How artificial intelligence could provide some respite for the NHSEmma Rich, University of Bath and Andy Miah, University of Salford
The NHS recently announced plans to trial an artificially intelligent mobile health app to a million people in London. The aim is to help diagnose and treat patients by engaging them in a real time text message conversation which will complement the NHS 111 phone based service (which was criticised by the Care Quality Commission watchdog). The app’s designers, Babylon Healthcare Ltd, use algorithms to make initial diagnoses which are then followed up with human consultations. It has already received a glowing CQC evaluation.
The app is likely to provoke a mixed response, with enthusiastic technophiles up against those concerned that more technology means a less human healthcare service. Yet, with the NHS being described as suffering from a humanitarian crisis, and with a growing healthcare burden and limited resources, some smart solutions are needed. It is hard to deny that problems of limited funding are enduring features of this unique public service. Perhaps AI has the answer.
In fact, providing effective healthcare is always a combination of systematised technological efficiency combined with patient centred human care. Polarised views on technology are often not helpful. It’s also necessary to recognise how this approach to healthcare is part of a wider technical revolution in which connected objects in the Internet of Things will change everything from healthcare to traffic maintenance.
The NHS app is really simple to use and has been likened to using the social messaging service WhatsApp – but with one crucial difference: you are chatting with a computer, not a person. Once the app is downloaded, you log your basic health information, and then start explaining your symptoms. The robotic “responder” will say things like: “I just need a few details from you before we get started,” and “nearly there” to keep the conversation going. After a more detailed exchange, it might come to a conclusion along these lines:
Ok so your symptoms don’t sound urgent, but I think they require further investigation. Make sure you arrange a consultation with a GP within the next two weeks. If left, symptoms like yours can become more serious, so book now while you remember and I’ll remind you closer to the time. If things change in the meantime and you become more unwell, speak to a doctor as soon as you can.
This digital diagnosis service intends to provide an additional communication tool between the NHS and patients. It it part of a broader ecosystem of digital health services which include online health tracking. Also, the app takes advantage of the fact that some people these days are likely to be more comfortable chatting by text than they are with talking on the phone.
This digital phenomenon is driven by the promise of a wider technological fix to social problems. Applications within healthcare could bring about big wins for society, where the functionality of the device is made all the more efficient by the aggregation of “big data” that it generates. Tech firm Babylon is joined by other big players seeking to do similar things, such as Google’s Deep Mind, which wants to mine NHS data to to enable earlier diagnoses for example, or to achieve more effective monitoring of treatments.
At the world’s largest tech expo in Las Vegas at the start of 2017, home AI systems have been one of the biggest hits. So perhaps the NHS has found an intelligent solution at just the right time. People may now be far more willing to have a “relationship” with an attentive machine than a call centre drone.
Driving these developments is the assumption that, within a digital knowledge economy, these forms of communication can offer more neutral and accurate responses, circumventing human error. Yet, scholars within the emerging field of critical digital health studies suggest that algorithms must be understood as part of a complex network of interconnections between human and non-human actors. A recent study comparing physician and computer diagnostic accuracy revealed that doctors “vastly outperformed” algorithms
So we need to ask some key questions about the assimilation of AI into healthcare. How do people make sense of the list of possible diagnoses they receive from the machine? Will people follow the advice, or trust it? How will AI need to be tailored to accommodate human variation, by geography, capacity, or cultural identity. Another important aspect of this trial will be the consideration given to the backgrounds of the users. Given enduring concerns about inequalities of digital access and digital literacy, trials for future digital health tech need to be conducted amongst those populations with limited resources, experiences, and technological infrastructure.
Perhaps the biggest question we face in a world where ever more of our data is locked up in the mobile app environment, is over the proprietary and privacy of our data. How can we ensure that we have the freedom to move our health data around, over time, and ensure that it is safe and secure? We may need a new Bill of Health Data Rights to underpin and limit their exploitation of our data, and work on this must start now.