Image: Jair Lazaro, Unsplash

Health Inequalities and Tech

Like other areas of life, research indicates links between inequality and poorer health outcomes for everyone, from life expectancy to infant mortality. And that isn’t the only inequality angle affecting health in this country: people have radically unequal outcomes depending on their socioeconomic background, race, gender, and which area of the country you’re seeking treatment in.

In 2010, Professor Michael Marmot published a landmark review of health inequality in the UK, called Fair Society, Healthy Lives. It found evidence of shocking inequality of health:

  • People living in the poorest neighbourhoods in England will on average die seven years earlier than people living in the richest neighbourhoods
  • People living in poorer areas not only die sooner, but spend more of their lives with disability – an average total difference of 17 years
  • The annual cost of health inequalities is between £36 billion to £40 billion through lost taxes, welfare payments and costs to the NHS

Ten years on, we found that this report’s recommendations had never really been acted on. But in that time, other, new problems have emerged.

That’s why we’re excited to launch a new report, Artificial Intelligence and Healthcare Inequalities.

Emerging technologies, especially AI, are rapidly being embraced as ways for thousands of minor or major decisions to be made across our institutions, analyse health data for us, or design personalised health plans. However, AI models that exist right now can only ever be mirrors to our society: they’re trained on the data and decisions we already make. This means they risk replicating the biases and inequities of our system. On top of this, the logic of AI-driven decisions can be harder for patients or workers using it to understand, with the source code often unavailable and impossible for non-experts to understand.

This report, created by researcher Daniel Guest, discusses the intersection of health inequalities in the NHS and growing deployments of AI, specifically around cardiovascular health, which remains the UK’s biggest cause of death globally. His findings are in three key areas: risks, opportunities, and questions to ask.

  • Risks: The report finds risks in the data used being unrepresentative historical data, and calls for clinical trials to be widened and representative. It also finds that cardiovascular disease is most attributed to socio-economic background, so AI must be implemented alongside measures to correct inequities in healthcare to avoid replicating existing structures.
  • Opportunities: In personalised risk management, telemedicine in underserved communities, and addressing bias in diagnostics, the report finds AI holds some opportunities to create a more equal health system.
  • Questions: The report finds that continuous questions must be asked by users of AI and by oversight mechanisms to ensure the technology is being used effectively and equitably.

Health Inequalities Webinar Recording

Event Summary

After introducing The Equality Trust, our host Priya Sahni-Nicholas set out the situation we face with health inequalities: a 16-year gap between the richest and poorest in her own borough caused by a wider unequal system: poor air, poor housing, poor conditions at work. This can have huge impacts, as our first guest, Dr Fran Darlington Pollock, speaks to. Healthcare itself can be unequal, with very different experiences for racialised people due to biased systems: biases that risk being replicated in emerging technologies like AI unless we’re careful, as explored by a new report from our second speaker, Daniel Guest. One way to improve this situation lies in democratising health policy through co-production, a topic discussed by our final guests, Hannah Turner-Uaandja and Sally Devine.

Dr Fran Darlington Pollock and Inequalities in Healthcare

Dr Pollock gave an explanation of William Beveridge’s 1942 blueprint for the welfare state and the five evils that Beveridge saw, including disease. The origins of the NHS give a lot of context to the inequalities of healthcare we experience now, as cited in a 1969 quote from Johan Galtung : “If a person died from tuberculosis in the 18th century it would be hard to conceive of this as violence since it might have been quite unavoidable, but if he dies from it today, despite all the medical resources in the world, then violence is present”. Health is much more than just freedom from disease, Dr Pollock argued, and this is important because it gives us space to look critically at the healthcare system Beveridge proposed relative to what we might propose, if we looked at health as including all the social aspects, like where you’re living, your sense of security and stress, and other things that dictate our wellbeing. Traditional healthcare systems prioritise diagnosing disease and injuries at the expense of wider experiences: in order to properly understand inequalities in healthcare, we need to consider everything.

Dr Pollock demonstrated the huge health advances for life expectancy and infant mortality in the years following the establishment of the welfare state, followed by stagnation or even a worsening of these metrics from 2014. This, she pointed out, is a canary in the coal mine for a wider systemic crisis and entirely avoidable. She compared this to work done in the 1990s by Dr Singer in Connecticut; Singer spent time looking at poor and high crime areas, and argued that the reason Puerto Rican migrants were suffering more than their White or Black neighbours in was because they had a combined social, economic, environmental, and political burden, and similar experiences were had in the UK during Covid-19. This can only be understood by looking at social models of health: the structural violence that harm communities in unequal countries.

She then turned to the cost-of-living crisis’ unequal impact. Much of the welfare state had been unpicked by the time the pandemic and cost-of-living crisis hit, meaning that Covid hit communities made more vulnerable by noxious social conditions. Discussions about individuals or groups with underlying health conditions missed the fact reason that underlying conditions were so unevenly distributed came down to policy choices and political inequality. This, she argued, is one of the reasons that the model outlined by Beveridge has struggled to keep pace with societal transformation. It wasn’t designed as a system for persistent inequality, diverse populations, or marketisation. Responsibility for poor health has been placed on individuals, rather than the systems that prevent you accessing healthy nutritious food or make you iller earlier in life, reducing your income and eventual pension.

The solution, she said, is to create a new healthcare system that (as well as being properly resourced) recognises structural violence and values care.

Daniel Guest and Artificial Intelligence

Daniel explained that he wanted to create a baseline situational analysis into the use of artificial intelligence in healthcare: a piece of work that examines the situation and potential, rather than making specific recommendations. The NHS is investing £21million in rolling out AI technologies across 64 Trusts, so it’s important to understand how this tool could entrench or combat inequalities. Within that, he decided to focus on cardiovascular disease, which remain the biggest cause of death and have higher than average levels of diagnostic errors. On top of that, there’s a high level of inequality within cardiovascular health: women remain under-represented in cardiovascular research through implicit sex bias, which shape gender biases and cause women to be under-medicated and more likely to be misdiagnosed. AI is also being used widely within cardiovascular disease, with a focus on risk detection.

There are some areas where technology is being used to reduce inequality in healthcare, such as telemedicine or wearables, but other areas have mixed results. An algorithm to detect health risks used on over 200 million patients in the US, for example, was found to demonstrate racial bias because it relied on faulty metrics. From this, we can learn that we need to have strong regulatory oversight of AI use, as well as to ask which tasks are the right ones for AI to take on. Overall, it’s clear that we need to make sure that programmes aimed at tackling health inequalities can’t be simply replaced with AI deployments, which could exacerbate the issues.

Hannah Turner-Uaandja and Sally Devine on Co-Production in Healthcare

Hannah and Sally introduced the organisations they’re working with at the moment, Vocal and Healthy Me Health Communities, both of which are aimed at reducing health inequalities by strengthening community involvement in the design and delivery of health services. For this particular project, they worked with residents in Gorton, Greater Manchester, to leaned what residents felt was most important for their health and make sure other organisations could use this information to inform how they designed their projects.

They way they did this was to bring clinicians, patient, and carers together to identify and prioritise their needs. Traditionally, that’s in one research area, but they decided to broaden the conversation out. This research was designed to make sure it was done with the public, not to the public. They’ve found that doing the work this way meant that residents shared important points and priorities that didn’t fit into the boundaries of pre-defined projects, which was frustrating for all involved. They wanted to make sure people were able to make these points and see that that have real influence over the work; the community also gave a lot of advice themselves on how to best engage with the community.

The way they did this is to build a steering group that identified partner organisations and locations for outreach. These outreach locations were set in places that had high footfall and community value, like local mosques, a local exercise group, and two food projects. Community representatives themselves identified themes to priorities and sorted them into buckets, and then the community voted on their priorities based on these themes in some of the outreach locations. Finally, they held a consensus-building workshop, led by community members, which finalised the top ten priorities. Community members were involved at every step of the process and had real influence over the results, which Hannah and Sally found a huge difference; the community members were able to contextualise or debate issues together in ways that shed a lot of light onto the issues.


  • Dr Pollock mentioned that we need a “revolution” in healthcare – how close are we to that revolution? And what impacts might we see in the short term from the cost-of-living crisis?
    Dr Pollock answered that we started to see the impact of austerity, particularly on women, in only a few years, so we’ll already be seeing the impact of the cost-of-living crisis. In rough sleeping, for example, we’re already seeing a big increase – things play out in real time with an accumulated lag, so things will get worse. But, she said, the revolution in healthcare appears to be quite close, particularly if communities facilitate this change. We’re all complicit in structural violence and our huge collective efforts to make things better haven’t managed to solve things yet; “Build Back Better” was only advocating for a stronger version of the status quo. We need to examine everything for inequality and create a structural change. Even Beveridge was racist and sexist, perpetuating unequal structures in his model. We need to put communities and co-production at the centre to avoid this in future.
  • In Daniel’s presentation, he talked about biased data and machine learning. How could this be prevented? Is there unbiased data we could be using? And are there examples of AI working for good?
    There’s definitely an element of commercial interest in the diagnostic fund, but, says Dan, there is an incredible opportunity to co-produce a plan for AI and data that works for everyone in society. Unfortunately, there’s no examples of unbiased data yet, but we need to move to a much more holistic view for mitigating bias in data sets and for health inequalities more generally. Driving inequality out of AI comes with driving inequality out of healthcare overall; AI could exacerbate problems or accelerate the solution.
  • Hannah and Sally talked about co-production being used more widely; if we wanted to adopt this, what would our next steps be? Have you seen the impact of your priority setting work?
    In terms of the effects it’s having, said Sally, we’re looking at ways to share our findings with partners and funders; some grassroots organisations that were also involved have been using the findings in their own work. Other research networks are developing training based on this process for their engagement teams – but at the time of the webinar, they’d only finished their project 14 days ago, so it was still very new! If people are interested in taking a similar approach, there’s lots of different ways they can democratise their approach, from citizens juries to deliberative democracy.
  • Ageism is another key source of health inequalities, from self-ageism to ageist practices. How can we address this?
    Dr Pollock said that Beveridge’s report talks about age from the point of view that all people get older and that needs resourcing, but stops there. There’s still a lot of ridicule for older people, and often work on their health considers them only as a homogenous group that is dependent on an allocated resource. Ageing is a fully intersectional experience with a lot of cumulative impacts from different socioeconomic situations; people who struggle are much more likely to say they’ve experienced ageism than wealthier counterparts. The ageing population is not being fully recognised by governments.
  • We have a general election coming up. What should we be asking of incoming governments?
    • Dr Pollock: We need to completely reform health and social care to fit the population we have and are expecting to have. We need active investment in the health and social care workforce; recruiting overseas cannot solve the shortage of workers. Politicians are working on a very short-term cycle, rather than the lifespan of the population; we need to call for policy to be based on the long term and hold politicians to account for that.
    • Daniel: I agree completely with Dr Pollock; the NHS is very based on five-year cycles and elections, which does have a big impact on how we operate.
    • Sally: Working in our area with a high level of deprivation, you can feel how people are struggling and the third sector is trying to fill the gaps. We’re very stretched, and this is an unsustainable model.
    • Hannah: We need to move away from the model that thinks about individuals who need blaming for life choices and towards empathetic, long-term services that can address the complexity of peoples’ needs. One of the things that can up the most in their community work was a particular 30-year GP who’d recently retired had had a big impact on many peoples’ lives and health. This really shows the power of a long-term relationship with services that care, and that can only happen with more funding.