Rush CIO Lac Tran on using predictive analytics to improve outcomes, efficiency
Rush University Medical Center counts on predictive analytics for a number of quality and efficiency improvement efforts--from reducing readmissions to treating patients at risk for stroke and cardiac arrest quickly and efficiently to reducing wait times, diversions and boarding in the emergency department.
Rush's CMIO, Julio Silva, M.D., will talk about these programs and share results at an executive breakfast panel discussion, on Wed. March 6, during the HIMSS 2013 annual conference in New Orleans.
Co-hosted by FierceHealthIT and the College of Healthcare Information Management Executives, the panel will focus on using predictive analytics to improve care and efficiencies. (Pre-registration is required, so be sure to sign up today.)
In the meantime, we caught up with Lac Tran, Rush's CIO, (left) to get a sneak peek of how RUMC is using predictive analytics to improve care for patients at risk for stroke and cardiac arrest and to improve efficiency and throughput in the ED.
FierceHealthIT: How is Rush using predictive data analytics to improve patient care in the emergency room?
Tran: We did a case study for three years of people who were admitted to the emergency room and ended up admitted to the hospital because of cardiac arrest or a stroke. We used that data to detect patterns. When patients check into the emergency room with symptoms that could indicate that they have a minor stroke or heart attack, the emergency room docs pass that on directly to the stroke specialists and cardiology so that they can quickly treat the patient. With stroke and cardiac arrest, time is the most important thing. And a specialist is available 24/7.
FHIT: How does using this historic data compare to other forms of decision support or even doctors' own diagnostic experience?
Tran: Typically during an emergency the patient doesn't know where they are. In true emergencies they can't tell you their symptoms. But through the data that's in their electronic records and through the data from the study we can predict outcomes and risk based on their condition.
Yes, it seems like a decision support system, but it's a little bit more than that, because the predictive analytics has built the decision support system.
All the symptoms that indicate a heart attack or a stroke have been gathered before. It is in a data warehouse with information about patients who have had a stroke or cardiac arrest. Then you do the analytics on that and you form the predetermined patterns. And that pattern will build your EMR decision support. It's a continuing effort. With very solid predictive analytics you build a better decision support.
FHIT: You've talked about the of benefits predictive analytics to quality of care. What's the business case?
Tran: There are a lot of benefits financially. Let's say that you look at the patterns of readmissions for the population. You can determine the common causes of the readmissions. You can really change that process and reduce readmissions.
FHIT: And how does it impact efficiency?
Tran: With predictive analytics you can zoom into the problem area and then fix that problem to improve throughput in the ED.
Here's the cause and effect: When the patient checks into the ED, it's typical that the nurse will start doing your vital signs. They move you to the exam room in the ED and you sit there and wait for the physician to come. And then the physician looks at you and reviews the vital signs and they probably order some tests, depending on the condition and diagnosis. Then you're lying there waiting for the results. Then the tests indicate that we need to admit you to be observed overnight or to be scheduled for surgery.
This is where the problem usually occurs. When they admit you, you sit there and sit there and you wait for the transport people to take you to admitting. If the communication between the three groups--the transport group, the nursing station and the emergency room--does not work out well, the patient as to sit there in the exam room for a long time.
That's one scenario where you can use the analytics to detect where the bottleneck is, focus on that, and see the resolution to reduce the time. We can predict when those bottlenecks are likely to occur.