Jerome Groopman’s book is an exploration of how medical mistakes are made. Not so much those that involve a slip of the knife on the part of a surgeon, or a misreading of a script on the part of a pharmacist, but errors of thinking – misdiagnoses being the chief instantiations of such errors.

Many of the mistakes he identifies – search satisfaction (stopping one’s search when one has identified a plausible answer, even when other solutions may be out there), “zebra retreat” (avoiding diagnoses of rare diagnoses, even though they could happen), diagnosis momentum (a preliminary diagnosis is made, which gets confirmed in doctors’ minds over time because they selectively emphasise evidence that confirms the diagnosis and ignore evidence that detracts from it), etc., etc – are common cognitive errors outside the medical realm as well, and it would be instructive for readers to reflect on how they apply to one’s own sphere of work as well.

But these cognitive errors, in medicine, should not necessarily be attributed to sloppy thinking on the part of doctors alone – Groopman points out that the risk of cognitive errors increases as doctors are forced to spend less and less time on more and more patients, affording them very little of the precious time that is needed to think. He also points out other problems of money-driven medical systems: drug companies and hospitals push treatments that have been shown not to have much utility, simply because they bring in the dollars, while multiple instances of the same test are discouraged as “wasting money”, even when there are reasons to suspect the first test, for example.

One aspect of medicine that Groopman frowns upon is the increasing use of diagnosis aids such as decision trees or classification schemes. He thinks that these circumscribe the physician’s thinking and aid cognitive errors, citing several anecdotal examples.

I think the issue here is not so much the concept of the tools as the validity of the actual tools and the ways in which they are used. Take decision trees, for example. They start off asking questions such as “Does the patient have a sore throat?” and, based on whether the answer is “yes” or “no”, lead the physician down the tree to a diagnosis. If the decision tree has only N possible diagnoses at the leaves of the tree, a doctor relying solely on the decision tree will certainly not think of a diagnosis outside of these N. In this case, yes, the decision tree is limiting the doctor’s thinking – but only because the doctor is letting it, and because the decision tree is flawed. Conversely, the decision tree may lead the doctor to consider alternative diagnoses if it contains possible diagnoses that were not within the scope of the doctor’s initial thinking. So decision trees can work both ways and can be useful tools, in my opinion. What is needed are well worked-out decision trees, and good guidelines for when to trust in them and when not to.

Another example Groopman gives is of a classification scheme called the International Prognostic Scoring System (IPSS) for a certain syndrome, which is used to determine whether only supportive measures should be taken, or actual treatment, based on various parameters such as the white blood cell count, platelet count, haemoglobin levels, etc. One fellow calculates the IPSS for a patient, which puts him at intermediate-II risk, indicating that supportive measures should be taken. The doctor says, though, that the patient has been deteriorating rapidly from one week to the next, and so he needs treatment right away. “The proliferation of these boilerplate schemes…has caused doctors to become so wedded to generic pofiles that they ignore the individual characteristics of the patient.”

But that’s because the first fellow took a static snapshot of the patient, instead of taking a dynamic view of the course of the patient’s illness. If he had taken the latter into account, he could have extrapolated to see that the patient was heading downhill fast, and would soon have been classified under the scheme not as intermediate-II, but as high risk, necessitating direct intervention. It’s not so much the tool as the way in which it was applied.

Groopman concludes by suggesting a way for patients to help doctors avoid cognitive errors – simply by asking questions, or going over their answers again. In particular, ask “what else could it be?” This alone, says Groopman, can boot the physician out of the rut of many cognitive errors.

Overall, fascinating book, though a little rambling. The anecdotes brought it a certain liveliness, though they seem somewhat suspect for me – after all, what if Groopman is himself committing a cognitive error – picking-and-choosing evidence to support his particular diagnosis of the state of medical diagnosis? Then again his book is backed by copious footnotes, and I suppose a book full of statistics would have been a huge turn-off to some people.

I remain optimistic, though, about the possibility of really good decision aids to help doctors make correct diagnoses. It seems to me that medicine is getting way too complex for doctors to be able to hold all the possible variables in their heads all the time. They need help, and medical cognitive aids would be fascinating to research.

P.S. A comment made at the end of the book by Groopman – “Each of us is unique in our biology, and there can be important differences in both the side effects we suffer and the benefits we gain from the same medication. We can share a single illness but not share its remedy…” This is what proponents of Chinese medicine often declare to defend poor performance in double-blind randomised clinical trials – picking a single Chinese herbal remedy and administering it to a whole bunch of patients with sore throats is not “the way” of Chinese physicians, who will tailor it to the constitution of the patient. Given Groopman’s comment, which I believe to be mostly true, how are clinical trials designed to address this issue? Does it pertain to other systems of medicine besides the Western one? Again, the issue of standards of proof in the complex world of medicine is a fascinating one, and I’d like to read and see more research about it.