There has been huge interest in Artificial Intelligence (AI), in particular how to maximize the benefits of AI for healthcare, while minimizing its risks and avoiding its pitfalls. But aside from theory, guidelines and principles, what practical progress has been made in the use of AI to improve patient care and outcomes? And whilst we may have seen AI being increasingly used to automate the reading of digital medical images, how has AI been used with other patient data?
“The development of artificial intelligence could spell the end of the human race”
“AI scares the hell out of me. It’s capable of vastly more than almost anyone knows, and the rate of improvement is exponential”
Our webinar attempted to demystify AI, beginning with an outline of what is actually meant by AI. Our speakers discussed how AI could improve patient outcomes and healthcare delivery, with some practical real-world examples including how the patient voice can be involved effectively and the hurdles which need to be overcome to realise the potential of AI.
The event was chaired by use MY data's Expert Data Adviser Chris Carrigan.
John Rigg, AI Specialist, IQVIA.
John runs an AI practice within IQVIA, developing various AI solutions across many disease areas. John described what is meant by the different terms "artificial intelligence" and "machine learning".
He suggested that we "think of AI as the application of machine learning".
John gave some examples of where AI has used patient data to improve health, including identifying undiagnosed patients with rare disease by linking primary and secondary care data, covering 5 million patients. He also referenced publications which described how AI had found undiagnosed patients with Hepatitis C virus (HCV) and a final example where AI was used to improve the clinical management of heart failure in the UK.
John emphasised the balance/challenges of data protection versus data use, noting that he had given up on a project as it took too long to obtain the data.
Anna-Grace Linton, Researcher, University of Leeds
Anna is a researcher at the University of Leeds researching the role of free-text comments of Patient-Reported Outcome Measures (PROMs). Her research is looking at how natural language processing can be used to explore how the comments provided by patients in survey responses can give a better understanding of cancer patient outcomes, compared to analysis of just the closed questions in the survey.
Anna highlighted that free text comments are rich in details of the quality of life, but often the data is too much to be analysed. AI can use the large amounts of unstructured data (in this case the free-text responses) and automatically process it into something insightful that focusses on the quality of life.
Sandra Irvine, Member, use MY data
Sandra illustrated how embedding the patient voice within AI work can improve the work and its outcomes.
She talked about her experience of integrating PPI into an AI project; a centralised AI based solution for tissue based delivery of personalised medicine testing for the NHS. The project use AI to look at markers related to colorectal or lung cancer.
The researchers had involved patients from the very start, with Sandra being involved in the writing and submission of the grant. She was also tasked with writing educational content and organising a public event at the end of the project.
A great question and answer session with delegates followed, with delegates putting a series of questions direct to the speakers. Some of the topics covered in the discussion were:
• Could AI/Machine learning revolutionise primary care – speed up diagnoses? Could it illustrate how primary care date be used for benefits, rather than focussing on risk?
• Patient & public involvement (PPI) – do the researchers involve patients?
Sandra responded with great advice, acknowledging that a lot of grant awarding body insist on PPI – this is a challenge as can be seen as a tick-box exercise and force researchers ‘to do’ PPI – however, most have been surprised at what they’ve gained from PPI.
Sandra advised that researchers should involve patients at all stages, right from the start of the planning. The involvement is useful:
• at the beginning of an AI project in terms of formulating the questions a researcher wants to ask
• in the middle in terms of looking at the answers the researcher is getting
• at the end, to help researchers work out how to present their research to the general public.
Patients could be the greatest supporters of researchers’ work!
As usual with our webinars, which are open to all, we saw a mix of Members and non-members. Around 61% of delegates came from outside of use MY data.