Specialty training: ethnic minority doctors’ reduced chance of being appointed is “unacceptable”

Dear Editor

In rapid responses of 18 April,[1] 1 May, [2]13 May,[3] 29 May,[4] and 6 June 2020 [5] regarding racial/ethnic differences in Covid-19 outcomes in the UK and US and the newly established NHS Race and Health Observatory, I discussed the impossibility of determining whether demographic differences are increasing or decreasing or are larger in one setting than another without understanding patterns by which measures of differences between rates tend to be affected by the prevalence of an outcome. One key pattern is that whereby the rarer an outcome the greater tend to be relative differences in rates of experiencing it and the smaller tend to be the relative differences in rates of avoiding it, which I previously discussed in many rapid responses between 2006 and 2016 with respect to interpreting data on health inequalities or in clinical trials.

One consequence of such pattern is that as health and healthcare improve, relative differences in favorable outcomes and relative differences in the corresponding adverse outcomes commonly yield opposite conclusions as to whether disparities have increased or decreased, as discussed, for example, in reference 4 with regard to the way the US National Center for Health Statistics and Agency for Healthcare Research and Quality would reach opposite conclusions about whether racial disparities in immunization increased or decreased over time.

Data in the following sentences in this article usefully illustrate the pattern: “Even specialties with large shortages, such as general practice, have large disparities. For example, in 2016 just over half of ethnic minority applicants to general practice were deemed appointable [52%] (1660 of 3212), but in white applicants the proportion was 75% (1935 of 2575). This gap narrowed slightly to 74% (2159 of 2922) versus 93% (1880 of 2029) in 2018.”

According to these figures, the ratio of white rate of being deemed appointable to the minority rate of being deemed appointable declined from 1.44 to 1.26, which is the way the 2019 NHS Workforce Race Equality Standard 2019 Data Analysis Report for the NHS Trust [6] would quantify the difference. That would support the statement that the gap narrowed. But the same figures show that the ratio of the minority rate of being deemed not appointable to the white rate of being deemed not appointable increased from 1.92 (48%/25%) to 3.71 (26%/7%), which observers relying on relative differences in adverse outcomes would regard as a dramatic increase in disparity. The method discussed in reference 1 that is theoretically unaffected by changes in the prevalence of an outcome would show an increase in the difference, from a probit d' value of .624 in 2016 to one of .832 in 2018.

A similar illustration can be constructed with 2018 minority and white appointable rates in the top and bottom rows of the article’s Figure 1. The former are for general practice where rates are comparatively high and the latter are for academic medicine where rates are comparatively low. The ratio of the white appointable rate to the black appointable rate for general practice is the 1.26 (93%/52%) ratio previously noted, while the ratio for academic medicine is 1.48 (34%/23%). This would lead some to believe there is a larger disparity for academic medicine than general practice. On the other hand, in accord with what is commonly observed in the circumstances, the ratio of the black rate of being deemed not appointable is larger for general practice than for academic medicine – that is, the 3.71 (26%/7%) ratio for general practice compared with a 1.17 (77%/66%) ratio for academic medicine. The method discussed in reference 1 would show a far smaller disparity for academic medicine (probit d' of .326) than the previously noted 2018 general medicine figure of .823.

While the NHS equity report measures racial/ethnic differences in outcomes involving being deemed qualified for or securing employment in terms of relative difference in favorable outcomes, it measures racial/ethnic differences in disciplining physicians in terms of relative differences in adverse outcomes (in both cases using the larger figure as the numerator in a risk ratio, as done above). And the report uses these measures both to appraise changes over time and to compare differences across trusts. Its authors do so while presumably aware that every favorable outcome has a corresponding adverse outcome and vice versa, but almost certainly while unaware that it possible for relative differences in favorable outcomes and corresponding adverse outcomes to yield opposite conclusions about directions of changes over time or from place to place (or with respect to different specialties), much less that this will usually be the case. And, like leaderships of the trusts themselves, the authors almost certainly are unaware that generally reducing physician discipline rates by adding circumspection or procedural safeguards to the discipline process or raising the thresholds for investigation – or even by improving training and guidance to make all physicians better able to succeed – will tend to increase relative racial/ethnic differences in discipline rates (though reduce relative differences in rates of avoiding discipline).

The above points do not address the role of racial/ethnic bias. To the extent that bias plays a role in observed differences, reducing bias will reduce all measures of racial/ethnic disparities. But, as discussed toward the end of references 4 and 5, it is when bias is at issue that it should be especially evident that normative or value judgements have no useful role in choice of measure. With respect to the data cited above, it should be obvious that it makes no sense to say that between 2016 and 2018 bias decreased with respect to decisions that general practice applicants were appointable but increased with respect to decision that applicants were not appointable, or that in 2018 appointable decisions were less biased, while non-appointable decisions were more biased, in general practice than academic medicine.

The more likely it is that bias is playing some role in either selection or discipline decisions, the more important it is that researchers and governmental entities employ sound methods for identifying and quantifying that bias and determining how it might be eliminated or reduced.

References:
1. https://www.bmj.com/content/369/bmj.m1562/rr-2
2. https://www.bmj.com/content/369/bmj.m1562/rr-5
3. https://www.bmj.com/content/369/bmj.m1562/rr-7
4. https://www.bmj.com/content/369/bmj.m2122/rr-0
5. https://www.bmj.com/content/369/bmj.m2191/rr-0
6. https://www.england.nhs.uk/about/equality/equality-hub/equality-standard...