Comment Blog 7 March, 2019

How A.I. could save your sight

Hospital eye care in England faces a crisis as demand for care outstrips capacity. Ophthalmology is now the biggest outpatient speciality in the UK, and we’ve recently seen alarming signs of backlogs causing unnecessary sight loss. The Royal College of Ophthalmologists has confirmed this is a nationwide problem. One study found that as many as 22 people a month in the UK are losing vision because of delayed appointments.

This problem will only grow as the ageing population suffers common problems like cataracts, and sight-threatening conditions such as glaucoma and age-related macular degeneration (AMD), on a greater scale. The Royal College says AMD cases will grow by 60 per cent over the next 20 years.

There is no sign of hospital eye service capacity growing to meet this demand. But there is good news on the horizon – it’s clear that artificial intelligence (AI) will have a massive role to play in future healthcare services, as the Topol Review recently highlighted. An ongoing partnership between Moorfields Eye Hospital and DeepMind Health has shown that AI can help people to diagnose eye disease as accurately as world-leading experts.

But to make a real difference, AI will need to be deployed in a way that makes efficient use of the whole eye healthcare workforce. Putting new AI tools in the hands of overstretched hospital consultants will bring some benefits – but there is a much bigger prize to be had.

Primary eye healthcare providers in the UK – the optical practices familiar on every high street which offer NHS and private sight tests – can play a major role in taking pressure off secondary eye care. Optometrists working in community practice are a skilled workforce, trained to detect eye diseases, and increasingly treating those diseases under NHS extended services contracts. And there are around ten registered optometrists for every ophthalmologist in the UK.

The recently published NHS handbook on transforming elective ophthalmology services shows the scope to grow capacity by moving eye health services into the community, and enabling eye care providers to work better together. But the success of this will depend on community optometrists making consistently excellent clinical decisions. Since most patients seen in high street optical practices have healthy eyes, it can be difficult for community optometrists to develop deep experience of the more complex eye pathologies.

This is where AI could be a game-changer. A 2018 study demonstrated that the DeepMind model enables ophthalmology consultants to diagnose patients more accurately than when relying only on traditional tools, such as OCT, fundus exams and notes. It also indicated that if optometrists equipped with AI would obtain the same increase in diagnostic efficacy, they could routinely deliver specialist-level diagnoses in a primary care environment.

That could enable the large, widely distributed optometric workforce to play a much wider role in delivering eye healthcare, reducing pressure on hospital eye services. To achieve that goal, policymakers will need to lay the foundations by:

  • Building the use of AI into optometrists’ training and professional development
  • Driving the design and commissioning of eye healthcare services in which community optometrists routinely use AI to make quick and accurate diagnoses and referrals.

A change on this scale will need a real willingness to work across organisational and professional boundaries for the good of patients. We are confident it can be done, and the dividends would be immense.

Dr Peter Hampson is Clinical Director at the Association of Optometrists.