Scan barcode
A review by politizer
Doing Harm: The Truth about How Bad Medicine and Lazy Science Leave Women Dismissed, Misdiagnosed, and Sick by Maya Dusenbery
3.0
The message of this book is a really important one and there is a lot of good it can do. However, the way Dusenbery argues for this message is often less convincing than it ought to be.
-- Labels vs. explanations --
One thing that crops up often here, and which is common in pop science writing, is conflating labels with explanations. It's surprising to see it here, because one of the central themes Dusenbery often reiterates here is that there are major gaps in our knowledge of medicine and there are many things that aren't explained yet; but, counterintuively, in other places she often overstates how well something is explained. Her discussion of chronic pain is an emblematic example. She begins by describing the modern model of what pain is, which recognizes four categories of pain (roughly, helpful/adaptive pain in response to a dangerous stimulus, helpful/adaptive pain in a part of the body that is healing, unhelpful/maladaptive pain caused by a problem in the nervous system, and unhelpful/maladaptive pain that is not explained by any observed injury or whatever). She contrasts this with an old model of pain which only has two categories: pain that's explained by an underlying medical condition, and "medically unexplained pain". She suggests that the new system is better and that chronic pain conditions fit into it, whereas pathological pain "just didn't fit into early models of what pain is". This argument makes no sense to me. Both the new and old models contain a category for "medically unexplained pain", just with different names; thus, it's not clear to me how the new model does any better at explaining or accommodating the kinds of chronic pain discussed here. There are other important differences between the new and old way of looking at things (e.g., nowadays there is a better [although still far from sufficient] recognition of the fact that "medically unexplained pain" is still real, and is not necessarily unexplain*able*, and that it might actually have a medical reason that juts has not been discovered yet), but these don't seem to have anything to do with the fact that the newer underestanding of pain includes more categories with different labels. This passage is a typical example of science writers assuming that putting a new name on some phenomenon is the same as providing an explanation of the phenomenon.
Another example: in discussing chronic "functional" (i.e., "medically unexplained") pain, Dusenbery cites evidence that people with these conditions show widespread hyperalgesia and allodynia compared to healthy controls" (hyperalgesia and allodynia are technical terms for feeling more pain than what one normally would) and fMRI scans of these people show BOLD activity in brain areas relating to pain. Dusenbery cites this evidence as if it vindicates that their pain is "medically real". But it's really just a bunch of fancy ways of saying "people with chronic pain feel pain". Which I suppose is relevant if you're arguing against the position that people are intentionally faking it to score drugs, but it's not relevant if you're arguing against the position that people are causing their own pain by being too stressed out or are overreacting to pain (and these latter two are the points she focuses on more anyway). There is plenty of other evidence that both those positions are wrong, but this is not a piece of that evidence; it's just one of these neuroscience "explanations" that looks convincing but actually doesn't tell you anything (see https://direct.mit.edu/jocn/article/20/3/470/4473/The-Seductive-Allure-of-Neuroscience-Explanations).
-- Questionable representation of statistical evidence --
Another issue is questionable ways of presenting data. For pretty much every claim made in the book, Dusenbery provides a ton of evidence; sometimes, though, evidence seems p-hacked or cherry-picked. For instance, in arguing that women take longer than men to get diagnosed for some condition, she will often raise evidence along the lines of "a survey found that 25% of women took over 5 years to get a diagnosis, compared to 13% of men who took over 5 years". This is a classic example of p-hacking, "garden of forking paths" type analysis (http://www.stat.columbia.edu/~gelman/research/unpublished/p_hacking.pdf). Sure more women than men took over 5 years to get a diagnosis, but how many took over 4 years? How many took over 6 years? Why the focus on an arbitrary cutoff like 5 years? (I'm not suggesting here that Dusenbery herself is p-hacking; it's more likely that she's just relaying the findings in whatever way the authors of those studies originally presented them.) There are two ways this could have been done better: one way would have been to show a graph of the whole data distribution (but that probably would not be feasible to fit into a normal-length book, given the huge number of studies that get cited, and maybe the presence of graphs would be a turn-off to many of her intended audience) and one would have been to just say "women take longer to get a diagnosis" and let the reader go read the original study if they want to see all the numbers. I guess I can see why Dusenbery instead chose to take a middle option -- presenting some numbers, but not all of them -- because of the sort of catch-22 women are in when they try to advocate in medicine (i.e., if she didn't bring numbers to back up her claims, she might just be dismissed out of hand). But, ironically, this middle option ends up actually resulting in a weaker argument than what either of the other options would have done.
Relatedly, sometimes there are claims which are supported by statistics that maybe look compelling at first but which, upon the slightest scrutiny, turn out to be irrelevant or tautological. For example, in the chapter on autoimmune disorders, Dusenbery frequently brings up findings showing that people who are misdiagnosed the first time they visit a doctor take longer to get a correct diagnosis, compared to people who are correctly diagnosed the first time. This is almost tautological (it's basically like saying "I always find my keys in the last place I look") and I don't see what it's supposed to convince me of. There are other statistics that are more important and meaningful -- i.e., the fact that this diagnostic delay is bigger in women than in men -- but the fact getting something wrong and then getting it right later takes longer than getting it right in the first time does not seem particularly enlightening.
It feels like I am nitpicking with the above issue, but it's very important. My experience when reading the book was that whenever I encountered an example like this, I still felt like I was convinced by the overall argument she was making -- after all, even if the way she summarized one particular study is problematic, she's also summarized 10 other studies before and after it that also support the same conclusion, and overall the whole weight of all the evidence discussed on all these topics is all consistent with the claim being made. But the problem, of course, is that the "overall weight of evidence" is also depending on accepting the conclusions she makes based on each individual finding, and if you start to scrutinize all of them they might not all hold up. It's ultimately disappointing, because I do feel like I believe the overall argument she is making, so I would have liked to see it argued better.
-- Devalorizing mental health --
Last but not least, I can't shake the feeling that the book often bashes mental health conditions / psychosocial illnesses. The discussions of many medical issues in this book take a form that can roughly be summarized as, "Throughout history doctors treated women as if this condition was all in their head. But it's actually a REAL medical condition". I totally understand the inclination to give evidence that these conditions are "real", given that women are frequently dismissed and denied treatment for these conditions and this has resulted in huge harm. But I don't think devalorizing mental health is a good way to go about that work. Mental health conditions / psychosocial illnesses are also "real" conditions, and deserve to be treated as such; indeed, many people writing within disability studies and mental health advocacy question the very notion of a divide between "mental" and "physical" illnesses. I don't think the message of a book like this should be "women shouldn't be dismissed and not-taken-seriously because often they have REAL medical conditions"; it should be "women shouldn't be dismissed and not-taken-seriously at all, regardless of whether their condition is 'physical' or 'mental'/psychosocial". In other words: the historical (and still currently widespread) practice of dismissing women's symptoms as "psychogenic" is bad, but it's not bad because "medical" conditions are somehow more deserving of treatment than "psychogenic" conditions; it's bad because "psychogenic" conditions don't get the treatment they deserve. The solution, therefore, is not to just throw mental health under the bus and give out more medical diagnoses instead, but to treat mental health conditions with the seriousness they deserve, rather than treating them as a "wastebasket" for patients that doctors don't want to deal with. (Of course, when people do actually have a non-MH condition they should be diagnosed as such, but the reason should be because *everyone* deserves to get the right diagnosis and get the treatment they need, not because "mental" conditions are less serious or less real than "physical" conditions.)
***
Ultimately, there is still much that is very good about this book. Not all the arguments are characterized by the problems outlined above; many of Dusenbery's points are compelling and convincingly argued. And throughout the book she often trenchantly and appropriately eviscerates the shoddy arguments the medical community has used to justify the mistreatment of women patients. And the fact that science and medicine are fallible and should be questioned is one that I hope everyone gets (while this book was written in 2018, I'm reading it in the covid "trust science" era, so this message is extra urgent now). In fact, several months on, I find myself frequently recommending this book to people. So in the end, the message she's putting out there is a message everyone needs to hear. It's just a shame that the book itself doesn't always do justice to this message.
-- Labels vs. explanations --
One thing that crops up often here, and which is common in pop science writing, is conflating labels with explanations. It's surprising to see it here, because one of the central themes Dusenbery often reiterates here is that there are major gaps in our knowledge of medicine and there are many things that aren't explained yet; but, counterintuively, in other places she often overstates how well something is explained. Her discussion of chronic pain is an emblematic example. She begins by describing the modern model of what pain is, which recognizes four categories of pain (roughly, helpful/adaptive pain in response to a dangerous stimulus, helpful/adaptive pain in a part of the body that is healing, unhelpful/maladaptive pain caused by a problem in the nervous system, and unhelpful/maladaptive pain that is not explained by any observed injury or whatever). She contrasts this with an old model of pain which only has two categories: pain that's explained by an underlying medical condition, and "medically unexplained pain". She suggests that the new system is better and that chronic pain conditions fit into it, whereas pathological pain "just didn't fit into early models of what pain is". This argument makes no sense to me. Both the new and old models contain a category for "medically unexplained pain", just with different names; thus, it's not clear to me how the new model does any better at explaining or accommodating the kinds of chronic pain discussed here. There are other important differences between the new and old way of looking at things (e.g., nowadays there is a better [although still far from sufficient] recognition of the fact that "medically unexplained pain" is still real, and is not necessarily unexplain*able*, and that it might actually have a medical reason that juts has not been discovered yet), but these don't seem to have anything to do with the fact that the newer underestanding of pain includes more categories with different labels. This passage is a typical example of science writers assuming that putting a new name on some phenomenon is the same as providing an explanation of the phenomenon.
Another example: in discussing chronic "functional" (i.e., "medically unexplained") pain, Dusenbery cites evidence that people with these conditions show widespread hyperalgesia and allodynia compared to healthy controls" (hyperalgesia and allodynia are technical terms for feeling more pain than what one normally would) and fMRI scans of these people show BOLD activity in brain areas relating to pain. Dusenbery cites this evidence as if it vindicates that their pain is "medically real". But it's really just a bunch of fancy ways of saying "people with chronic pain feel pain". Which I suppose is relevant if you're arguing against the position that people are intentionally faking it to score drugs, but it's not relevant if you're arguing against the position that people are causing their own pain by being too stressed out or are overreacting to pain (and these latter two are the points she focuses on more anyway). There is plenty of other evidence that both those positions are wrong, but this is not a piece of that evidence; it's just one of these neuroscience "explanations" that looks convincing but actually doesn't tell you anything (see https://direct.mit.edu/jocn/article/20/3/470/4473/The-Seductive-Allure-of-Neuroscience-Explanations).
-- Questionable representation of statistical evidence --
Another issue is questionable ways of presenting data. For pretty much every claim made in the book, Dusenbery provides a ton of evidence; sometimes, though, evidence seems p-hacked or cherry-picked. For instance, in arguing that women take longer than men to get diagnosed for some condition, she will often raise evidence along the lines of "a survey found that 25% of women took over 5 years to get a diagnosis, compared to 13% of men who took over 5 years". This is a classic example of p-hacking, "garden of forking paths" type analysis (http://www.stat.columbia.edu/~gelman/research/unpublished/p_hacking.pdf). Sure more women than men took over 5 years to get a diagnosis, but how many took over 4 years? How many took over 6 years? Why the focus on an arbitrary cutoff like 5 years? (I'm not suggesting here that Dusenbery herself is p-hacking; it's more likely that she's just relaying the findings in whatever way the authors of those studies originally presented them.) There are two ways this could have been done better: one way would have been to show a graph of the whole data distribution (but that probably would not be feasible to fit into a normal-length book, given the huge number of studies that get cited, and maybe the presence of graphs would be a turn-off to many of her intended audience) and one would have been to just say "women take longer to get a diagnosis" and let the reader go read the original study if they want to see all the numbers. I guess I can see why Dusenbery instead chose to take a middle option -- presenting some numbers, but not all of them -- because of the sort of catch-22 women are in when they try to advocate in medicine (i.e., if she didn't bring numbers to back up her claims, she might just be dismissed out of hand). But, ironically, this middle option ends up actually resulting in a weaker argument than what either of the other options would have done.
Relatedly, sometimes there are claims which are supported by statistics that maybe look compelling at first but which, upon the slightest scrutiny, turn out to be irrelevant or tautological. For example, in the chapter on autoimmune disorders, Dusenbery frequently brings up findings showing that people who are misdiagnosed the first time they visit a doctor take longer to get a correct diagnosis, compared to people who are correctly diagnosed the first time. This is almost tautological (it's basically like saying "I always find my keys in the last place I look") and I don't see what it's supposed to convince me of. There are other statistics that are more important and meaningful -- i.e., the fact that this diagnostic delay is bigger in women than in men -- but the fact getting something wrong and then getting it right later takes longer than getting it right in the first time does not seem particularly enlightening.
It feels like I am nitpicking with the above issue, but it's very important. My experience when reading the book was that whenever I encountered an example like this, I still felt like I was convinced by the overall argument she was making -- after all, even if the way she summarized one particular study is problematic, she's also summarized 10 other studies before and after it that also support the same conclusion, and overall the whole weight of all the evidence discussed on all these topics is all consistent with the claim being made. But the problem, of course, is that the "overall weight of evidence" is also depending on accepting the conclusions she makes based on each individual finding, and if you start to scrutinize all of them they might not all hold up. It's ultimately disappointing, because I do feel like I believe the overall argument she is making, so I would have liked to see it argued better.
-- Devalorizing mental health --
Last but not least, I can't shake the feeling that the book often bashes mental health conditions / psychosocial illnesses. The discussions of many medical issues in this book take a form that can roughly be summarized as, "Throughout history doctors treated women as if this condition was all in their head. But it's actually a REAL medical condition". I totally understand the inclination to give evidence that these conditions are "real", given that women are frequently dismissed and denied treatment for these conditions and this has resulted in huge harm. But I don't think devalorizing mental health is a good way to go about that work. Mental health conditions / psychosocial illnesses are also "real" conditions, and deserve to be treated as such; indeed, many people writing within disability studies and mental health advocacy question the very notion of a divide between "mental" and "physical" illnesses. I don't think the message of a book like this should be "women shouldn't be dismissed and not-taken-seriously because often they have REAL medical conditions"; it should be "women shouldn't be dismissed and not-taken-seriously at all, regardless of whether their condition is 'physical' or 'mental'/psychosocial". In other words: the historical (and still currently widespread) practice of dismissing women's symptoms as "psychogenic" is bad, but it's not bad because "medical" conditions are somehow more deserving of treatment than "psychogenic" conditions; it's bad because "psychogenic" conditions don't get the treatment they deserve. The solution, therefore, is not to just throw mental health under the bus and give out more medical diagnoses instead, but to treat mental health conditions with the seriousness they deserve, rather than treating them as a "wastebasket" for patients that doctors don't want to deal with. (Of course, when people do actually have a non-MH condition they should be diagnosed as such, but the reason should be because *everyone* deserves to get the right diagnosis and get the treatment they need, not because "mental" conditions are less serious or less real than "physical" conditions.)
***
Ultimately, there is still much that is very good about this book. Not all the arguments are characterized by the problems outlined above; many of Dusenbery's points are compelling and convincingly argued. And throughout the book she often trenchantly and appropriately eviscerates the shoddy arguments the medical community has used to justify the mistreatment of women patients. And the fact that science and medicine are fallible and should be questioned is one that I hope everyone gets (while this book was written in 2018, I'm reading it in the covid "trust science" era, so this message is extra urgent now). In fact, several months on, I find myself frequently recommending this book to people. So in the end, the message she's putting out there is a message everyone needs to hear. It's just a shame that the book itself doesn't always do justice to this message.