Storing Data in the Future of Healthcare: Federated Learning and Rules from the ONC and CMS
“When it comes to privacy and accountability, people always demand the former for themselves and the latter for everyone else.”
— David Brin
TODAY, I WANT TO TALK TO YOU about some of our options when it comes to the way that we’ll handle data in the future of healthcare.
Let me introduce you to a little-known AI technique called federated learning. It builds upon the same concepts behind machine learning and has the same advantages while offering some unique benefits when it comes to the security of data and the inherent privacy issues that such algorithms can raise.
Karen Hao covered the technology in a piece for Technology Review, where she explained, “In 2017, Google quietly published a blog post about a new approach to machine learning. Unlike the standard method, which requires the data to be centralized in once place, the new one could learn from a series of data sources distributed across multiple devices. The invention allowed Google to train its predictive text model on all the messages sent and received by Android users without ever actually reading them or removing them from their phones.”
Known as federated learning, we’re yet to see major applications of this new approach, but Hao argues that “its privacy-first approach could very well be the answer to the greatest obstacle facing AI adoption in healthcare today.”
The goal of federated learning is to achieve the perfect balance between privacy and utility, allowing us to benefit from the huge amounts of data that we create as a species without putting any single individual’s data at risk. Unfortunately, it won’t all be plain sailing, and the familiar problem of interoperability also rears its ugly head. “Current state-of-the-art algorithms require immense amounts of data to learn,” Hao explains. “In most cases, the more data, the better. Hospitals and research institutions need to combine their data reserves if they want a pool of data that is large and diverse enough to be useful. But especially in the US and the UK, the idea of centralizing reams of sensitive medical information in the hands of tech companies has repeatedly – and unsurprisingly – proved intensely unpopular.”
This leaves us with something of a chicken and egg situation in which federated learning could be the proof people need to open up their data – but it needs their data to prove itself in the first place. Still, there are plenty of advantages to federated learning and few drawbacks. For example, combining data from different models could create a master model that’s worse than its constituent parts. On top of that, for this approach to work in practice, each hospital would need to have the equipment and the technological knowhow to train machine learning models.
But the disadvantages pale into insignificance once you realize what a federated learning system can do. “You can’t deploy a breast cancer detection model around the world when it’s only been trained on a few thousand patients from the same hospital,” Hao explains. “All this could change with federated learning. The technique can train a model using data stored at multiple different hospitals without that data ever leaving a hospital’s premises or touching a tech company’s servers. It does this by first training separate models at each hospital with the local data available and then sending those models to a central server to be combined into a master model. As each hospital acquires more data over time, it can download the latest master model, update it with the new data, and send it back to the central server. Throughout the process, raw data is never exchanged – only the models, which cannot be reverse-engineered to reveal that data.”
At the moment, it’s still early days for federated learning, and we’re yet to see the full potential of what the approach has to offer us, not just in the healthcare industry but across the board. Don’t be surprised if you start to hear more about federated learning in the future.
Rules from the ONC and CMS
I try not to get too excited whenever I read about some new piece of legislation that’s designed to aid the healthcare industry. I like to think of myself as an optimistic realist, and I could be criticized for not being enough of a cynic, but when it comes to new regulations and legislation in the healthcare industry, it seems as though many of them do more harm than good.
Still, I have to admit that I was impressed when I heard about recent rules from the Office of the National Coordinator for Health IT (ONC) and the Centers for Medicare and Medicaid Services (CMS). The aim is to give people free and easy access to health data while making it easier for them to share that data with third parties, from healthcare providers to tech companies and more.
Writing about the new rules for the Harvard Business Review, David Blumenthal explained, “The ONC rule would require that healthcare providers and EHR vendors make patients’ health data easily and cheaply available to them electronically. Open APIs will make it easy for consumers – acting through authorized third parties [like Apple, Amazon and Google] – to gain direct access to their EHRs and their personal clinical data. The CMS rule aims to liberate patients’ data from insurers. Insurers would have to share data not only with patients and their authorized third parties but also with other insurers if requested by patients. Once again, the technical key would be the adoption of open APIs.”
If implemented correctly, these new rules could make a huge difference, helping us to overcome many of the flaws and inefficiencies in our current system. In my eyes, anything that shifts control of data over to the consumer is a good thing.
Blumenthal concludes his article by noting, “Even with data liberation, there is a substantial gap between theory and practice in creating functional consumer-driven healthcare markets. One problem is that data, by itself, is insufficient. Many consumers are ill-equipped to make sense of the reams of detailed information that populate their EHRs and their claims repositories. The solution: consumers could rely on IT companies to collect, manage and refine the data on their behalf.”
One thing’s for sure, though. Opening up the data and giving patients the ability to determine how it’s used and who it’s shared with will ultimately help to foster creativity and competition in the healthcare industry and help to usher in a brighter future, so long as the two rules work as intended. Only time will tell if they live up to their potential.
Want to learn more?
I talk more about new technologies and their impact on the healthcare industry in my book, The Future of Healthcare: Humans and Machines Partnering for Better Outcomes. Click here to buy yourself a copy.