By: Dan Hogan, CEO and founder, Medalogix

Gartner analysts predict by 2016 that 70 percent of the most profitable companies will manage their business processes using real-time predictive analytics.

A study by Singletrak Analytics and DataGen concludes that when structured well, bundled payment models can be effective and profitable for hospitals, especially those that can reduce the cost of each episode of care.

By combining the power of predictive modeling with the requisite awareness of the organizations clinical capabilities that bundled payment structures now hinge upon, health care providers are more likely to meet the needs of the Affordable Care Act, improve care, and be more profitable.
 
What is predictive modeling?
Predictive modeling is a statistical method that uses data mining to assess the probability of specific future outcomes. As health leaders search for the best technological approaches to improve care and adapt to ACA provisions, predictive modeling has been recognized as a highly effective solution. It is particularly effective in reducing hospital readmissions. For example, the application Medalogix's predictive modeling approach helped risk stratify Alternate Solutions HomeCare's existing patient census in a dozen of their agencies.  The more accurate identification of patients most at risk of readmission, and their subsequent intervention on those patients reduced Alternate Solutions' average rate of 30-day readmissions by nearly 36 percent in a mere nine-month period.

There are many ways this data mining technique can be used to predict patient risk and outcomes, one not often discussed, but extremely useful, is predictive modeling's application in the new paradigm of the Bundled Payments for Care Improvement initiative. 

What is BPCI?
In a nutshell, the Bundled Payments for Care Improvement initiative (BPCI) is a payment model where the Center for Medicare and Medicaid Services (CMS) reimburses providers collectively a lump sum by patient episode. According to CMS, the approach will lead to higher quality and more coordinated care at a lower cost to Medicare.

In a non-BPCI approach, providers are paid separately in a fee-for-service model without a quality of care measurement. For instance, if a patient had a lower extremity major joint replacement, he or she would visit a surgeon and a post acute provider and perhaps a physical therapist throughout his or her care regimen. Traditionally each provider would each receive a Medicare/Medicaid reimbursement based on the number of times the patient visited. Under BPCI, CMS allocates a single sum to all parties based on a previously agreed upon cost of episode - in this case, a major joint replacement of a lower extremity. If care costs are less than the previously agreed upon amount, providers may keep the difference.

In January 2013, CMS announced more than 500 providers who enlisted to test this payment approach.

How can predictive modeling positively affect BPCI?

• Assessing the best Diagnosis Related Groups (DRGs) to bundle - In laying the groundwork for BPCI, your agency can determine which DRGs to bundle. Orthopedic care is widely considered an advisable bundle because experts believe all parties involved will benefit: CMS is in favor of orthopedic bundles because the episodes are typically among the most costly DRGs; providers are in favor because they understand that orthopedic episodes could easily be made more cost effective through better care coordination; patients are in favor because they will benefit from improved care and efficiency.

Predictive modeling can mine your agency's existing data to determine if your data agrees with your assumption that it makes sense to bundle a certain set of DRGs. You'll know in advance if the bundle is statistically beneficial

• Determining the appropriate discount level - To set an agreeable and reasonable discount level, providers can use predictive modeling to forecast how much a specific episode will cost, based on the agency's data and state and nationwide benchmarks. Once the discount level is set, providers can continue to deploy predictive modeling to understand how a patient's unique medical history affects his or her personal cost of care, then compare it to the agreed discount level.
• Calculating transfer - One of the main objectives of BPCI is to better coordinate care among multiple providers. Predictive modeling can identify when a patient is ready to be moved to the next phase of care. For instance, predictive modeling can determine how long after surgery your specific patient is ready for release to post-acute care. This improves care quality and efficiency.

Overall, predictive modeling's finding can offer an objective point of reference for providers as they attempt to gauge their own internal capabilities on the front end of BPCI.  This type of data analysis creates a reliable framework that providers can act upon to improve patient outcomes and care efficiency within the new parameters of reimbursement, now and in the future. 

Dan Hogan is the CEO and founder of Medalogix, a Nashville, Tennessee-based health care technology company.