The PHE report on why BAME groups are hit harder by Covid has been badly misunderstood
The eagerly-awaited report from Public Health England on Disparities in the risk and outcomes of COVID-19 provides crucial information in the search for an explanation of why Black and Minority Ethnic (BAME) groups have been hit so much harder by the virus. But the way the media has written it up – and therefore the message that most people will draw from it – is based on a misreading of what the headline estimates at the top of the report refer to.
The key paragraph in the report’s Executive Summary is this.
An analysis of survival among confirmed COVID-19 cases and using more detailed ethnic groups, shows that after accounting for the effect of sex, age, deprivation and region, people of Bangladeshi ethnicity had around twice the risk of death than people of White British ethnicity. People of Chinese, Indian, Pakistani, Other Asian, Caribbean and Other Black ethnicity had between 10 and 50% higher risk of death when compared to White British.
The numbers in this paragraph are what people have latched onto. What they refer to are hazard ratios, namely the risk of death relative to that for white British. Values of these ratios include 1.10 for Black Caribbean, 1.44 for Pakistani and 2.02 for Bangladeshi.
But what exactly do these risks refer to? Do they describe the risk faced by a person picked from the population at random? That’s what any story that presents these numbers as plain “mortality rates” is assuming. But such an assumption is wrong.
The clue is in the opening words of the paragraph: “an analysis of survival among confirmed COVID-19 cases”. What these estimates refer to are not the relative risks of dying faced by someone picked at random but the relative risks for someone who has already tested positive for Covid-19. The two are completely different and for a number of reasons, mustn’t be confused.
First, these conditional mortality rates – conditional, that is, on a positive Covid-19 test – have nothing to say about the risks that different groups face in catching the virus (and being tested for it) in the first place. The report does deal with this. The chapter on ethnicity is clear about the two parts to the risk, namely the risk of acquiring the infection and the risk of poorer outcomes once the infection acquired. On both counts, it says, BAME groups are likely to be at increased risk.
What the report doesn’t do is give the two parts of the risk equal billing in the headlines. Although the hazard ratios deserve their prominence because they’re what’s new here, they only measure the risk of poorer outcomes once someone is infected, that is, the second part of the risk. This is confusing because previous risks by ethnicity, for example from the ONS, have referred to the overall risk.
Second, the fact that the hazard ratios are only part of the overall risk explains why some of these values are so much lower than other estimates. After allowing for age and geography, the ONS found overall mortality rates for BAME groups on average to be around double that for White. The graph below, showing hospital deaths per head of population for BAME compared with White up to 28th April, is in line with this. The particular point of the graph is to show how the relative risk varies depending on whether the population is adjusted to reflect the mix by age, sex and region of the people who have died. If age and sex are taken into account but not region, the overall risk of death for BAME is 2.6 times that for White.
Hospital deaths from Covid-19 to 28th April: the overall risk for BAME relative to the overall risk for White, using three different reference populations
 Table A1, adjusted hazard ratios.
 The ONS gave “odds ratios” (relative risk) for England and Wales after allowing for age, sex, region and rural/urban (model 2) including 2.4 for Bangladeshi, 2.3 for Black and 1.5 for Indian.
 The PHE reports (p39) shows an age-only adjusted risk of infection for black men almost three times that for white British men, with a comparable risk for black women of more than double.
 Source: NPI analysis of PHE deaths by NHS region and adjusted with ONS Population denominators by age, ethnic group, regions and countries