What will it take to make eye health innovation work for everyone?

New technology can reduce inequalities, or reinforce them.  Research needs to reflect the diversity of the people who will ultimately use new treatments and services. If some communities are missing from research and data, the technologies built from them may simply work better for some people than others.

That matters particularly for AI. Algorithms are only as good as the data they are developed and tested on. If some populations are poorly represented, the technology may perform less well for them.

Dr Shokat Adam MP, a practising ophthalmologist and Member of Parliament, highlighted the importance of thinking about who is represented in the data used to develop AI.

“If we have a wide core of people and the level of health inequality due to socioeconomic and health backgrounds,” he said, “we must ensure anything carried out represents all aspects of your community.”

Equity cannot be something we consider once an innovation is finished. It has to be designed in from the start.

Too often, lived experience is brought into a project once the important decisions have already been made.

It needs to happen much earlier.

Fola Tayo, a registered blind HR business partner and service designer with more than 10 years' experience in patient and public involvement, put it simply:

“People need to be part of the early planning.”

People living with sight loss can shape everything from the questions researchers ask to the way studies, treatments and services are designed. Their experience can reveal practical barriers or unintended consequences that might otherwise be missed.

Fola described the tendency for professionals to plan projects among themselves before deciding that patients need to be involved.

“Even if it's, say, an introduction stage to say we want to speak to someone about this. It's early on, we don't have all the answers, we want you to be part of it.”

A treatment can work clinically and still be difficult to access, impractical to use or poorly designed around people’s lives. Successful innovation has to work in the real world, not just in the lab.

AI is developing faster than many healthcare systems, regulators and workforces can adapt.

One of the biggest questions is what happens when something goes wrong. If an AI system contributes to a wrong diagnosis or decision, where does responsibility sit?

Dr Shokat Adam described this as a “responsibility vacuum”.

“If you have a practitioner, and if you have AI – and AI misdiagnoses something, who is responsible?” he asked.

There is also a risk of deskilling if technology starts to replace, rather than support, professional expertise.

Josef Huemer, a specialist and research fellow whose interests include medical retina imaging, AI and telemedicine, highlighted the need to think carefully about how AI is integrated into clinical training and practice.

“The more senior you are, the more you question it, the more reliant you are with this. That leads to another big problem… deskilling.”

The aim cannot simply be to automate more healthcare. Technology should strengthen expertise, improve decisions and ultimately deliver better care.

People cannot make informed decisions about technology they do not understand.

That means explaining clearly how AI and other technologies are being used, what they can and cannot do and what they could mean for someone’s care.

Fola described her experience of a project where patients were asked about AI without being properly taken through the process.

“If you are not explaining things to people, that's where the fear comes in.”

Building trust is not just about asking people whether they approve of change. It means giving them the information and opportunities they need to question it, shape it and influence how it is introduced.

As AI develops rapidly, taking the time to explain how it will affect people’s care will be an important part of making sure innovation is adopted responsibly.

Breakthroughs do not move from research into everyday healthcare by themselves.

Researchers, clinicians, charities, people with lived experience, industry, professional bodies, regulators and policymakers all have a role in turning promising ideas into practical change.

Josef highlighted the importance of bringing those different perspectives together, arguing that patient groups, diagnostic and pharmaceutical companies, AI technology organisations, representative bodies and policymakers all have a role to play.

“We need to talk about how we define grant applications, so they are as inclusive as we wish them to be,” he said.

That collaboration also needs to tackle the less glamorous parts of innovation: implementation, workforce, regulation, infrastructure and scale.

Connecting those different perspectives is why events like VisAbility matter. The challenge is not simply generating new ideas, but creating the conditions in which they can improve people’s lives.

There is no shortage of opportunity.

Existing infrastructure such as the UK’s diabetic eye screening programme could support new research at scale. AI could help us better understand eye conditions that have historically received too little attention. And making better use of the eye health workforce could help specialist services focus their expertise where it is most needed.

Josef pointed to diabetic retinal screening as a particularly important opportunity, highlighting the scale of the programme and the millions of images already being collected.

“We have the 12 million images that we take, and how can we use it, taking it from a diverse population to facilitate more research.”

Fola also highlighted the potential for AI to support research into conditions that may not result in blindness but can have a major impact on people’s lives.

And Dr Shokat Adam called for greater use of the workforce already present on the high street, potentially allowing ophthalmologists to focus on more complex cases and reducing pressure elsewhere in the system.

But every opportunity should come back to the same questions.

Does it solve a real problem? Has it been shaped and tested with the people who will use it? Will it improve access rather than reinforce existing inequalities?