Using Artificial intelligence in triage
Artificial Intelligence (AI) is an automated way of processing information. In this case, a computer system has been taught about medical conditions their symptoms. It asks questions to narrow down the possible causes and decide the priority the medics should give to it.
If you speak to someone on NHS 111 they will be using AI to take your answers and to ask the next question, until the system makes a recommendation. You can also do this yourself online at www.111.nhs.uk whenever you feel worried about a condition.
The at-home service will become even easier as the NHS has entered an agreement with Amazon so their Alexa platform will be able to answer your health queries. You will get advice just by talking to Alexa. (read more on this here)
There is some debate about how much Alexa should guide an individual’s health. Should it listen for indicators of suicidal thoughts (more here); and what does that do for individual privacy. Even the most ardent protector of personal privacy might soften if someone they loved was saved by an intervention because of ‘listening in’.
Some hospitals, such as Birmingham University Hospitals, are encouraging patients to undertake a two minute “artificial intelligence triage” before they consider attending hospital. This could instead direct people to their pharmacist or possibly advise people to call 999 for rapid assistance (read more here). Some hospitals have booths set up so when you first arrive you can enter your symptoms.
Spotting hidden urgent conditions
Many people have been surprised by the effectiveness of these AI systems.
A&E is often operating under pressure and humans naturally will go towards the obvious need. However, there are a number of conditions that don’t present dramatically but where speed of treatment is vital. For instance, strokes and sepsis are two conditions where early intervention can dramatically improve the outcome and where delay can cause life changing outcomes or even death. You may have heard of the ‘golden hour’ for strokes but many patients may present with cuts & bruises after a fall and not realise they have had a stroke. Likewise Sepsis (www.sepsistrust.org ), which kills 52,000 people a year, often presents as a general lethargy and occasionally people can be sent home to rest when they desperately need medical assistance.
With the pressures on A&E departments, many are missing their targets to see patients within 4 hours, when for some patients being seen immediately is paramount. Having pods where patients go upon arrival to fill in an electronic questionnaire may seem impersonal but it is a very efficient way of getting the information, perhaps because the machine will not get side-tracked by chatting about the weather, last night’s football scores or ‘Strictly’.
Other than the speed of collecting the right information, the major advantage of AI is it has machine learning and can scour through millions of individual’s records for combinations of symptoms that match. By tracking the eventual, confirmed diagnoses of patients with matching symptoms there is a stronger idea of your diagnosis. This is particularly impressive when separating out symptoms that may be due to overlapping issues.
Evin if your pattern of responses to the questions has only been seen 5 times throughout Europe, AI will find them all in seconds and include their end diagnoses and prioritise you accordingly.
Once you are in the NHS system you will then be cared for by professionals (humans) but AI can help get you to the right starting place faster and more reliably.