Machine Learning Times
Machine Learning Times

2 years ago
Wise Practitioner – Predictive Analytics Interview Series: Benjamin Cleveland, at UnityPoint Health


In anticipation of his upcoming conference presentation at Predictive Analytics World for Healthcare Las Vegas, June 3-7, 2-18, we asked Benjamin Cleveland, Data Scientist at UnityPoint Health, a few questions about incorporating predictive analytics into healthcare. Catch a glimpse of his co-presentation, From When You Arrive to After You Leave:  Integrating Predictive Models into a Cross Continuum Heat Map to Inform Care Coordination and Reduce Readmissions, and see what’s in store at the PAW Healthcare conference in Las Vegas.

Q. In your work with predictive analytics, what area of healthcare are you focused on?

A. Most of my work has focused on either clinical quality or operations applications. We’re starting to move into other core business functions like HR and Finance, which haven’t received a lot of attention in the health care data science community.

Q. What outcomes do your models predict?

A. The ones in production today include length of stay, appointment no-show risk, 30 day readmission risk, risk of sepsis, and expected probability of blood transfusion. Some in our pipeline are revenue projections, year-end contract shared savings, and nurse flight risk. I also use machine learning to risk-adjust and find variation in clinical care pathways.

Q. How does predictive analytics deliver value at your organization? What is one specific way in which it actively drives decisions or impacts operations?

A. It reduces the cognitive or operational load of our care team and then triggers processes at key decision points in the delivery of care. For example, our readmission risk models identify which patients might need more intensive follow up services post-discharge and their readmission heat map informs when over time they are most at risk and the services should be scheduled. In our clinics, using the model to drive patient appointment confirmations would require only contacting 25% of patients while still capturing 60% of those who will end up no-showing.

Q. Can you describe a successful result, such as the predictive lift of your model or the ROI of an analytics initiative?

A. Our readmissions models improve on the LACE score in discrimination ability by about 25-30%. Using predictive modeling to identify Sepsis patients in the ED in lieu of a heuristic model called SIRS we improved the sensitivity by about 20-25% while also decreasing the false positive rate. This literally helps to save lives by acting on Septic patients earlier while also resulting in less alert fatigue for our providers.

Q. What surprising discovery have you unearthed in your data?

A. While studying geographic location of 30 day readmissions predictions based only on clinical data in a high crime city we couldn’t really discern much. We expected to see clusters of high risk patients in high crime and poorer neighborhoods. However, when we zoomed out on the map, we found a big cluster of very high risk patients in rural Iowa. It was an Amish community!

Q. What areas of healthcare do you think have seen the greatest advances or ROI from the use of predictive analytics?

A. I think the use of predictive analytics to inform deterioration in a variety of settings and time horizons has and will increase our care team’s ability to manage populations more effectively. We need to ensure the outcomes of populations, but can’t actively manage each individual at all times. Like I said earlier, it’s all about reducing the cognitive and operational load of the care team while remaining effective. Predictive analytics is well suited to this task.

Q. Sneak preview: Please tell us a take-away that you will provide during your talk at Predictive Analytics World.

Deploying predictive analytics into production has been termed “the last mile challenge”. I would propose this is really the “last 1000 miles challenge” in that actually realizing value from predictive analytics goes beyond performance metrics and into enterprise adoption.


Don’t miss Benjamin’s presentation, From When You Arrive to After You Leave:  Integrating Predictive Models into a Cross Continuum Heat Map to Inform Care Coordination and Reduce Readmissions, at PAW Healthcare on Tuesday, June 5, 2018 from 10:30 to 11:15 AM. Click here to register for attendance.

By: Jeff Deal, Conference Chair, Predictive Analytics World Healthcare

Leave a Reply

Pin It on Pinterest

Share This