Russ Roberts

Oster on Pregnancy, Causation, and Expecting Better

EconTalk Episode with Emily Oster
Hosted by Russ Roberts
PRINT
Tyler Cowen on Inequality, the... Winston on Transportation...

Emily Oster of the University of Chicago and author of Expecting Better talks with EconTalk host Russ Roberts about her book on pregnancy and the challenges of decision-making under uncertainty. Oster argues that many of the standard behavioral prescriptions for pregnant women are not supported by the medical literature. The conversation centers around the general issue of interpreting medical evidence in a complex world using pregnancy advice as an application. Alcohol, caffeine, cats, gardening and deli-meats and their effect on pregnant women are some of the examples that come up. The conversation closes with a discussion of Oster's work on hepatitis-B and the male-female birth ratio.

Size:28.0 MB
Right-click or Option-click, and select "Save Link/Target As MP3.

Readings and Links related to this podcast

Podcast Readings
HIDE READINGS
About this week's guest: About ideas and people mentioned in this podcast:

Highlights

Time
Podcast Highlights
HIDE HIGHLIGHTS
0:33Intro. [Recording date: October 3, 2013.] Russ: Our topic for today is your book, Expecting Better, and maybe some of your other work if we have time at the end. The book is a guide to what we know and don't know about pregnancy. But it's also an example of how to think about data and how to think about causation, correlation, a long-standing theme of this program. When you were expecting, when you became pregnant, you initially thought--you say in your book--that the doctor was going to give you information, evidence about various phenomena that arise during pregnancy, the risks, the benefits. And you'd have to make decisions, along with--as everybody has to do under uncertainty. But that isn't the way it turned out. What surprised you? Guest: I think the biggest surprise for me was I seemed to be unable to get the numbers and the data from my doctor. So, there were very many times in which they would say, well, it's different for everybody. Or, well, the numbers are pretty small. Or, well, you know, we don't really know but to be safe, blah, blah. And I think, for me, the way that I think about decisions is I think about, okay, I need to get the data; I need some numbers; and then I will use that in conjunction with risks and benefits and some simple decision-theory to make these choices. And getting the actual hard numbers was just very, very challenging and wasn't something that was forthcoming from my doctor or from the standard pregnancy books that I started with. Russ: And why do you think that was? Guest: I'm not sure. So, I think that one issue is that there is a lot of conflicting data, often. So, when I actually went into the data to find out the answers, as I'm sure we'll talk about, it wasn't that easy. So there was a lot of subtlety. And I think that's one issue. And I think also because everyone is different in some sense, you need to explain to people: there's an average, and here's what the average is. And I think that that's something that doctors are sometimes reluctant to do, as opposed to just saying: I recommend x and that's what would work for most people. I'm not sure that they are typically interacting who want, really want the numbers. Russ: Yeah. I think there are two ways to think about it. There are patients who don't want the numbers, and there are doctors who don't want the numbers. Both groups are a little bit uncomfortable with statistics and uncertainty, in my experience. Guest: Yeah. I think that patients sometimes will just say: Look, I just want you to tell me what to do. I don't want to have to make this choice. And I think that in combination with doctors sometimes saying: Look, I don't have 45 minutes to go through all this with you; I'm just telling you, you should make this choice. I think those two things act in kind of a poor, they sort of interact poorly in a way that makes us not always make the best choices. And, like yesterday I was on the radio with a doctor who said: Well, statistics is really hard, and so we don't think, we shouldn't really expect people to be making a lot of these decisions with data. And I think, for me, it's like: Are you kidding? You have to make these decisions with data. Russ: As opposed to what? Guest: Right. What is your other system? Like, this is the only system that we have. There's no other, like secret other system. This is it. Russ: We have folklore. Superstition. Some of which is true. Some of the folklore is true, obviously. Guest: For sure. Russ: But in general, data help. And I think you make an interesting point there. When we talk about how people don't want the data themselves, then doctors, I think, sort of get in the habit of--and enjoy--telling people what to do. As opposed to saying: Well, this is your call. You are going to have to think about it yourself. Guest: Yeah. And certainly not all. I think one of the things that is definitely true is there is variation across doctors. And there is also this very clear set of constraints that I think doctors face, some of which just has to do with not having that much time and some of which has to do with legal issues. Russ: Sure. Guest: And I think for both of those reasons, it is in some sense incumbent on women to think through these choices for themselves. And this extends past pregnancy into other parts of medicine: that there just isn't really a choice--you have to do this, to some extent, by yourself. There isn't another option. Russ: When you said 'there isn't a choice,' you also manage, mention in the book that sometimes you just presume you are going to do x, just because "everybody does that." And that isn't really true. Guest: Yeah. Exactly. Russ: Did you ever find yourself saying, as I often do to members of the medical establishment, somewhat apologetically: Well, I'm an economist--to try to explain why you weren't like everybody else? Guest: Yes. Sometimes. I don't think I invoked that very often, but I did actually--my husband is an economist and I occasionally was afraid he would be too much like an economist. Like, he would come with me and I would say: Look, you just don't talk. This person is going to be delivering our child; like maybe you just don't say anything. We'll make this choice for ourselves and I'll get the data and we'll make the choice, but I don't think we need to get into a big hullabaloo with this person. So I think that certainly came up. Russ: Out of curiosity: Did you know you were going to write this book? Early on? Later? When did you come to the decision to write a book. Guest: So, certainly not at the beginning. There were a few things about pregnancy that I think I anticipated I would be thinking a lot about. Like, prenatal testing I kind of knew that I would want to spend some time understanding the options. Because I thought it was complicated I wasn't sure we would want to make the standard choice. I was surprised with the number of things that I ended up spending a lot of time researching. And the book--my idea about writing the book evolved slowly and at some point I was just playing around: I'll write a chapter on alcohol and see if I like that. And if I enjoy it and feel like it's a useful thing; and in the end I liked it and so did the publisher and so that was where the book came from. But I certainly didn't think: I'm going to get pregnant and then there's going to be a book. Russ: At first you thought you were going to write a pamphlet; and then it turned out you didn't. Guest: Exactly. Russ: It's good.
7:14Russ: So, what's hard about this? Let's talk about the general challenge of making decisions in these kinds of environments. And as you point out, it's not just about pregnancy. It's about almost all kinds of medical procedures. We are confronted with these decisions about testing, and the side effects of testing versus the benefits of the knowledge; false positives, false negatives. What makes it difficult in general, and in particular with pregnancy, to establish what's a good idea for a pregnant person, a pregnant woman to do? Guest: I think the first very big issue is in a lot of what I researched in pregnancy there is a very clear problem with causality. And it's in some ways insurmountable. So, I think caffeine is really the best example of this. So, the concern with having too much caffeine in pregnancy is that you might have a miscarriage. So, it's most about what are the issues involved with too much caffeine early in the pregnancy. And so the way that this is studied early in pregnancy is you look at women who drank some coffee early in pregnancy; you ask them how much coffee they had; and you look at their rate of miscarriage and you compare them to women who didn't have coffee, and you look across different amounts of coffee. Russ: And what do you find when you do that, on that just raw comparison? Guest: So, what you find is basically all of the studies suggest that up to 200 mg, which is about 2 8-oz. cups, is fine. So you really don't see any increased risk of miscarriage up to that level. When you start looking at, like, 8 cups of coffee, you see much of the evidence does suggest that there is an increased risk of miscarriage. When you look in the middle, like 3, 4 cups a day, the evidence is a little bit mixed. So it's one of those things like it's clear a little bit is okay; it's probably the case that a lot is a problem; but in this intermediate stage it's unclear. And part of the issue is that when you look at these studies, the kind of women who drink coffee are just different from the kind of women who don't. And this is a problem in any observational study like this. And they are different in ways that are also, themselves, linked to miscarriage. So, women who drink coffee tend to be older. Age is the biggest issue with this. Miscarriage is also linked to caffeine consumption. And so you can control for that, but you worry that there are other things like that, that are still going on. In addition, in the case of coffee--and this was in some sense the most interesting thing for me research-wise--there is also a problem with nausea. So, if you are nauseous in early pregnancy it's a very good sign about the health of the pregnancy. Women who are nauseous--good news, ladies: you are less likely to miscarry. But women who are nauseous are also more likely to avoid coffee. So when you look at women and you see some of them drink a lot of coffee, those women are also on average women who are less nauseous. And when you see that they then miscarry at higher rates, it may well be just that not being nauseous is a sign of miscarriage and caused them to drink coffee, but it's not that the coffee caused the miscarriage. And I think getting around that kind of challenge is very complicated. Russ: This is what is sometimes called 'spurious correlation.' It's also what my friend and long-ago EconTalk guest calls 'the dreaded third thing.' The thing that you can't measure or haven't measured that's actually the underlying causal variable. Guest: Exactly. And I think it's clear there, the direction of the problem, so that particular issue is going to cause us to condemn caffeine too quickly. And so we can say, look, we can see that there's no increase in risk of miscarriage up to 2 cups; we can feel more confident about that because the biases push against finding that. But then when you ask about what about 3 or 4 cups, there's like one study which shows maybe that's a little risky; and many studies which show it's not; and we then want to combine our data with what we know about the data and problems with the studies and try to draw some conclusion. I actually think that's a place where people may come to different conclusions. And so some of what I'm--kind of the point I'm pushing in the book is that not everyone will make the same choices with the same data. And that's okay. That's why you need the data--so you can make choices that work for you. Russ: But, Emily, then you can't look down on people you think are doing the wrong thing? Guest: No. It's true. It's a big issue. The judging. You really ramp down the judging. I agree. Russ: That's one of the things I love about the book. There's a subtle, extra point about people being different, which is of course not only does everybody make their own decision, but everybody's tolerance for caffeine and desire for caffeine and the impact of caffeine is different as well. So this is true in pharmaceuticals generally. We're talking about caffeine, but in pharmaceuticals, the dose that's right for you isn't the dose that's right for somebody else. And we haven't learned enough about that to be sure yet most of the time of what the right dose is for any one person. Guest: Exactly. Russ: So, there's a lot of uncertainty there.
12:42Russ: Talk about, in general, the type of evidence you looked at. So what you just referred to as an observational study--you have a group of people; you ask them retrospective questions. Which is also not ideal. These people might forget. And they also might feel guilty that they drank a lot of coffee when they were pregnant. So, if they've learned that that's not socially acceptable, they might lie, or misremember unintentionally about their past behavior. But there are other forms of evidence that you are going to be looking at and that we are going to be talking about. So, what are some of the other ways that we try to tease out the causation in these examples? Guest: There are actually a number of areas of pregnancy where we have randomized data. And so that's much better. So, a randomized study in this setting would say we have a group of pregnant women and we do one thing to some of them and one thing to another. This comes up much more commonly when we think about things around birth. So, if you ask a question like, are there any risks to the epidural? We actually have randomized data in which some women were kind of encouraged to have an epidural and some women were encouraged not to. And we can compare their outcomes and see what is different about them. And there you can be more confident that the effect that we are finding is causal. When those were available, that was nice. Now, those studies are also not without their downsides. They are run typically in one population, not in all populations. And sometimes the outcomes are not all the things that you would like. But it was easier to learn things when there were studies like that. That was nice. Russ: And as you point out, sometimes there were randomized trials. Which is nice. But they were a long time ago, sometimes just because they were done a long time ago; sometimes because you couldn't ethically get away with it now. It's not considered acceptable practice to shove caffeine down a bunch of pregnant women randomly and another group not get it. But those old times, those studies from long ago--things have changed. Technology is different; we have different ways of coping with things. So they are not always so informative. Guest: No. I agree. And there's also an issue that when even if something is significant in these randomized trials, you still want to ask the question: Is this important in terms of size? So, when we looked at the epidural, one thing people are always saying is: Oh, if you have an epidural it will really lengthen your labor. Russ: An epidural being a painkiller, a pain-reducing technique during labor. Guest: Right. It's the most commonly used pain-relieving technique in the United States. And so the concern is that it makes the process of laboring and delivering longer. And it's actually true that it does, on average, in randomized trials. But the length of time is about 15 minutes. And so, it is important there to know not only that this matters but also how much. Because whether you think this is a big problem or not, really I think is going to vary with: Is it 15 minutes or is it 6 hours? Like, if you told someone, look, you are going to get the epidural and you are going to be in labor for an additional day, that may sway people more than if you tell them, look, it's going to be another 10 minutes but by the way, it's not going to hurt; so you really are not going to care that much. So I think that, that times thinking about the magnitudes as well as the significance, is something that I think we often forget. Russ: Yeah. We forget that in public policy generally. I think that's an incredibly important point. Nassim Taleb, recently, in a podcast was talking about how well people predict. And the fact that people are right most of the time is no comfort if when they are wrong it leads to a disaster rather than a small unpleasantness. Guest: Right. Exactly. Exactly. Russ: So, you confess in the book that you love caffeine. So tell us what you did. Based on the evidence, how did you change, if at all, your caffeine consumption during your pregnancy? Guest: I actually did in the end cut down in the beginning, but not because I thought it was risky but entirely for the confounding, like, secret third factor, which is that I was pretty sick. And so I'm used to having--like now, I wake up at 5:30 before my kid and I have two cups of coffee before she's even awake. And when I was pregnant, I could barely get carbonated water down in the morning. And so I actually cut down a lot at the beginning--not on purpose, but then I ramped it back up. So I would say I was probably having 3, at least 3 cups, by the end. Russ: Is there an issue of caffeine during breast feeding, by the way? Guest: I actually, I will say, the book stops at the delivery room. In general, these diet things and breast feeding--I think most people are a bit less crazy about it. And so I think in general, caffeine is thought to be okay during breast feeding. Some people say that their kids react oddly to it. Mine seemed to like it fine. I didn't hear any complaints. Russ: That's because your daughter has a genetic predisposition to liking caffeine, like you do. Guest: Yeah, no, exactly. She does frequently ask for some coffee, but so far I haven't given in.
18:12Russ: So, let's talk about the more controversial case of alcohol. Most pregnant women I think believe, and are told, that they should have not a single molecule during pregnancy. And it's very dangerous. Talk about what the dangers are and what you found when you looked at the literature. Guest: The big concern with alcohol is that it can lead to birth defects, and in the most extreme version there's something called fetal alcohol syndrome, which typically has physically manifestations that you can see. Even lower levels of alcohol consumption can lead to, there's this series of disorders, under the, they're called fetal alcohol spectrum disorders: behavior problems, lower IQ, and so on. And so there's no question that those things are real and that excess exposure to a lot of alcohol in pregnancy is dangerous. I think that's certainly not up for debate; and in fact I think one thing that's probably less cited is, even a couple of episodes of binge drinking at the wrong time can be quite dangerous. So, that's one thing. I say that in the book; that's true. I think for many women the question that they are wondering about is not: would it be okay to binge drink; or even: would it be okay to continue drinking at whatever level I drink at now; but rather: would it be okay to have a glass of wine 3 times a week with dinner? And there, the advice--though certainly the Surgeon General says absolutely not--when you look at surveys something like 40% of obstetricians will say: Well, some amount of alcohol is okay. And clearly a lot of women, like myself included, are getting that advice, like, yeah, a couple of drinks a week is okay; don't overdo it. And so I think in the face of that kind of disagreement and in the face asking a question that isn't really the same as, Can I binge drink?--I wanted to look at the data which studied women who drink in moderation. And so, really there you want to look at women who, say, never have more than one drink at a time, and never have a binge-drinking episode. I tried to find data like that. And there is. There actually are a number of large studies. They mostly come from Europe or Australia, where drinking at this level in this more moderate way is just more common in pregnancy. When you look at those, the evidence shows that the children of women who drink in moderation perform as well, sometimes better, on IQ tests, in measures of behavior problems, as compared to children whose mothers abstain. So, taking from that evidence, I think what one would conclude is that there isn't any good evidence that drinking in moderation like that causes problems. So, that's what I say. Russ: And what did you end up doing? Guest: So, I had the occasional glass of wine, maybe half a glass, 3 or 4 times a week. I just took pretty seriously the idea that moderation meant not a whole glass or not 10 ounces. But I did have the occasional glass of wine. Actually, my obstetrician said that that was okay. Although in the end for me the data part of this was the most important part of the evidence. Russ: And that, I would call it level-headed assessment of the evidence has gotten a lot of people angry out in the world. Guest: Yep. Russ: What kind of reaction are you getting from that. Guest: There are different levels of reaction. So one piece of reaction that I expected was doctors who said, like very early on someone said: Yes, one drink a day is fine, but I caution women against doing that because if they have one then they'll have two or they'll have three, and that's not good. So, I agree that having 2 or 3 is not good. I'm not that excited about the implication that women are unable to, like, control themselves. I'm sure there are people in that category; there's no question, and I think if you are in that category you shouldn't start. I think there are a lot of women, like me, who, if you said you could have a glass of wine will just have one and understand what that means. So, I see that argument, but I'm not sure that I think it applies to everybody. Then, there has been a lot of pushback from groups who are associated with fetal alcohol syndrome, who I think feel either that it's just very important to say you can't have any because if you suggest that in moderation it's okay, that people will overreact. Or they genuinely feel that they disagree with my interpretation of the evidence. Unfortunately, it's been difficult to really engage on the evidence, which has been in some ways frustrating. But there you go. Russ: Of course they can counter--if they did engage on the evidence--that well, women who drink moderately, going back to our causation, dreaded third thing, are not like other women. As you point out, you and you know lots of people and maybe I do, too, know lots of people who can control themselves. But for other people, they are not in the data; they've got other problems, other behavioral issues; they may have other educational issues; and those may correlate with bad outcomes or good outcomes. And so you've got to be very cautious when you interpret the data. That would be, I guess, their response if they were thoughtful about it. Guest: Yeah. I think that is a thoughtful response, although not typically the one that I'm getting. I think what I'd say is--part of what's nice about these European studies is on average these women who drink in moderation look pretty similar to the women who abstain. You actually do see in those studies that women who drink in excess do have kids with worse outcomes. And so it's not that the study is underpowered to find an effect. So I think that's encouraging. We can engage on these kind of issues and I think they are important. I do think it's useful to separate the question of: Are there some people who struggle with alcohol--to which the answer is: Yes. And in many cases those women also have other issues, which society should be doing more to help with. I think we can separate that question from the question of: Are there women out there who can have an occasional glass of wine and stop there? Yes. And should we be telling them what the data says? I would argue: Yes.
25:10Russ: So, I'm a big fan of zero. For reasons of self-control. I think it's often much more difficult to have a little bit of something. In my case it's carbohydrates, not wine. Guest: There are carbohydrates in wine. Russ: It's not very large, though. I was thinking more of French fries. So, I understand the virtue of going to zero, personally. I don't like the infantilization of our culture that says: You are not mature enough to handle this. Even of course though it might be true. You may not be mature enough to handle it, but I don't like the idea of somebody deciding for me that I'm not mature enough to handle it. So I understand the virtue of this idea even though I don't like the paternalism of it. But of saying, zero. I like saying it for myself. What I'm curious about, though, is that for the binge drinking, or let's say larger amounts of alcohol than a few glasses a week--to go back to your earlier point: You are saying the studies find that those aren't good. It's bad. I'm curious: How bad? Do we have much evidence for how damaging excess--large amounts--of consumption are for the child? Guest: Yeah, we do. And it's very bad. I think it varies across people, so actually some people can drink to excess and have kids who turn out fine. But certainly that's not true for everyone. And there's a tremendous amount of data that suggests that drinking to excess is a problem. This is a major cause of birth defects in the United States. So I think it's not useful to under-weight that; that is a very significant issue. And I think, probably, again, and in some ways a more relevant thing to say, is: This can occur with even a few episodes. And so a lot of--in many cases people will say: Before I knew I was pregnant, four times I had six glasses of wine, but then as soon as I learned I was pregnant, at seven weeks, I stopped. That may be too late. And so I think helping people figure out that they are pregnant earlier, helping people curb drinking while they are trying to conceive: these are all relevant things to think about because the damage is so bad. I just feel like it's a different question than the question about moderation. Russ: So, when you were pregnant and you were publicly drinking a glass of wine--and as I think you point out in the book and it certainly happened with my wife--when my wife was showing, people wouldn't even offer her anything, either because they thought it was immoral or they just assumed my wife wouldn't accept it. But you must have encountered people who said: Come on; better safe than sorry. Guest: Kind of surprisingly, I never did. Now part of that is I think I only once or twice had a drink in public because I think I had exactly this thought: like, I don't really need to get into it with people. I did once order a glass of wine and the waitress was like: Good for you; my mom had an occasional glass of wine with me and I turned out fine. I was like, Okay; thanks; I'm glad to hear that; [?] I can't tell by that but at least, better than your telling me you're the worst person ever; all right, thanks. I enjoyed that. Russ: One thing you mentioned, and I always use this when I talk to people with these kind of issues, is that: it's nice to keep the mom relaxed. It's not a trivial factor--and I think correlated with the health of the baby--that you don't want to be stressed out during your pregnancy. So if you are uneasy, for whatever reason--because you've read too many studies about the damage from alcohol--having a drink cools you off. It calms you down. Guest: Yeah, and I think that is a real thing. I think the other side would say: Well, you should exercise or do yoga or find something else to calm you down. And maybe people should. But I agree that there is a stress-relief aspect to this kind of routine which for some people could in principle be helpful. We do know that stress is bad. Russ: You try yoga with a 10-pound basketball around you. Guest: Oh, I have, and it's terrible.
29:42Russ: Now, smoking is an area that you didn't find anything that went against the received wisdom. Guest: Yeah. And this is an interesting thing to explore for me because I sort of did it. Like, caffeine, alcohol, and then I come to smoking; and you do see evidence exactly of the same form as you see in the alcohol case, same form in the sense that it's women who smoke as compared to women who don't. You see a tremendous number of additional complications amongst smokers. You see that there's a lot of additional placenta problems; they are more likely to have babies with low birth weight; they are more likely to have their babies too early. So those are all kind of issues that come up. But then you ask the same question that one asks in alcohol or caffeine: Isn't it the case that the women who smoke are different from women who don't and maybe it's just these other differences that are driving this? What was very helpful in the smoking case was we actually do have some randomized evidence. And you may think: How do you ethically get people to smoke? And the answer is: You don't. But because we generally think smoking is really bad, there are studies which encourage women to quit. And so we have randomized studies in which some pregnant women are put into a program to try to encourage them to quit. Sometimes those are effective. When they are effective, we can look at what happens to their babies, and we see very large impacts, particularly on birth weight. So it looks like quitting smoking actually really improves the birth weight of your child. And so as a result, I think it's pretty clear that that's a goal that we should have. Russ: And birthweights--higher birthweight--is generally a good thing. Guest: Yes. Russ: I don't know why that is, by the way. I'm interested; there's always the issue that if you have lower birthweight because you are a smoker--it may not be as bad if you naturally have lower birthweight. Do you know--you don't talk about this in the book. Birthweight is correlated with many, many positive things. Why? Is it just a sign of health? Do we know? Guest: I think we think it's just a sign of health. There's a question of--we draw this distinction of very low birthweight versus medium birthweight. I think within the range of normal birthweight, like above 2500 grams, which is kind of the normal birthweight cutoff, it doesn't actually matter too much where your kid is. Very low birthweights are associated with additional problems, typically very early on, partly because they are also associated with being born very early. Which is something smoking puts you at risk for. And that can mean that lungs are undeveloped, and so on. So I think that basically this is a marker for the timing; it's a more precise marker for the timing of birth or a marker for the fact that you aren't getting enough, the baby wasn't getting enough nutrition in the womb. So certainly there are ranges of healthy birthweights which include quite low. Some babies are like great born at 5 lbs. But it's correlated because I think it's a marker for these other things. Russ: You mention a number of foods--and of course my wife was also told not to eat various things during her pregnancy. What makes that list? Why do they make that list? And what do you think belongs on the list and maybe what doesn't? Guest: So, for me this was one of the most confusing things. Because the list was very, very long and it contained a lot of random things that seemed quite random. And so it was like, there's deli meats. So at some point I went in and I said: Is prosciutto a deli meat? My doctor was like: Definitely not. I was like: All right, well. He seemed to think that was like a crazy question. It goes in a sandwich. Like what is the other definition of a deli meat? That I buy it at the deli counter? So, I think that from there, I came to think I really need to understand why all of these things are restricted so I can categorize: These are restricted for this reason and these are restricted for this reason; and then I can think about whether those are risky. Whether those are actually concerns. So, some of the foods that you see, like raw eggs or sushi, the concern is with salmonella. Which think we are all familiar with. It can give you a stomach flu. It turns out that actually those risks are no worse in pregnancy than they are outside of pregnancy. So, it's not the case, if you get a stomach flu from salmonella in pregnancy that it's particularly dangerous to the baby. Russ: Just no fun. Guest: It's unpleasant. It's unpleasant when you are not pregnant, it's unpleasant when you are pregnant; there's no question. But there isn't any excess reason to avoid some of those things in pregnancy. And so, you'd say, sushi, reputable sushi, not a problem. Then there are some things which are linked particularly to listeria, which are very dangerous. So, listeria is a very dangerous foodborne illness. It's particularly dangerous for pregnant women. Pregnant women are more likely to get it than the average population. And it can cause miscarriage or stillbirth. So it's a very bad thing to have. And it is the reason for the restrictions on soft cheeses, paté, on deli meats. But I realized that the question I needed to ask was not: Is listeria bad? Which it definitely is. But how much can I change my risk of getting listeria by avoiding these foods. Russ: Good question. Guest: So the story--I wanted to structure the decision. So: Let me go see what the last big outbreaks of listeria are due to. And the answer was: cantaloupe. The answer when I was pregnant was cantaloupe. That was the thing that was the big risk for listeria. It appears nowhere on the Restricted List. And this is because it is very, very hard to predict where listeria will come from. It is unfortunately just often quite random. So there was a big outbreak in cantaloupe. There's been a recent outbreak in celery. We don't know where the next thing will come from. And so, with a couple of exceptions, like raw milk, soft cheese, and deli turkey, I basically decided that avoiding many of the things on this list were not going to actually change my risk of getting listeria. Or change it only a tiny fraction. And so I then was faced with the question: Okay, I would like a ham sandwich; and there is a remote possibility that it might have listeria; I need to weigh those tradeoffs. And I think from there at least it made it easier to make that decision. Russ: When I mentioned to my wife the deli meat worry, she said, "Oh, sure--nitrates." But you don't mention nitrates. It's interesting. Guest: Yeah. There is some discussion of nitrates, but it's not quite the same as foodborne illnesses. It's also the case that if you want to generally avoid nitrates, you can typically get meats which do not have those. So that felt to me like a less--we don't know that much about that. Russ: Yeah. That's my feeling, too.
3 7:01Russ: What about cats? That's the animal. Not eating them, but being around them and changing the litter box. Which, we have had cats in our family but not during our pregnancies. What's the issue with cats? Guest: So the concern is that there's a parasite called toxoplasmosis, which if you become infected with it during pregnancy can result in birth defects. And cats are like a common vector of this parasite. Particularly when they are young, if they are living outside and eating a lot of raw meat, this is something that cats can be infected with. In practice, if you look at what is the cause of most toxoplasmosis infections, it is not cats. In fact, it is not common to get it from a cat, partly because most people's cats do not live outside; partly because many people have already been exposed to toxoplasmosis if they have a cat and once you've been exposed it's not a problem. And so in the end, this research about not cleaning the litter box is not particularly well supported by the data, something that many pregnant women have told me they wished I hadn't said-- Russ: Of course, because they get out of the litter box-- Guest: Exactly. Russ: I was the litter-box guy. Guest: Yeah. You've got to--no--you can still make your spouse do it. That's the message. Russ: It seems to me the message--and by the way, this is clearly a scurrilous campaign against cats from the dog lobby and we should try to get a cats' folks person on here for equal time. But it seems to me one of the lessons from this is you should get a cat early in your marriage, before you are pregnant, so you can get exposed to it and then the other forms that could expose you won't affect you. It seems like cats--I don't know. Let's see: I was going to ask about gardening, and then we're going to move on. How about gardening? Why would gardening be an issue for pregnant women? Guest: So, it's actually for the same reason as the cats, although in this one this is a real thing. So, it turns out one thing that is linked with toxoplasmosis is doing a lot of outdoor gardening. Probably because other cats--not yours--may poop in your garden. And then when you are gardening it stirs up the poop. And so if you are going to do a lot of outdoor gardening, the advice of wearing a mask or wearing gloves is not, although it seems a little crazy, is not actually crazy. Russ: Wow. Okay.
39:42Russ: Let's move away. One last thing about pregnancy in your book which fascinates me. I read an article when I was doing my background reading for this interview that--it was an angry screed; you may have seen it--suggesting that your book was awful because you are unqualified. You are the wrong kind of doctor; you are a Ph.D. in economics. You don't know anything about medicine. And it's irresponsible for you to go around making health recommendations. And people should ignore your book. What's your response to that? Guest: Um, yeah, I'm surprised you just found one like that. Russ: Well, I stopped looking after that, Emily. I thought one was plenty. Guest: One is enough. I think that if you read the book, it's very clear this is a book about data. And this is a book about looking at data and evaluating causality and thinking about what's a good study, what's not a good study. That is exactly what my training is in. I'm a health economist; I have a lot of training in statistics. I think all the time about what's causal, what's not causal. And so I think in many ways my training is better than a doctor's training for evaluating these kind of questions. The book is not going to deliver your baby for you, and so I think there's no question that this is the kind of thing that women will read in conjunction with also going to their doctor. I assume. And so I think it really is complementary. But I also think there's a very clear reason why someone who has a lot of training in statistics would be the person who writes it, a book which is all about data. So I think that's the--I think if people actually read the book they will see very clearly why someone with my training will come at it in this angle. Russ: Have there been a lot of criticisms, have you received a lot of criticism along those lines, sort of the circling of the medical wagons? Guest: Not as much as I thought. So actually I think a lot of doctors recognize that these are the way that their patients should be thinking about this stuff. And in part because there is absolutely no way that a doctor is going to be able to convey all of the information that someone needs about their pregnancy during their visits. There is a sense in which you have to rely on the patient reading something else. It is probably better for them to read something which at least discusses the studies rather than just reading the Internet. So I've gotten, actually, some pretty nice feedback from doctors. Russ: I would--not only do I believe that your training is particularly well suited to the questions that you deal with, so I second your defense. But I would also say that doctors are particularly ill-suited for these kind of issues. They don't typically--I think it's changing, but they don't typically get trained in data analysis. They are certainly not trained in statistics or decision-making. They don't have a very good appreciation of uncertainty. And they are prone to say things, as a friend of mine heard when he was a motorcycle rider. He broke his leg and the doctor put his leg in a cast and then said, I hope you learned your lesson. And my friend said, Yeah, well as soon as I get off my cast I'm going to ride my bike again. And the doctor was mystified. The doctor couldn't understand the idea that there might be a tradeoff. That life is dangerous; some things are dangerous. Sometimes it's worth it even though it's dangerous--this idea that economists have, that there is a continuum of risk, rather than safe-unsafe. And if you look at the pregnancy books and the guides and the other things you are reacting to, it's that people of course want to know: Is it safe? And the answer is: No; and it's not unsafe. It's complicated. And people don't like that. And doctors aren't trained to think other than that. Guest: Yeah. And some of what I get, like I get on the radio with doctors, you get: I just really care about the health of the baby. Yeah, well I really care about the health of the baby, too! Russ: No kidding. Guest: But I think we should also be making decisions which are correct. We shouldn't just not be doing things because we enjoy them. We should understand. And by the way: Do you ever allow your patients to take a non-essential car trip? Because that is very dangerous. Russ: Go skiing? Guest: Even if they don't let them go skiing. Getting in your car is the worst. We have to recognize that by not all living on a rural farm in Finland away from all motor vehicle transportation and only breathing healthy Finnish air, we are all putting our babies at risk. We have to recognize that and then make those choices correctly. Russ: But as you point out--and it's important that we mention this--as we live in this risky, dangerous world full of nitrates and cats and alcohol and Starbucks and all the other things that complicate our lives, pollution--our babies are safer than ever; our mothers are safer than ever. The maternal mortality rate is dwindling toward zero in developed countries and the infant mortality rate is dwindling toward zero. And the only reason it hasn't dwindled more is because we now count babies--I assume in the data--that are born at very, very early times with very, very low birthweight that wouldn't even have been in the denominator 50 or 100 years ago. Guest: Yeah. I think it's certainly worth noting that part of the reason that we can get so exercised about the question of is it appropriate to have 3 cups of coffee a day is because we are not getting exercised about questions like, 50% of babies are dying because we do not have good medical care for women. So I think that is a very valuable point.
45:49Russ: I wasn't going to bring it up but I want to now because we talked about--there's a lot of romance, and we talked about it for our children but we didn't end up doing it--but there's a lot of romance about home birth. And one of the arguments for home birth is, well, people have been having births at home for millennia. So it's obviously the natural, right way to do it. But as you point out, in developing countries, in poor countries, even a primitive hospital is better for the baby. Guest: Yeah. I think I do--the end of that clause is that women have been having babies at home for millions of years, but a lot of them die. I try, there's a lot of [?] about home birth; there are a lot of people on both sides of this in the United States. And I think it's a little bit of a different question to ask: Is it dangerous to have your baby at home in Park Slope, 5 minutes from a really, really nice hospital, with a really well-certified person? That's a very different question than: Is it safe to have your baby in a hut in rural Kenya-- Russ: In Finland-- Guest: In rural Kenya. Exactly. In Finland. I try actually in the book to do a more nuanced and balanced portrayal of this than I think sometimes we get in the media where we have sometimes the people on the home birth side on one side and the non-home-birth guys on the other side. I wouldn't do it, and I think we have to recognize that having a home birth has additional risks relative to a hospital birth. I think to not recognize that seems silly to me. Although I think in fact, like done correctly, it may be that those risks are relatively small. Russ: And offset, potentially, for some people by 100 other things that are nicer about home. Guest: Yeah. I have friends who have had their babies at home and said this was great and I understood that there was some risk but for me there was a really big benefit. And I think that acknowledging that people have different preferences, which take into account those risks and benefits, is very, very important. Russ: One of the advantages to home birth, by the way, is the avoiding of many of the things that you talk about in the book, things where, done to you or given to you, that you might not have been so crazy about. And sometimes there just is pressure to do these things. Whether it's a fetal monitor, an epidural. And I think you said the same thing--I've heard so many times, where I am fortunate enough to have a wife, I think fortunate enough, who had four natural childbirths. Didn't have an epidural. I think that's a good thing. Obviously it's complicated. I could be wrong. But there are a lot of times when someone said: Oh, you'll have one. With this smug, 'Oh, you'll have one.' And a lot of times the staff gets mad that you don't have one, because they don't want to listen to you yelling, or whatever it is--they don't want to deal with the complications. And so much of, I think too much of modern medical practice is: I'm doing this because, look, I've got to get out of here quickly so I'm going to give you Pitocin; I'm going to speed your birth along. Because I've got a golf game. I've got a trip to take. Etc. It's like: Why aren't I in charge of this? And I think there's a terrible problem that's not going to get better for a while, if ever, that the customer isn't in charge. Cause they are not the customer. Guest: Yeah and I think it's hard to advocate for yourself in that situation. I think one of the things that we worked--I also didn't have an epidural but most of my friends didn't think I'm a weirdo, and I think there are some real benefits to an epidural, clearly. But I think that this is a situation where it's hard to advocate for yourself, because the power dynamic is very skewed. And it's very hard if someone is telling you, look, you have to have a C-section [caesarean section] because otherwise it's dangerous for the baby. Most people, it's very hard to be like, well, are you just overreacting to this, this piece of information. I think that almost nobody is going to push back like this. Now, related to this is the very interesting paper by Erin Johnson at MIT (Massachusetts Institute of Technology) which actually looks at C-sections among doctors. And she sees among doctors in California, the doctors who themselves are giving birth are far less likely to have emergency C-sections. Which, certainly one interpretation of that is that in some other situations, doctors are overreacting and doing C-sections when they are not necessary; and when it's the doctor as a patient, they are better able to push back. That's kind of an interesting related fact. Russ: Yeah, that's fascinating. Or of course it could be that doctors are just less prone to C-sections. But that would be a tough argument to push. Guest: It seems hard, I agree. Russ: That would be difficult. I think that's another argument for having a doula, if you can afford one--a 'doula' being a person who helps out during the pregnancy, during the delivery. And it's really useful to have somebody--this is my small bit of advice for anybody out there who is expecting--it's very useful to have someone in the room who is not you and isn't your husband, who can represent you in these kind of situations, who has seen more than one pregnancy. More than one delivery. Which you often, on your first baby, have not. So, in our case we had the same issue that you talked about. We had a fetal monitor put in at one point; it showed that the heart rate was dropping, and our doctor hadn't arrived yet, and they decided to give my wife a C-section. Well, we really didn't want a C-section. We were really--but we weren't in a very clear emotional state. We didn't have a doula for our first child. And we were lucky; the doctor came shortly thereafter and said: Oh, that's just because there was a contraction; ignore it. Great, huh--Oh, we almost gave you a C-section, for nothing. But they were scared, too, by the way. Of course. It's a scary thing. But if you've seen 200 deliveries, you've seen that happen before and you react very differently than if it's your first one. But the more important point I'm making is that for my wife and for me, the emotional roller coaster that you are on at that point is not an easy place to make calm decisions. And you point out it's nice to have a list of things that you would like to have achieved during your delivery about what you will and will not do and at what point. And a doula is a nice person to help you do that. And if you don't have a doula, for whatever reason, just be prepared. You think: well, what good is it, this is all straight-forward; the doctor will do what's best. But what's best is very complicated. Guest: I agree. There's some randomized--I would go even farther on the doula thing. There's randomized evidence even that women given a doula when they arrive at the hospital, who haven't--just to have a labor support person in the room, [?] a C-section, [?] an epidural. I think in some sense that would be a cheap way for hospitals to decrease their C-section rate. Which is something I think most hospitals are trying to do. But I think certainly from the personal perspective that was probably the best decision that we made about labor and delivery, was to have a doula.
53:04Russ: So, let's move on. Let's move away from pregnancy for the last few minutes. I want to talk about some of your other work, not your pregnancy work. You wrote a very influential and much talked about paper on Hepatitis B. And you were trying to explain why the sex ratio in Asia was so dominated by males, why males were disproportionately represented. Talk about what you found in that paper and then what happened with other types of evidence that came along. Guest: Sure. So, when I first wrote this paper, in graduate school, there was some data from scientists basically suggesting that parents who are carriers of Hepatitis B had more male children. And I then put together a bunch of cross-country, cross-group data, which supported that idea: that places or groups that had more hepatitis had more boys. And I then took a next step and said, okay, one thing that we've noticed in developing countries is that places like China and India have more boys. And they also, China in particular, has a very high Hepatitis B rate. And I kind of did a little calibration and said maybe as much as 60% of this differential might be because of this Hepatitis B thing. And as you might expect, that claim was very controversial. It also, ex post, turned out to be wrong. So, maybe a year after I'd done this, some guys came along in Taiwan and actually had much better data, which actually had whether the mother, for a very large sample of women, like basically every birth in Taiwan for some period, they had whether the mother had hepatitis and what the gender of her kid was. And they showed there was basically no relationship. So then, people for whom this happens, I thought, is there any way to salvage this? And the thing I was asking was whether maybe it's the parental, maybe it's the fathers, the fact that the father has hepatitis is the thing that's actually driving this. Russ: That was clever. Guest: Yeah, well, one always wants to rescue one's work. So I actually went to China and ran, in conjunction with a bunch of doctors--they had a really big survey, in which we could see the Hepatitis B of both the mother and the father and look at their kids. And it turned out, at some point the data came in, and it turned out actually there was no link between either maternal or paternal hepatitis and the gender of the kid. So I wrote up a paper which was titled like 'Hepatitis B does not explain sex ratios in China,' and basically said this earlier argument was based on effectively circumstantial evidence. We have better evidence now and it doesn't seem to be what's going on. Russ: This puts you in a very small class of academics who have the privilege of conceding that you were wrong. And even better, publishing a paper showing it. That stuns me. Really--I'm not being facetious at all--how rare it is that when a paper comes out that counters what the original paper found, someone then concedes that they were actually wrong. Guest: Yeah. I think, relative to some of these other very long-standing disputes back and forth in economics about this stuff, there was a sense here in which it wasn't really that what I had done earlier was wrong. The conclusions were wrong but the analysis wasn't wrong. It was that this new data came out, and I think it made it, at least psychologically, a little bit easier to--although not that it was the biggest period of my professional life. Russ: It wasn't fun. Guest: But it was like a little bit easier to come back and say, new data comes out and that's how science works. And I think always in those situations you would like to be the bigger person, but it's very hard. Russ: Well, that's another interesting experience you got to have. These things happen constantly in economics--a paper comes out that makes a dramatic splash. People challenge it. And usually the people who get challenged don't concede an inch. In fact, they just say: That's true but I'll re-run the regression; I'll include this now; and look, it still holds up. Very hard. Guest: It's hard. I think we're all kind of--we all want--you get invested. Even if you don't ex ante care about what the answer was, once you've written it, you are invested. And it feels really bad to mess up. And I think that was definitely something that I learned. Russ: And you might have actually established something that's actually true. Let's close--I'd like to hear you speculate about the state of economics and econometrics. I'm a skeptic about our ability to tease out causation in many of these cases that we care about. Pregnancy is one example. We have some information; we've learned something. The question is how much. And in public policy, where the causal mechanisms are often much more complex, we don't seem to make much progress. Very few people concede they were wrong about the impact of the minimum wage, say, on employment when their study gets refuted, rejected. Well, it never gets refuted. A different study comes along and it finds something else. So, where do you think we stand? Do you think we are making progress? Guest: I think we are making progress. I see what you are saying and I think there are some fields in which the idea of doing things randomly, doing randomized stuff, has become more in vogue; and that has some downsides but it does help with some of these causality issues. But I do sometimes share your view that, particularly on some of these big things, it seems like we're never going to, there's just going to be just continual advances in methodology which lead us to kind of slightly different but not that different conclusions. And because none of these things are definitely, for-sure causal, it's going to be hard to feel like you are ever ultimately going to draw strong conclusions. But I do feel like as knowledge advances we do learn more. I hope.

Comments and Sharing



TWITTER: Follow Russ Roberts @EconTalker

COMMENTS (27 to date)
Brad writes:

Professor Oster carefully examined the data on drinking during pregnancy and concluded that there are no negative effects of moderate consumption. Earlier, Professor Oster had carefully examined the data on hepatitis and missing females and concluded that hepaptitis explained the missing females. It turned out she was wrong. To the best of my knowledge that error did not have any significant costs. Isn't it possible that she could be incorrect about alcohol as well? What are the costs if she is?

Greg Linster writes:

[Please remove and ignore previous comment with typos.]

This podcast reminded me of an idea Nassim Taleb has revived recently, i.e., anything (in the right dosage of course) can be "good" for you. In other words, the safe vs. unsafe dichotomy is silly -- what is "good" for you can be "bad" for you and what is "bad" for you can be "good" for you, again in the right dosages.

My wife and I are expecting any day now and I must admit (gasp) that she has broken a lot of the common rules about what she should and shouldn't consume. It's interesting to note, however, that there is a signaling mechanism at play with pregnancy too. While I'm somewhat of a skeptic and contrarian behind closed doors, I don't always signal that in public. What I mean by that is sometimes my wife would avoid certain behaviors in public (e.g., drinking any alcohol at all at a restaurant) in order to signal that we are "good" expecting parents. I suspect that many other expecting parents are doing the same.

[previous version removed--Econlib Ed.]

Myung writes:

It is extremely disappointing to listen to today's podcast referring to physicians as untrained in uncertainty and decision-making. The majority of decisions made by physicians are under uncertain situations with limited data due to individual differences and the various combinations of conditions each patients carry. Rarely do we deal with patients with only one condition such as pregnancy. I would have imagined that this podcast which is so sceptical on a broad range of data analysis would be so uncritical to such limited data regarding pregnant women. Also from a physician's point of view the most important endpoint would be the patient population as a whole. I believe that if such an approach of providing data and allowing alcohol in moderation were applied to all patients and that population resulted in a higher incidence of adverse effects of alcohol that would be sufficient reason against such an approach. This kind of approach has recently been expanded in comparative analysis studies. I would recommend that you look into the changes in recommendations for prostate cancer screening. For the benefit of the larger population screening is currently not recommended for the general population although a number of patients who develop metastatic prostate cancer would have benefited from screening. I believe that denying healthy women alcohol for 10 months is much less significant than missing the diagnosis of cancer, if it can prevent alcohol complications in a larger population

Philip writes:

I was extremely interested to hear what Oster had to say about pregnancy. It seems that no area in our modern lives is subject to as much tradition and folklore (not to mention superstitious type thinking). I think Oster did a really great job outlining what are difficult topics.

Given that Russ brought up the point on Taleb, I thought it was odd he didn't then come to the conclusion that: "why even drink during pregnancy if the risks are monumental?" It doesn't really matter what the probability is, it only matters what the severity of the outcome is and if a parent could reduce those odds quite easily (barring alcoholism) then it just makes sense to do it from a risk management perspective. Although its not exactly the advice most people would like to hear, it seems to run along the lines of logic as wearing a seatbelt. Even though a crash is unlikely, should it occur it will be highly impactful, so just wear it anyway, especially if the cost of participation is trivial.

Greg Linster writes:

Philip: How far should one extend that line of thinking though? Should one get in a car (or airplane) knowing that there is a small probability of dying in a crash, a very severe and harsh outcome indeed? I've grappled and struggled with the idea you mention for quite some time. I don't think it's quite right to suggest that a good risk management strategy entails avoidance of all things that potentially have severe outcomes without regard to their probability though.

brian writes:

Dear Russ,

The whole section on cats and Toxoplasmosis is bad advice. I have included a paper at the bottom of this message that addresses the issue on how cats act as a vector. The experts in this issue disagree with your guest.

A cat will shed cysts once during it's life. If you are unlucky, your cat might only shed the cysts during your pregnancy. This means your first exposure happens at the worst time. Better not to take that chance.

I also question how your guest can make the statement "if you look at what is the cause of most toxoplasmosis infections, it is not cats".

Most of the literature I have read is more nuanced. For instance in the paper below: "direct contact with cats is not thought to be a primary risk for human infection."

My impression from the literature was that it is difficult to know where or when the infection occurs because most people only get slight flu-like symptoms. They don't even know they were infected. No one can be sure when or where the infection was obtained. The only certain thing is that cats are the primary vector of transmission, hence the advice not to handle cat poo.

This statement from Oster is wrong: "partly because many people have already been exposed to toxoplasmosis if they have a cat and once you've been exposed it's not a problem". Infection of a foetus has been documented even when a women was previously exposed. Basically the parasite lay dormant, and then became active during pregnancy. It has happened

This is the only part of your overall discussion of which i have some knowledge. Given that her advice in this section is potentially dangerous I have reason to doubt her other assertions.

http://www.sciencedirect.com/science/article/pii/S1471492210000218

James Drake writes:

Epidurals are one thing, but what is the upside of moderate alcohol consumption during pregnancy? Drinking is a totally avoidable risk. For nine months, why take the chance?

The stigma on drinking during pregnancy is part and parcel of an overall societal "policy" where in essence we put a big sign on pregnant women that says "handle with care." It's something that we as a group have chosen to make unambiguously clear and universally understood. There are legitimate reasons why.

For every woman with great discipline who is able to limit it to a glass of wine every now and then, how many benefit from that societal pressure? Suppose there were no more stigma on drinking during pregnancy. What's the best case scenario for society? Now... what's the worst case scenario? Does it seem worth it?

To walk away from a sensible custom like this is hubris. It's this sort of foolishness that so many of our social problems are born from... And "nanny" government is the end result.

Michael writes:

James wrote:

"The stigma on drinking during pregnancy is part and parcel of an overall societal "policy" where in essence we put a big sign on pregnant women that says "handle with care." It's something that we as a group have chosen to make unambiguously clear and universally understood. There are legitimate reasons why."

Who is "we"? I personally know more than one [i]obstetrician[/i] who says the occasional drink during pregnancy is fine. There is no evidence to support the claim that an occasional drink during pregnancy is risky.

"For every woman with great discipline who is able to limit it to a glass of wine every now and then, how many benefit from that societal pressure? Suppose there were no more stigma on drinking during pregnancy. What's the best case scenario for society? Now... what's the worst case scenario? Does it seem worth it?"

You are making a lot of assumptions here. Why is "great" discipline required to avoid binge-drinking? Most people are not binge drinkers. Most pregnant women care a great deal about the health of their baby.

Why should people not be allowed to make their own, informed, decisions?

Listen to the Gary Tabues EconTalk episodes to see some of the harmful effects of stigmatizing the wrong things.

Ben writes:

Myung, it was people with statistical training like Prof. Oster that lead to the recommendations against prostate cancer screening. If it was up to physicians/urologists, population wide prostate cancer screening would still be the recommended - mostly due to their lack of training with statistical decision making but also due to not wanting to be sued and vested financial interests. In fact urologists demonstrate how easily thinking can be distorted and biased due to viewing only part of a whole population - those who would have been helped with screening and not the many more that would be harmed.

With the link between infant gender and hepatitis I would have thought there was an obvious confounding factor in places like India and China - gender selective infanticide for cultural and legal reasons (e.g. one child policy).

Michael writes:

Ben writes:

"With the link between infant gender and hepatitis I would have thought there was an obvious confounding factor in places like India and China - gender selective infanticide for cultural and legal reasons (e.g. one child policy)."

This was, in fact, the proposed explanation for the missing women. Oster's early work on this topic offered an alternative explanation that turned out not to be supported by studies evaluating patient-level data.

Myung writes:

Ben, the point in my post was that it is difficult, no matter how trained you may be in statistics to directly compare risk with benefits. Russ often mentions is other podcasts the illusive nature of numbers such as the multiplier, GDP or discount rate. With small changes to these numbers the results of a statistical analysis may be completely different. How do you compare the benefit of preventing 1 death with complications of sepsis and hospitalization from a biopsy of several patients? (in prostate cancer screening) What is considered equal? 1 to 3? 1 to 5? It is very difficult to know if a patient is prone to alcohol abuse. Even known alcoholics ( if no family is with them) will admit only to 1-2 drinks a week. This is especially so for obstetricians who do not have a long standing relationship with patients prior to pregnancy. By allowing moderate alcohol, a certain number of patients were to fall back into excessive use and resulted in a higher incidence of fetal alcohol syndrome, how many people must benefit from the relaxing effects of alcohol to justify that? 10? 100?

I believe that statisticians played a role in cancer screening recommendations but without initiatives from physicians It would have been impossible to fully understand the controversy and the implications to each patient. That is why different societies have different recommendations and different physicians have different opinions. Although you mention that professional societies are focused on only 1 group this difference acts as a check and balance mechanism not to become biased by one set of data. The is only limited information you can obtain from certain statistics and the first mandate is to do no harm.

Jerry John writes:

Russ,

Fantastic podcast.

By way of background, I'm a heart failure trained cardiologist contemplating conceiving. Moreover, as a physician I have 2 years of dedicated statistical training and study design and as such, perhaps may not typify the average doctor discussing data with patients.

I am not a teetotaler, enjoy wine and coffee, but do not smoke.

I had a very simple thought that was not brought up during the podcast, that stems from a Nassim Taleb philosophy and moreover, moral philosophy.

Simply stated, what leads toward the good, and diverts from the bad.

Not drinking alcohol or caffeine during pregnancy will have no bad consequences for the child, i.e. zero risk.


It is the rare parent who serves their infant alcohol or coffee in the bottle; i.e. zero risk with abstaining and no definite benefit toward the good by administering the agent.

I understand and agree with concept of confounding variables and interactions in observational studies.

However, the physiologic effects of caffeine and ethyl alcohol are well characterized and represent a dose effect, rather than a binary effect.

Apart from purported effects of hormesis, I cannot ascertain a good effect for children in utero or during childhood of administering ethyl alcohol or caffeine.

Ethyl alcohol or caffeine are natural compounds with physiologic effects enjoyed by many individuals, similar to cocaine, though the latter has more restrictions in the United States.

Risk is definitely a continuous variable, sometimes linear, and at other times non linear in response to an independent variable.

Balancing the pleasure of the mother against the health of a child will always remain a personal choice and though alcohol and caffeine are a bit more nebulous with respect to risk, relative to lead, mercury or smoking, my premise stands on embryologic developmental and that a developing child's brain and organs may not have the ability to detoxify (hepatic, pulmonary, or immunologic clearance) compared to an adult.


As Nassim Taleb points out, its not merely the risk prediction, but the magnitude of effect. Lead poisoning, fetal alcohol syndrome, mercury toxicity, ionizing radiation exposure , iodine deficiency, are devastating, long lasting outcomes for children.

Not all medical knowledge needs to be tried in the crucible of a randomized trial. I'm a big fan of physiology and common sense.


Look forward to more podcasts.

[email address in url removed--Econlib Ed.]

Aaron writes:

She sounded reasonable but I'm skeptical since EconTalk has a habit of giving air to reasonable sounding guests who come in and basically proclaim an entire field to be mistaken (Gary Taubes comes to mind along with the exercise guy who hated running).

I think it's an issue of expertise. No matter how skilled you are if you enter an unfamiliar field you're going to pick a lot of low hanging fruit that doesn't really exist. That doesn't mean you can't make a contribution but you need a subject expert to help you distinguish the real insights from the things that only look insightful because you don't understand them.

That's my concern with this guest, she may have done a fine job of communicating accumulated medical wisdom about risks during pregnancy. But she also spent a lot of time suggesting that she had a better understanding of medical research than the MDs and epidemiologists who created it and that's a bit of a red flag to me.

Amram writes:

I appreciate how confusing and frightening having a child is, in this and every age. That said, I found Ms. (Dr.?) Oster's claims, discovery, decisions, and conclusions various levels of inaccurate, needlessly adversarial, and/or self-serving.

1. I do not know what kind of medical care Mr. (Dr.?) Roberts and Ms. Oster received, but my obstetrician was immediately forthcoming with statistics, links to extended information, studies, his reservations about reproducibility, probabilities, and the difference between advice predicated on 'better safe than sorry' and solid statistical evidence.

2. All power to people who feel the need for a doula if the woman/couple want more support than her family, doctor, and a bevy of nurses are able to provide. But it is needlessly adversarial and, in my multiple birthing experience, completely unfair to paint a picture where the doctor and the nurses are not there representing the interests of the parents. My doctor was well known and respected in the hospital, focused on our birthing needs, and able to get what he wanted done, when he wanted it done in the byzantine bureaucracy of the hospital in hop-too time. As far as an 'advocate' I can think of no better. Backhands like 'scheduling his golf-game' and 'you are in no condition to advocate' imply there is some conflict between the doctor/nurse's goals and requirements and the expectant mother's. In my and ALL of the parents I know's experience, this was nothing near the case. The hospital staff and personal doctors put the needs of the patient as not the leading, but the only issue of concern. The children were born healthy and, if complications arose, the parents were provided necessary information to CHOOSE the direction of remediation.

3. Glass of wine, piece of sushi, cup of coffee? Really?!? These are the necessities Ms. Oster went to the mat over? We were informed about dozens of genetic and environmental diseases, influences, and mitigating factors. My doctor taught us about natal nutrition, exercise, sleep positions, sound and pollution limits (we live in a city), and countless other facts relating to bearing a healthy baby. I notice Ms. Oster did not bother going over any of those issues because they are obviously to the benefit of the mother/child. If the research is trending in a direction, it is better for the doctor to tell his patients to abstain until further information is available. That is their job. They are health care providers, not medical researchers. Sure, if pressed, they will tell you "the research is unclear", but their basic job is to ensure the health of the baby, not the full education of the mother.

The general tone of Mr. Roberts and Ms. Oster was one of deep suspicion and caustic dismissal with the medical profession. As people who, thankfully, have, I assume healthy children with reasonably pleasant birthing experiences, I wonder at the basis of their mutual distrust.

That said, keep doing what you are doing...I listen every week.

Amram writes:

While deciding what about Ms. (Dr.?) Oster's various claims triggered an annoyance response in me, I came up with an analogy rather than analysis.

When a pregnant woman enters my train car, I get up and offer my seat. Whether she takes it or not is her choice. But if she declines the seat and the train lurches and she falls down or bumps against the pole, then I have to wonder to myself "why didn't she just take the seat?" Sure, the chances of a mishap are exceedingly small, but the dangers are concomitantly large (thanks Taleb). And if I were invested in the well-being of the child beyond that of a casual observer (say husband or family member), I would feel great anger if something so readily preventable caused harm to the baby.

We, as a society, try to place a protective fence around pregnant women indicating our respect for the high-risk delicacy of the condition. In fact, we do not go near far enough. Legislated pre-birth leave would be a nice start. Nothing about our collective nod in the direction of expecting mother is coercive, other than in social ways ("Are you going to have a DRINK?"). If a woman, such as Ms. Oster chooses to put her coffee addiction above the admittedly dubious risk of birth defects, so be it. But forgive the rest of us for thinking her decision is completely related to a low tolerance for self-denial and little to do with the statistical indefensibility of her doctor's recommendations.

I notice the one vice she did not partake in, cigarettes, came in for blanket opprobrium. Perhaps there was a sampling selection bias in the tests she chose to review. When it was coffee or wine an "out" was found. For ciggies...not something Ms. Oster particularly cared about...the doctor's word was law.

So, by all means Ms. Oster, keep standing. Perhaps you are off at the next station. But I will dutifully rise for the next mother-to-be. When she takes the seat, I will not think "does she know there has never been a problem with pregnant women falling on the subway". I will think "at least she will not be the first".

Dmitry writes:

I was a little bit disappointed with this one.

It seems like Emily was continuously self-contradicting:
in the first part of the podcast she complained about doctors not giving her information for making her own decisions; in the second part she spent a lot of time talking about difficulties of inferring right decisions from complicated and entangled medical data.

This looks like a joke to me - if the data is so controversial why the hell do you want to make decisions based on it?

It seems that in this particular area tradition and folklore are much better for decision making than standard data-based approach; it is a very Hayekian idea that in a situation of complicated phenomenon traditions aggregate much more information than any particular agent, even with data-mining skills Emily has

Krishnan writes:

A good conversation ... "data driven" as Oster mentioned a few times ... and thus, if better data comes along in the future that may challenge some of the conclusions, then so be it ... What came through for me is the fact that women (and men) are frightened by all the messages they get about this and that WITHOUT data - based on "traditions" or "folklore" (Sure, there may be some basis for traditions and folklore, but tough to figure out what is true and what is simply held to be true even it were false) ...

I have learned to distrust publications when it comes to "tobacco" - I remember examining data about "second hand smoke" and concluded that there was indeed NO effect on non smokers from close, second hand smoke - but "tobacco" is one of those things that today's society considers "evil" and "must be gotten rid off" and so on - there does appear to be a strong correlation between smoking and certain diseases - there is also some more direct, in-vitro evidence showing changes in DNA from some of the drugs in tobacco - but because of the insanity with which society treats "tobacco" I have cause for concern about the scientific process ...

I also suspect that genetic predisposition may have more to do with what does happen to women and their children during pregnancy than their behavior (drinking, smoking - in moderation) ...

I had not known about the hepatitis B paper and the reworking of that ... yes, amazing to see an author retract the conclusions by examining an older study - rare indeed.

Michael writes:

Dmitry wrote:

"This looks like a joke to me - if the data is so controversial why the hell do you want to make decisions based on it?

It seems that in this particular area tradition and folklore are much better for decision making than standard data-based approach; it is a very Hayekian idea that in a situation of complicated phenomenon traditions aggregate much more information than any particular agent, even with data-mining skills Emily has"

The problem is that there is no one, definitive "tradition". Right now, in the US, we have a cultural norm that pregnant women should not consume even one drop of alcohol.

However, somewhere in the ballpark of one-third to half of obstetricians say that the occasional drink is not harmful. Our cultural taboo is not based on anything close to a universal opinion among our medical professionals. And it's a relatively recent taboo, and one that doesn't hold in every country, even though every country does have pregnant women and doctors who care for them.

If a recomendation is made, it it completely sensible and reasonable to ask... on what basis is this recommendation made? The answer could be data, could be tradition (but whose tradition/which tradition/what happens when there are conflicts between traditions)?


Kevin Smith writes:

As a physician and PhD in Epidemiology I thought the discussion was very interesting. It is always very helpful to get an outside perspective and to illustrate where we have work to do and remind us that our data has incredible limitations. I agree with the author that a general skepticism about the "risk factors" identified by epidemiologists and transmitted by doctors is warranted, but as others have pointed out the risk itself is not enough it is important to appreciate the harms. And, as another above pointed out almost all the risks described are continuous not binary.

As stated by Myong above I thought the dismissal of physicians handling data and uncertainty was silly and lacked any appreciation for the physician training and actual practice. Maybe you had really bad doctors. The impression is perhaps formed because you are pushing physicians on what is information of little expertise for them - physicians cannot be experts on the literature of risk factors because doing that requires an incredible amount of time and energy. I spent 20 pages per risk factor in my PhD to develop the nuance and criticism necessary to understand risk factors for a single outcome, that level of working knowledge is not useful for physicians day to day. That is the purview of epidemiologists who work with physicians and produce recommendations. Physicians can learn those recommendation and transmit them to patients but expecting them to have the level of intimate data knowledge the author wants neglects that talking to you about risks is a very small part of their day to day decision making, which is often done in a haze of uncertainty because of incomplete knowledge (not all things can be studied) interacting with incredible complex systems (the human body and its interaction with the environment).

The patient and Russ laughingly dismiss other forms of decision making besides being data driven, but basic economics dictates that for most patients getting the level of data analysis the author did is very unlikely on the margin to make any difference in the outcome and as such a complete waste of time. An alternative decision making strategy is to trust the expertise of the person you are paying (within reason) and make sensible choices. What if a patient said - well I could read that book or I could just do what my doctor recommends, avoid coffee, avoid cats, avoid EtOH, and smoking. They might miss out on a few drinks, a little coffee, and some other food items, but will save time analyzing the data/reading the book to gain a more nuanced appreciation of the data - time they can spend doing much more interesting and enjoyable things for them (unless that is what they enjoy, as Mrs Osler does). And at the margin they will have no worse outcome than someone who spends the time.

Ok, two more problems - both briefly mentioned in the podcast. As compensation falls for physicians they sadly have less and less time to spend with patients to discuss these issues. Secondly, if they tell you some EtOh is ok, and you drink and have a bad outcome they will be sued and in all likelihood lose. That drives a lot of discussion about risks, being nuanced can lead to legal problems if there is a bad outcome, especially with babies. That is a sad reality in the US.

A very tangential side note, Myoung notes that there is now a recommendation to no longer do PSAs. There currently are recommendations, looking at the same data, to do PSAs from the NCCN, ACA, ACR and others (none of those are urologist groups, although they recommend it as well). Both sides of this debate are stacked with people who know statistics and math. One organization suggests not doing them, several others look at the data and disagree. I suspect the government organization is driven more by cost than the science - one of the largest randomized controlled trials ever done shows an overall survival benefit with PSA screening. Cost is likely the driver here for the recommendation - not medical science. I am not a urologist, but even if I were that does not mean the academic urologists don't know how to do data analysis. What the new recommendations from the government have made me wonder, is if this is the kind of work they do in cancer recommendations, what else should I be suspicious about?

Emily Oster writes:

These are great comments, I appreciate everyone weighing in. A few reactions.

1. To the question of why focus so much on alcohol and caffeine, when there are bigger issue in pregnancy -- I agree. And most of the book is on other things. For me, issues of prenatal testing were very salient, and much of the early research I did during pregnancy was on this. The book covers this, along with home birth, drug safety, other birth interventions and so on.

2. On the criticism of doctors. There is no question that many doctors are great, and do go through the statistics carefully with their patients. It's also true that statistical training in medical school -- that is, training in how to evaluate causality in observational data, in particular -- is done less than in a graduate program in economics or epidemiology or public health.

But fundamentally, no matter how good your doctor is on this, most women have to make these decisions on their own. When you think about prenatal testing, your own preferences are crucial for deciding what to do. Your doctor can explain the data to you, but you have to make the decision. And with limited time available with medical practitioners, most women look outside their doctor for advice in pregnancy. My goal in writing the book was to give them a place to look which was rooted in the data.

3. On alcohol. I think we need to be clear on the distinction between "what is a good recommendation for government to give" and "what does the data say." Even if the data says -- as I argue it does -- that occasional drinking is fine, it still may make sense for the government to suggest abstinence as a policy. This depends on whether we think that recommendation is more effective at reducing heavy drinking than a recommendation of moderation. This isn't something we have much evidence on, and not an issue I go into in the book.

It's also clear, as other commentators have pointed out, that many OBs will tell their patients than an occasional drink is fine. This is what mine said, but I wanted to see the data behind that . I ended up agreeing with her, but I wouldn't have felt comfortable making this choice without seeing the data for myself. Others may find that they'd rather just do with their doctor says, which I think is fine, but in this case you have to then accept that you'd do something different if you had a different doctor.


Ben writes:

Kevin makes some very good points. However I don't believe the reason the USPSTF gave for recommending against PSA tests was costs in dollar terms. Even if this was true that would be a good thing as the USA spends 18+% of GDP on medical services for worse outcomes than other countries (e.g. Australia, Scandinavian countries, etc) that spend half as much. Making recommendations based on "medical science" without consideration of economics is the wrong way to look at things (unless you have a vested interest) as there aren't an infinite amount of resources that can be allocated.

Urology seems to be a field where hip-pocket based medicine is practised rather than evidence based medicine. Not only was prostate cancer screening done for a long time without evidence of its efficacy, there is also little evidence that robotic surgery is any better than traditional laparoscopic surgery.

Joe writes:

I enjoyed this podcast - my wife and I just had our first child and there was so much information to process that we typically didn't have the time or resources to understand some of the recommendations.

The one item I wish was covered was the mental psyche of the couple in the unfortunate event of a miscarriage. Our first attempt at starting a family did not go as planned and my wife was adherent to the recommendations from her doctor. Even with that, she wrongly blamed herself. Had she indulged herself to a drink once in awhile, I think there would be a nagging feeling that she caused a miscarriage to happen even though it is a common event early in pregnancy and potentially independent of alcohol or caffeine consumption. Based off our experience, that alone was enough for her to adhere to the recommendations the next time she got pregnant.

As for the doula, I could not be a bigger advocate for them. Ours offered support in the delivery room by translating what was going on after the nurses/doctor left the room to make sure we were prepared what was coming next. She also coached us at home to make sure we didn't arrive to the hospital early - if it hadn't been for her, we would have arrived at the hospital hours earlier but instead got to spend that time in a comfortable environment.

Stephen Valder writes:

I enjoyed your recent podcast regarding pregnancy. I am a pediatrician, and an avid listener to your podcast, and others like planet money and freakonomics.

I agree with how the author used the data to understand the risks of caffeine, alcohol, etc...

I think some of the critical feedback is because the physicians and other advocates are asking a different question. Her question is how does this data inform the expectant mother.
The physicians question is how does this data inform me, and how I instruct my patient. I need to know if my telling women to not drink at all decreases adverse outcomes more than advising moderation. I don't know if that data exists. But I would guess that those that have reacted would agree with the data she gives, but have a different application for public policy

Steve writes:

That was a good paper. A topic related to pregnancy and health decisions that I was curious about is how one evaluates decisions such as whether to pay the $2,000 plus a small yearly fee for the uncertain future benefits of banking a baby's cord blood. Here in CA, we receive numerous pamphlets from the state and advertisements from numerous companies that perform this service, but do not know what framework to use to evaluate this since the potential benefits are not well defined.

Chris Stone writes:

Ms. Oster was definitely not "nauseous" (during this podcast, at least.)

nauseous
Causing nausea; sickening

nauseated
To be feeling, or having been caused to feel nausea.

So, next time you are tempted to say "I feel nauseous", understand that you are saying "I feel that I make other people sick", or basically "I feel nauseating".

I enjoyed the podcast with Emily Oster, still think it's best for pregnant women not to drink alcohol, coffee, and smoke. A side point, in much of Asia women still don't drink alcohol in public (this may be changing thanks to marketing) and traditionally don't drink at all; it's considered behavior that characterizes prostitutes. I think there is well documented science that alcohol does have a negative effect on the ova. Is it possible that this tradition is based on folk wisdom and experience, thus women should err on the side of caution? Given that they carry the future, this tradition may constitute genuine wisdom. No data needed for this decision.

megan writes:

I just ordered Ms. Oster's book (Dr. Oster? Not sure)! I had been meaning to for awhile and the podcast reminded me. I am very excited to read it. I am also a big fan of Econtalk and have to thank the program for making economics accessible to laypeople like me -- my husband is an economist and this is one of the few economics-related things we enjoy together. I find nearly every podcast interesting, informative, and (usually) easy to understand even though I have no formal training in the field.

One question related to this podcast: near the end both Russ Roberts and Emily Oster mention that they did not use epidurals (well, in the case of Russ, he was talking about his wife). And I was curious whether this was merely a personal preference or whether there were reasons related to Ms. Oster's research that led her to make this decision? I am pregnant with my second. I had an epidural with the first and plan on doing it again, but would (potentially, ha!) reconsider this position if there was enough evidence that it is truly beneficial to go without one and/or detrimental to have one. I believe the book may go into this more, but thought I would add it to the comments only because I was surprsied they discussed the use/non-use of epidurals without stating the reasoning behind the decision for each party.

Dr. Duru writes:

I think Oster minimizes the potential range of explanations for the results of the MIT paper on C-sections.

Firstly, it is a well-known fact that C-sections are generally more risky than natural birth simply because they are surgical procedures. Natural births are also *easier* for doctors, and they take on more of a role of an assistant in helping the baby through the birth canal. So, doctors do not recommend C-sections because they are trying to get to the golf course on time. They generally recommend them out of true medical necessity or what they otherwise conclude is the better risk/benefit decision.

So, it is odd that doctors would seemingly have lower C-section rates than the general population....until you examine the behavior of doctors over a range of medical procedures. I can only speak from anecdotal and personal experience here, but the doctors I know are notorious for minimizing the need for doctor intervention when they themselves may need medication attention. Doctors have an adage that doctors are the worst patients...yes, they push back, but it is because they think they know better about their own health and the intimate familiarity with risks makes them very comfortable in assuming risks for themselves are much lower than they might really are.

So, in the MIT paper, I think it is very possible that the lower C-section rate is a result of doctors pushing back, but NOT in a good way. In other words, it very well may be the case that these doctors were ignoring very sound, medical advice and instead chose the path of higher risk out of false familiarity with the risks they themselves faced. Sure, I guess you could argue that ALL patients should have the complete freedom to make such potentially hazardous decisions...even under the duress and immediacy of childbirth...

(For the record, my wife is an OB/GYN. Our first child was born out of a VERY urgent, emergency C-section. In *hindsight*, my wife continues to wonder whether we could have just taken the risk of the natural birth, but I for one am just relieved that we have the technology to detect the potential issues that our baby had in utero...and I am of course even more relieved that everything worked out OK for Mommy and baby!).

Comments for this podcast episode have been closed
Return to top