Machine Learning Times
Machine Learning Times

2 years ago
Wise Practitioner – Predictive Analytics Interview Series: Louis F Rossiter Ph.D, College of William & Mary


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

In anticipation of his upcoming conference presentation at Predictive Analytics World for Healthcare Las Vegas, June 16-20, 2019, we asked Louis F Rossiter Ph.D, Professor of Public Policy at the College of William & Mary, former Secretary of Health & Human Resources for the Commonwealth of Virginia, former Deputy for Policy to the Administrator of the Centers for Medicare & Medicaid Services, a few questions about their deployment of predictive analytics. Catch a glimpse of his keynote presentation, Healthcare Market Competition: The Changing Players and the Hand of Government, 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: The short answer is we try to focus on what our clients want. Rather than picking key performance measures we think are useful and perhaps easy to analyze, we listen to the customer. Our hospital and physician customers tend to emphasize the regulatory and payer performance measures they are required to monitor. These tend to be simplistic, basic performance measures that get stale after a while, or top out. I am thinking of the Centers for Medicare & Medicaid Services Hospital Quality Assurance and HEDIS measures from the National Committee for Quality Assurance. Unfortunately, it is still rare for a hospital or health system to measure in order manage cost of quality unless an external entity has required them to do so. Avoidable readmissions has probably been the most frequently predictive analytics request.

Q: What outcomes do your models predict?

A: We just published in Healthcare Financial Management Magazine, the professional journal for hospital and other Chief Information Officers across the country, a study of what we called hospital Overstays. Overstay patients were defined as any patient with a length of stay (LOS) that exceeds the case-mix-adjusted expected LOS by two or more days. The expected LOS was the Medicare average LOS for that diagnosis-related group plus two standard deviations. We use predictive modeling to find the seven-hospital system’s five highest-volume major diagnostics categories (MDCs) with the most patients with overstays. That led to a concerted effort on an ongoing basis to get those patients out of the hospital sooner, breaking down all the barriers to faster discharge. The lost financial margin for these stays was enormous. The patients also had higher all-cause 30-day readmission rates because they were overstaying in the hospital.

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: At the Predictive Analytics World conference in June I plan to talk about how our modeling has enabled and supported pay for performance and value-based purchasing for our customers. For hospitals, so much of their efforts, in reducing overstays for example, has tremendous overlap with nursing. To get the results I report above, nursing had to provide greater support for weekend discharges and/or services. The discharge planning nurses had to engage the patient’s support network early and monitor patient medication compliance. Nursing was important in reshaping patient and family exceptions using a protocol to help predict the day of discharge. We are hard at work on a prototype to do traditional forecasting using patient census trends, course per patient day, and budget vs actual performance. This is the first step in building a more dynamic machine learning model of nurse supply and patient care demand.

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

A: Yes. Regarding overstays, in the preintervention period, 12,104 discharges were associated with 120,121 days of care. After the intervention, 10,475 days of care were saved by reducing overstays among the types of patients who had tended to experience them previously. Moreover, the readmission rate fell 8 percent, and it, too, has been trending downward ever since.(Cite: Masiulis, K et al. “New Approaches and Technologies for Improving Cost Performance in Health Care, HFM Magazine, January 2019)

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

A: We did an analysis of a census of claims data from New York to determine whether patients from Long Island were being re-admitted to a New York City major medical center after discharge from a Long Island hospital. The theory was that if things did not go as planned after an initial discharge, patients or their physicians were being referred to the major medical centers. It turned out not to be the case. Our predictive analytics showed that the likelihood of readmission to New York City was actually quite low. Although the patient did not always return to the index hospital.

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

A: ROI will be the theme of my presentation. Pay-for-performance, value-based purchasing, pricing optimization and transparency, wellness/disease management, and evidence-based medicine, have been, and will continue to drive patients from the hospital. Most hospitals actually today derive barely half their revenues from inpatient services, and the rest from outpatient services. I do not mean the hospital will disappear. It is that hospitals will become less the major cost center to the payers and play a much smaller role. Predictive analytics will be a major enabler by short circuiting the traditional patterns of care, and helping the payers to steer patients toward insurance products and clinical services that best meet the patients needs, there by cutting back on unnecessary or poor quality care.

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

A: The disruption I am talking about will have an impact in three areas:  clinical, operational, and financial. The successful supply chain companies combine their technical horsepower with analytics and mastery of operations. Predictive analytics is one important tool in operational improvement that will put the traditional expensive hospital at a disadvantage going forward.


Don’t miss Louis’ keynote presentation, Healthcare Market Competition: The Changing Players and the Hand of Government, at PAW Healthcare on Tuesday, June 18, 2019 from 1:30 to 2:10 PM. Click here to register for attendance.

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

This content is restricted to site members. If you are an existing user, please log in on the right (desktop) or below (mobile). If not, register today and gain free access to original content and industry news. See the details here.

Comments are closed.