An app a day keeps the doctor away

Healthcare is poised for an upheaval, with the arrival of artificially intelligent health apps that could replace visits to the doctor.

By Douglas Heaven, New Scientist 19 July 2017

Stiff neck, headache, tingling in your fingers. You list your symptoms, answer a few questions about how long they’ve lasted and whether they seem to be getting worse. Then, without ever leaving home or queueing at the clinic, you get the diagnosis: a strained neck. Or, at least, eight out of 10 people with those symptoms have one. (You also learn that 1 in 1000 people with these symptoms turn out to have degenerative cervical spine disease. Would you like to have your case reviewed by a human doctor? That will be £14.99.)

This isn’t some future scenario. The free app Ada, which offered up this diagnosis, was launched in the UK in April. It took six years for 100 data scientists to train the artificial intelligence behind Ada, using real medical records. So far, more than 500,000 people have used it. And with each case it sees, it gets a bit smarter.

Ada is one of several AI-driven apps to hit the market in the last few months, and they are set to shake up healthcare.

Back to basics

For Ali Parsa, founder of the company behind one of these apps, Babylon, it’s about making healthcare a basic utility. “How do I make it so damn cheap that everybody can afford it?” he says. Before his Babylon venture, Parsa spent several years running UK hospitals. That experience led him to a simple solution: stop people wanting to see a doctor.

Babylon Health is available to subscribers for £5 a month. Other apps, such as Dr AI and Your.MD, offer similar services for free, with the option to pay for premium services such as phone access to prescribing physicians. They all aim to make credible medical advice more accessible – as simple as Googling your symptoms. But unlike self-diagnosis online – or using apps like WebMD or iTriage, which just repackage information from the web – these AI-backed services lead you through a clinical-grade triage process.

Several studies have now found that AI may be more reliable than even seasoned medical professionals for certain diagnoses. That may bear out in user confidence: when Babylon added AI triage to its app a year ago, its doctors’ workload eased considerably. “Half of our patients disappeared on us,” says Parsa.

Healthcare providers are watching closely. The UK’s National Health Service runs a telephone triage line to help manage patient numbers. Now a North London hospital trust is running a trial in which callers have the option to wait for a human to pick up or talk to Babylon’s AI. Many are opting for the AI. “It’s the largest ever deployment of artificial intelligence in medicine,” says Parsa. “They’re seeing a 40 per cent work reduction.”

The underlying tech knits together several strands of AI: the ability to process natural language, including speech, so that you can be understood when you casually describe your symptoms; expert systems that trawl vast databases of the world’s medical knowledge in an instant; and machine learning software trained to spot correlations between millions of different complaints and conditions.

What if the AI makes a bad call, however? Some doctors have voiced concern about the rising popularity of these apps. When Ada launched in Australia, leading GPs advised people to be wary of both false reassurances and false alarms.

App developers say that’s why they take pains to involve doctors in refining their software. Ada uses both unsupervised and human-supervised learning to train the app, and Babylon makes sure its doctors agree with the app at least 99 per cent of the time. “But we’re finding when the doctors and machine disagree, the doctors are wrong as often as the machine,” says Parsa.

Immune to stress

If anything, he says, there are grounds to have more confidence in the computer than humans. “A machine cannot be negligent,” he says. “It doesn’t get stressed. It doesn’t get hungry. It does exactly what it is meant to every single time.”

Others are more cautious. “An app that can hold a back-and-forth conversation takes you most of the way there, but you might not have full confidence in the diagnosis,” says Basil Harris, co-founder of Basil Leaf Technologies in Pennsylvania and an emergency department doctor.

Harris says that trusted diagnoses are ones supported by tests. “I see patients all night long,” he says. “They come in and the very first thing you do is send them for tests.”

His company recently won $2.5 million in the Qualcomm Tricorder XPRIZE competition to build a fully automatic home diagnostic device, and is now continuing that work with an AI-driven app called DxtER that works alongside home test kits such as non-invasive blood glucose monitors and compact ECG machines.

But even with test info to hand, Harris does not think an app should offer a confident diagnosis by itself. “It should say, ‘Hey, let’s consider pneumonia and let’s talk to you doctor about that,’ rather than ‘You have pneumonia, you need this antibiotic, I’ll get an Amazon drone to drop it off at your front door.’ ”

“That type of AI has to earn its place, it has to earn our trust,” he says. “Until then, they need to be kept on a nice tight leash”.

Source New Scientist

Also see
Pre-Primary Care: An Untapped Global Health Opportunity Your.MD
Apple’s hush-hush foray into personal health records in iMedicalApps
Boston Children’s new health app Caremap helps parents track their kid’s health iMedicalApps