Doctors Know They Prescribe Too Many Antibiotics. Why Don’t They Stop?

In 1928, the Scottish biologist Alexander Fleming made an accidental discovery. It turned out to be one of the most revolutionary medical breakthroughs in history, adding around 20 years to the average human life expectancy across the globe. It may have even saved your life.

Fleming discovered Penicillin. In the early 1940s, it became the world’s first mass-produced antibiotic.

For all of human history before then — even in an industrial nation like the United States — a simple cut could be a death sentence if it got infected. Diseases like pneumonia also killed, often. There simply was no cure. Until antibiotics came along.

OLD T.V. CLIPS: Industrial monument to the miracle drug. These invaluable products of microorganisms have been responsible for saving millions of human lives. Pioneers in a struggle which promises now to drive back that ancient enemy, disease.

If only the story of antibiotics ended on that high note. Because of evolution, the bacteria that cause disease are a moving target. When confronted with a threat — in this case antibiotics — they evolve to resist these drugs. And when we overuse antibiotics, we encourage that resistance. The more humanity has leaned on antibiotics in our war against disease, the less effective they have become. So, just how much do physicians lean on these drugs?

Jeff LINDER: Studies that I and others have done show that at least 30 percent, if not up to about 50 percent of antibiotics prescribed in the United States in outpatient practice are either for not a good reason or for no discernible reason whatsoever.

That’s Dr. Jeff Linder. He’s a leading researcher on antibiotic prescribing practices. Which, as his work suggests, are not so good in this country.

Health care providers have known for a long time about the risks of overusing antibiotics. The C.D.C. even publishes guidelines for doctors to follow, detailing which diagnoses do and don’t indicate that antibiotics would be appropriate.

LINDER: This is not a knowledge problem.

So, what kind of a problem is it?

From the Freakonomics Radio Network, this is Freakonomics, M.D. I’m Bapu Jena. Today on the show: why do doctors still prescribe so many unnecessary antibiotics? How can we get them to stop? And — what if we don’t?

LINDER: I’m Jeff Linder. I’m the Chief of General Internal Medicine at Northwestern University Feinberg School of Medicine here in Chicago, and I have a particular interest in how we can improve antibiotic prescribing.

JENA: Let me just actually take a step back here. What are antibiotics and what are they supposed to be used for?

LINDER: These are drugs that work against bacterial infections. The big ones: pneumonias, urinary tract infections. And then life-threatening things: sepsis, meningitis — that is what they’re for. The big problem is when somebody comes in with a diagnosis of a cold or acute bronchitis or the flu, and they get an antibiotic for it.

We’ve all been there. You or your child don’t feel well, it’s been a few days, nothing seems to help. The doctor says, “Let’s try an antibiotic,” and you agree. What could go wrong? Well, a lot. To start, antibiotics are meant to treat bacterial infections, as Jeff said. Not viruses, like a cold or the flu. Doctors know this, and so do a lot of patients. These medications are so widely used that we might perceive them as being benign, but they have real, and risky, side effects.

LINDER: People are putting a chemical in their body for something where it has no chance of helping and it can hurt them. You can get adverse reactions like rashes. Women can get yeast infections. People very often get diarrhea. And then you can also get more serious complications. There’s a bacterium in our guts called Clostridioides difficile or C. Diff. Somewhat paradoxically, taking an antibiotic kind of kills off all the good bacteria and the C. Diff overgrows, and that can actually be a life-threatening infection caused by antibiotics.

So that’s how using antibiotics when they’re not needed can harm the patients who take them. But at scale, the unnecessary overuse of antibiotics hurts everyone for the reason I talked about earlier: antibiotic resistance.

The bacteria that antibiotics were designed to kill are becoming resistant to the drugs, and quickly — precisely because we use them too much. This makes infections increasingly difficult — or in some cases impossible — to treat. The World Health Organization has declared antimicrobial resistance “one of the top 10 global public health threats facing humanity” today. Think of it this way: however much the discovery of antibiotics changed health outcomes for the better, they’ll change for the worse if antibiotic resistance persists.

This is a global issue, but in the U.S. alone, more than 2.8 million antibiotic-resistant infections occur each year, and more than 35,000 people die as a result. And there are other costs. A 2013 report from the C.D.C. estimated that the burden of antibiotic resistance on the U.S. economy could be as high as $55 billion a year, with $20 billion in direct healthcare costs and another $35 billion for loss of productivity.

This is what we call a tragedy of the commons: one doctor’s decision to prescribe an antibiotic that isn’t necessary doesn’t create a global public health crisis. But when millions of doctors make these treatment decisions for many more millions of patients — well, we have a big problem. A lot of Jeff Linder’s work has looked at how to change these decisions, one patient interaction at a time. So, why do doctors overprescribe?

LINDER: It’s for a whole host of reasons, but the main driving one we always hear from doctors when we talk to them about why they’re prescribing antibiotics to patients is they think patients expect them. This is not a knowledge problem. And so, I was not having success in discouraging doctors from using inappropriate antibiotics until I started actually working with behavioral economists and social psychologists, recognizing the contextual and psychological and social and emotional aspects of antibiotic prescribing that lead doctors to inappropriately prescribe antibiotics. All of the reasons to prescribe an antibiotic are right there in the room and emotionally salient and all the reasons to not prescribe an antibiotic, so things like guidelines, an adverse event later — resistance — happen after the visit and often we don’t find out about it. We very much remember the rare patient who gets really angry with us when we didn’t prescribe them antibiotic, and that is an uncomfortable situation for us.

JENA: In those settings where patients want antibiotics, or the doctor perceives that they want antibiotics, where does that come from? Because I think patients are probably aware of this idea of microbial resistance. They’re probably aware that medications can cause harm. So, what is it that leads them to request antibiotics in settings where they’re being told it’s probably not going to help you?

LINDER: There absolutely are a minority of patients who come in and explicitly want antibiotics, but it is the minority. Part of the problem is us assuming the patient wants an antibiotic. So, it turns out, we’re terrible at discerning which patients want antibiotics. And not surprisingly, by far the strongest predictor of inappropriate antibiotic prescribing is the doctor’s perception that the patient wants an antibiotic, not the patient’s actual expectation. We often just don’t ask patients whether or not they want antibiotics. We over generalize those uncomfortable interactions with a very demanding patient to all patients — you know, I don’t remember the hundreds and hundreds of patients I’ve had who were perfectly fine with an explanation that what they had was viral, they don’t need an antibiotic, and here are some things you can do to feel better, versus a patient I actually wrote a paper about, a lawyer I saw, who got very upset and wrote a very nasty letter. You know, I can recall that interaction pretty easily.

JENA: This reminds me of this study that I was part of with one of our colleagues, Ateev Mehrotra, where we looked at patients who had gone to urgent care facilities and some of these patients, by chance, had seen an urgent care provider who just had a high rate of antibiotic prescribing. And so, not surprisingly, they’re more likely to walk out with an antibiotic for an upper respiratory infection. Now, the interesting thing is fast forward a year later, those patients who saw a doctor who was a high prescriber are more likely to then go back to seek care for an upper respiratory infection in the future. It’s sort of this confirmation bias that a patient goes to see a doctor, they happen to walk out with an antibiotic. They get better, and when they look back, they might attribute their improvement to the antibiotic, whereas they would’ve just gotten better anyway. And so it creates this sort of self-fulfilling prophecy. 

LINDER: Absolutely. And it’s important to know that acute respiratory infections make up about 10 percent of all of the ambulatory visits in the United States in a given year. If somebody has a cold and they come in to see the doctor, that is time that the patient could have spent doing something else and time that the doctor and the entire healthcare system could have been focused on doing something that was productive for patients.

JENA: What do we know, if anything, about how overprescribing has a differential burden on specific populations?

LINDER: The two areas I would say this comes up most often is we usually see white patients get more inappropriate antibiotics than Black patients, and that’ll sound very reminiscent of the opioid epidemic — sort of this idea of our white patients just getting more care, even where it’s unnecessary. And then the other area that comes up over and over again is that antibiotics on a population basis are more prescribed in the south of the United States as compared to the West, Midwest and East — just higher than everywhere else in the United States.

JENA: Is there any theory as to why that is?

LINDER: The theory is that it’s cultural, that that’s just the norm that you get an antibiotic for respiratory symptoms and that that perception and the practice is more prevalent in the south of the United States than elsewhere.

JENA: Jeff, I grew up in Richmond, Virginia. I had antibiotics for breakfast.

LINDER: Yeah. I can tell from your accent that you’re a high antibiotic user.

These cultural differences in antibiotic prescribing practices, even within the U.S., suggest that overprescribing could largely be a social and emotional issue. So, how do we move the needle?

LINDER: It turns out that we’re powerfully influenced by what we think others think of us.

That’s after the break. I’m Bapu Jena, and this is Freakonomics M.D.

*          *          *

Before the break, my guest Dr. Jeff Linder explained why doctors prescribe so many unnecessary antibiotics, despite knowing the costs — things like rare but dangerous side effects and growing antibiotic resistance. Jeff has done a lot of research testing different ways to reduce inappropriate antibiotic use. Three approaches have stood out.

LINDER: The first one is this pre-commitment poster. So, recall I said overwhelmingly, a reason why doctors prescribe antibiotics is that they think the patient wants antibiotics. So, to try to short circuit that dynamic, we developed a poster that has a message like, “I am dedicated to only prescribing antibiotic when you really need it.” Or: “I’m dedicated to only prescribing antibiotic when it’s more likely to help than hurt you.” And then ideally, we would have a physician sign this, put their picture on it to sort of personalize it, blow it up to poster size and put it up in the practice — the idea being that the patient sees this commitment poster, sees that the doctor is committed to only prescribing antibiotics when they think it is more likely to help than hurt. More importantly, the doctor knows the patient knows. And hopefully that short circuits the assumption that the doctor has that the patient wants an antibiotic. And we found that putting up those posters led to about a 20 percent reduction in inappropriate antibiotic prescribing.

JENA: That’s really large. Has that been adopted into guidelines? Are other practices doing that now?

LINDER: This was sort of one of the more gratifying things in my career. So, this actually winds up in the C.D.C.’s core elements for antibiotic stewardship. So, if you are a practice or a practice manager or a clinician and want a version of the commitment poster, look on the C.D.C. website.

JENA: Okay, so that’s the first intervention. What’s the second one?

LINDER: Let me talk about accountable justifications. So, this one is based in the electronic health record where, if a doctor goes to prescribe an inappropriate antibiotic, they would get a warning saying, “Hey Doc, antibiotics are not generally indicated for, say, a cold. You can go ahead and prescribe that if you want, but before you do, we want you to write a tweet-length justification for it that we’re going to put into the E.H.R. as an antibiotic justification note. And if you don’t, we’re going to put in a note that says no justification given.” So, a couple things are going on there. First, there’s a little bit of a pause to force the doctor to think, “Okay, well, somebody thinks I shouldn’t be doing this.” But then we’re kind of socializing their decision. We’re telling them we’re going to tell their colleagues or make available to their colleagues that they may be prescribing antibiotic for inappropriate reasons. That resulted in an inappropriate antibiotic prescribing rate justification from about 23 percent to 5 percent.

JENA: That’s a huge effect. It’s amazing.

LINDER: If you think about the social mechanism going on there, by and large, a lot of what we do in practice is invisible to our colleagues, and we know it’s invisible to our colleagues, but it turns out that we’re powerfully influenced by what we think others think of us.

JENA: Has that accountable justification been used in other areas where there’s concerns about doctors perhaps inappropriately prescribing medications? I’m thinking here about controlled substances, things like that.

LINDER: You’ll be shocked to know that this and other interventions, we’ve generalized to other examples of overuse, like over-testing and overtreatment of the elderly, opioids. We’re talking about doing a trial for sleeping medications that are overprescribed. Any area of kind of overuse, we’re going to test this out.

JENA: This seems like a pretty simple strategy to implement. I mean, practices have E.H.R.s. It would not be difficult to embed something like this within an E.H.R. Do you have a ballpark of how many practices might be doing something like this. Is it like 2 percent? 20 percent?

LINDER: I would think it’s closer to 2 percent because this does actually take work and it takes somebody who’s motivated to kind of force through all of the work that needs to happen to re-jigger the E.H.R. to do these justifications. But hoping to show that it works in other circumstances too, and it can be done efficiently and effectively across various examples of overprescribing.

JENA: Okay. What’s number three? What’s the third intervention?

LINDER: Number three is a peer comparison, and this also has a very social aspect. Once a month we would send prescribers a report about how they were doing for non-antibiotic appropriate diagnoses. So, this is a cluster of diagnoses for which we thought the right antibiotic prescribing rate is zero, so that’s a nice target to drive towards. And it turns out the top performers had an inappropriate antibiotic prescribing rate of 0 percent, so we got to tell those doctors, “You’re a top performer.” The remainder of the doctors — about 80 percent of the doctors at the beginning — we told them, “You’re not a top performer.” And that has a couple advantages.

A common mistake when it comes to peer comparison feedback is telling people where they fall on a curve, because it sets average performance as the standard. Even though some of the bottom performers might improve, there may also be a boomerang effect where the top performers start performing less well, since they don’t want to be outliers, even in a good way. One advantage of Jeff’s strategy is that it sets top performance as the standard, moving the whole pack in that direction. Ever notice that your home energy report compares your energy usage to that of your “efficient neighbors”? Well, that’s why.

There’s a second advantage to this strategy:

LINDER: We’ve given them a very simple dichotomous feedback that gets your attention right away. In feedback, giving people way too much data and hoping they’ll figure it out doesn’t work. Everybody’s busy — doctors and otherwise — so we need to give feedback that’s immediately understandable. So, we told them, “You’re not a top performer.” And we saw a reduction in inappropriate antibiotic prescribing from 20 percent down to 4 percent by the end. And it turned out when we closed up shop and kind of quietly went away, stopped sending the reports, over the ensuing year, there was a little bit of backsliding, but there was the least backsliding among this peer comparison group as opposed to any other intervention we tested.

JENA: Why do you think that is? What is it about this that led to the sustained response?

LINDER: I don’t know for sure. I like to think that it’s this sort of social effect. We’re normalizing top performance. Showing that you have peers that have a rate of 0 percent — they’re doing it and you can too — and just kind of changing what the norm is among the group.

JENA: Where has that gone? Has that been implemented into guidelines or are organizations using it to try to curb inappropriate antibiotic prescribing?

LINDER: Most stewardship programs and then the guidelines from the C.D.C. definitely have measurement and tracking or audit and feedback as part of it. But it’s funny how people who do stewardship are really touchy about giving doctors as stark feedback as I described before. Even my own organization — so, at Northwestern Medicine — I get a quarterly feedback report about cancer prevention, antibiotic prescribing, opioid prescribing, and a couple other domains. There’s a lot of information there and I will use every psychological defense mechanism available to me to convince myself I’m doing a pretty good job. I think we need to counter those psychological defense mechanisms with pretty stark feedback. Like I need a red light saying like, “You’re doing a bad job on this, and we see that.”

So, looking across the board at all of the interventions that Jeff and his colleagues have tested, what’s had the biggest effect?

LINDER: I told you about the trial that involved peer comparison and justifications, and those were effective. It turns out, though, that in absolute terms, the most effective thing we found in the trial was being in the trial. Of all the doctors we invited to be in the trial, 75 percent agreed. And those that agreed had inappropriate antibiotic prescribing rate that was about 10 percent lower than the doctors who didn’t agree to be in the trial. So, right there we’re sort of skimming off the better performers. It turns out their inappropriate antibiotic prescribing rate dropped another 10 percent just right at the beginning of the trial, and then it continued to drop throughout the trial. So, I said that the peer comparison group and the justifications group, very roughly had inappropriate antibiotic prescribing rates drop from 20 percent to 5 percent, but the control group had an inappropriate antibiotic prescribing rate drop from about 20 percent to about 10 percent. And so, bang for your buck, if you want people to change their behavior, make them know somebody’s watching their behavior.

JENA: That’d be incredibly cost effective. I mean, Medicare could literally just tell doctors, “We’re going to monitor your antibiotic prescribing,” and then they literally could do nothing. So, Jeff, there is one study that you did not talk about, which is sort of right up my alley. It doesn’t require any actual intervention. Just a little clever thinking. Can you guess which study that is?

LINDER: That is the time-of-day study.

JENA: Oh, there can be only one. Tell me about it.

LINDER: Yeah, so we looked at this study of Israeli judges giving parole.

The short version of that study is that you don’t want to come up for parole right before lunch or right before the end of the day. We should note that the findings have been called into question by some other researchers, but the idea is that at those times, when judges are more tired and hungry, they’re more likely to opt for the “easier” and “safer” option of denying a prisoner parole.

LINDER: Colleagues of mine and I got the idea that, well, maybe we see the same thing in antibiotic prescribing. Do doctors default to doing the easier, wrong thing at different times of the day? And so, we looked at antibiotic prescribing for respiratory infections over the roughly four-hour morning session from 8:00 A.M. to noon, and then the four-hour afternoon session from 1:00 P.M. to 5:00 P.M. and I think your listeners can sort of anticipate what we found. We saw a slight rise in antibiotic prescribing over the morning. There was a drop over lunch, and then the antibiotic prescribing rate went up over the afternoon, such that even when we adjusted for the patient’s diagnosis and adjusted for clustering by clinician — so statistically this is the same doctor seeing the same patient — they’re at least 5 percent more likely to prescribe an antibiotic at the end of the day than at the beginning. This gets attributed to decision fatigue. You’re either worn down or getting behind, or as the work you’ve done stacks up, we kind of default to the easier, quicker thing — in the case of antibiotics, prescribing more inappropriate antibiotics late in clinic sessions and later in the day.

JENA: So, what are you working on now on these issues related to antibiotic prescribing?

LINDER: Most of the stuff I mentioned to you is very focused on a narrow set of acute respiratory infection diagnoses where the right antibiotic prescribing rate is 0 percent, and so to me, the next frontier is broadening the target of our stewardship for squishier diagnoses like sinusitis and pharyngitis. And by squishier, I mean some people with sinus infections and sore throats should get antibiotics, but it’s harder to ferret out which ones are when we’re looking at large populations of patients. And then, can we get our hands around all antibiotic prescribing? Can we implement a measure that compares across clinicians, practices, groups of practices about their overall antibiotic use? Becayse from a public health standpoint, that’s kind of the ultimate goal. The other area is patient-focused stewardship. Many of the interventions are happening when the doctor is in the room with the patient and the patient has already gone out of their way, taken time off from work, traveled to the clinic. And so, I and colleagues have proposed doing a study looking at high antibiotic utilizing patients to see if we can change their behavior.

Antibiotic overuse and resistance are big problems. Jeff’s work suggests that part of the solution are small, but meaningful, interventions. And the data indicate these approaches might be working. According to a C.D.C report from 2019, deaths due to antibiotic resistance in the U.S. are still in the tens of thousands annually, but they’ve decreased since 2013. There’s still more work to do, but progress of any kind on this issue is encouraging.

Thanks to my guest, Jeff Linder, for playing a part in those efforts — and for joining me today. And thanks to you, of course, for listening.

Before I tell you about what’s coming up next week, here’s an idea I had based on my conversation with Jeff. Wouldn’t it be interesting to study whether doctors, when they have a cold, get prescribed antibiotics? It might tell us something about how important knowledge is, compared to emotion, when it comes to wanting an antibiotic when you’re the one who doesn’t feel well.

Also, if you’re interested in learning more about combating antibiotic resistance, check out episode 60 of another podcast in the Freakonomics Radio Network, People I (Mostly) Admire, hosted by my friend, the economist Steve Levitt. That episode is called “Cassandra Quave Thinks the Way Antibiotics Are Developed Might Kill Us.”

And on that note, coming up next week: Could a new study about breastfeeding help us answer longstanding questions about its effects on children’s health?

FITZSIMONS: We do find evidence that it may be contributing to persistence in intergenerational transmission in inequalities.

We’re going to talk about these compelling findings, which offer some of the best evidence to date about the possible benefits of breastfeeding — and also why good research on this topic can be so hard to do.

OSTER: There’s a huge literature which does, the basic thing, which is compare the children who are breastfed to children who are not., and that literature has a lot of problems.

That’s all coming up next week on Freakonomics, M.D. Thanks again for listening.

*          *          *

Freakonomics, M.D. is part of the Freakonomics Radio Network, which also includes Freakonomics Radio, No Stupid Questions, and People I (Mostly) Admire. All our shows are produced by Stitcher and Renbud Radio. You can find us on Twitter at @drbapupod. This episode was produced by Lyric Bowditch and mixed by Eleanor Osborne, with help from Jasmin Klinger. Our senior producer is Julie Kanfer. Our staff also includes Neal Carruth, Gabriel Roth, Greg Rippin, Rebecca Lee Douglas, Morgan Levey, Zack Lapinski, Ryan Kelley, Katherine Moncure, Jeremy Johnston, Daria Klenert, Emma Tyrrell, Alina Kulman, and Stephen Dubner. Original music composed by Luis Guerra. If you like this show, or any other show in the Freakonomics Radio Network, please recommend it to your family and friends. That’s the best way to support the podcasts you love. As always, thanks for listening.

JENA: So, would it be safe to say that if any listeners email you to prescribe them antibiotics, you’ll be happy to do that?

LINDER: Uh, that is incorrect.

JENA: Oh, okay.