Cloud technology is becoming a crucial solution for healthcare systems of all types. No provider is too large or small to benefit, says Peter Lee, corporate vice president of Microsoft Healthcare.
“From a one-nurse clinic in Kenya to Kaiser Permanente and everything in between, we have the privilege to see these organizations making this shift of their fundamental IT infrastructure from on-premises systems to the cloud,” Lee says, “and a real desire to embrace machine learning and AI and data science.”
The challenge comes in helping providers embrace the shift and in developing new tools and interventions to facilitate seamless, intuitive care. But much of the relevant data to do so comes via electronic health records — a platform Microsoft wants to assist, not displace.
“What the EHR companies do is really important and has a level of complexity and expertise Microsoft would never take on,” Lee says. “But we know that more and more of those systems will be migrating to the cloud. Even the ones that don’t will have a need for cloud services.”
The goal: using cloud technology to help providers collaborate, cut costs and cure disease.
Lee recently spoke with HealthTech at the 2019 Health Datapalooza in Washington, D.C.:
HEALTHTECH: What cloud-related security concerns do healthcare systems have?
It has been a learning process for us to realize that the healthcare industry isn’t aware of the security benefits of the cloud. There’s a level of robustness and knowledge and global view we have that really can’t be matched.
To give you a sense of scale, we have over 270 data centers around the world dispersed in 54 geographical regions. All of those centers are on our own private global network. It’s a level of security that is really impossible to match on a smaller scale.
As little as two years ago, cloud systems weren’t really capable of onboarding large amounts of patient data in accordance with regulatory requirements. That has changed dramatically. Our cloud is now to the point where it is fully compliant with all requirements around HIPAA. You can now bring health data to the cloud, and you can analyze it using machine learning.
HEALTHTECH: How might the roles of AI and the cloud evolve in healthcare?
The one that’s nearest and dearest to our heart has been developing AI technology that empowers people, enabling them to be more productive and improving workday satisfaction and healthcare experience.
For example, we’ve been working very hard on understanding doctor and patient speech and language to reduce the burden of clinical note-taking — a system that would listen to doctor-patient encounters and automate a large portion of that conversation.
Another effort, Project Hanover, is ingesting more than 3,000 research papers in oncology every day and doing automated machine reading to build a knowledge graph being used by tumor boards around the world.
HEALTHTECH: What function can predictive analytics play?
AI can look at health data from large populations and extract insights and make predictions. It’s one of the primary parts of our collaboration with Walgreens, which has more than 9 million customers entering their stores every day.
The company has the opportunity to see an extremely large amount of trend data. So, we might be able to project the demand for, say, the shingles vaccine in different regions of the world. That’s not only a public health benefit but a revenue benefit for Walgreens.
The other side of predictive analytics is being able to look at claims from a large number of people. From that, you’re able to develop models that can make a good prediction; for example, what’s the next health event that a person with a given history most likely could have?
HEALTHTECH: Where is precision medicine headed in this context?
We’ve been working with organizations such as St. Jude Children’s Research Hospital to sequence all of their patients to open up the world’s largest pediatric cancer gene database. We did that in close collaboration with the top researchers at St. Jude to make a foundation for what we hope to be a source of AI-powered research tools to really help get at these rare cancers.
We’ve also been focused on work with adaptive biotechnologies, sequencing and understanding the adaptive immune system. We’ve been building a vast machine learning pipeline, similar to what we do in translating different languages.
The long-term dream is a test for everything: Every year, you take a simple blood test and you get a snapshot of what your adaptive immune system is dealing with. The rate of training data we’re generating for machine systems is about 1 trillion labeled data points per year. We’ll need several trillion for this to work.