Technology / What’s app, doc?
What’s app, doc?
14 December 2017
Joanne Frearson talks to Stephen Brobst, CTO at Teradata, about how data-driven healthcare applications could make for more accurate – and far cheaper – diagnosis and treatment by 2030.
Stephen Brobst, CTO at Teradata, turns up to our meeting at their UK headquarters in London Bridge in a summery mood, wearing a Hawaiian shirt and jeans. For good reason – the native Californian is a big name in the technology industry, currently ranked the fourth most influential CTO in the US, above those of Amazon.com, Tesla and Intel. I’m here to talk to him about how data is driving the next industrial revolution, and how it will be used in the UK by 2030.
“Improving the quality of life should be the number one priority,” he says. By 2030 in the UK, Brobst believes sophisticated deep-learning techniques will be able to analyse our personal data and our individual biological make-up to provide us with the best treatments for illnesses, thus helping to reduce the costs associated with providing medical assistance.
“Treatments by 2030 will be much more individualised to a specific person based on the data about them,” he says. “That data will include their genomic structure and family history of illness, not just their own history. It will also include the results from the treatments of lots of other patients with similar ailments so you can figure out what works best.”
As the commercial world uses data to make better decisions, says Brobst, so it also has the potential to help us make personal decisions about many different aspects of our lives. Progress has already been made, with technology currently being developed that can capture data about a diabetic patient’s blood sugar levels via a sensor inside a contact lens. “By monitoring that, you can decide how and when we can intervene with a diabetic to keep them out of the emergency room,” says Brobst.
But despite this progress, some hurdles need to be overcome until data can be used in the ways Brobst suggests. The delivery models and the way we are analysing data, he explains, will have to change.
He says: “With the data continually doubling in size, eventually we will have to get to a new generation of massively parallel processing technology, which will likely use quantum computing.” The problem here is that quantum computers are only in their infancy, although Brobst is confident they will exist by 2030 and be capable of processing exponential amounts of data quickly.
Another concern Brobst thinks we will have to tackle relates to the use of data in healthcare – as machine learning gets more sophisticated, he explains, it could make patients uncomfortable because they do not understand how the decisions are being made. “We need to be able to explain why we make decisions,” he says. “Why are we recommending the doctor do this surgery instead of this other surgery?”
He explains that doctors could need to have a greater understanding of the medical research behind a procedure, to demonstrate why it is the best course of action. “We need to be able to reverse-engineer the decision in order to explain the decision,” he says. “That is not doable yet, but by 2030 I believe it will be.”
Although there are concerns of a Big Brother-type scenario emerging, with virtual doctors watching our every move and telling us not to eat the last slice of pizza, Brobst thinks the data should simply provide patients with a constructive set of choices. By 2030, he believes, the concept will have been normalised, and people will see the benefits of using data to improve healthcare. He envisions a scenario where the average person will be able to know when to see their doctor based on the data. “In the future devices will be able to tell you, when is the right time to go to the doctor,” he says.
He offers a vehicle maintenance analogy: “If you open up the car manual, it will tell you should change the oil every X miles, or change the tires every Y miles. This is called milestone-based maintenance.
“There is a lot of data analysis to figure out whether the distance is 15,000 miles, not 10,000 miles, but in reality there is a distribution within which things are going to fail. This is where most people need to do maintenance.” However, he points out, the problem is that everyone is unique and so drives differently. What might suit one driver as a maintenance check-up might not be necessary for another person. And doing check-ups when you don’t really need to could end up being expensive.
In much the same way people’s individual genomic make-ups can be used to tell whether we need to go to the doctor for a check-up. And once we get there, instead of treating everyone who has a disease with the same treatment, they can be tailored to that genomic structure.
Brobst says: “It may mean some people go to the doctor more than once per year and some people go less. It is based on the individual healthcare needs that optimise quality and cost of care. The annual physicals are exactly when you need them to happen. Not too late and not too early.”
At the moment, he says, it is too complicated to develop treatment for each individual person. “But with the kind of data they might have available by 2030 and the kind of learning algorithms and computing capabilities, we will be able to do that work. We absolutely can have unique treatment for every individual – even to the point of unique drugs for each individual.”
In the long run, Brobst believes that however things develop, the cost of healthcare will fall as people will not be going to the doctor unnecessary or be given treatments unnecessarily. And by processing data far faster than we are currently able, people can be provided with the individual treatment they actually need.
This article was published in our Business Reporter Online: UK2030.