“Many of us take for granted that we live in an age of medicine where, to put it quite simply, we know what we are doing,” writes Steven Hatch in his new book Snowball in a Blizzard: A Physician’s Notes on Uncertainty in Medicine.
Doctors’ tools, knowledge, and treatments have improved since the bloodletting days, and we now have the ability to scan and analyze the body down to the cellular level. But “precision is not the same thing as certainty,” Hatch writes, and often, doctors are just making guesses based on the best evidence they have—a measuring of risks and benefits and probabilities that can be easily influenced by their preconceptions.
Medicine is a high-stakes game of uncertainty, complicated by the fact that people are naturally predisposed to seek certainty whenever possible. If you don’t know what something is, it could be a threat, out there on the ancient savannah of evolutionary psychology logic. That goes for patients and doctors alike, and if both parties are in agreement that certainty is best, it’s possible that they’ll just blow past the risks of a treatment, or the dubiousness of a diagnosis, for the sake of having an answer.
I spoke with Hatch, an assistant professor of medicine at the University of Massachusetts Medical School, about how understanding and acknowledging the uncertainty in even the best available tests and treatments could improve the practice of medicine. Below is a lightly edited and condensed transcript of our conversation.
Julie Beck: There seems to be a real disconnect between the way the public (and doctors sometimes, too) think about medicine—as a field that gives you answers—and what the field really is, which seems to be more just trying to minimize uncertainty as best we can.
Steven Hatch: I think one of the reasons why we have this issue in medicine is: To become a doctor you go through this weeding-out process where you go to your chemistry classes and biology classes and take the MCATs. All of those, for the most part, are situations in which the person who gets the most right answers is rewarded. By the time you get to med school, you’re already primed to think that everything is about a right answer. Then what happens when you get into the practice of medicine is, it’s a lot of fuzzy variables.
Beck: Does the Socratic method, and the focus on having the answers ready at hand whenever you're asked, instill that in some ways?
Hatch: Yeah, I think that’s right. In med school that gets primed, especially with the Socratic method and this old term we have called pimping.
Beck: What's that? Literally pimping, like P-I-M-P?
Hatch: Yeah. I have no idea where the term comes from, but to pimp a student is basically to ask them some arcane medical question, typically as a gentle hazing process. You’ll say “Dr. Smith, can you tell us the three cardinal features of pancreatic cancer?” and the student is expected to produce those answers. And it’s always a right answer, as opposed to “What do you think’s going on?”, which are more of these open-ended questions. By the time you get to residency, you’ve already got that downloaded into your mindset, and it takes a couple years to realize that you're doing something else. And some doctors, I think, never really completely become aware of it.
That’s where doctors find themselves in trouble. They don't even realize that, for instance, guidelines and recommendations are actually the synthesis of a lot of studies that are kind of fuzzy. But they just want to know the punchline.
Beck: I think that's an attitude patients have, too, when they're looking for care: “Something's wrong, I want an answer.”
Hatch: I definitely think the problem is bilateral. The culture of medicine prizes it, and part of the reason why the culture of medicine prizes it is people come to expect it in us. Sometimes it's very jarring for patients to hear, “Well, I’m not completely sure what’s going on.” Sometimes patients can become very upset with their doctors. I can tell you I've been on the receiving end of this. Sometimes there really is a problem that nobody has put together in the right way, and sometimes it’s because nobody knows. In the book, I have a chapter that deals with chronic fatigue syndrome, which is really an unknown unknown. We really don't understand why these people come down with this horrifying debilitating illness.
Beck: When people are seeking out these alternative medicines or diagnoses, with people who think their chronic fatigue is caused by Lyme disease, or anti-vaxers, things like that, how much of that comes from the fact that science isn't able to offer them much certainty at all, whereas the false explanations can?
Hatch: I think that is the central explanation for their appeal. What they offer that we don’t is the comfort of certainty. And I wish I could say that the story ends there. But I think the other thing that they offer that mainstream medicine often fails to deliver in those situations is sympathy, understanding. When I see patients in the Lyme-disease clinic who are these chronic-fatigue patients, the first message I give isn’t about Lyme disease, it’s, “I believe you and I hear you.”
The other thing I often try to tell my patients is, “I don’t think you’re crazy.” Because a lot of times, they go to doctors and the doctors don't know what's going on, and they eventually get the message, either literally or nonverbally: “Not only do they not believe me, but they think I’m just totally insane.”
Beck: In the context of testing, we end up with a lot more false positives than false negatives, and the scale skews toward overdiagnosis rather than underdiagnosis. Why is that? Why do we run in that direction?
Hatch: Some of it is, we're really hardwired to react more to positive tests. There’s this sense that doing more is better. It seems logical. What we’ve found just by looking at data and doing studies, is that more doesn't always equal better and more can often lead to harmful outcomes. It’s an assumption shared by both patients and doctors that you should always do more testing because it shows you that you’re doing something. One of the messages I was hoping to get out in the book is that you can flip the old statement on its head. Don’t just do something, stand there. Let's slow down and let's think about what the implications of the testing that we're doing are. The more tests you do, the more likely you are to find some result and then feel compelled to act on that result.
Beck: I think that's been starting to switch a lot recently, especially in the realm of women's health. It seems like the recommendation is don't do anything anymore. Fewer mammograms, fewer pap smears, don't do self-breast exams, all these things. Can we talk a little more about this famous example of mammography? How did a misunderstanding of uncertainty lead to too much screening and the overdiagnosis of breast cancer?
Hatch: Boy, that's a big question. The first trial started in the mid-‘60s. By the mid ‘70s they’d accumulated data, and what the data showed was that there was about a 33 percent, one-third reduction in the number of deaths from breast cancer [with screening mammograms]. And you say, “Wow, one-third reduction is huge!” In reality what the reduction was, was there were 128 deaths in the group that didn’t get mammograms, and 91 in the women who did, but the overall number of each group was 31,000. What you realize is, you’re squinting. It turns out the number can be portrayed in two different ways. One is the relative reduction, which is 128 versus 91, that turns out to be 33 percent. Or the absolute reduction which is 91 over 31,000 versus 128 over 31,000, and that comes out to be one-tenth of 1 percent. So you actually have a very, very small absolute benefit to mammograms. Neither number is lying. They’re both accurate descriptions. But obviously, they have very different emotional impacts.
Beck: When they eventually revised the guidelines to recommend that younger women don’t get annual mammograms, there was a very emotional response. Do you think uncertainty played a role in why people got so upset?
Hatch: Let me first say, I do want to be careful as a man talking about a very emotionally hot topic for women. I don’t want to presume to lecture women about this particular issue, but I do want the public to see the numbers. With that caveat in mind, I think there’s a psychological effect about false positives. So I'm just going to shift gears for one second. Back in the ‘90s, when people were using [the prostate-specific antigen test, or PSA] to screen for prostate cancer, if you had a PSA that was elevated and you went in for a biopsy, and the biopsy showed you had prostate cancer, and you went back and got a prostatectomy and you’re living five to 10 years later, it’s an entirely reasonable assumption to think that your life was saved because of the PSA. That just makes sense.
But when you actually do the population studies, up to 90 men in 1,000 are getting diagnosed with prostate cancer, and are either getting radiation or chemotherapy or prostatectomy, which are very serious interventions. And of those 90, 70 would have gotten prostate cancer that would have come to attention eventually. So that means there are 20 men [who didn’t have cancer] walking around who have an absolute belief that the PSA screen saved their life. So it's a weird psychological effect. It's very hard to convince people that if they didn’t get screened they’d still be alive anyway because they never would've gotten diagnosed with a cancer that they didn't have.
So [with mammograms] people become much more confident that this is a lifesaving tool, because at some level we want to believe we’re doing best by women. Particularly because medicine hasn’t done so great by women in the past. So when mammograms became associated with women taking control of their bodies, I think it made it a little more difficult to appraise the data. I want to be really careful about saying this in a way that does not come across as condescending or misogynistic, it’s just that we load our data interpretation with our desires for what the technology can do for us.
Beck: Can you explain the concept of predictive value, and how even accurate tests can be wrong if you don't use them the right way or on the right people?
Hatch: Just [as an example] to keep things simple: If you have a disease where the incidence is one per 100, and you have a 95-percent accurate test, if you do the test on 100 people that means, of the 99 people who don’t have the disease, you’re going to pick up 5 people and indicate that they have the disease. It's going to be wrong 5 percent of the time. Then the 1 person in the 100 who does have the disease, odds are very high that you're going to pick up that disease. So now you have six people who think they have a disease, of whom one actually has the disease. That is, at its core, the problem with false positivity.
The less common a disease is in populations, no matter how good the test is, the more likely you are to get false positives. That really is why the U.S. Preventative Services Task Force recommended in 2009 that women under 50 shouldn’t get mammograms. It was really that principle. They were looking at the numbers.
Beck: I do want to talk about guidelines and recommendations and what the best way to make those is. Did you see the CDC guidelines that came out recently about women and drinking and birth control?
Hatch: I've been buried in trying to stay up with Zika, so I quickly saw it, but I didn't look at it in depth.
Beck: Basically what they recommended—or what the language that they used made it seem like they recommended—was that women should not drink unless they are on birth control, because they might maybe be pregnant and they might maybe harm the baby. As you might imagine, there was quite an outcry about that. And I've thought about it a lot since it happened, and come to the idea that what they were trying to do was be like, “Yes, this is a preventable thing. There are these risks, there are these benefits; if nobody drank alcohol when they could possibly be pregnant, we would get rid of fetal alcohol syndrome.” And that would be the benefit. But the way it was presented was as, “This is what you should do,” and this is what got people upset.
It's interesting because you wrote in the book that "guidelines … could be understood as policy statements of relative benefits versus relative risks," rather than saying, "Do this, don't do that."And because that was in my mind, I was wondering what, in your opinion, would have been a good way to go about writing those recommendations while embracing the uncertainty? Both the research on drinking and pregnancy—which is very uncertain—and the uncertainty as to like, people aren't going to stop drinking.
Hatch: I think it’s a great question. It gets at some things that go toward my philosophy of medicine that I try to teach my students all the time. We as physicians should really avoid the word “should,” because it implies a moral judgment. What we should do is try to provide the best information that we can on what the evidence shows, and let intelligent patients make their own decisions. That comes down to whether they should take an antibiotic if they’re sick, whether they drink alcohol, whether they smoke cigarettes. Obviously, you won’t find any card-carrying physician who doesn’t think that it’s best for patients not to smoke, but you don’t get very far if you say, “You shouldn’t smoke.” I can cheer you on, if you want to try to quit smoking. But I'm not going to sit here and lecture you. I’m not going to be your moral scold.
And then when you get something like drinking alcohol and the risks that it might carry for the fetus, nobody’s going to do a double-blind randomized controlled trial on women who have one drink during their pregnancies per week versus women who have don't drink at all. That would be an unethical trial. So all the data that comes in on the relationship between alcohol and fetal defects is all fuzzy data. So for any policymaking body to say, or to even imply, “You shouldn’t do this”—that’s where we start getting into the very kinds of problems where people start reacting negatively toward medical organizations. They resent being told stuff like that, and I think they rightly resent it. We should be in the business of laying out what the evidence is about a given topic. It’s either strong, moderate, or weak.
I think part of getting away from the simple yes-or-no answers, both for patients and physicians, is to try to get to this model of what constitutes a strong recommendation, what constitutes a moderate recommendation, and what constitutes a weak recommendation. And what constitutes, like, “Psh, I don’t know.”
Beck: You talked a lot about what doctors should do to manage uncertainty, but what should patients do to manage uncertainty as they're talking with their doctors?
Hatch: Both in terms of when they talk with doctors and when they read media, they should really look at strength of evidence. And if they can’t figure out, either from talking to their doctor or from reading a news story, what the strength of evidence is, then they should just discount it.
We want to hear what you think about this article. Submit a letter to the editor or write to firstname.lastname@example.org.