The NHS faces mounting financial pressures from an aging population, rising treatment costs, and slowing increases in government funding. Moorfields Eye Hospital believes a new chatbot will help it fill the gap.
Its Oriel Assistant takes its name and initial use case from the Oriel project, a £344 million proposal to relocate Moorfields and the UCL Institute of Ophthalmology from their separate sites in Old Street, London, to a new integrated facility in King's Cross that will offer eye-care treatment, research and education services.
Users can ask the chatbot about the plan, find information about events and discussion groups, and provide their own feedback on the proposals, which will be fed into a broader consultation that closes on 16 September.
The Oriel Assistant also contains a feature that reverses the traditional virtual assistant model. The chatbot uses natural language processing to analyse the speech of users and ask them relevant questions tailored to their needs, creating a two-way conversation between the user and virtual assistant.
Peter Thomas, director of digital innovation and a consultant paediatric ophthalmologist at Moorfields, told Computerworld UK that the Oriel relocation provides the perfect test for the chatbot before it's expanded to queries about conditions and treatments.
"We'd start with things like interactive patient information leaflets," he said. "We have a system where when you receive a diagnosis, we give you a patient information leaflet, and you want to know one thing, but you have to read through five or six pages of information to get to it. It would be nice if patients could go to a chatbot, and just ask a general question about their condition."
Thomas believes the chatbot will prove extremely useful for NHS service users, who currently often struggle to navigate the health service's complex array of communication channels.
Moorfields investigated how this applied to the experience of its own patients by analysing the content of phone calls to the various hospital departments. They found that numerous calls requested computable data on subjects such as appointment times and contact information, which a chatbot could respond to with greater speed and efficiency than a human.
The hospital selected IBM Watson for the project after reviewing a range of vendors, from tech titans such as Amazon and Google to chatbot specialists including Nuance and Flow.ai. Watson stood out for its record of developing NHS chatbots for the likes of Alder Hey Children's Hospital and Velindre Cancer Centre, its ease of use, and the agnostic IBM Cloud management platform.
"We wanted to ensure that we were able to be relatively platform-neutral," said Thomas. "One of the nice advantages of Watson was that it's less locked into the Amazon or the Google ecosystem, and it has the capability to move some of the Watson Cloud tools onto hosting from different vendors."
The Oriel Assistant was trained on more than 500 stock questions gathered through surveys of staff and the public, on issues from directions to the proposed site to its impact on services. The chatbot will continuously improve over time as it learns from the interactions it has with users.
To make the interface accessible to patients with visual impairments, a feature was added to the chatbot window that allows them to adapt font sizes and colour contrasts to their individual requirements. Thomas plans to provide further support by adding voice integration.
"We've tried to do what we can to make the text interface acceptable to patients with visual impairments, but we really want to go that extra mile and make sure that we have it integrated with voice so that patients who have a really severe visual impairment are able to use it," he said.
Thomas has been encouraged by early feedback, but warns others exploring chatbot technology that they will not be able to predict everything that users ask it.
"My advice is to understand that it needs to be a dynamic process," he explained. "When you launch it on day one, you'll find that some people ask questions you hadn't anticipated, and you need to be able to respond to that to make sure that you're constantly making sure it suits the need of the user."