From Bias to Balance: How to make your digital health solution truly inclusive.
We know there are benefits to the use of digital and technology solutions within the NHS and other healthcare systems in the UK which provide more flexibility for patients and staff, generate greater efficiencies, enable care to be delivered closer to home and enable earlier disease diagnosis[1]. But there are well-reported challenges around digital exclusion for patients to access healthcare when inclusion hasn't been considered robustly. As a consequence there is the potential to worsen existing health inequalities for certain populations based upon ethnicity, gender, age, disability, sexual orientation and social-economic factors.
Alongside there is increased focus on the risks of racial and gender biases being built into medical devices and artificial intelligence (AI) tools. For example, using unrepresentative datasets or clinical diagnosis guidelines that potentially discriminate based upon ethnicity.
The UK's Medicines & Healthcare product Regulatory Agency (MHRA) and other regulatory authorities are starting to focus on their approval processes to request evidence that any digital or technology solution is not worsening these historic biases thereby increasing inequity and inequality within healthcare provision.
If you don’t know where to start don’t worry I’m here to provide guidance through my 10 point plan. This plan will help you to both reduce the risk that your solution increases health inequalities, whilst potentially assist you to demonstrate how your solution could actually improve access and outcomes for groups of patients who face greater health inequalities.
Build co-design activities at the design stage (preferably) by actively involve all people who will be impacted by your solution. This protects against poor design by gathering the experiences and opinions of diverse end users. Also look at using creative ways to meaningfully involve them.
Follow the NHS England’s inclusive digital healthcare framework and make your solution simple and intuitive to use for ALL users. For example, simplify the login process whilst meeting data security requirements. Build in personalisation (eg. push notification timings, display settings to magnify text), for hearing impaired people use captions for video content and transcripts for audio content.
Design any clinical trials or research / feasibility studies with activities to widen participant recruitment and thereby ensure a diverse range of participants based upon race, gender, age etc. This is particularly important if your solution is relevant for a disease or condition where there is a higher prevalence amongst certain ethnic groups e.g. diabetes and South Asian / African / Caribbean. Use innovative methods to recruit participants e.g. financial incentives, working with specialist agencies (e.g. People for Research, and promoting via social media targeted campaigns.
With any co-design, research or user testing make sure you record and analyse the demographics of participants, as a minimum ask participants to self-disclose their age, gender, ethnicity, sex and disability status (explain why you are collecting this information).
At the testing stage involve representative users including neuro-divergent individuals but don’t assume that the needs of people with dyslexia, autism or ADHD are all the same, they are not a homoegenous. With both beta and alpha testing, involve those with self-identified lower digital skills and health literacy ( you could commission local voluntary organisations to find these testers). Compare performance across protected groups e.g. age, sex, race and socioeconomic status. Make sure you place equal weight to the feedback received from ALL end-users – place value and respect to recognise their opinions.
Look at creative ways to reduce digital exclusion and digital poverty at the deployment stage. If access to devices and / or a stable internet connection is an issue look at ways to loan devices and / or wifi (or signpost to free wifi). Make sure your solution doesn’t require large data usage for the user e.g. background updates and devise strategies for data usage optimisation.
Through co-design of the deployment stage with your healthcare customers follow a two-tiered strategy to ensure there are always alternative pathways or access routes (e.g. face to face, telephone etc) for those who can’t or won’t access their healthcare care through the use of your solution.
As some individuals don’t trust the sharing and protection of their health data outside of NHS organisations, clearly articulate your privacy policy in all communication with target end users to reduce their barriers to using your solution.
Provide written guidance, training and onboarding in plain English with step-by-step instructions to use your innovation (for both staff and patients) with support via email, chatbot or telephone. Offer training online in real time or in-person to allow for questions and this offer should be continuous not a ‘one-off’. Give contact details of local digital skills training courses or digital champion programmes.
Communication during deployment is key to successfully addressing digital exclusion. Use representative images within any marketing material and display in places that your target group will go to e.g. library, GP surgeries, places of worship, shopping centres, pharmacies etc. Look whether marketing materials need translating into local languages and avoid technical jargon. Emphasise the benefits of the digital service for individuals and give examples where your solution has helped similar individuals. Display how users can easily feedback any issues or concerns.
If during your design and testing stages you haven't undertaken this actions, don't worry it is not too late as you can build them into future technical iterations, ongoing testing and new adoption of your solution.
You may be thinking that all these activities will add unjustified costs and resources but if you are targeting NHS customers, they are increasingly prioritising programmes that reduce health inequalities and digital exclusion in their purchasing decisions. You will also be future proofing against requirements by healthcare regulatory bodies to generate evidence that provides assurance that health inequalities will be reduced or mitigated against. Below are links to useful resources.
Useful Resources and examples
This research provides useful tips for the implementation of medical AI systems, to reduce potential biases within these systems.
Information Commissioner's Office - guidance on AI and data protection
MHRA recommends if ‘AI as medical device’ to use ISO/IEC TR 24027- Information technology — Artificial intelligence (AI) — Bias in AI systems and AI aided decision making as a quality assurance accreditation.
King's Fund - Moving from exclusion to inclusion in digital health and care.
Digital apps and reducing ethnic health inequalities - NHS Race & Health Observatory
Evidence of disruptive innovation and health inequalities
Health Equity Assessment Tool (HEAT) - Office for Health Improvement & Disparities
NICE Equality Scheme - EIA for digitally enabled therapies for adults with anxiety disorders
Accessible Information Standard required for all those providing NHS and social care services
National Voices Addressing inequalities in clinical trials
Historic inequalities within diagnosis of chronic kidney disease.