First up Coleman's, well, he's a crank. There's no other word for it - he's an AIDS denialist, a COVID-denialist, and the lion's share of his career has been spent writing lurid tabloid "advice" columns in such bastions of facts and integrity as The People and The Sun, running premium-rate sex advice phone lines and writing (mostly) self-published books. He hasn't been a licensed doctor for some five years and he hasn't been actively practicing medicine for longer.
None of this means he's necessarily wrong, mind you. Although it would suggest approaching any of his, er, proclamations with a substantial dose of skepticism.
The book you refer to, and in particular it's pivotal claim was so sketchy and poorly evidenced as to attract an advertising ban from the Advertising Standards Authority.
But it's not only Coleman's book that's making this sort of claim, if it were then I doubt anyone who didn't buy their headgear in the baking supplies aisle would give it a second thought. As you mention there's actual papers, published in serious medical journals making some pretty scary-sounding claims about the levels of medical error.
The Johns Hopkins paper you mention is perhaps the most often touted (and IIRC the highest figure that doesn't come from someone like Gary Null or Mike Adams) at ~250,000 but before that there was the Institute of Medicine's report "To Err Is Human" in 1999 which had a lower-but-still-alarming estimate of 44,000 to 96,000 per year.
But these analyses aren't without issues, the IoM report was perhaps a little too keen to mark things as "errors" when there wasn't really anything to back that up and scaling up from two studies that looked at specific areas of the US to the whole country has a bit of a "finger in the air" feel to it.
The Makary and Daniel paper takes a couple of relatively small studies (Landrigan et al and Classen et al) which looked at adverse events and then performs some pretty sweeping up-scaling. Classen is structured around every adverse event being "preventable", and this is what Markary and Daniel scale up from the 1,000 hospital admissions in the study (vs. ~37million a year total in the US) to produce their upper estimate of 400,000 deaths per year.
The biggest study included was the 2004 HealthGrades Quality Study which looked at "Patient Safety Incidents" which include:
- Complications of anesthesia
- Death in low mortality Diagnostic Related Groupings (DRGs)
- Decubitus ulcer (bed sores)
- Death among surgical inpatients with serious treatable complications
- Iatrogenic pneumothorax
- Selected infections due to medical care
- Post-operative hip fracture
- Post-operative hemorrhage or hematoma
- Post-operative physiologic and metabolic derangements
- Post-operative respiratory failure
- Post-operative pulmonary embolism or deep vein thrombosis
- Post-operative sepsis
- Post-operative abdominal wound dehiscence
- Accidental puncture or laceration
- Transfusion reaction
Some of those you can clearly see could be but not necessarily are the result of medical errors. And as the authors of the report point out the numbers of fatalities resulting from some of these are heavily skewed by the fact that this is coming from Medicare data - which means it's predominantly the over 65's (Is the notion that the elderly are more frail and consequently more likely to die after surgery really news to anyone?) Which makes Markary and Daniel's scaling of this to the whole US population to get their ~250,000 figure, shaky at best.
A recurring problem for studies trying to examine this is that there's an awful lot of estimating going on, a great deal of use of indirect measures to signal an "error" may have occurred and a veritable cornucopia of differing methods for determining what truly counts as a "medical error".
A recent (2020) systemic review from Yale - Rodwin et al was published in the Journal of Internal Medicine and attempts to address this indirectness. They did so by examining studies of in-patient deaths rather than admissions and then looking to see if they were preventable. To quote:
Studies that review series of admissions and determine whether adverse events occurred, whether the events were preventable, and what harms resulted have been criticized for indirectness when used to estimate the number of deaths due to medical error.5, 6 In contrast, studies of inpatient deaths offer a more direct way of estimating the rate of preventable deaths.
They attempted to address some of the scaling issues of previous efforts thusly:
Studies limited to specific populations such as pediatric, trauma, or maternity patients were excluded because our primary research question was to determine the overall rate of preventable mortality in hospitalized patients and these populations are less generalizable.
This review went on to establish that the likely figure is far lower than suggested by the Johns Hopkins paper, estimating it to be around 22,000 per annum and this would also place it far, far lower on the "cause of death" leaderboard:
While the rate of preventable mortality in hospitalized patients is lower than is often reported, it still represents what would be the 15th leading cause of death in the USA and deserves the continued attention of clinicians, hospital administrators, and policy makers.
There's still some pretty big limitations to the Yale study - The determination of whether a death was preventable or not is still essentially subjective - determined by physician review. And there was a dearth of suitable studies from the US for example so you're trying to metric the rate of errors in the US from data in other countries, and the US medical system has some pretty significant differences to the ones in use in places like Canada and Europe.
Another US-aimed study attempting to determine the role of medical errors in deaths is Sunshine et al from 2016 suggests that the numbers of deaths where an "Adverse Effect of Medical Treatment" (AEMT) as they call it was the underlying cause of death is low (~5,200) and higher where the AEMT was in the cause of death "chain" (~108,780) that the estimated proportion of these that is classed as "misadventure" (i.e. medical error) is only about 8.5% so the figures are pretty low (about 9,572). The data underlying the study isn't going to capture everything, as they authors point out:
The GBD approach for estimating mortality associated with AEMT also has limitations. First, ICD-coded death certificates have been shown to have varying degrees of reliability in identifying medical harm. They may have limited ability to distinguish between variation in completeness of death certificate reporting and variation in the occurrence of AEMT events. It is also probable that many deaths involving AEMT are not captured either because of motivated misreporting or unintentional omission.
So it's probably a lower-bound at best - but still interesting when compared with the Yale findings.
If the Yale figure is accurate is this still too high? Probably. Does this mean there should be continuing efforts to reduce this figure as much as possible? Absolutely. Eliminating errors entirely is..unrealistic. But physicians and other medical professionals are human and humans do make mistakes, it doesn't make them monsters and it doesn't mean that seeking medical treatment is more dangerous than not.
Sure if you go into hospital with appendicitis there's a non-zero chance the surgeon is going to slip, slice something they shouldn't and you might die, but I think it's a fairly safe bet that your chances are a hell of a lot worse if you stay home.