Can AI save the NHS?

It's going to be a lot harder than our politicians seems to realise

The NHS has lasted more than 75 years. It has, by most metrics, been a success. But it is now under considerable strain. It faces chronic staff shortages, record numbers of emergency attendances, an ambulance wait-time and workforce crisis, and much more. It’s not surprising that there is an intense focus on increasing NHS productivity. Of the many mooted solutions, none has received as much attention in either policy rhetoric or press coverage as Artificial Intelligence (AI). The Prime Minister is a notable advocate – regularly arguing that AI can solve the biggest social problems of our time.

The hope reflected in numerous policy documents, is that by combining the NHS’s data riches with the latest techniques in AI, data-driven healthcare can be brought into reality. Let’s imagine what that might look like in an ideal world:

Olivia, 25, has for the last few months been suffering from tiredness, cold hands and feet, dizziness, and occasional shortness of breath. Last week she used the new NHS 111 chatbot to see whether her symptoms were serious and whether she should seek medical advice. 111 reassured her, suggesting her symptoms were most likely caused by anaemia, but recommended that she book an appointment with her GP to discuss treatment.

As Olivia had logged into 111 using her NHS login details, to book the appointment she simply had to consent to 111 sharing the information from their exchange with the NHS App which conducted a form of automated triage, and then booked an appointment at a convenient time. The NHS App, accessing her exchange with the 111 chatbot, said she needed a face-to-face appointment, but not urgently. It offered Olivia 3 different appointments for the following week.

Arriving at her GP, Olivia checked-in using the NHS App which reminded her she was there to talk to her GP about a possible anaemia diagnosis and provided a short list of topics that were most likely to be discussed, along with several suggested questions. Oliva was greeted by her GP who already knew her symptoms and was able to tell her that, based her previous medical history, and the latest data from her Apple Watch, Olivia was indeed most likely suffering from anaemia. The GP then sent Olivia to get a blood test, to inform the treatment plan. After the test the GP explained that she would make her treatment recommendation once she had the results, and that Olivia would receive a push notification.

Finally the GP asked Olivia if she would be comfortable discussing health conditions she was potentially at risk of developing in the future so as to take preventative action now. Olivia agreed and the GP opened the ‘preventative advice’ tab on her screen. Based on Olivia’s family history, weight, and current lifestyle she was at moderate risk of developing hypertension (high blood pressure) in the intermediate future. A personalised preventative action plan was attached to Olivia’s health record which she could view within the NHS App. The next day, Olivia received a presciption for ferrous fumarate, as well as an iron rich diet tailored to Olivia’s genetic make-up and lifestyle. The prescription was sent to the pharmacy nearest her house and she was also sent a discount code she could use to buy specific iron-rich foods from her supermarket. Whilst in the App, Olivia also reviewed her hypertension preventative action plan and saw she had one month’s free access to an online exercise platform and recurring appointments to have her blood pressure checked every 6 months.

Prompted by the interaction with Olivia, the GP decided to check whether any other patients on her list could be suffering from undiagnosed anaemia or at risk of developing hypertension. The identified patients were each sent a personalised notification, alerting them that a change in their health status and inviting them to an appointment to discuss the change. The aggregate results of this search were sent back to the commissioning team at the relevant Integrated Care Board who, noting an increase in the number of potentially hypertensive patients decided to launch a local public health campaign.

The benefits of realising this vision are clear and worth pursuing. But we are a very long way from it and, at the moment, we are not doing the right things to make it a reality. Instead we are in danger of taking a confused and poorly implemented approach that could worsen the quality of healthcare.