because our body doesn't differentiate we are you know biased something or intentionally or perceived so then you come up with the stress causing this and then it's likely that would equally produce it in men but yet we have not found what maybe causes it may be depression which end up to lack of appetite and additional and if they come to become a truth then you come up with some issue which is ethical or moral implication in causing a stress whether individually or for a large population. That's certainly true there have been a lot of studies of hypertension and stress and men do get hypertension part of the reason that birth outcomes are kind of appealing is because of that critical period that you can so neatly identify but I mean that framework that you describe is of course there I mean that is the implication. It was a wonderful, wonderful discussion I'm glad I encourage you to present this paper about the Arab American community don't hear many such discussions in this seminar series and I think it was really a delight to hear the paper. Thank you so much for coming. Well.