What have you seen change in the amount and range of health data over the past 10 years? Is it making your job easier?
Within the actual practice of medicine, there is more and more data available, but there aren’t more and more ways to interpret it or use it. Even though we have this new information, it's too much for one person to process. Additionally, the way we practice medicine was not created using this type of information.
Take monitoring blood pressure, for example. We now have devices with the ability to measure it 10 times a day. It's highly variable throughout the day, so people are engaged and want to use that technology, but we don't know how to use that data. Our current practice is not designed for that, nor are our treatment guidelines. We are awash in data, but the ability to process it and the knowledge to use it correctly is not there yet.
Do providers sometimes feel overwhelmed by all the data?
Yes. Burnout rates continue to climb because more data is in the electronic health records, but there's less support to deal with it.
Is there a solution down the road? Is someone going to figure out how to use all this data, or is there always going to be more than we can really deal with?
There are a few things that need to happen. One is that the definition of sound evidence — the evidence upon which we make decisions — needs to evolve in medicine. If you’re going to use any of this new data, any of these new analytics, or AI, the type of evidence sufficient to make clinical decisions must evolve. If you require using an old model, we'll never get there due to the size and complexity of the data available. AI has to be part of it. We need mass analytics to deal with this huge amount of inputs.
The third thing is that we need to get over the division between academics, health care operations, and industry. Each bring something unique to the table, and it's something the others do not have. None of them can put it all together without the others. I live in all three of these worlds, and I firmly believe that academics will need to collaborate with industry partners to build appropriate tools. And of course, the hospital system is the care delivery vehicle.
How does that relationship work when it comes to working with data? It seems like industry might be embracing it more, at least initially.
There's this tension, because traditional data science says that you don't need to be an expert in what you're doing. The data will speak for itself and tell you what to do if you set up your system properly. That's true in a lot of cases, but in health care, it's a little different. The liability and regulatory constraints are different. What I see industry doing is having spectacular engineering, but not vetting it in any traditional way, because it's proprietary. You instantly lose a big chunk of physicians or providers, because you don't have any evidence of whether it works or not.
If the industry engineers were able to get to a place where they could leverage expertise in implementation science or some kind of clinical research, the product would be better, credibility would be better, and they would be less vulnerable to the mistakes that inevitably happen. But that's currently not happening. It's very hard to do in the current environment. That's a problem.
What about the security aspect? How big of a worry is that?
I think a lot of it is appearance. We can be transparent that data is shared in many ways — that's just a reality. In a way, it’s a generational phenomenon. I suspect that people who came of age in the early 2000s or later have a different sense of what data privacy means or what the loss of data means compared with more recent generations. Now, there's an assumption that everybody knows everything about you, which allows you to use more technologies. Members of older generations like myself struggle much more with that exchange.
The main challenge right now is how do you use cloud computing? Especially in health care, the popular opinion is that the cloud is less secure than something on premises — something on your desk. That's not true, but that's the initial reaction by almost everybody. So there's a lot of restraint to experiment or try to develop out capacity using cloud services.
Once people get more comfortable with that concept and the fact that it can still be secure going to and from a cloud platform, it opens up all these possibilities for the AI part of it. You can store a lot more data. You can capture a lot more data. You can integrate things like wearables. In terms of data security, I would say that’s the biggest difference from 10 years ago. The idea of a security breach and inadvertent disclosure of data, or misclassification of data so that someone could get hold of it when they shouldn't — those are still issues, and we still battle them constantly.
