Ask users for ethnicity

[sarawilcox]

What

A pattern to ask users for their ethnic background. See GDS pattern: https://design-system.service.gov.uk/patterns/ethnic-group/

GOV.UK Design System Backlog issue

Why

In our work on the guidance on asking good questions for forms, we found that teams working in health need to know how best to ask for this information.

Anything else

Bear in mind the data set used in the NHS data dictionary: https://datadictionary.nhs.uk/data_elements/ethnic_category.html?hl=ethnic
Also consider SNOMED codes and how they map to the NHS data dictionary: https://phinvads.cdc.gov/vads/ViewCodeSystemConcept.action?oid=2.16.840.1.113883.6.96&code=372148003

See also ONS data categories: https://www.ons.gov.uk/methodology/classificationsandstandards/measuringequality/ethnicgroupnationalidentityandreligion

Related issues

• BAME, ethnic minorities, ethnic groups #295
• Ask users for equality information #423

[henocookie]

The current iteration of the ethnicity question on the NHS coronavirus status checker follows the GOV.UK Design System component guidance whilst matching the NHS Data Dictionary: https://citizen-survey-tool-prototype.herokuapp.com/sex-and-ethnicity/ethnic-group u: love p: nhs

[sarawilcox]

Note that we followed the GDS pattern, rather than the NHS data dictionary - as far as users are concerned. Behind the scenes, we mapped the GDS pattern to the NHS data dictionary categories.

The NHS data dictionary puts Chinese in other but our (limited) user research as part of our work on forms questions suggested that some people expect to find it under Asian, rather than under Other.

We added the words "South and East Asian" to the hint text on the home page for "Asian" as some other people of East Asian backgrounds weren't sure whether to count themselves as Asian - they saw "Asian" as more "South Asian". We haven't tested this further though.

We mapped the GDS categories to the NHS data dictionary categories as follows:

Asian or Asian British
Bangladeshi - K
Chinese - R
Indian - H
Pakistani - J
Another Asian background - L

Black, African, Black British or Caribbean
African - N
Caribbean - M
Another Black background - P

Mixed or multiple ethnic groups
Asian and White - F
Black African and White - E
Black Caribbean and White - D
Another Mixed background - G

White
British, English, Northern Irish, Scottish, or Welsh - A
Irish - B
Irish Traveller or Gypsy - T

Another ethnic group
Middle Eastern - S

All other options are ‘Other’ or ‘Prefer not to say’

We originally thought that it might be better to include Middle Eastern under Another ethnic group. (GDS includes Arab.) But in retrospect, I think we should make "Another ethnic group" a free text box only. I don't see why we would capture "Arab" only (in line with GDS) and, looking more closely at SNOMED, "Middle Eastern" doesn't include Arab, Iranian or Israeli. I would leave "Another ethnic group" open - for any ethnic groups not covered in the other screens. Another ethnic group as a whole would be S.

Also, if we're following the data dictionary set, we should probably allocate Irish Traveller to C but note that SNOMED (the more detailed list) would class it as T.

[sarawilcox]

Further reading:

• https://www.ons.gov.uk/methodology/classificationsandstandards/measuringequality/ethnicgroupnationalidentityandreligion#ethnic-group
• https://designnotes.blog.gov.uk/2019/01/29/researching-how-we-ask-users-about-their-ethnicity/
• https://design-system.service.gov.uk/patterns/ethnic-group/#research-on-this-pattern

[sarawilcox]

https://uxdesign.cc/rethinking-the-design-of-the-race-and-ethnicity-question-on-surveys-6f9066c69392 (Thanks @amyj2110)

[Laurenlukasiewicz]

Just sharing some work around the GDS ethnicity question......

The Sign up to be contacted about the coronavirus vaccine trials team have used the GDS ethnicity question in their service. Service users experience of using the question was researched in interviews and surveys during Jul and Aug 2020. There was also some analysis of data around how the question was answered in the live service.

Users did have issues with the question but this did not result in them to dropping out of the service., or feeling that they had given a greatly misrepresentative answer.

However the main issues reported were the following

• because the question was being asked in a health context (and in a pandemic context where it is unclear the role your genetic history could play when experiencing Coronavirus) users were concerned that they needed to be quite detailed in defining their ethnicity as they saw it as analogous to their genetic history and genetic history as analogous to nationality. So for example White users may want to put they are both White Irish and White British in the Sign Up service if that is their background. They think they need to provide a full genetic/nationality history of their family but the answer options don't allow for that. (There is also an argument that we could have managed better users understanding of how much detail they need to supply) These users reported that In more transactional services they would normally just put White British but here they assumed they had to detail everything in their family history.

• Similarly a user cannot indicate they are both Caribbean and African if they want to say that or more than one ethnicity within the Asian group (e.g. Indian and Pakistani, Indian and Bengali etc). They do not see the mixed group option as a way to express this, as they see the different ethnic backgrounds they want to select as part of one ethnic group.

• We also had users indicating that they wanted a Black British option within the Black ethnic group answer options.

This confluence page goes into more detail around our findings
https://nhsd-confluence.digital.nhs.uk/display/PTC/Ethnicity+question

[Tosin-Balogun]

This was ranked joint 5th from the in progress group at the content backlog prioritisation workshop held on 24-Nov-2020

[GrilloPress]

Observed a few users inputting their details to either sign up for a test and trace account or registering their test kits.

General response has been that the first screen, of the two-part ONS style pattern, seemed to raise no concerns [particularly as there is a prefer not to say] but the second part they find either redundant “you already know enough” or they find what they see themselves as is not present.

Other people from Test have reported:

• some users not being sure why they were being asked and having to interpret why
• Some users found it didn't include them [first or second screen]
• some users do not think it important to getting a test and "yet another question" in an already long journey

First question:

Variation of second question:

[henocookie]

@GrilloPress you mention for the second question that "what they see themselves as is not present". Did few (or no) users notice or use the "Another [X] background" option?

[amyj2110]

Contributed by @samanthasaw and Lucy Winder.

Screenshots taken from the organ donation service within the NHS App.

Research was carried out by the NHS app team to find out if changes to organ donation legislation were being communicated to users effectively. Users are asked to optionally provide details about their ethnicity at a point in the organ donation journey. This is in order to understand the make up of the NHS Organ Donor Register and will not be stored against a person's registration.

Only 1 out of the 9 participants understood why this information was being collected. None of the participants read the information saying that this question was optional.
Most participants assumed this was being asked in order to match you with an appropriate donor.

Full research findings can be found in this document:
Organ Donation - Findings April 2020.pptx

[GrilloPress]

@GrilloPress you mention for the second question that "what they see themselves as is not present". Did few (or no) users notice or use the "Another [X] background" option?

From my recollection they used prefer not to say instead

[sarawilcox]

The GDS team is retiring the pattern to 'ask users for ethnic groups'
In March 2021, they published a new pattern to ask users for equality information - based on data standards set by the Government Statistical Service (GSS).

To meet these standards, this new pattern replaced the pattern to ask users for ethnic groups.

[TJFDM]

Register with a GP has been using the following pattern for asking about ethnicity. I thought I would update with what our research has found. We have done 3 rounds of research with 10-12 participants each time, with each round having no more than 3 White-British participants.

Our service allows people to register online with a GP - often viewed as the 'gateway to the NHS'.

Round 1:

• Overall users understood the question that was being asked, felt represented in the options provided and that this was information that is important for a GP to know about you on registration
• 2 users preferred not to say: User 1 (Black African & Irish) preferred not to say as they had experience bias based on their ethnicity in the past. They believe that the NHS does not need to know what ethnicity they are "They're either going to provide care for me or they're not" User 2 (White British) felt it was something to discuss directly with the GP and not within an online form
• 1 user (Indian) felt confused by the term 'Asian' and that this was a very broad and vague name for the group. A person who is from Bangladesh is 'very different' to someone who is from China, but both are grouped under the same name. Didn't feel like it allowed them to accurately describe their particular ethnicity or allow the doctor to be able to pick up on certain illnesses that may affect Indians more than others.

Round 2:

• 5/12 users felt that the option provided were sufficient and the groupings meant that the page was no overlong and complicated
• People have many assumptions about why this information is collected, we need to better explain it
• 4/12 users told us that having the options for 'Other' and 'Prefer not to say' were good to see and made it feel more inclusive
• 1 user didn't feel represented in the options provided as they are from Cornwall which has recently received 'protected national minority' status as per EU ruling (https://www.cornwall.gov.uk/people-and-communities/equality-and-diversity/cornish-national-minority/) - so they can now identify themselves as Cornish as the Welsh, Scottish and Irish do. However, I believe this mostly translates to Nationality, and not Ethnicity. That being said, this user would not have answered the question as they did not feel represented in the options
• It was identified that this is a 'common question in forms' and is important information for the GP to have
• The placement of the question in the wider registration process is equally as important - in this round, users identified that the question was out of sync with the questions that they previously answering and going on to answer as it was in a wider set of 'health' questions and they believed it belonged more upfront in the process with demographical information questions like 'Name' and 'Address'

Round 3:

• Most people felt represented in the options
• The ability to select 'Other' or 'Prefer not to say' was again welcomed by users
• 1 user felt the term 'Ethnic group' was particularly nice, however the same users also didn't feel represented in the options (White Eastern European) and doesn't agree with labelling. They selected 'Other' and expected to be able to type something in there - the prototype did not allow this and I will check the pattern to see if it should. The same user also didn't feel that terms like 'White' and 'Black' should be included in the options here, as they pertain to race and not ethnicity
• 1 user (Asian) was pleased that 'White' was not the first option 'as it usually is'

[sarawilcox]

We've had some feedback about the BMI calculator on the NHS website which suggests that some people may see alphabetical ordering as discriminatory. For example, we have a small proportion of the population (the Bangladeshi community) at the top of the list on the BMI calculator with the larger White population at the bottom of the list. Esp if other elements of the form/calculator are not listed alphabetically, e.g. sex.

[sarawilcox]

This issue is an epic on the service manual team's roadmap.

[vanessapereira-nhs]

My team is working on the redesign of the BMI tool, and one of the questions we ask the user is their ethnicity.

Why do we ask for it
We need to know the user's ethnicity in order to give a more accurate result with the proper thresholds. This is because the BMI thresholds for someone from a ethnic minority (Asian, Black or Middle Eastern) are 2.5 lower for overweight, obese I, obese I and obese III.
For example, a white person with a BMI of 24.5 would be a healthy weight, but a person with an Asian, Black or Middle Eastern ethnic background would be in the overweight category.

What we tested
We tested 3 different versions of this question, and came to a decision based on 2 rounds of usability testing and multivariate testing.

How are we asking
The option that worked best follows the GDS pattern, but we added the "Middle Eastern" option since this is a recommendation from the NICE guidance on obesity (1.2.8). See screen attached below to have a more accurate picture of the pattern.

Evidence
The usage of radio buttons with those particular options to choose from, compared to the other options we tested, performed better in terms of sentiment, inclusivity and providing users with adequate options. The pattern was clear to the users and there wasn't any particular issue while choosing their ethnicity.

-- Screenshot:

[sarawilcox]

Note that GOV has now archived their old ethnicity pattern and instead includes the ethnicity question in their equality info pattern.

They say:

These patterns are based on the harmonised standards developed by the Government Statistical Service.

Collecting equality information in a consistent way across the public sector makes the data more useful. For example, an organisation can benchmark its own services against other public sector services or the population in general. And it can adjust its approach if it finds a particular group is under-represented.

Do not use this pattern to collect information for operational reasons - especially if you’re legally required to ask for the information in a particular way. For example, you’re asking about the user’s gender or sex to work out how much State Pension they’re entitled to

When we are collecting ethnicity data to improve access to services, we recommend using the GOV equality info pattern as it is.

When we are collecting ethnicity data for operational (or clinical) reasons, for example to give people health advice based on the ethnicity they enter, we diverge from the pattern where needed. See above example from the tools team. The team added Middle Eastern in order to give clinical advice in line with NICE guidelines.

[vanessapereira-nhs]

We have some updates related to this pattern mentioned on top.
As mentioned before, we took this design into multivariate testing and came across some new insights for this pattern:

• Overall, 98% of participants completed the ethnicity question by providing their ethnicity (i.e. selecting options other than 'prefer not to say' or 'other').

• 80% of participants felt the expander contained the right amount of information however 69% failed on a measure of comprehension of the link between BMI and ethnic background. This could be explained by the fact that only 1% of participants clicked the expander - highlighting that majority of users are missing out on this key information. As a results, recommendations were made to have some information pulled out from the expander to provide users with some context and having the expander solely for additional information if users need that information.

• We tested the wording "family background" vs "ethnic background" and the results were pretty similar, there was no difference in disclosure or sentiment between either. However using the word "ethnic" is closer to the Office of National Statistics recommendation and to the GOV.uk pattern.

If you want further information, you can go to the NHS.UK coversheet in the audit trail in SharePoint.

-- Updated screenshot:

[sarawilcox]

Further to Vanessa's comment above, here's some more info about the 2 other options that the Tools team tested alongside radios.

One was checkboxes - to see if it helped users of mixed ethnic background to select more than one box, but 1 in 4 users were confused by it.

The 3rd variation was a yes/no radio to a question about whether users belonged to the NICE-specified ethnic groups, which didn’t test well.