By: Jeff Deal, Program Chair, Predictive Analytics World Healthcare
In anticipation of his upcoming conference co-presentation at Predictive Analytics World for Healthcare 
Boston, Sept 27-Oct 1, 2015, we asked William Wood, VP, Medical Affairs at St. Joseph Healthcare, a few questions about incorporating predictive analytics into healthcare. Catch a glimpse of his co-presentation, Improving Care Coordination and Reducing Readmissions Using Real Time Predictive Analytics, and see what’s in store for the second annual PAW Healthcare conference in Boston.
Q: In your work with predictive analytics, what area of healthcare are you focused on?
A: We are focused on managing at risk patients to provide better care coordination to reduce unnecessary utilization. This includes reducing unnecessary inpatient readmissions and emergency room visits for patients managed by our primary care physicians. We also use the analytics to identify patients with high mortality that might benefit from services like palliative care. We also use the analytic tool to identify uninsured patients who are accessing our emergency department and primary care practices so we may connect them with the healthcare exchange, MaineCare (Medicaid), and community resources. We are able to focus on specific chronic diseases, identify gaps in care, and opportunities for interventions.
Q: What outcomes do your models predict?
A: There are two basic types of clinical risk models we use, population based risk models, and event based risk models. All the models are updating nightly providing near real time risk for all of our patients under management.
This population risk models are person based and are used by population health managers to understand each patient or member’s risk of an event in the future 12 month period. This includes the following models:
- Emergency Department (ED) Visit Risk
- Inpatient (IP) Admission Risk
- Predicted Future Cost
- Risk of a Stroke
- Risk of Diabetes
- Risk of an AMI
- Risk of Mortality
The event based risk models are encounter based and are triggered by an inpatient admission or an emergency department visit. These risk models are applied upon admission showing the likelihood of a 30 day readmission for inpatients and a 30 day return to the emergency department for emergency patients.
- IP 30 Day Readmission
- ED 30 Return Visit
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: We believe we are better managing inpatient readmissions and appropriate emergency room utilization by proactively targeting the at risk population for proactive care management. Predictive risk scores help us use our limited care resources more effectively by targeting the higher risk populations. The tools have also helped us design a care management model that covers the care continuum. We are able to focus our efforts on the high risk population, working more efficiently to allocate our resources to those that are in most need, and at highest risk.
Q: Can you describe a successful result, such as the predictive lift of your model or the ROI of an analytics initiative?
A: The implementation of computerized predictive risk models immediately eliminated the manual work effort of our nurses assessing risk during discharge planning. This was a savings of 1000s of nurse hours that are now spent on providing the interventions to patients. For readmissions and emergency room utilization the recent trends are showing a decrease which we expect to translate into hard ROI from reduced penalties. We are embarking on ACO type population risk contracts which would provide the financial incentives and ROI for better population management, but that is longer term.
Q: What surprising discovery have you unearthed in your data?
A: We are identifying at-risk patients that were not obvious to the clinicians previously. Although many patients are well known to care management, we routinely identify “unknown” patients at risk for mortality or MI/CVA/DM through the tool. We have improved our follow-up for patients as they have transitioned from other healthcare organizations/skilled facilities to primary care. We now have the ability to view patients who are currently in a facility along with their potential risks. This allows us to better plan for their needs upon discharge and to see when they have left the facility. We are not dependent upon a call to alert us from the facility. We are able to be proactive and not reactive in care. This not only increases our productivity but decreases the risk of patients experiencing negative outcomes and readmissions during that critical 72 hours after discharge.
Q: What areas of healthcare do you think have seen the greatest advances or ROI from the use of predictive analytics?
A: We are currently and primarily still a fee for service hospital, which is financially incented to manage volume and throughput based on what happened to a patient retrospectively. As we participate in ACO and population financial risk contracts along with bundle payment models, we have implemented the risk based tools to start managing patients prospectively.
The biggest benefit to our organization has been the shift in thinking to better understand what will happen to a patient after they leave our 4 walls. We have also been better able to identify gaps in care as we refine the care management model.
We are starting to see early results from this mind set change in reduced readmissions and ED visits. As the financial incentives continue to change, we will see continued investment in staffing, training and systems to better manage patients proactively which inevitably requires predictive risk analytics.
Q: Sneak preview: Please tell us a take-away that you will provide during your talk at Predictive Analytics World.
A: Predictive analytics is only a part, an important one, but still only one part of the improvement cycle. An organization needs to invest in the adoption of risk tools to best understand how to apply them. This includes workflow integration and staff training. The risk tools gave us better information that we didn’t have previously, and we needed to then understand what to do to a patient given their risk profile. The predictive math doesn’t provide the second step, and we needed to invest in the development of care interventions tied to patient risk profiles and then train our staff accordingly. The adoption, workflow integration, and training are a necessary part of the total cost of ownership to take advantage of predictive risk scores.
—————————–
Don't miss Williams’s co-presentation, Improving Care Coordination and Reducing Readmissions Using Real Time Predictive Analytics, at PAW Healthcare on Tuesday, September 29, 2015 from 10:05 to 10:50 am. Click here to register for attendance.
By: Jeff Deal, Conference Chair, Predictive Analytics World Healthcare