The Government’s Incubator for AI (i.AI) has developed a tool to help pharmacists assess which patients could be at risk from multiple prescriptions of medication.
Dr Laura Gilbert, director of i.AI, said this has come in response to worries that many people are being harmed, even dying, due to complications in having a large number of prescriptions.
She outlined the project as an example of i.AI’s work in providing AI solutions for public services while speaking at UKAuthority’s AI and Data4Good conference last week.
Gilbert said that academics have estimated that each year up to 20,000 people die in the UK because of problems with their prescriptions. As they get older they are given different types of medication which can interfere with each other and have side effects that are harmful to health.
In addition, the NHS spends about £750 million on prescriptions that are doing direct harm and another £250 million on some that are not useful.
Triage, review, detection
This has prompted i.AI to develop an AI tool named MedGuard. It provides a triage process to identify and prioritise high risk patients, uses CoPilot to review guidance from the British National Formulary, and detects errors in the use of medications.
It has been granted provisional medical device status by the Medicines and Healthcare Products Regulatory Agency and i.AI is hoping to use it in clinical trials soon.
Gilbert said it has been developed for use by pharmacists as they have more time than GPs to assess the combination of prescriptions.
“It works with the pharmacist, who has an overview of the patient’s medical history such as all the drugs they are on,” Gilbert said. “Also, in some ways they are less busy than GPs.
“The pharmacist can go through patient data, figure out using AI which patients need help and adjust the medication; and the AI will come back and give advice on how to adjust the profile and look out for which symptoms to look out for in the patient..
She pointed out that the NHS has the Structured Medication Review under which patients can go through their medication with a pharmacist, but the limited number of pharmacists means that only about 100,000 of the one million people who qualify for these are able to go through the process each year.
“So this is a really good use case of where AI can find all the people who need it,” she said.
Principles for assurance
She also described i.Ai’s assurance framework for the development of products, saying it is based on five key principles: that products are safe, secure and robust; they are appropriately transparent and explainable; they do not introduce additional bias; they are improvable and open to challenge; and the organisation’s governance processes keep it accountable.
“We have a set of assurance standards and open sourcing is an important part of that,” she said.
“We call it ‘radical transparency’ – we’ll tell anybody what we are doing pretty much all the time – is very important in that it allows academic and industry experts to pretty much mark our homework, and if they are so inclined they can check improvements and fixes as well.
“We’re also very interested in making sure the public knows what we are doing…. By being really transparent about what we are building we hope the public will believe us when we say we are not trying to disadvantage them or take decisions without humans.”
Watch Dr Laura Gilbert's presentation:
