Last year I attended a Nuance showcase in London and found myself extremely impressed by a demo of the Dragon Speech Recognition product for use in hospitals. What was striking was that the system understood the most complicated medical jargon – spoken in a regular way, at speed – and added them to existing Electronic Medical Records.
This technical triumph did not emerge from nowhere. Nuance has a long pedigree in voice recognition technology and is probably most famous for being the company which helped develop Siri. Yet this is still a relatively recent development in its 20 year history. In its company brochures it boasts that 75% of Fortune 500 companies use its technology from divisions as a diverse as automotive, legal and mobile. However, its biggest source of revenue – at close to $1bn a year – is healthcare.
For the most part – as per the demo I saw last year – Nuance’s software is used to by doctors to make notes. This is something which is both difficult to do well from a technology standpoint – especially when you consider different terminology, languages and accents – but it is also difficult to deliver in a way doctors will use. The first part now appears to have been cracked with constant iterations and developments, such as the recent introduction of deep learning. The second part, on the other hand, is an ongoing challenge.
Gerhard Grobauer, senior director of engineering, who works on healthcare product R&D tells me at a partner showcase event in Barcelona this week that if you’d asked him 20 years ago he’d have thought that the use case for voice recognition in hospitals would have been solved by now, but in fact, it is still the same ongoing problem. “Technology is fast. Healthcare is extremely slow,” he says. In five years’ time he anticipates “at the most basic level” he’ll still be trying to do the same thing: increase penetration of speech recognition in European countries.
This technology is currently most widely deployed in the US and the UK’s NHS (National Health Service). It is also most commonly used amongst radiologists as this is the largest and most labour intensive department in a hospital. Nuance is dominant in this space and claims to be in 72% of US hospitals and pretty much every NHS hospital in the UK (along with quite a few private ones). Yet the story across Europe is far patchier.
As with all market dominance, of course, there is a possibility this won’t last forever. As the FT suggested at the end of last year, companies like Google may present a wider threat. Google already has a big stake in voice recognition and has made huge inroads into translation in the consumer space. So, if it was to set its sights into the healthcare market, Nuance could face a challenger.
Either way, the case for voice recognition seems fairly clear cut. The simple reason for this is that it decreases paper volumes and so improves accuracy and timeliness of records. This in turn makes information easier to locate in times of crisis.
John Rayner, the regional director for Europe and Latin America for HIMSS Analytics – a company which is focused on using data to improve healthcare outcomes – quotes the statistic that 250,000 US citizens die each year from medical errors. “Some US colleagues think this figure should be higher,” he says “others rubbish the data”. Overall though he feels the number itself “doesn’t matter” because “people lose their lives” due to poor documentation.
A very small UK study conducted by an independent body (on Nuance’s behalf) supplements this picture. It suggests that 90% of practice managers feel patient documentation is a burden on their practice. While 76% of the teams involved spend over half their time on documentation. Another Nuance statistics adds that a sixth of doctors’ time is spent on admin.
None of this seems especially surprising. Healthcare is essentially a two part process comprising of medical expertise and enough timely, easy to access documentation, to ensure patients can be treated effectively whenever needed. As Tony Brown, practice manager at The Bondgate Medical Practice points out many GPs spend years training to be a doctor then basically inherit a business with no experience in this area whatsoever, so anything that makes the admin side faster and provides clear guidelines, is an asset.
So, what are the down sides to all this? Well, the use of medical data always raises privacy concerns. All this is hard to avoid of course, but it is a particular problem in Europe. And although Nuance’s has cloud solutions in the US and UK (a pan-NHS solution should be completed this Summer, although it has been dragging on for some time) there are clears barriers to adoption in Europe.
Difficulties partly arise from the fact this is such a big disparate market with a lot of very separate jurisdictions but also stem from fundamental European concerns about a US firm holding all that sensitive data. The fuss last year around Google Deepmind medical trials in the UK shows that despite the march of digitisation this is not an area that looks likely to go away anytime soon.
The other key thorn in the side of all this progress might be the secretaries. These individuals look likely to be the casualty of automation in healthcare. And as Brown of the Broadgate Medical Practice put it in very human terms, although most people see the advantages of this technology “my secretary didn’t like it – it made her feel vulnerable”. Yet the role of secretaries itself is complicated one because depending on how processes have evolved these individuals are far more than ‘typing machines’ and provide an integral support facility for doctors.
“Speech recognition is not a technology, it is a change in the way people work,” concludes Nuance’s Grobauer. When it fails it typically comes down to poorly orchestrated change management.
Jon Collins’ in-depth look at tech and society
Phil Muncaster reports on China and beyond