Readings: Doctors, DNA, Democracy, etc.

How many patients do doctors accidentally kill every year? This turns out to be a more difficult question to answer than you might expect. 

It might seem easy to answer this question. You need just three things, more or less ordered in time:

  1. A live patient
  2. A medical error
  3. A dead patient 

The problems start here, however. How do we know that the error killed the patient? Sure, sometimes it’s obvious — doctor gives patient wrong medication, patient dies, etc. — but most of the time it’s less clear than that. But even in that case, if there was no way to know the medication was going to have that effect, is that a doctor-caused death? If we did this experiment, ahem, on patients, say, 50,000 times to get statistical significance, would they all die? Inquiring minds want to know, even if it’s not an experiment we’re likely to run.

And how do we know the error was a preventable error anyway? Many times patient die as a result of cascades. Here is one case: “aspiration led to respiratory failure, acute renal failure, shock, and cardiac arrest”. The allegation is that the aspiration was preventable, so the resulting series of unfortunate events leading to a dead patient were also therefore preventable. But was the aspiration preventable?  I supposed it depends on what they aspirated, when, and how quickly everything went badly afterward. If they ingested a sponge left in their mouth by a doctor, that’s bad; if they somehow choked on saliva during the night, then that’s trickier.

Critics go on and on about this sort of thing, about poor data; about headline-hungry researchers; about devious doctors hiding errors; about the difficulties in post-death preventability assessment; about whether sick people (hey, they were in a hospital after all) would have died anyway; about our inability to run proper randomized experiments; and so on. The result is wide variation in estimates of how many patients are killed by medical errors every year, from 25,000 to 400,000, and pretty much every number in-between.

There can be no doubt that hospitals are nothing like the slaughterhouses they were in the 19th century and earlier. Back then the safest thing you could do, if sick, was stay as far away as possible and maybe die of something else. With no viable theory of infection, for example, or at least none that had anything to do with how infections actually happened, hospitals were petri dishes for post-operative bacteria, with instruments shared across patients, sterilization non-existent, etc.

But not all medical progress is toward safety. As Lindsey Fitzharris described in her wonderful “The Butchering Art”, the arrival, for example, of general anesthesia, while welcomed by patients who previously had to be strapped down so they could suffer through grisly procedures, didn’t initially have the desired effect. Instead, doctors, no longer under time pressure, began attempting riskier and more complex procedures, or being more exploratory during what should have been less risky procedures, causing death rates to initially increase somewhat.

As weird as this will seem, we have no idea how many patients that doctors kill per year. We only know it’s highly non-zero, that it’s higher than it should be, and that it’s unlikely to ever fall as far as it could, given the nature of risk, of uncertainty, and of causality.


Here are some articles and papers worth reading: