Machine Learning: Patient Groups Evaluate

How can patient groups evaluate machine learning?

Let’s begin with machine learning. It touches all aspects of our lives. It is how Amazon, eBay, Google and so on `learn’ how to recommend different products. This is based on websites we visit or what we buy online. Following on, machine learning is becoming embedded in all aspects of healthcare, from clinical research, through to diagnostics and personalising treatment and management.

Patient groups are increasingly being invited to give their views on aspects of machine learning, for example:

  • The British Standards Institution (BSI) and Nesta ran a recent workshop. Its aim was to explore the feasibility of introducing standards or guidance for patient-centred machine learning in healthcare.
  • Two of the UK’s leading organisations for patients with sight impairment, the RNIB and The Macular Society, have supported Google’s DeepMind machine learning research project at the Moorfields Eye Hospital.
Machine learning: Hidden in plain sight

Machine learning touches all aspects of our lives. It is how Amazon, eBay, Google and `learn’ how to recommend different products. This is based on the websites we visit or what we buy online. Banks can pick up on transactions different from our usual patterns, to detect possible fraud.

With this in mind, in healthcare, machine learning can analyse the results of clinical research to pick up important clues. For example, how say, a medicine behaves in our body. In medical imaging, machine learning can use algorithms – steps, questions and actions to solve problems. Also, machine learning can offer support for clinicians and patients when making decisions.

It is also going to be a fundamental part of eHealth and mHealth solutions, including apps, given that much of digital health is about providing tailored support that the patient needs and wants.

During account development, how are the patient views being taken?

But how are patient views being taken into account as machine learning systems are developed and implemented? The types of risk for patients vary. It varies from higher risk activities around diagnosis and clinical decision-making, to lower risks. This is around targeting and personalising information. Plus, there is the privacy risk. Patients have a different view on what they want machines to learn about them and their treatment, and just as important, what they do not want machines to learn.

Myhealthapps was invited to take part in a recent British Standards Institution (A UK-organisation that helps set standards) and Nesta (a UK foundation dedicated to innovation) workshop. It aimed to explore the feasibility of creating guidelines or standards for machine learning in health. Getting the right mix of stakeholders is always key. This is so the workshop had a mix of clinicians, NHS Digital, the UK MHRA (the government body that regulates medicines and medical devices), and the developers of machines learning (large and small). Many had already worked with the BSI on developing guidance for health and wellness apps (PAS 277).

Can patient organisations fill the `trust’ gap?

Who to trust in setting standards? A breakout group in the workshop felt that the motives of the National Health Service (NHS) in England are now perceived as being more about cost-cutting and health-rationing than care delivery. There is a breakdown in patient trust. For this reason the group argued that patients/patient associations should be charged with working on future guidance for machine learning because people have “lost trust in the NHS”.

But if trust in the NHS is being undermined, can this trust `gap’ for machine learning be filled by patient groups?

This raises three key questions:

  • Can patient organisations of all sizes develop the capability to influence the potential impact of digital technologies on their therapies like machine learning? Or even evaluate their impact?
  • How can there be a level playing field between well-resourced stakeholders such as big developers, venture capitalists, and the payers of healthcare like the NHS, and patients and carers?
  • How far up the agenda should digital health and machine learning be for patient groups stretched for time, people and resources?
Connecting patients? What myhealthapps argues…

Just as even well-motivated health app developers find it challenging to involve patients throughout the development process. As do developers of new technological challenges like machine learning.

Some, mostly large patient organisations, already have well-established track records in digital health such as in diabetes, or some cancers. Treatments already have technology built in. As Asthma UK put it recently, the idea of `connected’ inhalers. Some of which use machine learning, makes sense because asthma patients are likely to always have their inhalers and smartphones together.

Digital projects and new workshops are something patient organisations do not have the time or money to get involved with. It is challenging to identify which projects to dedicate time to. For these other organisations, it’s a longer journey to play an informed role when sharing patient views on digital health. However, as digital health becomes all the more commercial in its approach, and healthcare systems becoming increasingly expert in commissioning and using elements of eHealth and mHealth, it is all the more important that patient and carer needs are kept at the heart of all development.

Patient organisations which are pushed in margins, as the digital health bandwagon moves on. In a climate of complex healthcare decision-making, how can patient organisations influence the roll-out of emerging technologies like machine learning?

DeepMind: a `textbook’ approach to patient group involvement

Google’s DeepMind is a world leader in artificial intelligence research and its application for positive impact. The company is working with the NHS, accessing patient data. Key to progress has been DeepMind’s investment involving patients and patient groups. It is clear that this is central to building confidence in its partnership for machine learning, in particular a project with the UK’s Moorfield Eye Hospital NHS Foundation Trust.

The project focuses on one of the most promising areas for machine learning. The ability to identify patterns in data, such as in data imaging. Part of the research involves applying machine learning to one million anonymous eye scans. This is looking for early signs of eye conditions that humans might miss.

On its website, DeepMind puts the support of the most relevant and trusted patient groups centre stage:

“This is an exciting development towards early detection of eye disease and finding a cure for conditions including age-related macular degeneration (AMD). AMD is a devastating condition and delays due to pressure on eye clinics have resulted in some people suffering unnecessary sight loss. This technology could ease that pressure if it can accurately diagnose conditions such as wet AMD resulting in urgent referrals for only those that need them.”

Cathy Yelf, Chief Executive of the Macular Society 

“AI technology that can check retinal scans and detect eye disease at a much earlier stage could play a big role in tackling avoidable sight loss. In many cases, once sight is lost it cannot be restored, so earlier detection that leads to rapid treatment will be hugely beneficial. We look forward to seeing the results of the work as the research progresses”

Clara Eaglen, RNIB Eye Health Campaigns Manager 

Other activities to build patient confidence in the machine learning project include:
  • A DeepMind Health patient involvement event at Google’s offices publicised by the patient groups. This attracted more than 130 patients, carers and members of the public. This helped to decide the best way to involve patients in the project.
  • A screencast available from the involvement event, to reach more patients.
  • A well-designed section within the Moorfields Hospital website, profiling the benefits of the research, addressing concerns such as data protection, and including videos and news updates on the research project

Finally, it is a sign that investment in gaining patient and carer acceptance is as important as building the clinical and value case.

In the meantime, we will keep in touch with BSI and Nesta to see:

  • machine learning guidelines and standards developed for healthcare
  • patient and patient groups invited to contribute


Find out more about the machine learning project at Moorfields Eye Hospital…

Click here



Find out about DeepMind Health and research collaborations…

Click here

Download the BSI PAS 277, Health and wellness apps – Quality criteria across the lifecycle – Code of Practice….

Click here 


Leave a Reply

Your email address will not be published.