Emily Oster of the University of Chicago talks with EconTalk host Russ Roberts about why U.S. infant mortality is twice that in Finland and high relative to the rest of the world, given high income levels in the United States. The conversation explores the roles of measurement and definition along with culture to understand the causes of infant mortality in the United States and how it might be improved.
Play
Time: 59:31
How do I listen to a podcast?
Download
Size:27.3 MB
Right-click or Option-click, and select "Save Link/Target As MP3.
Readings and Links related to this podcast episode
Related Readings
HIDE READINGS
About this week's guest:
Emily Oster's Home page
@ProfEmilyOster. Emily Oster on Twitter.
Emily Oster's articles at Slate.com.
Emily Oster's Articles at FiveThirtyEight.com.
About ideas and people mentioned in this podcast episode:
Books:
An Essay on the Principle of Population, 6th edition, by Thomas Robert Malthus. Library of Economics and Liberty. [Starting with his 2nd edition, Malthus began traveling to individual countries, cataloguing birth and death records taken from sources such as church records and foundling hospitals. Notably, he was the first economist to classify countries not by region but by their statistical similarities, incidentally making his research the first to identify developed versus undeveloped countries.--Econlib Ed.]
Articles:
"Why is infant mortality higher in the US than in Europe?" by Alice Chen, Emily Oster, and Heidi Williams. PDF file.
Poverty in America, by Isabel V. Sawhill. Concise Encyclopedia of Economics.
Web Pages and Resources:
Apgar Score. Wikipedia.
Podcast Episodes, Videos, and Blog Entries:
Oster on Pregnancy, Causation, and Expecting Better. EconTalk. October 2013.
Martha Nussbaum on Creating Capabilities and GDP. EconTalk. September 2014.
Scott Atlas on American Health Care. EconTalk. July 2012.
Highlights
Time
Podcast Episode Highlights
HIDE HIGHLIGHTS
0:33
Intro. [Recording date: November 11, 2014.] Russ: Our topic for today is a paper you've coauthored with Alice Chen and Heidi Williams on infant mortality. In particular you are looking at why the United States higher rates of infant mortality than many European countries. And I thought it would be interesting to talk about this particular issue, which I'm very interested in, in itself; but I'm also interested in the general question of how we use health data to evaluate public policy and think about ways to make things better. So, I hope we'll get into those issues as well in today's episode. Let's start with what is to be explained. What kind of differences are we talking about, between the United States and Europe? Guest: They're very large. If you look at the baseline, like what is the World Development Indicator (WDI) say are the differences in infant mortality rates, the United States ranks something like 50th internationally. And the difference between the United States and a frontier country like Finland or Sweden is something like 3 deaths in 1000. So, the infant mortality rate in the United States is about 6 in 1000; in these other places it's about 3. And you know, if you think about that, that amounts to something like 12,000 excess deaths a year, among live-born infants, so that's a reasonably large number. It's certainly very large as a share. Russ: Right. It's double, which seems like a very large number. Three seems small; 12,000 seems horrible. Right? So it is an interesting example in itself of the challenges of trying to assess what's big and small. But I think the 100% part--the fact that our U.S. rate is double that of, say, Finland, is surprising and interesting to think about. Guest: Yeah. I think that's a lot of how people have thought of it. On the one hand, the good news is that in all of these places, the infant mortality rate is very low relative to historical norms or relative to developing countries. On the other hand, clearly there's a long way to go; there are many places, most of Europe, even places outside of there, which are doing much, much better than the United States in this dimension. There's no particular reason that we think that should have to be. Russ: And to put it into historical perspective, the number I have used in the past for infant mortality in 1900--and infant mortality, we're going to get into the definition, but usually when you just say 'infant mortality' it's death within the first year of life. In 1900, the data I have seen says it was 1 in 10; and now it's about .5 per 100, or 6 per thousand. So, it's about a 20-fold improvement in the last century. Guest: Yeah. That sounds right. Those changes should not be forgotten. There's enormous amounts of progress that have been made on this dimension, and I think that's really wonderful. Russ: So, let's start just the fact itself. Which seems like a fact, but of course all facts have context and there's issues of measurement and definition. So, how is infant mortality defined in the data that is usually used that we're talking about at the national level--say, 6 per 1000, 2 or 3 per 1000 in, say, Scandinavia? What is the definition? Guest: The definition is deaths in the first year among live-born infants. So, that is the definition. And when you see numbers that are reported by the World Development Indicators, or the CIA (Central Intelligence Agency) has some numbers, that's the number that you're going to get. And that's reported by the country. And so one of the issues that I think often comes up, that we talk about in the paper, is that how you define a live-born infant actually varies. And so, in particular, human gestation is intended to be approximately 40 weeks; infants born before 22 weeks effectively never survive. But there is a lot of variation across countries in whether you ever report infants born in that earlier period, before 22, as live births; and those kind of reporting issues actually do potentially bias some of these comparisons. So, because if you report any live births from that period, there are certainly going to--that is going to count as a death. You can inflate the numbers. So, the first thing we do in the paper, which is sort of possible because of the kind of data that we have, is limit to what we think of as a comparable sample were kind of all of the countries are reporting infants in that range as live births. And that actually turns out to make some difference. So, the difference between the United States and the comparison countries we use--which are Finland and Austria--shrinks from 3 to 2 deaths per 1000, once you adjust for these differences in reporting. So that's certainly something to start thinking about. Russ: I want to come back to that. But just to make it clear, when you say the difference shrinks from 3 to 2: so, right now, the raw difference is 6 per 1000 in the United States versus 3 per 1000 in, say, Finland; so that's a difference of 3. And a third of that difference is data-driven, is what you are saying. Guest: Exactly. A third of that difference is reporting [?]. Russ: Before we get to that, we should give listeners some feel for the range outside of the developed world. There are countries tragically with infant mortality rates over 100 per 1000, still, correct? That's the equivalent of a 1900s-level of infant mortality. Guest: Yeah. There are certainly places like that. Afghanistan would be in this category; some countries in Africa. If you take a country like India, which is of course very large, they report infant mortality rates something like 40 in 1000, so it's still very big; and then this goes all the way down to about 2 in the places that are doing the best, which would be like Sweden, maybe Japan. Russ: But I'm thinking, when you talk about the definition and how live birth is defined, in some countries, particularly the poor countries, I assume there's variation in a pre-term infant, whether it's defined as a live birth or now. Guest: Yeah, I think so. And the truth is, in these developing countries where these rates are so high, those differences are not going to be very important in this scheme of the overall level. So, part of what makes these reporting differences potentially important when you are comparing, say, the United States to Europe, is that already the rates are pretty low. And so things that kind of matter a little bit are going to matter relatively more as a share. When you are talking about a place that has an infant mortality rate of 150 per 1000, small differences in reporting are going to be vanishing in the scheme of those comparisons. Russ: Right. But I do want to mention, in my quick look at the data, there is improvement in the last 5 years, even in those poor countries. Guest: Yes. All of this has been getting better, much, much better over time; and I think that that's a function of better health care and more vaccinations, some better sanitation, and other policies. And so I think things have really been improving. Which is, again: broadly this problem is getting better, which is very encouraging. Russ: One of the challenges of thinking about this is there are specific things, like you just mentioned: sanitation, health care. But my first thought, stepping back from this is that resources generally are really important per capita. Some measure of standard of living is going to be extremely important. And that exactly is why it's such a puzzle that the United States, which is a very rich nation, still is twice the rate of, say, Finland. Guest: Yeah. In general, like almost everything, this is going to line up with income. And it broadly does line up with income. And it is puzzling that the United States is down, like next to Croatia, even though the United States is much richer than Croatia. Russ: About 3 times, if I remember correctly from your paper. Guest: About 3 times.
9:41
Russ: So, let's talk about some definitions and a little bit about how data get collected in this area. So, when you talk about the World Development Indicators, I assume every hospital in the United States has some requirement to report births, deaths; and those numbers get aggregated in some way and then sent to some international source; and that's where we get the 6 from. Is that correct? Guest: Yeah, that's correct. So these are going to come, in the United States, from what are called the Natality Detail Files, which are collected and reported by the CDC (Center for Disease Control), so they are basically like short-form birth certificate data: you can actually see, not people's names and addresses, but for each birth you see the characteristics of the birth--things that happened: the gestation, the birth weight, and so on. And then that's linked to deaths in the first year. So that's like a nationally available, nationally curated data set. And that is how--those are the numbers that will then be reported as international statistics for the United States. Russ: So, that's the birth side--and that word in there was 'natality,' correct? N-a-t-a. Guest: Natality. Correct. Russ: It's a tough word--don't use it that often in everyday conversation. Guest: Yeah, it doesn't come up. Russ: But that's the birth side. How do we know that a child--if an infant dies, tragically, in the hospital, that gets counted pretty straight-forwardly, I assume. Or if it's born and doesn't survive the delivery, again there's a question about whether it was a live birth or not, depending on how many weeks it was, maybe. But an infant that dies at 6 months or 11 months, how does that get into the data? Guest: So, there are mortality reporting files in the United States, also. And so deaths get reported in the same kind of systemized way. And then in the back room somewhere at the CDC these things get linked. So, you can actually get access to the birth data linked to the information on death, including the recorded cause of death and the standard stuff that would be on a death certificate. Russ: And that's going on in every country. Guest: Well, that's going on in many developed countries. Let's put it that way. Russ: Okay. Why do you say 'many'? Guest: So, as part of this project actually we tried to get this equivalent data for as many countries as we could. And we got some. I think some of them are, like, not sort of linking this in quite the same way, such that you could use it like this. But some version of this where you are recording births and you are recording deaths, some version is happening in all developed countries, for sure. Russ: What's not happening that would make it hard to use? You say it's not happening in every country: What's going on? Guest: I think in many poor countries, births and deaths are not reported in a consistent way, and so these numbers for mortality are taken from smaller field surveys or from estimates from parts of the population, or from various kinds of inference. And I think there it's very complicated. Even in developed countries, there's no particular reason that you need to have the micro individual birth data linked to individual death data except for research purposes. And so I think in many countries that doesn't happen. Russ: Okay. So let's go to your data. You have a different kind of data. You call it 'micro data.' Describe it and, what are its advantages? Guest: So, we've got this data that I described for the United States, where we see every birth. We're looking at 2000-2005, so we see every birth linked to information about death in the first year, if the infant died. And then we have effectively the same data from Finland and Austria. So it's every birth linked to information about the birth, the birth weight, the gestation, some things about the mother; and then also linked to information about deaths in the first year, including exactly at what age the infant died and some information on cause of death. So we have quite a rich and comparable data set across all three places. Russ: Why did you choose those three? Guest: So, the truth is that we actually tried to get this data from every place that we could. So, we knew we had this from the United States--this is very commonly used. For the European countries we basically just mass called and emailed all the European countries in an attempt to get this, and these are the two that worked out. We also have some data from the United Kingdom and from Belgium, where it's a little more aggregated, so we can't do quite the same analysis, but we can do some of our analyses there. It turns out, I think--these are good comparison countries because Finland is in this frontier of places that are doing really, really well; and Austria is kind of right in the middle of the U.S. distribution, but I can't say that we did this in some way that was that well thought out. Russ: No, you took what you got. Which is fine.
14:56
Russ: And let's talk about the other data that's in there, besides births and deaths. So, you have information about the mother. Is this a sample? Or is this purporting to be exhaustive. Guest: This is the universe. This is purporting to be exhaustive. Russ: So, when my wife gave birth, the hospital had a bunch of information about her. What kind of information would they have? And obviously when the baby is born they do a bunch of tests. The baby is weighed. The height, length--it's not really height because they are not standing up, but it's the length of the baby. There's Apgar scores, which--you should describe what those are. And what other stuff do we get, that's required? Guest: Yes. We have everything. So, an Apgar score is the most commonly used measure of just how is the baby doing at birth. It's on a scale from 1 to 10--or 0 to 10, I guess. And so, like, a 9 would be the sort of standard--like, everything is looking good kind of score. So that's the kind of summary of how well the baby is breathing and so on. So, we see that. We see the weight, the length. We see the gestation. We see a bunch of information about complications of labor and delivery and complications of the baby. So if the baby has a birth defect, that's recorded. A heart problem or Down's Syndrome, that goes in this data. If the labor was very fast or very slow or there was a C-section (Caesarean section), all of that stuff goes in the data. So it's quite a rich set of information about the circumstances of the birth and characteristics of the infant. Russ: What do you know about the mom? Or the dad? Guest: We know that--we know some. We know the education; we know their race; we know their birth country. Obviously we know the location of the birth and the location of residence. And assuming that there's a father around, his education is also recorded, and marital status, and that type of thing. So, what you think of as sort of basic demographic information. Russ: How do you get the education? Where does that education number come from? Because that's important, obviously. Guest: So, these is all in these natality files, and is I think intended to be reported in this monoform[?] birth certificate information. I mean, I remember being asked--when I had my daughter, I remember being asked at the hospital, like, how much education do you have? Russ: Hmm. I don't remember that. Guest: I remember writing that down. Russ: It's the kind of question I don't like. Guest: A lot of other stuff was going on, so [?] was not--you know, had I not been thinking about this research project it probably would not have been in the front of mind. Russ: Yeah. I just don't remember anything remotely like providing that information. But of course I have health insurance--but not everybody does. I'm just wondering how--you say you have the universe of all births, say, in the United States, say in 2000-2005. There must be missing data. Lots of missing data. Guest: I think it's very likely that there are births that are missed. Births are supposed--in order to get a Social Security card and be able to work and be a U.S. citizen you must have a birth certificate. This information is collected as part of the birth certificate. So, like at least some of this information. So, if you are born in a hospital, you are getting this. If you are born at home, you know, you are supposed to get this information; this information is supposed to be collected so then you can get a birth certificate. Now, what is almost certainly true is if there is a birth at home, all of the details about the medical procedures that occurred during the birth and so on are likely to be less well reported. And you know--but basically anything that happens in the hospital, most of the complicated information here would have nothing to do with the individual having to report. It would be reported by the hospital, by the doctor. In that sense. Russ: So, just out of curiosity, do you know what proportion of births are home versus hospital? I assume it's very small, the home births. Guest: Very small. It's like half of one percent. Russ: Okay.
19:22
Russ: So, you have that for the United States, you have that for Finland, you have that for Austria. And so you've got a lot of micro-information about the birth that's not in the national numbers that we normally. And now, tell us what you found when you looked more carefully, given that data, about what might explain those differences. Guest: Yes. So, I think we were pretty focused on trying to think about the kind of accounting problem. And we broke it down into kind of four parts. First is this issue of reporting. And so, just kind of limiting to a sample we thought was comparable. The second issue is looking at differences in prematurities. This is an issue that gets a lot of attention in the United States. The United States has a high rate of prematurity, meaning a large share of babies born prior to what we consider a completed gestation. And so we find that both reporting and prematurity matter a fair amount. So, as I said, about a third of this gap between the United States and elsewhere are closed by this reporting issue. And we do find that also, particularly compared to Finland, but also to some extent compared to Austria, the differences in prematurity matter. And [?] the thing that's most interesting, that is most interesting, that is most different from what people have found before, is in these data where it will just separate the neo-natal period--so the very early period after birth, the first week or the first month--from what happened after that. And this distinction is kind of very central, because the things that caused death in those two periods are very different. And what we find is actually, conditional on birth weight, the United States is doing very well early on. So, in the first week or the first month, we're actually doing better than, like Finland, and very, very comparable to Austria. And then when we move from the period of a month to a year of life, when infants are typically at home and a lot of deaths occur in the household in some way, that is the period in which the United States is really lagging behind other countries and really that is like driving a lot of these differences. And I think that was sort of the most surprising thing we say. Russ: Yeah, that's fascinating. I want to come back to that. I want to start with the two issues you just mentioned, you just started with, which are the reporting issue and the prematurity issue. Let's talk about the reporting issue. I'd like to give the listeners a feel for how you actually "controlled for" that. What did you do to your data to take the reporting differences out of the measurement? Guest: So, we did--it's pretty straight-forward. So we did basically three things. We took out infants where the gestational age was less than 22 weeks. We took out infants where the weight at birth was less than 500 grams--because these are kind of standard reporting things. All countries report babies in the categories of later than 22 weeks, larger than 500 grams; and there's variation below that. And we also, in this case we took out singleton births--sorry, we took out plural births. Not so much of issues of reporting but because there are many more of those in the United States, probably due to [?] reproductive technology-- Russ: Yeah. Guest: And so in some sense because those infants are more vulnerable, you would like to compare apples to apples and look at singleton births. So, that's all we did. It was not an especially sophisticated way to approach this. We just took out things we thought where the reporting was not consistent. But what's interesting is that the United States, then, has more deliveries at less than 22 weeks, and more deliveries under 500 grams birth weight. Is that right? Guest: No, that's not so clear. So that's the issue. Because if this is supposed to be recording of live births. So I think the way you want to think about it is, if the infant is born at 20 weeks and kind of moves around a little bit, that will sometimes in the United States get reported as a live birth; but it will not get reported as a live birth in these other places. So it's not--it is possible the United States has more births in that period. But that is not something we can see in the data because we simply are not seeing those observations elsewhere, because they are getting reported as miscarriages. Or stillbirths. Russ: And 500 grams is, according to my crude effort--is that about a pound? Guest: Yeah, it's about a pound. Russ: So that's a tiny--the average birth weight of full term baby is about 7-8 pounds? Guest: It's about 7. Russ: Okay. So that's remarkably small. So, that's the first one. Did we talk about prematurity while I was doing that? Did you talk about, while I was doing my conversion, just that? Guest: We didn't. I mentioned prematurity. Russ: Yeah. Tell me what you did to deal with that. Guest: So, what we do to deal with that is we put in basically adjustments for either gestational week, or in fact what we use is birth weight. So, it turns out birth weight in these data are much better reported than gestational age. Which is not surprising. The way you record gestational age is by asking the woman, like, when was your last period? People don't--this is a hard thing to remember. Birth weight is very precise. It's measured in the hospital and the correlation is obviously very, very, very high. So, we use birth weight. And we basically--one thing is we just look at the comparison across the countries, and we see basically if the United States adopted the birth weight distribution--like magically acquired the birth weight distribution of one of these other countries, but kept the mortality conditional on birth weight the same, how much of the gap would that close? So it's sort of like a counterfactual. Like, if that was the only thing that we changed, how much of a difference would that make? And the answer there is it would close much of the gap with Finland, like 75%; but only about a third, less than a third of the gap with Austria. So, it clearly matters; but it matters a lot more relative to Scandinavia. So, state that again. If the United States--how different is the birth weight distribution? That's what's so surprising, right? Guest: It's interesting. So if you look--we have some graphs in the paper. Which of course are difficult to look at on the podcast; but you can look at them. Russ: We will put a link up to the paper, of course. Guest: Excellent. So check the [?]. One of the things you can see is actually the United States and Austria have really, really, really similar birth weights. The only place they differentiate is a little bit at the very bottom. The United States has a few more babies in the kind of 500-1000 grams, but they're very similar. The big difference is the Finnish babies are much larger than either of the other places. Like, much, much larger. And I don't think there's like not a great explanation for that one. One possibility is just like the Finns are tall and they have giant babies. That's something that's a little hard to tell. But for example they are more than 200 grams heavier on average than babies in the United States. So, that turns out to matter. And they are also--the rates of prematurity are just lower. And that's actually something we know to be true. Like, that's a commonly observed fact that is not very well understood, because we don't know really what causes infants to be born prematurely, which is making it very difficult to understand why it might differ across space. Russ: Right. So, and, you said we don't know very much. Do we know anything? Guest: Basically--we know a few things. So, for example, smoking increases the rate of prematurity. Using drugs like meth, also not good. But beyond those things, that basically--that's it. So, 12% of births or 10% of births in the United States are pre-term. And obviously 10% of women are not smoking or using methamphetamines, and so, you know, there's a tremendous amount of variation and we just have a very, very poor understanding of it. Russ: There's a random component that may be related to other things, but we don't know what they are, is what you are saying. Guest: Exactly. There's some random component; it's probably related to some other stuff. We don't know what it is. This is related to the fact that we don't know what causes women to go into labor in the first place. Or we have only a very poor understanding of that, even at term, full term. We just don't know very much about what are the biological mechanisms that prompt that to happen. Russ: There's not a switch, strangely enough. Guest: There's not a switch. Russ: Or if there is, we don't know where it is. Guest: Yeah. Probably there is, but we haven't figured out how to turn it off or not.
28:33
Russ: But we do know--I guess the only thing I know about this--and I assume there's some evidence for it--and of course Emily, in her previous EconTalk appearance was talking about what we know about pregnancy, which is sometimes less than we'd like to know. The thing that I think we know, is that not being active slows it down. At least, women who are at risk of delivering earlier are told to take it easy. Is that--do we know that that actually works? Guest: No, that doesn't work. Russ: They get put on bed rest. Guest: Yeah. That doesn't work. Russ: Bed rest does not work? Guest: No. Bed rest does not work. Russ: How do we know that? Guest: Randomized trials[?]. Russ: Because people--women get put on--I say 'people'--women. Women get put on bed rest all the time. Right? Guest: No, this is one of the like--when we did this earlier, this stuff about pregnancy, this is like a very striking fact. A lot of people are put on bed rest; they think it is increasingly well accepted. Bed rest is definitely not useful for preventing premature babies, or doing anything. It's actually pretty bad. It has some other, like, negative consequences. So, systems, like, just lay around: not effective. Russ: It does have an 18th century ring to it. Which makes you nervous. Guest: Sure, yeah. I mean, everyone likes to relax. I think it turns out--it sounds great, when you're pregnant, to be like, oh-ah, just lay in the bed. But actually being forced to lay in your bed all the time-- Russ: Horrible-- Guest: It's very unpleasant. Russ: Good to know. So, let's regroup here. Let's summarize what you've told us so far. We've got, controlling for birth weight, prematurity. Is that the only two we've talked about so far? Guest: Yes. So, adjusting for reporting differences and adjusting for prematurity, you find that the United States is still substantially worse off than elsewhere. So, even if we had the same levels of birth weight as these other places, things would still be much worse in the United States. Russ: But not relative to Finland? Guest: But much less, relative to Finland. That's true. Slightly worse. Russ: Butt relative to Austria. Guest: But relative to Austria, much worse. Slightly worse relative to Finland.
30:59
Russ: So let's turn to the surprising finding, which is fascinating. Which is that neonatal--which I guess means close to birth--as you said, first week or so, maybe first month: U.S. neonatal mortality is very, very good relative to these two countries, at least. Whereas the, say, months 2-12, it's where the United States is doing much worse. And this, of course is not visible in the standard national data because that just looks at any time within the first year. Is that correct? Guest: Yeah. And it's actually--and it's even in some sense harder than that. Because sometimes in these national data sets they do separate, like, the first month and the later period. But what they are not able to do is do that controlling for birth weight. So, like from a policy--so economists, like I think we're interested in what we would do about policy. [more to come, 32:00]