American Enterprise Institute
February 22, 2006
[Edited transcript from audio tapes]
Proceedings:
Many experts today insist that a patient’s race profoundly affects how the medical-care system deals with him. The notion that physicians are biased against minorities––overtly or subtly––has acquired considerable weight in both academic literature and the popular press. In their new book The Health Disparities Myth (AEI Press, 2006), authors Jonathan Klick, a legal scholar and health economist, and AEI scholar Sally Satel, a physician, critically assess recent research bearing on this question and discuss other factors that contribute to health-care disparities. They also suggest strategies for improving the health of all underserved Americans. They presented their findings at a February 22 AEI book forum.
Sally Satel, MD
AEI
Good morning. I’d like to get started. My name is Sally Satel. I am a scholar here at AEI. Our topic today is health disparities, and the occasion is the publication of this monograph that I’ve written with my colleague Jonathan Klick called The Health Disparities Myth: Diagnosing the Treatment Gap.
Jon is going to present our major findings today, but I wanted to take a few minutes to orient you as to why we wrote this. First, to be honest, I wanted to write this to set the record straight. I’m a physician, I’ve been around other doctors and medical centers for almost a decade, for nine years in Chicago, in New Haven, places where there are a lot of poor folks, a lot of minority patients. And I was very surprised within the last few years when I started to read that physicians were biased as a class, that there was prejudice and discrimination in the way they dealt with patients and that this manifested in a treatment gap--the notion that minority patients frequently got fewer procedures and poorer quality care, and of course poorer health status.
Now what I saw when I worked as a doctor were lots of hard working men and women, nurses and physicians, who gave everyone, it seemed to me at least, the best care no matter what their race. But it really didn’t matter that I was offended that there were accusations of bias and discrimination, because perhaps there were data that could confirm this. So what were the data? So every time a paper would appear in a peer-reviewed medical journal alleging doctor bias, let’s say in the use of cardiac procedures, I would read the original. Frequently these were papers that found their way into the mainstream media often on the front page, reported in very sensational ways, you know, “Bias in the Health Care System,” “Healthcare is Better if You’re White.” Well, it’s better if you’re rich, we know that, and whites often tend to be better off than blacks. But it was alleging actual racial discrimination. So I’d read the paper, the original, and I’d find that it didn’t demonstrate discrimination of bias. Frankly, that’s kind of a hard thing to demonstrate empirically when you think about it. It did find often a difference in care, a difference in treatment, which is significant, but it did not find bias. And also, the authors of those studies did not make those conclusions that bias was the cause. But still, this notion that physician prejudice and discrimination was creeping quite efficiently into the atmosphere.
In 2002, when the Institute of Medicine published its report Unequal Treatment, it explicitly claimed that bias, prejudice, and discrimination on the part of physicians was a significant cause of treatment differentials. Not the only cause, but a significant cause. And it subsequently became almost accepted wisdom.
Meanwhile, in another universe, Jon Klick, a health economist, was doing his own research on the factors behind health care differentials by race. So the monograph then was an outgrowth of my wanting to set the record straight about the claims of bias in the maintenance of a treatment gap and how I felt they were unfounded by the literature, and Jon’s work on what in fact is driving the treatment gap.
I want to say also a quick word about the title. A few people have commented to me about the word “myth.” The title as you know is The Health Disparities Myth. Now, there is nothing mythical about the difference in health status that minorities have: it is poorer on average, higher infant mortality rates, lower life expectancy, etc., and also frequently they will get poorer quality of care and less care. This is true. This is not the myth; this is a fact. But what I am fixing on is the word “disparity.” If you look it up in the dictionary, it is a perfectly neutral word. It means difference. But over the last few years, it has taken on another connotation. It has taken on the meaning of that of injustice, a difference that is perpetuated by unfairness or an injustice, or, in this case, a racial prejudice. And that’s what Jon and I are skeptical about. We acknowledge that health care certainly varies by race, but we challenge the idea that it varies because of race.
And this is very important of course because it dictates what our policy solutions are going to be—our policy solutions for improving minority health, which is something we need to do. So we’ve invited several outstanding people to comment on our monograph today. We’ll start with Jon discussing the book, then we’ll have our lady and gentlemen respond, then Jon can respond to questions they raised, questions between all of you, and then the audience. So now I’d like to introduce Jon Klick, who was here at AEI last year. That’s how I had the pleasure of getting to know him. But now he has moved on to Florida State University in Tallahassee, where he is the Jeffrey A. Stoops Professor of Law. He’s published widely on healthcare economics and issues related to individuals’ access to care. Jon . . .?
Jonathan Klick
Florida State University
Thanks, Sally. I’m going to talk basically about the framework of our book and of our argument. I’m not going to dwell too much on some of the original research that we rely on because I’m actually flanked by folks who have done that original research and, you know, I’d rather hear them talk about it than myself.
But, Sally and I started out and sort of think that one of the most important issues involved with the whole health disparities discussion, and particularly the way the IOM report and those who have jumped off the IOM report have presented the health disparities issue, comes down to a fairly simple statistical issue—something that people talk about as an omitted variable’s bias. In practice, when you’re looking at correlations or relationships between lots of variables, you’re never going to be able to control for everything. In practice, the data don’t exist; in practice, it may not be particularly obvious what should matter, what shouldn’t matter; and sometimes, you’re not even going to be able to control for things that you think really do matter. Maybe the data don’t exist, maybe in the data set that you’ve got, you don’t have all the information you’d like to have.
But, when the omitted variable’s bias comes up and turns out to be particularly important is when the omitted variable that you can’t look at and you can’t look at it’s effect actually tends to relate to something that you do look at. And in the health disparities area, this turns out to be huge. The health disparities literature, and certainly what the IOM folks focused on, was race. They looked at differentials that related to race and said: “Ah ha! There looks as though there is something going on here, there’s some sort of bias going on here at the doctor level.”
But it turns out, as pointed out by the research and by the folks around me right here, that race correlates quite highly with residence. And it turns out that residence matters quite a bit for both health treatments, health treatment patterns, and probably eventual health outcomes. So, the fact that most of the studies that the IOM relied on focused on race but did not control for residence has the potential to generate questionable conclusions.
Now, it turns out that why does residence matter? Well, black and white patients don’t go to the same hospitals or even see the same doctors. So, to infer physician-based bias from differential treatments is, at least as a first cut on things, unfounded and potentially harmful.
So let’s give you a simple example. Let’s assume that everyone, black and white, who lives in Town X receives good care; and everyone, black and white, who lives in Town Y receives bad care. Well, if more black people live in Town Y, then if you do some correlation analysis or some regression, if it doesn’t control for residence, and if it turns out that, once again, these races aren’t uniformly distributed across these towns, well, you’re going to pick up what seems to be a negative health treatment effect relating to someone being black, when in fact, by the simple example, it comes from being from this particular town that provides sub-optimal care—not having anything to do, at least as the first order question, with that person’s race.
Now, if you look through the studies that the IOM relies on and the folks who have been trumpeting health disparities and a biased-physician model of health disparities, these folks seem to be uninterested in, or at least unconcerned with, this residence effect and seem to be entirely consumed with the race effect.
But, it’s interesting, in studies done after the IOM report came out, a lot of them done by economists but some done by folks in the public health world, they show that really residence is a big, and maybe the main, part of the problem. So, for example, Kate Baicker, John Skinner, and Amitabh Chandra and other folks at Dartmouth have found that residence matters very much, specifically if they look at health markets across the country, as they define them. It turns out that only 10 percent of these health markets, these local health markets, have what we might consider to be a representative, a nationally representative, population. The rest of the health markets have either disproportionately many minority residents, or disproportionately many white residents. And what they find is that you see big residential pattern differences, systematic differences by residence, in both terms of health treatments and in terms of health outcomes. And so, for example, some of the work done by the Dartmouth folks finds that some particular treatments and the likelihood of getting a particular treatment, so for example, eye exams for diabetes patients, declines as the representation of black folks in the particular health market goes up. It’s a residential effect, at least in terms of the exact micro-relationship as the number of black folks in the local health market goes up the likelihood of getting certain kinds of care goes down.
Angus Deaton at Princeton has also found, looking at mortality rates, for both white and black people, mortality rates go up as the number of black folks in your given area, your local area, goes up. Jeannette Rogowski and folks at RAND have found that black babies are systematically born in lower quality hospitals. So, hospitals that tend not to have neonatal intensive care units, hospitals that disproportionately tend to be government run hospitals. And they’re able to correlate that with worsening infant mortality. If you happen to deliver in a low-quality hospital, your chances of the baby’s survival go down, and it turns out that the preponderance of lower quality hospitals is disproportionately found in minority neighborhoods—and specifically black neighborhoods.
So, the local health care market matters significantly in terms of what kind of treatment you get, and what kind of outcomes you’re going to have. One of the most important things that correlate with how good or how bad your local health care market is the racial makeup of the market.
So if we move down from the local health care market, and we just look at hospital differences as well, it turns out that some public health people have found that if you look at racial health disparities, but then control for what hospital the person was treated at, it turns out that the hospital differences matter much more than what your particular race is, in terms of what treatments you get and in terms of what your eventual outcomes are.
Peter Bach, sitting to my left, and some colleagues have found that black patients are much more likely to see doctors who are not board certified in their specialties and have less access to high-quality referral networks. And so, what we find is there are lots of differences that do happen to vary by race but not because of race. It turns out that it matters where you live much more so than what racial background you’ve got. So at the end of the day, it looks like when white people and black people see the same docs in the same hospitals in the same area, their treatments do not seem to differ systematically.
Now, you might say, why does it matter? Why does this matter? At the end of the day, if it turns out that black people and minorities are getting screwed in the health care markets, why do we care what the actual causal mechanism is? Sally and I aren’t claiming that there aren’t disparities. Sally pointed out that we call this The Health Disparities Myth not to say that the disparities or the differentials by race don’t exist. What we’re calling the myth is what is the causal mechanism. And you might say, “Who cares? This is sort of a trivial question. If there are disparities, we need to fix them. We need to help those folks who are getting poor care.” That’s right. But how this is actually happening matters quite a bit. If it is the biased doctors who are causing the problems, and so we take the prescriptions laid out by the IOM, you know, things like we need more cultural competence training for doctors, or perhaps we need affirmative action in medical schools, or things like that. If doctors really are driving the health care disparities, then those are perfectly sensible solutions. Right? But if the causal mechanism is different, if the reason these disparities arise is something more systematic, something more systemic than actual doctor-patient relationships, that has to do with socio-economic factors, that has to do with why do we see black people and white people living in different areas, if that’s the causal mechanism that’s going on here, at best, the IOM prescriptions are wasteful. The cultural competence isn’t going to do very much, affirmative action isn’t going to do very much at the medical school level, because we’re still going to have different residential patterns and this is still going to largely lead to differential care. And so those resources are going to be wasted, at best. At worst, these kinds of prescriptions could actually lead to health care losses for both blacks and whites alike. And so it’s our position that understanding the real causal mechanism here is the first step in coming up with useful policy prescriptions. We’re never going to get anywhere in this discussion if we don’t search for the causal mechanism.
Maybe at the end of the day, if we were able to make some gains on these more systemic issues, well then maybe we might find there is some residual doctor bias or something like that. But as best as we can tell from existing literature, that question or that problem is by far, by orders of magnitude, a second order question to this more important question of why is it that people in poor areas get worse care. And so if we take that as the most important question, the policy prescriptions that flow from that are quite a bit different.
What we need to worry about is how we can get people in these bad neighborhoods, black and white, or in these poor neighborhoods, black and white and other minorities, how can we get them better care. And so the solution may involve subsidizing higher quality doctors to go into these areas, or it may involve subsidizing transportation to get folks to where the better medical resources are. It may be something even simpler as funding night hours for clinics. That’s an idea Sally has talked about quite a bit. If we fund night hours for clinics, maybe we can get better access for folks in these poorer neighborhoods to higher quality healthcare. And so we think that’s where the question needs to start before people start making policy prescriptions and jumping off the sensational reception that the IOM report and following literature have gone. We need to worry about what is the real causal mechanism or else we’re not going to make very much headway on this question.
Now that’s the main part of the book. We also discuss a handful of other issues. One issue that we pay some attention to is the IOM report focuses on health care treatment differentials, and we echo some other people who have pointed out that treatment differentials may be irrelevant at the end of the day. What we should really be worried about is outcome differentials. It turns out that in a handful of studies you can look at, it may not be that black and minority patients are under-treated; it may well be that white patients are over-treated. That is, they get treatments that have no objective effect on their own health outcome. And you might think, “Well, why might that be?” Well, we’ve got some evidence from the RAND health insurance studies in the ‘70s and ‘80s that suggests that there’s a big consumptive or consumption aspect to health care. People consume health care not necessarily to improve their objective health, but they treat it like going to the movies or something like that, where at a minimum, they may treat it as an extra insurance function that doesn’t really provide anything more than some kind of a placebo effect. So, if that’s the case, if we see health treatment differentials, well, at the end of the day, we might not be so worried about that.
There also may be some effect where there’s wasteful health care, or wasteful treatments, that are produced as an effect of some worries about litigation. There’s been some work done on defensive medicine, and if it turns out that doctors feel that they’re more likely to face litigation from white patients than from black patients, they may over-treat the white patients for purely defensive reasons. It’s not for anything that has to do with actual health outcomes. And if that’s what’s driving health treatment differences, we shouldn’t be so worried to the extent that we think minority patients are getting short-changed; we should be worried about why do we have this wasteful treatment.
And so these kinds of questions, I think, are important as well. And we need to reframe the issue that the IOM has given to us, reframe it in terms of what really driving the differences. And at the end of the day, do the differences really matter? Or are they something that doesn’t make all that big of a difference in the health care or in health outcomes and are really red herrings that we’re wasting time and energy on. So that’s the crux of our book. As I said, Sally and I are really doing much more synthesizing than adding very much to the original research. And so I’m really interested in hearing from the folks on this panel who actually have done the original research in this area.
Sally Satel
AEI
Thank you, Jon. Well, that leads us to Peter Bach, who has done a fair amount of original research. Peter is a physician and a specialist in pulmonary medicine. He is associate attending physician at Memorial-Sloan Kettering Cancer Center in New York, but within the last year has also been working at CMS, the Center for Medicaid and Medicare Services as a senior advisor to Mark McClennan. He is published widely on health disparities, and in fact, I read his first article on lung cancer in 2000, 1999, Peter? And that really was a seminal article in the disparity debate. And now, Peter.
Peter Bach, MD
Memorial Sloan-Kettering Cancer Center
Thanks, Sally. Thank you for having me, and good morning. I’m going to talk a little bit about research and what research I’ve done looking at health disparities. I want to back up a little bit from Jonathan’s comments and explain that at least my interest in research, but I think most researchers focus on the active empiricism and the act of counting things and measuring things, and trying to disentangle the complex social issues into numbers that can at least be interpreted in a way that policy makers can then act on, as Jonathan pointed out.
But there are challenges there. As complexity becomes distilled, there is the potential for misinterpretation. There’s also the potential that particular things are focused on in excess of at least what the research or the data suggest should be the focus. And I guess one of the intriguing things . . . I’ve been in this disparities business, or whatever, for about seven years now, and it is intriguing to me the sort of dichotomy that sort of belies the complexity of our health care system in our society between we have disparities because doctors are prejudiced and biased or facilities behave in prejudiced or biased or racist ways hence we have disparities, versus in the dichotomy, we don’t have disparities because of providers or physicians, we simply have them because of geography or distribution of patients or health care resources lining up with the distribution of races of patients. It is important to try and clarify the relative contribution of each of these different forces, but I don’t think research, at least as it’s currently constructed with the available data and the available measures and the issue of omitted variables, as Jonathan mentioned, can actually determine whether or not the effects are X percent this and Y percent that. And certainly no research can determine that it’s 100 percent of one phenomenon versus 100 percent of the other.
So let me walk you through this. Sally’s correct. I was one of the authors of a study in 1999 demonstrating that essentially the entire difference in lung cancer mortality between blacks and whites was due to differences in surgical rates between blacks and whites. Blacks and whites who are diagnosed with lung cancer mostly have advanced stage disease, both blacks and whites, which has a poor outcome. And that’s a bad thing obviously. But the fraction of blacks and whites who have early stage disease, which is basically curable through surgery, is also about the same. But we showed that there was a very large treatment gap between blacks and whites in terms of getting this curative surgery; and that treatment gap, if you ran the numbers, really did explain the difference in cancer mortality rates between blacks and whites with lung cancer.
And so some disparities may not induce outcome differences; some disparities may actually be harmful to whites because of over-use/over-treatment and poor selection of patients. But that was an example of a disparity we thought was important in terms of cancer mortality statistics and translated into thousands of unnecessary deaths a year. And so we thought that was an important jumping off point to ask why would this exist.
And, as Sally and Jonathan have already pointed out, a lot of theories have arisen that these disparities exist because of differential treatment of patients by particular physicians or by particular facilities. And we got interested in this question because, really because of some of the work by Amitabh and other people at Dartmouth asking, (Amitabh’s no longer at Dartmouth anymore, but he used to be until, like, six months ago, so I guess he’s a defector.) But, the research suggested what we already knew just from traveling around the country, looking at a map of the US census that the populations of blacks and whites were not distributed evenly and we knew that about cancer statistics as well. And so, we said, if we’re going to make practical policy initiatives, we have to measure some other stuff. We have to measure the distribution of patients and doctors and their races, so that we can start asking some of these very basic questions about, well, if we are going to intervene, where do we intervene and how do we intervene.
So let me talk a little bit about a study we published in 2004 where we asked some questions about primary care physicians and which patients they treated. And then, to follow up on that, what those physicians were like. So the aim of this study was to assess whether the widespread associations that have been observed between patient race and quality of care might in part be to patients of different races receive care mostly from different physicians who in turn differ in their ability to provide high quality care. We asked this question only for blacks and whites for purely methodological reasons. Most of the disparities literature focuses on blacks and whites and blacks and whites are the two racial groups that are easiest to identify and most reliably identified through claims data. When Medicare data or Social Security says “black” or “white,” it is quite likely that when you contact that individual, they self-identify the same way. That is not true for Hispanics, and it’s not really true for Asian Pacific-islanders and other groups.
So let me lay this out. Jonathan presented this as an omitted variable problem; I’m going to present this as a confounding variable problem. They really are the same idea. Lots and lots of literature, hundreds of articles have demonstrated an association between patient race (black/white) versus quality of care. In truth, there is a difference in treatment quality. You can measure it in cardiac disease or cancer; it doesn’t really matter. The question we had is where do physicians sit in this association between race and quality of care. We had this hypothesis that there was an association between patient race and physicians; to make this more technical, in other words, patient race was somehow predictive of which physicians patients saw--patients aren’t randomly distributed across physicians. This is in line with the comments already. But then we had this other hypothesis that those physicians were in turn associated with the quality of care that patients would receive. So that, to some extent, the mechanism between patient’s race, physicians, and quality of care was underlying health disparities.
So these are the questions we asked. You’ll see the little diagram is now up in the upper-right-hand corner. Arrow number one is this question: Is the care of black and white patients clustered among distinct groups of physicians? Geography of the United States in and of itself would suggest it is. Arrow number two: Are these groups of physicians in association with patient race different in their ability to provide high-quality care? And in our study we measured only their qualifications and their access to resources. Those are things we could measure from a survey. And then: Are these differences due to geographic availability of physicians?
We looked at a survey of practitioners as our core data set. They were respondents to round three of the Community Tracking Study, an ongoing longitudinal study, a survey study of practicing physicians. We limited our analyses to the primary care physicians who were respondents to the survey, who’ve been over-sampled are nationally representative of primary care physicians. And this survey includes self-reported measures of physician characteristics and resources I’ll get to. We then took those physicians and linked them to the patients they actually saw from the 5 percent Medicare claims data, for those of you who are into Medicare claims. But those are records or claims submitted by doctors seeing patients. And we limited our analyses, just to do a little Medicare speak here, to evaluation and management visits, which are the kind of doctor-patient visits you think about where you talk to the doctor about the problem, they examine you, and they write you usually 2.3 prescriptions. It’s the sort of visit Dr. House is supposed to do but he never shows up and does, for those of you who watch that show. Beneficiaries aged sixty-five or older, which are the bulk of our population, black or white patients, and the sample size is listed at the bottom of the slide.
We used logistic regression to look at these analyses, and I won’t dwell on this slide, but for those of you who are methodologically intrigued by this, the outcome of the variable of the analyses was actually the race of the patients of the dyads. We looked at these patient pairs, and we said, well, “If the physician has X characteristics, what’s the likelihood that a particular patient they see is black rather than white?”
And we looked at the variables listed on this slide, such as the physician characteristics I’ll talk about, the practice setting variables we controlled for. And then we weighted our results for national estimation at the patient visit level and sampling design and variance correction so that you could look at the numbers and actually say, “Okay, this is representative of the U.S.”--at least for Medicare patients
And then we did our geographic analyses. Once we had our results, we asked the simple question, “Well, okay, here’s a black patient’s visit with a particular doctor, and now we know something about that doctor’s characteristics. Does that doctor look like they are just like all the other doctors in that neighborhood? Or to put more technically, do they appear to be randomly sampled from the neighborhood where the patient got their care? Or is there some additional selection going on that we don’t understand?” And we looked at the Health Service Area, a relatively small geographic unit, and the Hospital Referral Region. And I think that Amitabh will probably touch a little on both of those metrics; but, if not, they’re simply geographic measures.
So, study number one: For those of you who like graphs, you’re really going to be thrilled. Are patients clustered amongst doctors, blacks and whites clustered amongst different doctors? This graph addresses this issue; it is scaled to be representative of the 84,000 primary care physicians who take care of indemnity Medicare patients. Along the X axis are displayed these primary care physicians, placed in the order of the ratio of black-to-white patients that they see. So the doctors furthest on the left . . . (I don’t know if this has a laser, it has something but I don’t know what it does, so I won’t turn it on.) The doctors all the way on the left take care of predominantly white patients, and as you move to the right, the populations of patients. (Oh thank you. It says “Danger.” Just don’t look at me.) As you move to the right, the proportion of blacks rises relative to whites. Each block, again for those of you who are interested, represents about 3,300 docs, with these docs seeing mostly white patients and the ones on the right seeing more black patients.
Well, what does this graph mean? Well, if you add up all the white patient visits accounted for by these doctors, you’ll see, this is a cumulative account of white patient visits across the doctors. By the time you get to all the doctors on the far hand side, you’ve of course accounted for all the white patients’ visits. But you’ll see that each chunk of doctors really accounts for about the same number of white patient visits. That’s why that line is basically straight. That’s not true for the black patients. The care of most black patients is clustered amongst these physicians who are on the right-hand side of the graph where the practices are predominately or where the proportion of blacks is higher. And you can see: it is not that doctors contribute equally to the care of black patients. To illustrate this just numerically, an arbitrary line drawn here gives you these kinds of useful statistics. If you take the first 80 percent of physicians who see not very many black patients, they account for about 78 percent of white patient visits and about one-fifth of black patient visits. The last fifth of doctors account for about a fifth of white patient visits and 80 percent of black patient visits.
Arrow 1: Yes, there is clustering of patients. Arrow 2: Do the characteristics of the physicians differ? Now listed here on the first line is the characteristic “male”--a physician characteristic. And you can read this table in the following way: the typical white patient visit is 86 percent of the time with a male physician; the typical black patient visit is 82 percent of the time with a male physician. Now the gender difference or sex difference between physicians probably isn’t very important in terms of aggregate health outcomes, and it’s also not statistically significant, but I put this here just so you can orient yourself to this table. The question then is: What is the likelihood that the physician is white? And the answer is, 85 percent of white patient visits are with white doctors; 60 percent of black patients. What is the likelihood that the doctor is black, in the next row? It’s less than 1 percent for a typical white Medicare beneficiary; it’s about 22 percent for a black Medicare beneficiary.
What is the likelihood they’re board certified? 86 percent, versus 77 percent—a statistically significant difference, a 9 percent difference. The clinical implications of this are uncertain; although, some research has suggested that board certification predicts things like adequate adherence to delivery of preventive services, follow-up care that’s diagnosis-oriented rather than symptom-oriented. And we’ve actually published a subsequent study linking board certification to the delivery of preventative services, in line with other research in this area.
What’s the mixture of charity care and Medicaid, and what’s the income in the area? And exactly what you would expect, you see. Black patients are more likely to go to doctors who are providing more free or charity care, which take a larger proportion of the practice from Medicaid, and practice in poorer neighborhoods—all things which contribute to the lower quality of care these doctors may be able to deliver.
Well, what do the physicians tell you about their access to resources? Black patient visits . . . This table can be read in exactly the same way, although here on the left-hand side are questions asked of the doctor. Are you able to provide high-quality care, in general? I am always or almost always able to access blank: specialists, or imaging or elective admission for my patients. And you can see for each of the metric, the likelihood that a black patient is seeing a doctor who is saying, “No, I can’t provide high-quality care” or “No, I cannot routinely provide access for these services to my patients” is higher (oh, pardon me, this is actually phrased in the positive). So, for black patient visits, it is less likely they are seeing a doctor who affirms that they have access to high-quality care or these various resources listed—all statistically significant, controlling for all the geographic and other factors.
And then lastly, a result I won’t dwell on, because Amitabh is going to talk about this in more detail. But, simply put, if you ask the question, “Is this the doctor the patient is seeing or is it the neighborhood where they are getting their care?” the answer in almost all cases is it’s the neighborhood. The doctor they’re seeing isn’t any different from all the other doctors practicing in the area in terms of their survey responses. The only exception is physician race. And the black patients and the white patients are more likely to select somebody of their race than would be expected if they just, sort of, randomly open the phone book and selected a doctor from that neighborhood. It’s higher than chance.
Important limitations of this study: first of all, each survey question, the doctors answered about all of their patients, not about their Medicare patient. I think it still has important implications, and the racial clustering of younger patients parallels that in the Medicare population. People have shown that through some other practice level data. But, we have to think about the potential impact on the Medicare population, where everyone has Medicare indemnity insurance versus these doctors telling us, “Well, I can’t always provide this access to care” because they have a larger percentage of uninsured black patients. And so they can’t access specialists for them--that may not affect actually their ability to deliver care to Medicare patients. Although, I think our follow-up study suggests that in fact there is a correlation there.
And then we adjusted for this thing about pay or mix, and socialized socio-economic status, and geographic differences in income, but those are inadequate markers of these differences in wealth and health care infrastructure. I can’t read the next limitation, so it’s probably not important. And then the last one is that no outcomes were observed, and outcomes are obviously what matter.
So, care for black and white Medicare patients is clustered among different groups of physicians. The visits for blacks are with physicians who are less likely to be board certified in their primary specialty, less likely to be able to access needed referral services and less likely to report being able to provide high-quality care. The patterns of the physician characteristics are mostly explained by local availability, not by selection of patients—the exception being the race of their physician.
And there are some possible implications. To some extent, racial disparities in health care may emerge from differences in the quality of care that different physicians are able to provide because the differences between the physicians that we’ve just shown on the previous few slides are correlated with the race of the patients that they see. And therefore, we may not need to invoke differential treatments of patients by the same doctor, for example, bias or racism to understand unequal treatment. And in fact, our findings suggested only a few doctors could really substantially influence disparities--only those doctors who treat both blacks and whites. I do want to point out, I want to make sure that the negative on this second bullet is clear. The fact that we have observed this differential distribution of patients that tracks with doctor qualifications does not rule out the possibility that these doctors in these practices still do treat blacks and whites differently. We haven’t measured whether or not that occurs or not. The only way that you could exclude that is if that graph I showed you, that distribution had 100 percent whites treated by 80 percent doctors and 100 percent blacks treated by the other 20 percent, or something like that—a full division or schism in the health care system. We don’t have that; in fact, our graph shows that there is in fact overlap, which may be important.
So here were the concerns that were raised for us. Blacks cannot necessarily find physicians with alternative characteristics locally; and so, we’re not happy that they see doctors who can’t provide access to specialists. That may affect things like access to cancer screening by a colonoscopy or maybe a diabetic eye exam. But those physicians, if they opened up the yellow pages aren’t in their neighborhoods. Therefore, interventions to improve the training and resources for those physicians who treat blacks should be considered as an approach to reducing health care disparities because these physicians can actually be identified through claims data, as we’ve just shown. So you could actually target your resources and your interventions at the doctors who treat blacks. That would probably be the logical way of improving the care of those blacks. But, I’d argue that further research is needed to make sure that the measures that we’re looking at and the correlations we’re looking at are in fact predictive of things we care about, like health outcomes and health services delivery. Thanks very much.
Sally Satel, MD
AEI
Thank you very much. Next is Linda Gottfredson, who is a professor of education at the University of Delaware in the honors program there. She is published extensively on the impact of general education, on personal functioning in different domains, including school, work, and most recently, I think you’ve added health to your portfolio. She’s on the editorial board of various journals, including the journal Intelligence, an area in which she is an expert. Thank you.
Linda Gottfredson
University of Delaware
Today I’m going to try to do three things: first, agree that there are more plausible explanations for health disparities than treatment bias; and then, I want to spend most of my time amplifying a point that Jonathan and Sally brought up in one of their chapters but haven’t talked about here today, which is patient-side factors. We’ve talked only about doctors and hospitals so far, but what about the patient’s contribution to care? And I’m going to focus on the patient factor of general reasoning ability, or more technically, you may recognize it as the g-factor or the general mental ability factor. And then I want to describe how this information can improve health and save lives.
We’ve seen a lot of the standard disparities model today, and Jonathan and Sally have a third variable’s model, in the sense that there are other things that have not been taken into account, primarily geography is what most of the speakers have focused on. And therefore, we cannot assume that differences mean discrimination. But, as I said, what about patient attributes, which they talk about to some extent.
Now the patient’s role is really primary. You may not realize it, but you are really your own primary health care provider in many ways. Scary thought, isn’t it? Lifetime of self care is key to good health. Preventing injury, preventing chronic disease, taking care of yourself after you’ve sustained an illness or injury and it’s a very complex job, as I’ll try to show, requiring lots of independent judgement.
**At this point in the conference, the recording encountered technical difficulties. What follows is a summary of the panelists’ remarks.**
Linda Gottfredson
University of Delaware
Although health disparities are largely attributed to differences in geography (a manifestation of wealth and social status), health scientists have noted that differences in an individual’s cognitive abilities may explain why some patients receive better care than others. This theory suggests that the variation in effective treatments may result from an inequality of reasoning capabilities among patients. Health literacy research and related studies have shown that an individual’s mental resources are essential for successfully exploiting available care.
Patients with lower general reasoning abilities are less likely to seek preventive care, to know signs and symptoms of disease, and to adhere to treatment regimens. Therefore, further investment in health care will lead to an expansion of inequality in health outcomes as a result of some individuals’ inability to take advantage of new treatments and resources. Health educators advocate that health materials be written at no higher than the fifth-grade reading level. Because this goal is unachievable for complex treatments, health care workers may find it beneficial to explain to patients more completely the procedures for effective treatment
A high level of intelligence is useful in all aspects of life; it is essential when tasks are novel, untutored, or complex; and when situations are unclear, fluctuating, or capricious. Because individuals will interpret treatment procedures differently, identical treatment will not yield the same results. Thus, equalizing access and quality of health care does not, and can never, close the health disparity gap. Instead, health professionals must improve their patients’ understanding of the treatment procedures and labels need to be clearer.
Christopher Foreman
University of Maryland
Claims of disparity among health conditions cast blame widely among citizens, their respective genetic endowments, medical providers, economic forces, and government policies. The claim of disparate quality of and access to health care narrows the blame considerably, often to a single factor: the biased or insensitive health care professional.
Jonathan Klick and Sally Satel’s monograph tries to determine what policies offer realistic traction against real problems. A purely racial story line may not be entirely appropriate where underlying or complicating economic, regional, or other forces are at work. The motivation for concentrating on racist health care seems to be a political reinvigoration of civil rights. Such a political strategy doubtless has its uses, but it also has limits and costs.
One striking issue in the monograph is chapter six’s implication that civil rights organizations ought to challenge the trial lawyers. Here we have the possible seeds of a new wedge issue dividing one element of the Democratic Party from another. A second issue is that Klick and Satel are not blaming the victim in their monograph. If society is to progress we absolutely must get beyond the impulse to interpret these kinds of arguments in that light.
AEI research assistant Jonathan Stricks prepared this transcript.
---
