Using AI to Improve Health and Care

Call overview and agenda

An online discussion on using artificial intelligence to improve health in North West London was hosted on Wednesday 4 August 2021 (4.30pm to 6pm) via Zoom Pro and was attended by 19 members of the public from a wide range of backgrounds.

The aim of this particular online session was to:

• Introduce the Imperial Biomedical Research Centre and its proposed Digital Health Research Theme
• Provide an overview, and an example, of Artificial Intelligence and how it’s used in care and research 
• Give attendees an opportunity to ask questions 
• Facilitate small group discussions around the topic

Key Insights Summary

Attendees of the discussion session on 4 August 2021, provided perspectives about both the benefits of using artificial intelligence within their health record to identify possible health risks as well as concerns.

Benefits
The benefits included the ability to identify potential health risks especially in relation to genetic conditions within families. They also considered its ability to improve resourcing in healthcare (i.e.. rare blood types) and to improve healthcare more generally e.g., for rare diseases and prevention to also be benefits.

Qualifications
The benefits of using artificial intelligence within their health record to identify possible health risks were considered to be subject to certain things being put in place or existing including their consent to use artificial intelligence within their health record in this way, the need for the results to be accurate, the need for transparency about artificial intelligence being used in this way, medical professionals having the confidence and skill of to use artificial intelligence and the need for artificial intelligence to be used as a tool rather than to make a final decision.

Concerns
Attendees expressed the following concerns about using artificial intelligence within their health record to identify possible health risks: lack of trust and understanding of artificial intelligence, the limitations in communicating about artificial intelligence to diverse communities, the negative reputation of artificial intelligence from other contexts e.g. school exam results and the quality/quantity of the data used in the artificial intelligence e.g. the fact that medical records are not always correct. Other concerns were the relationship between healthcare professionals and artificial intelligence not being clear i.e. does it benefit both doctor and patient?, data security and confidentiality e.g. what if the data is used against you or is released through error, the need to retain a relationship with the healthcare professional especially to discuss results identified by artificial intelligence. The consequential burden on the NHS from what is identified by artificial intelligence and too much reliance by doctors on artificial intelligence was also seen as a concern e.g., will the doctors default to just using artificial intelligence.

Explanations for poll responses
Attendees provided more details about the reasons for their responses to the poll question: “Which of the following areas do you think would benefit most from the use of Artificial Intelligence and should be prioritised by the Digital Health Theme? (Please choose your top 3).”

• Monitoring responses to treatments: to stop treatments that are causing more problems or harm, this presents is a lot of benefit and potential and the fact this is a logical/measurable task which can be delegated to a non- sentient function.
• Identifying at risk patients: the doctor would need to confirm the patient was at risk after identification by artificial intelligence. This would assist with prevention and encourage a healthier lifestyle. It is a logical/measurable task which can be delegated to a non- sentient function and digital tech is less likely to mess up and exacerbate an illness.
• Improving patient experience: making artificial intelligence accessible to all especially by using simple language, rectifying a bad experience through use of artificial intelligence, using artificial intelligence to learn patient availability for appointments and improving the patient experience would mean patients are more inclined to seek help early on (and improve outcomes) and digital tech is less likely to mess up and exacerbate an illness.
• Assisting with clinical decision making: to be used as a support tool only by healthcare professionals and used properly and responsibly. The large amount of data available means it can be used extensively and it may encourage the doctor to explore certain diagnostics or treatment pathways which may not have been done otherwise and help reduce human error. However, it was noted there is a risk of depersonalising medicine or diminishing the patient/doctor relationship. It was also seen as a logical/measurable task which can be delegated to a non- sentient function.
• Monitoring responses to treatments: to be used as a support tool only by healthcare professionals.

Artificial intelligence was seen as good for yes or no answers but that it depended very much on the algorithm to get reasoned answers.

Challenges with artificial intelligence in everyday use

Attendees identified the following challenges with artificial intelligence in everyday use: lack of accessibility to digital devices and wifi to access artificial intelligence, the fact that voice recognition does not always pick up accents and the risk of racial profiling. Another challenge was that artificial intelligence has been incorrect as it doesn’t always understand what is meant, suggest things accurately and changes choices. Other comments made to this question were the lack of understanding of artificial intelligence in the population including underrepresented communities, the need to engage the older population about artificial intelligence and the risk of hacking and ransom attacks being a concern in light of recent examples e.g. the Irish HSE.

How we used the insights

This insight report summarising key points from the session was made available to Theme leads and the BRC Executive in order to shape the BRC application. The report was also provided to the members of the public who took part in the involvement activity. A full report on all public involvement activities undertaken in preparation for the BRC application can be found here. Through the process of conducting this and other public involvement activities, we have established a wider and more diverse network of contacts for ongoing involvement. We would like to thank all those members of the public who gave their time and thoughtful insights through these activities, and the researchers who engaged enthusiastically in the process.