The health Ai revolution is coming — just not yet

There’s been a lot of news about Ai in Health Tech over the past couple of years. The prospect for artificial intelligence in health is very exciting. Both in helping enhance the performance of doctors, or even to get rid of them entirely! After all, much of the typical doctor visit revolves around a series of checks to narrow patient’s symptoms. All this is something very achievable by software in theory, but it’s the practical aspect I’m skeptical about.

The problem with health Ai

One evening a few months ago, I experienced an unexplained drop in body temperature. Instead of immediately calling for help like a sensible person, I felt it was a great opportunity for a ‘live fire’ test of these Ai health services. But to my disappointment the Ai wasn’t up to scratch. I ended up calling the NHS hotline instead (turns out I was fine, thankfully!).

Although this may have been a rather complicated case, I’ve heard similar anecdotal experiences from friends and colleagues testing these services. It’s not that health Ai lacks potential, but it’s still missing two key elements to make it ready for prime-time.

• Consultation data (accumulated data from the interactions between users and the Ai platform)
• User diagnostic data (health data from patients themselves)

The Ai chicken and egg

An Ai platform may have a lot of raw ‘intelligence’ — in that it can be very clever at storing and retrieving information, but it can’t really mature ‘till it has sufficient user data to crossreference and conduct ‘deep learning’.

But at the same time, the service will likely be less useful till more people use it (and generate sufficient data for insightful connections to be made). This creates a ‘catch-22’ scenario that makes it difficult to get the required mass of users in the first place.

This dilemma hasn’t been lost on service providers, which have offered their ‘Ai Check-ups’ free of charge. A trend that will likely expand to capture more user-data. Although in an age of heightened sensitivity about data security, this could become more challenging.

Of course there could be a role for national health services to help with this data accumulation, but the regulatory aspect of this could be equally challenging.

Despite the difficulty this lack of data poses, it seems like a more realistic issue to overcome in the near to mid term. Especially when compared to the diagnostic challenge.

Half the story

A fundamental part of the health management equation is patient health data. In a traditional consultation, doctors will check our temperature, blood pressure, heart rate and possibly request a battery of additional health tests. But Health Tech services haven’t really improved on this. Both telehealth and Health Ai solutions are offered remotely with no factoring-in of patient health measurments. Which means they’re missing half the information they need to offer a tangibly useful service.

Imagine how much more powerful Health Tech solutions would be if they had access to live and historic user health data. Blood pressure, heart rate, temperature and more. This could be directly incorporated into any ‘Robodoctor’ consultation to better inform diagnosis and recommendations. Maybe even offering pre-emptive health advice. The impressive stories of Apple Watches saving heart attack victims hint at this potential future.

But despite being one of the most popular wearable devices, Apple Watch sales pale in comparison to the number of iPhones sold. Meaning that Ai health services can’t rely on users owning these devices. Health Tech providers can tip the odds in their favour by subsidising wearable technology for customers. But even then, they’ll need these techologies to be much more sophisticated — the Apple Watch can’t event track temperature for example.

Of course these devices will likely get smarter, we’ve even seen reports of ‘intelligent mirrors’ that check your vitals. But in the near term it’s harder to imagine a credible diagnostic component to Health Tech services. Again limiting the usefulness of these services.

The long hello

Ai has been one of the hottest trends in Tech over the past few years. Health is one of the areas where it could have a huge impact. Dramatically changing the way healthcare is delivered and our everyday health is managed. But anyone that’s uttered the phrase ‘Siri’, ‘Alexa’ or ‘Ok Google’ to their phone — knows that there’s still a ways to go for Ai in general.

It’s important that we have businesses making serious steps into the world of Health Ai, but don’t expect the services to be revolutionary till they solve these inherent challenges.