Take a more disciplined approach to QI efforts for better results

CHICAGO—Quality improvement efforts should “practice what they preach” by following processes similar to those designed to improve workflow and patient care, said George Hripcsak, MD, MS, director of medical informatics services at NewYork-Presbyterian Hospital/Columbia, speaking at the 2014 Healthcare Leadership Forum.

A lot of quality improvement is driven by an impending Joint Commission review, an internal event or an incident at another nearby facility, he said. At Columbia, “we’ve prioritized all the projects and made a big difference in how we do things from a haphazard thing to a little bit better.” Before, people called their favorite IT person to get something implemented, he said. As a result, it wasn’t always the best ideas that were implemented.

Columbia has created a platform that sits underneath all the systems and has data that can be shared among them to drive decision support. They estimated that one in 300 orders are for the wrong patient when the doctor writes it. “We measured that in our institution and set out to reduce that.” A “Swiss cheese model” of systems prevented that one in 300 incorrect order from going through but they implemented a semi-interruptive alerts that questions whether the provider is sure he or she has the right patient.

They also wanted to implement a simple way to help clinicians check those orders. They learned that the patient name doesn’t help, nor does medical record number or gender. The most effective factor turned out to be the chief complaint. In fact, the chief quality officer designed the screen and when they went live two weeks later, Hripcsak said, he ordered a head CT for patient with knee pain and noticed the discrepancy from the alert.

The institute reduced the wrong order rate by 30 percent and two years later they maintained a 25 percent reduction.

Hripcsak discussed NewYork Presbyterian/Columbia’s provider-facing efforts to improve patient handoffs. He said a pediatric attending told him he doesn’t read notes anymore, he just writes them, because he gets better information from lab values, vital signs and the resident’s sign-out notes. Hripcsak said that when you have two sets of books—one you want the IRS to see and one that includes what’s really going on—you get arrested. “We’re trying to get those to be one set of documentation that’s effective.”

His organization also is using a quality checklist because you get to have 50 alerts. They found that users override standard alerts 90 percent of the time if they are not hard stops. He said they designed a flu shot alert as a floating pop-up which providers follow two-thirds of the time. While he said the answer is not to turn on more alerts, he “doesn’t really have the answer on how to do it.”

For the issue of 30-day hospital readmissions, Hripcsak said they want “to predict not just who will be readmitted but why so we can intervene.” To achieve that goal, they’re working on pulling in new data sources including ICD-9 codes, lab values demographics and text documents.

Visit history and lab results are the biggest factors when it comes to predicting readmissions, he said. And, while “we’ve proven from the literature how important social and behavioral determinants of health are, we’re not mentioning them enough in the record. That’s probably why they’re not adding as much value as I would have hoped.”

Despite the hurdles, there are solvable challenges, Hripcsak said. Speaking of his participation in the federal Meaningful Use workgroup, he said it was a complicated effort. “You don’t get many chances to put $30 billion into health IT.” He pointed out that had the rate of adoption continued at the same rate as prior to 2009, the U.S. would have achieved 100 percent EHR adoption in 500 years. “We had to do something different.”

The issue, he said, was that if only 4 percent of the market is using EHRs, the vendors have no reason to invest capital in building better EHRs. “Now, there is a lot more money in the market and it’s worth building better systems and getting more customers. Now is when we’re going to see real improvement.”