Could artificial intelligence create virtual GPs?
An interview with Paul Brown, Stowhealth, conducted by April Cashin-Garbutt, MA (Cantab)
Could you please outline the vision behind the new partnership between Orbital Media, the University of Essex and Innovate UK?
The vision for the partnership is to incorporate artificial intelligence into online GP consultations by providing accessible information that is certified and verified by Health Commissions so that we reduce the demand for GP services. The underlying aim or objective is to shape demand for healthcare services.
A lot of current technologies look to provide access to healthcare clinicians quicker, but what we are looking to do is to change the way people access information about their own health.
Rather than relying on a doctor, they would get self-care information from artificial intelligence through an automated online service utilising photo realistic avatar technology or video technology to engage with patients and provide them with certified information that comes from a trustworthy source.
In partnership with Innovate UK, the project will exploit Artificial Intelligence (AI) and Machine Learning technology to provide medically approved health advice to the increasing numbers of users who seek medical information online.
The technology developed, via the partnership, will be used to serve interactive advice via groundbreaking, photo realistic avatars. Information supplied by the platform will focus on minor ailments such as colds, coughs, flu and hay fever.
The resulting technological solution will provide a visual, reliable and robust online health advice service, to meet the rapidly growing demand for online ‘symptom searches’.
People talk about doctor Google doctor where people Google their symptoms only to find they get a million different answers. We want to cut through that and provide people with trusted, reliable information that allows them to treat themselves in a safe and low-cost manner.
Will the platform only focus on minor ailments?
What it is not geared up to do is to diagnose patients. We're not looking to diagnose; that's still very much what a doctor does. Perhaps it could do that in time.
Since we're balancing risk with providing alternative solutions to people's ailments, we are starting with self-care minor ailments so that we are providing people with safe, reliable information that will allow them to make informed choices about their own care, as opposed to us saying "What you've described is this. You must do that."
You would ask the virtual platform questions about a condition or ailment. You might say "I have a cough," or "I have a cold and I've got blood in my phlegm. What shall I do?" It would then give you medical advice about how best to treat yourself, or if appropriate, tell you the best thing to do is to book an appointment.
There are red flags built into the system, so that if people ask an alarming question, it would immediately tell them they need to seek further assistance. Since you're looking at minor ailments, that shouldn't happen very often, but it is built in there.
The skill of the doctor is still very much in the background, but what we know is that there are so many appointments being requested of doctors that aren't necessary. We're targeting that, to free doctors up to perform the stuff that they're highly skilled in, not stuff where the advice is "Rest in bed for two days and take some paracetamol to alleviate the pain."
How much GP time do these self-treatable conditions currently account for?
These self-treatable conditions currently account for a significant proportion of GP time and are identified by the Proprietary Association of Great Britain (PAGB) as one of the five examples of wastage in the NHS system, amounting to a cost of £2 billion per year.
How will Stowhealth be involved in testing the system?
Because we're looking to replicate the manner and the thought processes of a doctor, our input is to form how that artificial intelligence behaves and allow it to know what the correct questions are to possibly ask and how it should respond to certain things.
In earlier iterations, we've been providing the answers to questions. Now, the latest iteration is moving beyond that and is stringing together different aspects to make the decision-making processes a lot slicker and to provide a greater range or depth of interaction. That's our role - to support the engineers in getting it to feel right. We will also facilitate real time patient testing.
What do you think will be the main challenges to overcome?
The main challenge is to make it work the first time, every time. We need to manage that, because, if it doesn't work, then people aren't going to come back and use it. It needs to be simple. Our approach is not trying to do too much in one go, but before it goes much broader, it needs to flow smoothly.
One of the challenges we found is that people ask the same question in many different ways. For example, we've learned that people use slang for different symptoms. Someone might say, "I feel grotty,” but, across the country, people have different ways of describing the way they are feeling. The way they interact with the machine is not as simple as just plugging in a thesaurus. There's a whole slightly deeper level of interpretation that we need to overcome. If we can overcome that, then that would allow the interaction to be more seamless and smooth, rather than asking questions every time someone gives a slightly odd description of themselves.
What role do you think artificial intelligence and machine learning technology will play in the future of health services?
I think the potential is colossal. It's quite telling that there are so few people in this field at the moment. There's lots of talk about it and lots of ambition, but no one really knows how to take the next steps.
One opportunity, for example, is assisting people trying to understand how the health system works, and all the different services and team that are available to support you in your need. Then there are applications in hospitals such as accessing pre-operative information that people can go to when they’re due to have their knee or their hip replaced, for example. Time is limited and actually you can find out everything you'd want to know about a knee or hip replacement during a virtual consultation. The same principles could apply to every single ailment and hospital procedure.
While we're focusing on minor ailments at the moment, that is only the tip of the iceberg. It's not just a GP environment: it is consultations with consultants, it is the follow ups post-surgery. Then you've got community providers and social care and so on. There are a number of artificial intelligence applications in the field of mental health already.
It's colossal and it's just a matter of when the tools will be developed in a way that people feel comfortable with and feel confident using.
Where can readers find more information?
The partnership with the university has only just begun so there will be there will be information on our website in due course. In the meantime, further information can be found via Orbital Media, the company partnering with the university and Innovate UK, and specialies in pioneering technology solutions using Virtual and Augmented Reality.    Further details can be found at http://www.orbitalmedia.com/contact/
About Paul Brown
Paul Brown has been a senior business manager at Stowhealth, Suffolk for the past 6 years and involved in many innovative healthcare technology applications in the Primary Care setting including wearables technology for weight loss, patient led blood pressure monitoring and healthcoaching for patients with long term condition.
Stowhealth was rated as Outstanding by the Care Quality Commission in 2016 and Award General Practice Team of the year in 2015 and the national General Practice Awards.