Skip to content Skip to navigation

Live from the iHT2 Miami Health IT Summit: BI, Analytics Key to Reducing Readmissions

February 10, 2015
by Rajiv Leventhal
| Reprints
Panelists, from L to R: John Christly, Sara Rushineck, Steven Christoff

Predictive analytics is the next frontier in reducing avoidable readmissions, but many challenges remain to keeping patients out of the hospital, according to panelists participating in the Health IT Summit in Miami, sponsored by the Institute for Health Technology Transformation (or iHT2, a sister organization to Healthcare Informatics under our corporate parent, the Vendome Group, LLC).

During the Feb. 10 panel, "Reducing Readmissions with Analytics and Business Intelligence," moderator John Christly, CISO, HIPAA Security Officer at Nova Southeastern University opened the discussion at the Ritz-Carlton Coconut Grove, Miami, by asking, "What is the goal to reducing readmissions, and who is doing this well?" To this end, an audience member rephrased the question to "Who is doing anything?" Any effort to look at the data is better than nothing, he said, referencing the 80/20 concept, in that getting to 80 percent is easy,  but the last 20 percent is the challenging part. "I encourage everyone to study your populations. It will change your whole way of thinking and how you deal with healthcare," the member of the audience said. 

Everything works and nothing works, but activity will certainly lead to things," added panelist Steven Christoff, executive director, clinical integration, at the Ocala, Fla.-based Physician Health Partners. Christoff noted that while Medicare accountable care organizations (ACOs) can collect data easily from the Centers of Medicare & Medicaid Services (CMS), it's not like that everywhere. "[Our payer] is willing to share the data, but I'll get a raw data dump that I often don't know what to do with," he admitted. Christoff said he relies on the payer's portal, which tracks things such as which patients are "frequent flyers" to the ER. However, that claims data is five months old, he noted. "That's the challenge and that's where the business intelligence comes in," he said. "There are three legs to this: the claims data, which is historical; the clinical data, which is today; and the predictive analytics, which we need to move towards. How do we get in front of the cost curve and create patient registries?" Christoff asked.  "We are a reactionary healthcare, and there needs to be a mindset change from both patients and providers. Right now, it is counter intuitive for hospital systems to understand the value of this in a fee-for-service world. It's a paradigm shift," he said. Panelist Sara Rushinek, Ph.D., professor, health informatics, University of Miami, agreed with Christoff on the notion of being more proactive. "We want to know how to deal with the patient that is here or the ones we will have in the future, as opposed to what's in my past," she said. 

Another audience member from Memorial Healthcare System based in Hollywood, Fla., shared his organization's readmissions story, which began, not by its own desire, but rather due to CMS pushing the agenda a few years ago. "We were in the seven stages of grief—in the denial stage—when it came to heart failure, pneumonia, and chronic obstructive pulmonary disease (COPD). We weren't where we wanted to be," he said. "But we started pulling the data and now our readmissions are down significantly," he noted. He did add, however, that successful readmissions strategies involve patients staying in the hospital a little bit longer, and that means someone is losing financially. 

To this end, another audience member, a professor of accounting, brought up a question on leveraging analytics for the purpose of revenue and cash flow. Christoff  replied that the industry and community might bot be mature enough in knowing true costs when it comes to determining revenues and potential ROI. "I'm not an advocate of a single payer system or anything, but we need some conformity. Medicine is fluid, and to figure out that cost of a static number, well I don't know if that's possible today," Christoff said. 

Moving on to the quality of the data, Rushinek noted that it's not the cleanliness of the data that is the issue, but rather the predictive analytics that's needed in the future. "Once we have our predictive models, we need to test it, back test it to ensure that it's correct, and improve it along the way," she said. "Look at what's avoidable. When someone is coming in and is sick, there is a reason for that sickness. How can we do better? What are the best interventions? When a person comes in and says there is no one at home to help, what technology can we use?" she asked. The data is clean enough since it's used for doctors to get paid, added Christoff. "It's a good starting point, but it doesn't tell the full story. But there are a whole lot of unknowns once we let the patient loose from the hospital, and that's where need to get in front," he noted. 

Another issue with BI, analytics and readmissions revolves around data sharing, Rushinek said. "Data is power, but what is the incentive a physician has to give away that data?" she asked.  Rushinek told an anecdote explaining how this lack of an incentive could be a deterrent to better patient care. "I was at University of Miami at the Coral Gables campus and our data people from the medical school visited us. After 10 minutes, I asked someone where the data was. He said that he could only 'see it' on the medical campus, not on the Coral Gables campus. Well, that is a fine line that we must address,"Rushinek  said.

Towards the end of the discussion, the panelists were asked about what is needed in today's analytical word. Rushinek said that IT people aren't needed per se, but the tech savvy people do need to understand healthcare. Christoff added that the analytics need to look at what is different, what the anomalies are. "It's more than just looking flatly at the data," he said.