Earlier this week the GMC at long last released data on the number of doctors who died while undergoing fitness to practice investigation and monitoring. The number of deaths should be extremely small — the individuals at risk are working age adults with good socioeconomic status — and the numbers are small, on average about ten a year in recent years. But are these numbers really that small? For comparison, there are currently around twenty deaths in police custody annually in all England and Wales, a conceptually similar but in other respects very different so-called never event that should never happen. When Dr No last wrote about what we might loosely call deaths in GMC custody, in 2013, the annual number of deaths, revealed after a FOI request, was very similar, around eleven, a number subsequently confirmed in a 2014 internal GMC review, which also revealed that around one in four of the deaths were suicides. In the data released this week, around one in five or six of the deaths were suicides.
The high proportion of suicides sent a signal that the harshness of the GMC’s long drawn out processes, which are notoriously punitive, obnoxious, adversarial and Kafkaesque, might be partly to blame for the deaths. The GMC duly did some public hand-wringing and bed-wetting, and promised to clean up its act. And yet here we are, almost a decade later, and the annual number of deaths is pretty much unchanged, though the proportion of suicides may perhaps be slightly less. On the face of it, the GMC’s claim that it has made its processes more humane appears not to be borne out by the figures — the never events are still happening at the same annual rate. But, as ever, the devil is in the detail, in this case the number of individuals at risk. If, for example, the annual number of doctors investigated by the GMC has doubled, and nothing else changed, then we might expect the number of deaths to double. We need to determine how many doctors underwent fitness to practice investigation a decade ago, and how many are subject to investigation these days. We can then compare rates, to determine whether things are getting better or worse. We can also, albeit very crudely, compare the number of deaths we might expect in a similar population not undergoing fitness to practice investigation to the observed number, and determine whether our doctors are indeed dropping like flies, or dying pretty much at the rate we would expect.
The denominator, the number of doctors at risk, is surprisingly difficult to determine. Not only do the number of doctors reported to the GMC vary year on year, changes in the way the GMC handles complaints have also changed over time, meaning considerable variations in the number of doctors who end up under investigation. And then there is the question of when does an enquiry become an investigation? If there is a preliminary enquiry, as there often is, does that amount to being under investigation, or is that stage only reached when the GMC commits to a full investigation? And finally, there is the duration of the investigation. The GMC is notorious for the bovine slowness of its processes, which can take years to conclude. A doctor whose investigation starts in one year may well still be under investigation over a year later, and so is still part of the at risk population, but will only be counted once, in the year the in which the investigation started.
The GMC now has, like every other organisation you would rather not know about, a dashboard which has a chart showing the annual number of investigations since 2007. These are so called full investigations, which by and large happen when the doctor faces allegations that could put his or her registration at risk. The chart shows that, apart from a sharp peak in and around 2013, the number of full investigations was similar in the late noughties to that seen in more recent years. However, the average annual number of full investigations for the period 2007 to 2013, which overlies the 2005 to 2013 period covered in the GMC’s 2014 report, is 2,105, compared to 1,404 in the period covered by the GMC’s latest 2018 to 2020 figures. This means annual crude mortality — average annual deaths divided by annual number of new investigations — has marginally increased, not decreased, in recent years, from 6 per 1,000 to 7 per 1,000. The crude confirmed suicide rate has remained pretty much the same, at a little over 1 per 1,000, though the actual rate is likely to higher, because not all suicides end up as confirmed suicides.
These rates come with all the usual caveats. The annual numbers of events (deaths) are extremely small, there are classification problems (when does a doctor become ‘at risk’, how many suicides are not recorded as suicides), The duration of fitness to practice investigations may also have changed over time, thus increasing the number of person-years at risk, but that does not appear to be the case. For the period 2013 to 2020 (the only period for which Dr No has been able to find comparable data), the total number of cases still open after one year varied from 477 in 2015/16, to 919 in 2013/14 (and was 800 in 2019/20), while the median time in weeks from initial complaint to final hearing ranged from 80 in 2018/19 to 107 in 2016/17. By and large, although there is considerable year on year variation, it appears GMC investigations still take an unacceptably long time to reach conclusion.
There remains one other problem in the denominator in the GMC’s latest report. The total three year number of full investigations in this week’s report appears to agree with the dashboard (around 382 of the preliminary enquiries became full investigations, which if added to the 3,834 always full investigations gives a total of 4,216, compared to 4,213 from the dashboard), but the report then obfuscates things by breaking down the deaths into deaths while under investigation (n=19) and while being monitored (n=10). This appears to be a lump of mud thrown in to muddy the waters, but there is a way round it. Any doctor being monitored has also been under investigation. He, or more rarely she, has just had the double hit of being investigated and monitored. We can keep the denominators as they are, and just bear in mind that the doctors being monitored will very likely have had extended periods of exposure to GMC duress.
Finally, and perhaps most importantly, how do these numbers of deaths compare to the number of deaths we might expect in a similar population not subject to GMC investigation? But what reference population to use? Annual working age specific mortality rates can be calculated readily enough from standard ONS data for all individuals (very roughly, 3/1000 for men, 2/1000 for women), but that is not much use, given the wide range of mortality experienced by different socioeconomic groups. Age standardised mortality rates are available for many professions, including health professionals, whoever they are, but those rates (2.25/1000 for men, 2.13/1000 for women) include all ages. What we need is the intersection of these two rates: what is the mortality rates in working age health professionals? Unfortunately, the data is not available, and so we are going to have to use a proxy. Proxies always carry a statistical health warning, but we do have the above all working age and all health professional rates as sanity checks.
The closest proxy Dr No has found is a study that reports age and sex specific mortality not by occupation or socioeconomic group, but by geography and deprivation, for 2017. The research is up to date and thorough, and contains age band and sex specific mortality rates by deprivation, from the most deprived area to the least deprived. To allow us to use the mortality rates in this study as a proxy, we are going to assume that the mortality of doctors is broadly in line with the mortality rates found in areas of low deprivation, regardless of where the doctor actually lives. Dr No does not think this is an entirely unreasonable assumption: a doctor, wherever he or she lives, will enjoy many of the health benefits available to those who do live in the affluent areas, notably high income and generally healthy lifestyles. Turning to Figure 2 in the study, we can eye-ball the right hand ends (the areas of least deprivation) of the orange (2017) lines, for males and females aged 40 – 54 — see the screen grab below — (and perhaps the charts for ages 25 – 39 and 55 – 64; note the captions for individual charts are above, not below, the charts) and come to mortality estimates of 1.75 per 1,000 for males, and 1.6 per 1,000 for females for these affluent areas. These figures are not exact, but neither do they fall foul of the sanity checks above: they are, Dr No suggests, good enough.
Screengrab 1: Death Rates (Deaths per 1,000 population) for 40 – 54 year olds, by ventile of deprivation and sex in 2003 and 2017. Note the very sharp gradient moving left to right, from the most deprived areas to the least deprived. Source: see text
The GMC dashboard shows visually (unfortunately, the data download option doesn’t work…) that around three quarters of those investigated are male, and the majority are middle working aged (which is why Dr No used the 40 -54 year old charts, noting that the ones on either side probably cancel each other out). Of the 4,200 doctors under investigation over the recent three year period, around three quarters, or 3150, are male, and 1,050 are female, giving an expected number of male deaths of 5.51 (3.15 x 1.75 = 5.51) and an expected number of female deaths of 1.68 (1.05 x 1.6 = 1.68), making the three year total number of expected deaths in age/sex matched peers from a similar socioeconomic background but not undergoing fitness to practice investigation a grand total of 7 (5.51 + 1.68 = 7.19, rounded to 7). Among the doctors undergoing fitness to practice investigation, that number was 29, making their mortality rate compared to those enjoying similar life experience, but without the weight of a GMC investigation on their back, four times higher.
Dr No has taken pains to underline the fact these findings rely on a proxy measure, and there may be other confounding factors, not least the possibility that the doctor under investigation came to the GMC’s attention because he or she was ill, and so was already at higher risk of death, though in a sane humane world these doctors would be managed and indeed cared for well away from the GMC, and never come near its fitness to practice procedures. Nonetheless, on the basis of the data and results presented here, which rely on a methodology that Dr No considers is an improvement on the method he used in 2013, we find there is a fourfold increase in mortality among doctors under GMC investigation, compared to comparable individuals not undergoing investigation.
A decade ago, Dr No characterised the mortality of doctors dying while under GMC investigation as ‘dropping like flies’. Ten years on, after a lengthy period in which the GMC has been at pains to say it has made its fitness to practice processes more humane, we find that little has changed. Doctors under those processes are still dropping like flies.