3 min read
This blog post has been co-authored by Slawek Kierner, SVP of Enterprise Data & Analytics, Humana and Tie-Yan Liu, Assistant Managing Director, Microsoft Research China.
Using AI models to make real-world impact
Trips to the hospital happen. And while everyone in the industry strives to deliver world-class care for in-patient experiences, everyone—patients and care teams alike, would prefer to avoid those stays at the hospital. The teams at Humana believed they had enough data to explore the possibility of proactively identifying when patients were heading toward a high-risk event, and they put Microsoft Cloud for Healthcare and AI technology to the test.
Humana’s questions were straightforward: How do we take the data we have today and use it proactively? How do we use AI to identify signals in our existing ecosystem that tell us someone might be experiencing a scenario that puts them at risk? And most importantly, how do we engage proactively, meeting our members in their own environment before they end up in an emergency room?
The first approach to monitor chronic patients is often focused on remote patient monitoring and IoT devices, but to approach this challenge, we wanted to take a different, and much bigger, approach with AI. By combining clinical data, key event triggers that might indicate a patient was experiencing deteriorating health, and a combination of predictive models, Microsoft Research and Humana data science teams collaborated on research to explore whether they could develop a system that would identify potential gaps in care among patients and engage high risk patients with care teams that could reach out and offer support.
The power of AI model refinement
The result of the research was a glimpse into the future of AI in health. Health organizations like Humana have spent the last several years developing powerful, single focus predictive models. Humana had existing models that predicted the likelihood of acute hospital admissions in the near future across their 4.9 million Humana Medicare Advantage members, as well as additional models that predict the cost of care and the likelihood of readmissions. Microsoft Research and Humana data science teams brought those models together with structured data to create and test a combination of neural networks and tree-based models with the Microsoft cloud technologies.
Cloud scale tooling was critical to develop the multivariable model, as well as technology in the Microsoft Cloud for Healthcare to unify the variety of patient data streams. Moreover, Microsoft Research designed an advanced deep learning based sequential modeling approach to capture the dynamics of health status which is crucial to accurately predict the likelihood of readmissions. To further increase the robustness of the learned research model, Microsoft Research developed self-paced resampling techniques to address the sample imbalance challenge in this readmission prediction scenario. The research demonstrated that by integrating all these technologies together, the model’s precision improved by over 20 percent. And most importantly, the advanced models were developed using de-identified data, protecting patient information.
Empower care teams to help patients when they need it most
“Model precision is critical here in identifying at-risk members,” shares Mike Hardbarger, Director of Data Science at Humana and a contributor to this project’s research. “Our members deserve personalized, proactive care. Using this model in conjunction with others, not only can we help them avoid hospital readmission, but care teams can have the necessary data to follow-up with a custom plan.” From effective prescription management to addressing food insecurity, a care manager can then work directly with the member to set the next best action into motion.
Proactive problem-solving like this relies on collaboration and innovation. Deep learning allowed research teams including Sean Ma, Lead Data Scientist at Humana, to get an inclusive scope of both science and industry considerations. “Working directly with algorithm authors significantly accelerated progress. I am excited for what’s to come,” says Ma.
Using Microsoft Cloud for Healthcare to do more with your data
This research project is just one step in the evolution of the Humana analytics engine. Enhancements will continue over time as additional research is conducted the model continues to be validated.
Learn more about Microsoft Cloud for Healthcare.