Photo by Hush Naidoo on Unsplash

Sharing eye health data: Putting patients at the heart of decision-making

Thu Aug 20, 2020
$download_content = get_field('download_content');

We’ve been involved in a project to put patients at the heart of data sharing for eye health, and helping to increase knowledge around the leading cause of blindness in the UK. Researcher and consultant Walter Pasquarelli explains how we’re doing this

We’ve been involved in a project to put patients at the heart of data sharing for eye health, and helping to increase knowledge around the leading cause of blindness in the UK. In the second of four blog posts, Researcher and consultant Walter Pasquarelli explains how we’re doing this

Last autumn, the ODI –as part of its partnership with the eye health charity Action Against Age-related Macular Degeneration – set out to support the INSIGHT Health Data Research Hub to put patients and the public at the heart of data sharing. 

The NHS says that it’s committed to involving people in decisions that affect them. From front-line service commissioning to health data collection and research, a wide range of governance structures and processes exist to define the balance of power in decision-making around health. 

But researchers report that there’s a lack of access to data may restrict the opportunity to progress research that may benefit patients and the public, and across the NHS and other health systems, ophthalmic data is held by a very large number of different data controllers, with no single, streamlined application process.

Improving data for eye health research

One of the most powerful tools for the diagnosis and monitoring of sight-threatening diseases is a form of eye scan, called optical coherence tomography (OCT). These scans are now standard in hospital eye services – such as Moorfields Eye Hospital and University Hospitals Birmingham – and are routinely taken on most patients with diseases affecting the retina (the light sensitive layer at the back of the eye).

OCT scans only take a few minutes to do and provide very high resolution images – down to a few thousandths of a millimetre – and so can be used to detect the earliest signs of disease, such as age-related macular degeneration (AMD), the leading cause of blindness in the UK.

OCT scans can be combined with artificial intelligence (AI) to provide rapid, automatic detection, in some cases comparable with human experts

This is probably just the beginning. Researchers around the world are showing how OCT scans can be combined with artificial intelligence (AI) to provide rapid, automatic detection, in some cases comparable with human experts. In the near future these kinds of innovative techniques will be important to help the increasing number of people affected with these diseases which cannot be covered through traditional capacity alone.

In a study using anonymised images, Moorfields Eye Hospital worked with Google Health to develop AI-enabled diagnostics across a wide range of retinal conditions. Building on this work, Moorfields has now shared approximately one million anonymised OCT scans to investigate how AI can predict the course of age-related macular degeneration and help inform when they may need to be seen. Generally, however, researchers report that it is difficult to access such datasets, and this raises the concern that a lack of access to data may restrict the opportunity to progress research that may benefit patients and the public. 

Across the NHS and other health systems, ophthalmic data is held by a very large number of different data controllers (usually NHS hospital eye services) and there is no single, streamlined application process.

INSIGHT – the UK’s Health Data Research Hub for Eye Health – is part of Health Data Research UK, and has the mission of using data to ‘enable discoveries that improve people’s lives’. The Hub has a particular focus on ophthalmic imaging, and brings these scans and related data together in a systematic way, with robust processes of data curation, assessment of data quality and governance.

Moorfields has now shared approximately one million anonymised OCT scans to investigate how AI can predict the course of age-related macular degeneration and help inform when they may need to be seen

Through INSIGHT, academic, industry and NHS researchers and organisations will be able to apply to access anonymised collections of OCT scans and related data from across Moorfields Eye Hospital and University Hospitals Birmingham. In the long-term the ambition is to build a platform able to include other data controllers’ datasets, enabling Hospital Eye Services from anywhere in the UK to participate. This will allow for diverse datasets to be used to unlock new insights into retinal and other diseases.

Involving patients and the public

There’s a growing body of research to show that people feel strongly about who should have access to data about their health and for what purposes. For example, a study by Genomics England showed that patients and the public are generally comfortable when data that relates to them is used for research purposes and by charitable organisations. Pharmaceutical and other for-profit companies are eyed with a degree of suspicion, but there seems to be cautious acceptance so long as the work is focused on generating benefit to the public.

Ophthalmic data is held by a very large number of different data controllers

As well as understanding the circumstances in which people tend to be comfortable with sharing patient data, we’re interested in how they want to be involved in the process.

In Foundations of Fairness: where next for NHS health data partnerships’, published in March 2020 Understanding Patient Data found that patients want:

  • fairness, transparency and accountability. While people care deeply about NHS data and how it’s used, it can be unclear to the public how decisions are made. Decisions about how data is used should be subject to transparent process and external oversight.
  • to be involved in decision-making. Deliberative participation methods can help embed citizens and patients into the process. Involving people early on ensures decision-making is informed by public views, values, concerns and expectations.

As we discussed in our first blog in this series, there are a number of arrangements in the UK health data research space that involve patients and the public, and enable them to share their views.

It was in this spirit that the organisations behind the INSIGHT Hub included a range of plans for public and patient involvement and engagement (PPIE) in their initial bid to Health Data Research UK. Their plans included a mix of different deliberative methods, such as making INSIGHT’s work visible in the hospitals involved, and running workshops and events for patients, the public and other stakeholders to share their views. 

In addition, they proposed the introduction of an innovative governance structure to include the public, patients, and other stakeholders, to be led by INSIGHT’s charity partner Action Against Age Related Macular Degeneration (AAAMD). The essence of this proposal was based on a previous AAAMD-ODI project exploring what a transparent and legitimate construct for stewarding eye scans and related data would look like.

The ODI was well-placed to help AAAMD and INSIGHT develop and implement their vision, given our interest and expertise in the concept of data stewardship – collecting, using and sharing data – and involving people in it who may not otherwise be included. At the ODI, we also wanted to test our ideas around ‘data trusts’ – which we described at the time as involving one party authorising another to make decisions about data for the benefit of a wider group of stakeholders. 

Initial steps

There were some important decisions made in the first few months following the INSIGHT Hub being announced and work started in October 2019. 

Firstly, while the group of public, patient and other stakeholders that we would help to convene would be responsible for providing input into and oversight of the Hub’s stewardship of the OCT scans and ophthalmic health data, the hospitals involved – the data controllers – remained legally accountable for the data. The role of the group would therefore be advisory during the first stages of establishing the INSIGHT.

In practice, requests to access the data would be assessed to ensure that these would be legally compliant and feasible from a technical perspective by appropriate INSIGHT staff under strict operational rules. This meant that, for now, we wouldn’t encounter some of the challenges we’d anticipated if the group was to take on full responsibility and decision-making power – these included, for instance, issues related to legal liability over the scans and data.

It was striking that the data controllers involved in INSIGHT were themselves keen to learn from this process, and asked that the group would regularly give feedback to them so that the NHS Trusts could share in the learnings from this work

Secondly, although the set up of the group would be informed by the similar groups and processes we’d come across in our research, we’d try to give this group an even stronger influence over the hub’s governance. We’d do this by asking it to produce its own criteria for assessing applications to access INSIGHT data, and we’d also aim to address what we felt were shortcomings across other groups – such as a lack of transparency around their operating terms and decisions. It was striking that the data controllers involved in INSIGHT were themselves keen to learn from this process, and asked that the group would regularly give feedback to them so that the NHS Trusts could share in the learnings from this work.

Lastly, given that the group would be a central part of the Hub’s plans to empower the public and patients’ into governance, we recognised the need to work closely with the hospitals’ existing PPIE specialists. Conceptually, we’d see our work alongside AAAMD as supporting direct input and oversight over requests for data access , as well as providing a bridge to  the hub’s other involvement and engagement activities and communities (which provide broader understanding of stakeholder attitudes and expectations). In practice, we’d work closely with the PPIE teams and seek their input and advice, and further down the line, use their existing channels to recruit and convene members of the group.

In our next post, we’ll describe some of the implications of these decisions, as well as talk more about the term ‘data trust’ and its relevance here.

Read more about this project