In Defense of Mistakes

Martin Samuels“How much of the medicine that you now use, did you learn during medical school?”

My answer may be surprising. It is not the response given to me by my professors, when they were asked similar questions. I recall them telling me that virtually nothing that I was learning in medical school would be correct 20 years later.

I have thought about this since and will reveal my answer shortly, but before I do, we should pause for a moment to reflect on the process of medical education. I will refer here to natural selection as an analogue of this process, a concept that I have adapted from some ideas gleaned from David Dawkins and Susan Blackmore.

Darwinian natural selection is based on the concept that replicators (eg genes, viruses, prions) compete for their locus based on the phenotype produced. In the case of genes, these replications are done with high fidelity, but not perfectly, so that there are a few imperfect copies (mutations) produced, such that there are alternative genes meant for the same chromosomal locus (alleles). It is the competition among the alternative alleles, measured by their phenotypic expression that is the basis of natural selection.   This process accounts for all of the dramatic variation seen in nature, including the present state of the information processing hardware (the brain), but it does not account for more rapidly changing behaviors and beliefs (cultures).

For example, as David Dawkins has observed, women’s right to vote swept across the world as a cultural norm in just about a century (New Zealand in 1893 to Kuwait in 2006), much too fast to be explained on the basis of genetic evolution.

The cultural analogues of genes are memes. If a gene is a bit of nucleic acid that is capable of exactly (and occasionally inexactly) replicating itself, then a meme is a unit of cultural inheritance that can reproduce itself with varying degrees of fidelity.

The analogy to genes is not perfect in that memes have no easily visible structural components, such as chromosomes, but memes are replicators that can compete with alternative versions based on the phenotypic outcome, namely that the meme will successfully reproduce itself.  Successful memes can spread rapidly from brain to brain, producing a zeitgeist (spirit of the times), an example of which is the spread of the cultural norm of women’s right to vote, at a rate (only a century) that is much too fast to have occurred by genetic natural selection.

A “successful” meme is one that is more likely to reproduce itself. The word successful in this context has no other higher moral meaning. Memetic evolution cannot be engineered any more than we can engineer genetic evolution, because the conditions upon which the phenotypic competition will depend are not known prospectively. Indeed, those cultural conditions are themselves determined by memetic competition. Like genes, memes do not exist in isolation, but may be influenced by other memes, producing memeplexes.   Examples of memeplexes include clubs, universities, professional societies and religions.

Because of their relative youth, changes in medical educational methods cannot have evolved by genetic evolution (i.e. it is very unlikely that the brains of Hippocrates, Maimonides, Charcot and Osler were substantially different from each other or from ours). A partial list of successful memes for medical education might be:

  1. Doctors are servants of their patients
  2. Doctors are teachers who pass their knowledge to their successors
  3. A doctor-patient relationship is private
  4. Doctors are skeptical scientists who use evidence to apply current knowledge and to devise new methods of diagnosis, disease prevention and treatment

So-called core competencies (medical knowledge, patient-care, professionalism, interpersonal and communication skills, practice-based learning and improvement, and system-based practice) are not memes. This is precisely why they are so counter-intuitive and difficult to apply independently when thinking about a given doctor’s performance. They have not evolved through memetic competition, but rather have been imposed by committees and bureaucrats. Arguing that these competencies are necessary for evaluation because we have done a “bad” or “incomplete” job educating doctors in the past is analogous to arguing that wings are superior to scales. The comparison is absurd. It clearly depends on the environment and the environmental conditions that will determine the success of future memetic mutations is at best questionable and more likely totally unknown. The current medical culture is obsessed with perfect replication and avoidance of error. This stemmed from the 1999 alarmist report of the National Academy of Medicine, entitled “To Err is Human,” in which the absurd conclusion was propagated that more patients died from medical errors than from breast cancer, heart disease and stroke combined; now updated by The National Academy of Medicine’s (formerly the IOM) new white paper on the epidemic of diagnostic error.

For many years now, I have been collecting my own errors and intermittently categorizing them according to various causes of the mistakes, such as misuse of shortcuts or heuristics (eg framing, anchoring, availability, representativeness, blind obedience), premature closure, inadequate knowledge, such as failure to grasp Bayesian principles, or hubris. Some of these ideas were gleaned from the work of Daniel Kahneman and the late Amos Tversky, summarized in Kahneman’s book, Thinking Fast and Slow. With the help of my colleagues, Allan Ropper and Barbara Vickrey, I have made it a habit to present various versions of this errors talk in many settings including the national meetings of our professional societies, various grand rounds venues, resident and student education sessions and as a visiting professor in numerous institutions. The exercise is valuable as a teaching tool as it personalizes the experience of trying to come to grips with complex medical decisions, but there is actually no convincing evidence that studying these mistakes and using various contrivances to focus on them, reduces their frequency whatsoever. For example, there is absolutely no reason to believe that a comprehensive medical record will reduce the frequency of cognitive errors, whereas it is evident that efforts to populate this type of record can remove the doctor’s focus from the patient and place it on the device. Doctors are actually very open to thinking about their mistakes, but one of the worst mistakes is the belief that one can eliminate mistakes.

The process of genetic evolution is not random, though the individual mutations occur by chance. The same may be said of memetic evolution. The evolution of organisms, ideas, religions and customs occur in tiny steps each one of which is the result of competition between a perfect copy of the prior generation and one that is slightly altered. The alteration itself is a random event. So-called “intelligent design” no more works to explain the nuances of culture than it does biological diversity. If a slight change in our way of educating future doctors is a successful meme, it will exactly reproduce itself and the zeitgeist will be altered, just as we see in retrospect about women’s right to vote. If not, it will disappear into a sea of non-improvements, the nature of which we cannot even remember. Our brains have not evolved to design the future. Attempts to make them do so will be futile.

The miracle of the rapid conversion of ordinary people, most with only four years of medical education, into doctors occurs because of a memetic process whereby the doctors in training rapidly copy the clinical method of the generation before them. Thereafter the process of change becomes extremely slow and is fueled only by mutations in the perfect copies (errors). Without these errors, there would be no further evolution of medical thought and we would have the same views of medicine as our forbearers. Note that change does not make us “better” than our professional ancestors; only more adapted to the current environment. My great great great great neurological grandfather, Jean Martin Charcot, would undoubtedly not be able to recognize AIDS, but he could spot a case of tabetic syphilis literally at fifty paces, something that I certainly cannot do. The futile crusade for the eradication of error by ever more perfect guidelines and computer programs will only produce robots, which can only do exactly as they are told.

Therefore my answer to the initial question may no longer surprise you. At least 90% of what I use as a physician today I learned by copying my teachers. The rest has been very gradually altered by innumerable errors, some of which, by chance, changed my thinking or practice, little by little. One may like to imagine that we have made enormous advances over our ancestors, but, alas, this is not the case. Even some of the seemingly enormous changes that have occurred during my career, such as CT and MR imaging and genetic testing, have only slightly altered the clinical method that I now use to try to help real people. Of course, I try the best I can to avoid errors, but there is absolutely no way that I can succeed. Doctors must make myriad mistakes, recognize them for what they are and change gradually as a result of the tiny percentage of them that make one better adapted to the environment, whatever that may be. This process is not controlled by our free will. We all try to avoid errors but none of us will succeed. This is fortunate as errors are the only road to progress. Focusing on the evil of errors takes our attention away from the real enemy, which is illness. We should relax and enjoy the fact that we are lucky enough to be doctors. Future generations can judge what changes were or were not worth making.

Martin Samuels, MD is Chairman of Neurology at Brigham and Women’s Hospital.


Dawkins R. The Selfish Gene
Blackmore, S. The Meme Machine
Kahneman D. Thinking Fast and Slow
Vickrey BG, Samuels MA, Ropper AH. How neurologists think: a cognitive psychology perspective on missed diagnosis. Ann Neurol 2010; 67: 425-433.

Categories: Uncategorized

7 replies »

  1. Certainly, mistakes have to happen for physicians (and everyone else) to learn, but that doesn’t mean the same, patient-harming, mistakes have to happen repeatedly so that each physician can learn on his or her. An “I’ll do it myself” approach is built on the premise that one is aware of those mistakes and, in situations where timing matters, is aware while the opportunity for recovery is still available. Unfortunately, physicians are often unaware of diagnosis-related errors because the patient has moved on to another provider or stops seeking care altogether. That leaves the initial provider with no feedback from which to learn and (if necessary) change practice. Until we have a feedback system that gives physicians the opportunity to learn in real time from their own mistakes, employing case studies and aggregated data from peers is a worthy proxy.

  2. To be clear:

    1) I am not only part of the patient safety movement, but had examined all the medical literature on patient safety from the 1950s through the mid-1990s as part of my book, “Demanding Medical Excellence.”

    2) No reasonable person would disagree with Dr. Samuels that doctors should be allowed to make mistakes as part of training. And I do not disagree.

    3) If one looks at Bosk’s “Forgive and Remember,” the classic on medical training, it talks about different types of mistakes. I apologize for using the word “superficial” to refer to dr. Samuels’ concern about To Err is Human and the new IOM report on diagnostic error. I’m sure his worries are sincere. However, if one looks closely at these reports (particularly the former; errors of execution are a lot more clearcut than “diagnosis”), the authors bend over backwards to address that concern. I see no evidence whatsoever in the writings of Berwick, Blumenthal, Leape or others that would support the worry of “demanding perfection” — as opposed to medical excellence.

    On the other hand, I see quite a bit of evidence of opponents of systematic quality improvement using the fear of demanding perfection as a way of avoiding needing systemic changes. Dr. Samuels may not be one of them, but, with the best of intentions, he is, to my mind, raising a straw man used by opponents of needed change. I invite commenters here to cite chapter and verse otherwise in an evidence-based dialogue.

  3. Having read through several of the comments (I know and like Michael Millenson and know he is well meaning), I found the article of interest. Many of my mistakes have been influential forming me as a physician I still see the faces of people I cared for who had an unexpected outcome. Some of these faces haunt me. I wish I could have done a better job, made different decisions, but medicine is an imperfect science.

    Dr. Samuels comments are spot on. The majority of our learning comes via watching/listening to our mentors, with another variable amount through experiential learning. For me, the question is how we might take more value away from each poor outcome. Regrettably the 90% taught to us by our peers is fraught with black pearls, knowledge that has been past down, but never truly validated, information that is wrong. And when we unfortunately have an event, experiential learning, it is often transactional rather than representative of best practice. For example, a patient responding poorly to a medication does not mean the drug shouldn’t be used moving forward, but rather the complication should be carefully monitored. We respond to our last big mistake rather than having an overview of our practice and how it compares with expected outcomes.

    If physicians are guilty of anything it has been more of an oversell of what we can do for our patients. We have done a poor job of managing expectations. Medicine will never be perfect. However, better tools to learn from our mistakes are sorely lacking.

  4. It’s usually a bad idea starting off with ad hominem. I think Dr. Samuel’s point is that stamping out errors in a contrived manner, as opposed to a more natural process, could have consequences of their own. We know that judgment & heuristics have their shortcomings. Gutting judgment could be a monumental & irreversible error.

  5. I thank Michael L. Millenson for reading my essay “In Defense of Mistakes.” It apparently had high limbic valence for him, possibly because of his own personal attachment to the patient safety movement. Please see my earlier essay entitled “The Antihypocricy Rule,” as in it I advocated that any pundit who opines about medical care should reveal his own conflicts of interest as identified by the manner in which he himself obtains care or has a business or books related to the subject.

    Mr. Millenson appears to have authored a book entitled: Demanding Medical Excellence: Doctors and Accountability in the Information Age. Perhaps Mr. Millenson should reveal his own personal attachment to that movement as it is clear that he stands to gain by the medical profession accepting his cynical view of the way medicine is practiced. Labeling someone’s ideas as “superficial” is unnecessarily pjorative, as simplicity may be much more profound than complexity (e.g. F=MA). Ad hominum statements and sarcasm implying that one has not read the literature are also not very collegial or helpful. I am very aware of the important work of Lucien Leape, David Bates and David Blumenthal. In fact, I know Drs. Bates and Blumenthal personally as they are or have been associated with the Brigham and Women’s Hospital during my long tenure here. I don’t know Dr. Leape personally, but, of course have read much of his work with great interest.

    My ideas are not in conflict with those of Drs. Leape, Bates and Blumenthal. My essay was about the adaptation of memetic principles, a concept developed by David Dawkins and Susan Blackmore, to the evolution of medical ideas over a very long time span. As I mentioned, none of us wants to make mistakes and we certainly don’t try to take out the wrong kidney, prescribe the wrong drug or operate on the wrong side of the brain. Systems, like the ones advocated by leaders of the patient safety movement, may reduce operational errors of this type, though there is little objective evidence of this yet. I was referring to the evolution of medical ideas and the mechanisms whereby randomly occurring currently “wrong” ideas change the direction of thought. As I clearly stated, the core concepts articulated in my essay were not mine. They belong to Drs. Dawkins and Blackmore. Time will determine whether they are “superficial” or not.

  6. Dr. Samuels’ superficial equating of “you need mistakes to learn” (true) with the 1999 and 2015 IOM reports, and his dismissal of them, is absurd. Were he to talk to Harvard colleagues such as Dr. Lucian Leape or Dr. David Blumenthal or Dr. David Bates, he might learn something about the real definition of medical error and its true impact.

    Instead, he is as superficial as a guy standing out in the street during a snowstorm proclaiming that it disproves global warming.
    Dr. Samuels: you are desperately in need of a curbside consult with some experts at your own institution.

  7. What a wonderful essay! I wonder what Dr. Samuels would say of the current “value based payment” meme for doctors. Is this a manifestation of “a culture of perfect replication and avoidance of error”?

    As for “The futile crusade for the eradication of error by ever more perfect guidelines and computer programs…” Today’s clinical computer programs are almost universally secret software. The meme of introducing secrecy into medical practice as a necessary incentive for progress is now pervasive among my physician colleagues. Will medical education evolve to teach about our relationship with information technology as well as our patients?