Data Ethics Club: ‘The data was there – so why did it take coronavirus to wake us up to racial health inequalities?’#

What’s this?

This is summary of Wednesday 2th November’s Data Ethics Club discussion, where we discussed the Guardian article ‘The data was there – so why did it take coronavirus to wake us up to racial health inequalities?’ written by Angela Saini. The summary was written by Huw Day, who tried to synthesise everyone’s contributions to this document and the discussion. “We” = “someone at Data Ethics Club”. Nina Di Cara and Natalie Thurlby helped with the final edit.

Who Killed Hannibal Meme, "400 years of colonial oppression" shoots "Health Equality across races" and then "Journalists covering disproportionate covid deaths" ask "Why would vitamin D do this?". Meme made by Euan Bennet.


In this installment of Data Ethics club we discussed a 2020 Guardian Article discussing how persisting inequalities in healthcare were further exposed when covid data was being so heavily scrutinised. For example, it is an established fact that “black women in the UK are five times more likely to die in pregnancy than white women, while Asian women are twice as likely to die”. This is not just in the UK either as the author notes: “A US study found last year that black patients were 40% less likely than white patients to get the medication they needed to relieve acute pain.”

It took the covid-19 pandemic to push this to the point of acknowledge, but despite this being an established body of research, many (particuarly those not in minorities) were surprised to learn of this inequality. We discussed this surprise, what responsibility the public, the media and those working in the field have to report and acknowledge these inequalities and potential consequences of the way we record race.

Did you know the maternal mortality rates discussed in the article? Were you surprised?#

Some of us were unaware of this, especially pre-covid but came to be aware of it during covid. Many of us were vaguely aware before but seeing the scale of the difference was shocking. Some knew and weren’t surprised due to personal experiences or were aware through friends who have personal experience/insight such as midwives, doctors or black mothers. Some famous people such as Serena Williams have already spoken publicly about her experiences of racism in healthcare. This is depressing but perhaps not surprising, with medicine, examples, and diagrams/explanations geared/aimed at towards white men.

Research from 1970s has shown black males having disproportionate restraints/drug doses in mental health care. Why has nothing been done about this? Stories stick around longer even when they are disproved with research - the example of vitamin D was mentioned.

Many of us were aware of the research bias. For example; how psychological research studies are done on psychology students or how medical text books show e.g. skin conditions on white people or autism diagnoses. Even within biology (for example reproductive biology), people thought of how important sperm was until female scientists came along, then they thought about the eggs e.g. The Egg and the Sperm: How Science Has Constructed a Romance Based on Stereotypical Male-Female Roles. or this article.

60% of people who died from covid were disabled - this was disproportionate even when those disabilities were nothing to do with respiratory diseases e.g. lung function/immune system. This is partially because anyone with a learning disability had a do not resuscitate order as a default. Some of these people were on the one hand “essential workers” and on the other hand disposable. And there was a similar talking point to the one in the article, “genetics must be causing this issue”, in the media with the “underlying health conditions” framing.

How are minorities covered by clinical trials? Some people find seeking medical help quite difficult because they’ve had issues with doctors in the past. There is already lots of racism throughout society and that will be reflected in medical professionals. Lots of communities that don’t trust healthcare professionals. This isn’t helped by language bias/barriers.

Diversifying medical teams leads to reduction in inequalities across all intersections. Entrenched systemic attitudes (at the top of hierarchies) mean that despite 70%+ women nurses (for example), the NHS is not necessarily a women-friendly place. Even in applications for public health, there are reports that highly qualified people are filtered out so that most people actually interviewed end up being white.

Patriarchy is another problem! Everyone is harmed by this, except for the small group of white men at the top of the hierarchies. People also need stories to explain and people to blame and it seems too easy to blame smaller, already disadvantaged groups.

Saini discusses how race is misunderstood as a biological concept when it is actually a social one. What implications should this have for the way we measure and record ‘race’?#

Some diseases are more common in different races/ethnic groups and it could be dangerous to ignore these correlations. But it would be important to not let this take away from the social issues from things like pregnancy or covid.

People have stopped talking about monkeypox because they (wrongly) think it only affects gay men. There’s a similar issue with HIV. We don’t want this to be stigmatised but if it’s popping up among gay men (or any other particular group of people) quite a lot so it would be good for people to get protection for it. It’s really hard to encourage treatment whilst minimising stigmatisation of certain groups.

It’s telling that people wanted to engage with scientific responses first, not the social ones. You can’t ignore the biological side of things. Risk factors are scientific but are caused by social factors. There were people arguing that German people were genetically better equipped to deal with covid in the media. One thing that happened at the start of covid was a lot of bad theorising with limited information. It transpires that observations were explained mostly by the NHS being underfunded and also behaivourally in Germany people seem to be more community focussed, potentially leading to better practises to prevent the spread of covid.

Looking at covid deaths, the exposure was the issue, but for some reason the first thing people/media latched onto was “race”, without any thought about why disproportionate numbers of people of some races were more exposed to covid?

How people are racialised by society is distinct from their identity. Most ethnicity data that is self-reported is probably based on their own experience of how society has racialised them throughout their lives. But NHS ethnicity data is not self-reported at the moment. And it’s very difficult to change once someone else has recorded for you. Should UK also record class?

Census has evolved over time, but still requires people to sort themselves into boxes defined by someone else. We should make sure a self-reported ethnicity option is available alongside the prescribed boxes! In healthcare the second most popular category is “not known” (after “white british”). Need to break down ethnicity to include things like background, individual history etc, and not just base it off a superficial judgement of skin colour. Scapegoating of minorities is such a fundamental part of British politics/media that it’s difficult to tackle these issues properly.

How do we actually measure and record factors such as race?#

“How do I identify” is a pretty good measure. How do you see yourself? It depends on what you want. Do you want a 23andme type of thing which is genetic based? Or is it based on appearance on social factors? Should we establish what are the risk factors for illnesses and then just check for those? If I’m checking for covid, maybe I just need to know “are all your friends vaccinated?”. Whilst that may be affected by social factors, why not go straight to the source. What do I mean by this when I ask by race? What do I want when I want this information? In UK Census there was a weird thing where you could be White English but only Black British.

By doing categorisation, you sort of make it a thing when it is less of a thing. Is it reification or objectification perhaps? Intention is important. Why are you gathering this data? People tend to be more forgiving for healthcare though.

Some suprise that race really isn’t a genetic construct whilst gender and sex do have guidelines: SAGER Guidelines. These aren’t perfect, but does the same thing even exist for race and ancestry?

Culture shock difference of “treating everyone the same” versus appreciating and understanding difference. Equality versus equity. Can be difficult when you don’t always get to choose what data is available. The data that is available is also tied to cultural norms, shame, distrust - and therefore sometimes difficult to account for. How someone reports their own race may be different to how they are perceived by staff and also different to how well they are represented by healthcare research.

Case Study: Data Collection Practises in the EU (in particular, Sweden)#

In Sweden, no ethnicity grouping is allowed to be recorded. The intention is to prevent biases being built into the system.

In a more comprehensive Analysis and comparative review of equality data collection practises in the European Union, this is noted in context amongst other country:

“Sweden: Race is not used as the term is considered to be closely linked to racism. Ethnic origin (etniskt ursprung) is more accepted, used predominantly within the anti-discrimination area. The politically correct term used in public documents is foreign origin as opposed to Swedish origin. These two terms are officially defined by Statistics Sweden as follows: Persons of Swedish origin are persons born in Sweden who’s both parents were born in Sweden or persons born in Sweden who’s one parent was born in Sweden and one parent was born outside of Sweden. Persons of foreign origin are persons born outside of Sweden or persons born in Sweden who’s both parents were born outside of Sweden.”

Finland agonised over inclusion and equality and talk about it openly, and have had great success. The UK has been really bad with this. UK doesn’t collect data for White Other, for example. This needs to be intersectional - look at ‘black women’ vs ‘asian women’ for example.

Saini also refers to the overlapping of different inequalities and privileges in different contexts. How does this intersectionality of identities impact the way we do or should analyse data about demographics?#

Intersectionality is important, but the more intersectional a group you want to look at, the more difficult the data is to get (especially in big enough numbers to apply machine learning to).

We touched on this earlier, but splitting categories seems sensible (although numbers get small in increasingly narrow categories). Ethnicity recording across the National Health Service (NHS) has improved dramatically over the past decade. This study profiles the completeness, consistency and representativeness of routinely collected ethnicity data in both primary care and hospital settings. The second greatest category in ethnicity recorded is second after White British. The question is if this group is still predominantely White British or not.

As we mentioned earlier, perhaps the most useful thing to do is just to ask people what you need to know for each context at a time. If we need to know if someone is privileged or disadvantaged in a particular context regarding a particular characteristic, we should probably just ask about that instead of trying to find a catch-all measure that will never work all of the time. That way we can gather information that is relevant to what we need to know, and not rely on proxy values. After all, humans are not one-dimensional, and our identities aren’t either!#


Name, Role, Affiliation, Where to find you, Emoji to describe your day

  • Natalie Zelenka, Data Scientist, University of Bristol, NatalieZelenka, @NatZelenka

  • Nina Di Cara, Research Associate, University of Bristol, ninadicara, @ninadicara

  • Huw Day, PhDoer, University of Bristol, @disco_huw

  • Euan Bennet, Lecturer, University of Glasgow, @DrEuanBennet

  • Zoë Turner, Data Scientist, Nottinghamshire Healthcare NHS Foundation Trust, [@Letxuga007]

  • Amy Joint, Content Acquisition Manager, F1000Research [@amyjointsci]

  • Kamilla Wells, Citizen Developer, Australian Public Service, Brisbane