Skip to main content

Developing an EDI strategy for a research funding application

Here you will find key questions to consider when developing an EDI strategy for your health and social care research funding application.

Population – Who is this intervention seeking to benefit?

Are there any obvious types of patients that should feature in this application as specific groups which ought to benefit from the research, ie, elderly, homeless, particular ethnic or cultural groups – any groups with protected characteristics or characteristics that are not protected by law – what do we know about this population.

Example: A cardiac trial might need to consider that there are considerable numbers of smokers who are in the population who could benefit. The background needs to describe the population characteristics.

Does this population (or groups of people) have a charity, social club, pressure group that might be useful to include in a strategic manner while developing this application. Is there the potential to include someone right from the beginning who works with that population? This might be separate from the PPI as this might be another health professional or this might be part of the PPI strategy.

Does the PPI group include people with backgrounds that are relevant here – are all affected populations represented.

Example: If you are looking at those with diabetes you must acknowledge that rates are higher in those from Southeast Asia and seek to have representation from an organisation that supports people from Southeast Asia and a PPI group that contains people from South East Asia.

Intervention – Will this intervention work the same for all these patient groups?

Will you need to do some development, refinement or validation work with the intervention to make sure you have a version suitable for all the relevant populations.

If you need to validate it for different populations that will need separate sample size calculations – you may need to over sample from those key populations.

Examples: Dietary interventions for diabetes will need to reflect more than one cultural perspective. Interventions which produce genetic risk scores that say they are only validated on white populations will need validating on those who are not white.

Accessing the patient group – Does access to standard care vary for these groups?

How will you ensure that you can recruit the population who need the intervention if they do not normally turn up to receive standard care.

What additional measures could you put in place to recruit from the hard-to-reach groups.

Example: Consider going to social clubs and cultural centres to tell people about the study, consider local radio, social media, posters, etc. Include GP practices from areas of need. Put in extra resources where they are needed most.

Outcome measures – Consider access issues for your populations of interest.

Translations, transport, child-care, parking, taxi’s, peer support to attend appointments, payments for completed questionnaires etc.

Consider relevance for your population groups, cultural issues such as barriers to uptake and who the decision makers are and how to collect confidential information about appearance and function, for example, when the family might be doing the translating.

Talk to the PPI group about how to get unbiased data collection of relevant outcomes for example – not to use words which have a different meaning for some people, or which don’t translate well – there is no word for cope in Spanish – there is no recognition of the phrase “life is not worth living” in some cultures

PPI – Have you involved a diverse group of people?

Have you ensured that additional resources are available where necessary to ensure inclusivity and that opinions are sought from all those groups the clinical research seeks to benefit?

Do you need a separate PPI group from populations in which you have tried to over sample or validate the intervention – consider allowing a PPI group just for the hard to reach population if the research is all about increasing uptake from that population.

Dissemination – Will this get to the people who need it most?

Will this help “level up” to reduce inequalities or do the opposite? You can reduce inequalities by making the healthy unhealthy as well as making the unhealthy healthy but designing something that only makes the healthy even more healthy by being elitist in the dissemination is not the answer to fixing inequalities.

Is the PPI group involved in helping you share the results. Are you producing separate lay summaries for those who need it most to use on websites, or as short videos. Are you putting things on health talk online from a diverse group of individuals. If you have patient facing materials you need to make available to all patients, how are you going to get them translated or accessed for those who need them most. Each case will be different – think outside the box and use the PPI to support this.

Aim of your research.

*Is your aim specifically about reducing inequalities? – if yes – good! *

Example: Older Asian women do not attend exercise classes for stroke rehabilitation. We wish to develop patient facing materials for this group of patients to make it easier for them to access the same quality of care.

*Or is it effectively going to increase inequalities? *

_Example: “I want to see who would benefit the most from this very costly intervention and who benefits the least?” Question “why? If the answer is so that I can offer it to the right people – no – this will mean that those who are not doing well will get less. This will most probably increase inequalities. If the answer is “so that I know who needs more support” – then yes – this will reduce inequalities! It’s the same statistical analysis but the justification is different.

Is your aim about screening and diagnosis?

Example: Can you use the same artificial intelligence algorithm to detect skin cancer on people with white skins as you can on people with dark skins. If your tool will only ever work on white middle class, middle aged, professionals, then this will make inequalities worse – we are not doing much of this research right now.

Is there a chance that you perpetuate a bias in the way disease is diagnosed?

Example: Doctors believe that disease X is more common in people with a particular protected characteristic. You want to use AI to find the cases sooner so that they can get treated faster. When you put this protected characteristic into the computer it just screens for all those who have that characteristic. Its becomes blind to any other symptom. The AI has become biased. To get round this you need to analyse blind to this protected characteristic or validate all the diagnoses and test the controls too.

Contact us

This guide was developed by the RDS South West. If you would like more help from the RDS South West with this or with your funding application in general then get in touch.