SIIM: Decision support pioneer shares lessons learned

DALLAS—An enterprise decision support implementation often presents an uphill climb. Savvy adopters can learn from the experiences of pioneers such as New York Presbyterian Medical Center-Weill Cornell Medical Center in New York City. Keith Hentel, MD, vice chairman for clinical operations, detailed the organization’s trials and tribulations during an educational session at the annual meeting of the Society for Imaging Informatics in Medicine (SIIM).

The current healthcare model demands providers focus on high value care to patients, Hentel said. The imaging ordering process represents an opportunity to meet this goal, he continued.

An assessment of New York Presbyterian Medical Center’s process indicated it worked fairly well, with clinicians ordering appropriate exams, according to the American College of Radiology (ACR) appropriateness criteria, almost 80 percent of the time.

Hentel and his team realized reducing inappropriate exams presented an opportunity to improve patient care. They observed that few providers used ACR criteria in the ordering process. Thus, the center embraced clinical decision support and joined the Medicare Imaging Demonstration (MID) project in 2011.

The project required that clinicians order 80 percent of 33 high-cost CPT codes through clinical decision support in the first year, and raised the bar to 90 percent in year two.

Hentel and colleagues ambitiously decided to apply the same criteria to all patients and learned several key lessons in the process.

How decision support is implemented is critical, he said. He emphasized the distinction between imaging utilization and appropriate imaging, explaining utilization can be controlled by deductibles or radiology benefits managers. However, neither method can ensure appropriate imaging. “Understand what you are trying to achieve and have realistic expectations,” he said.

Workflow also is critical. Decision support should not disrupt the physician workflow and should require few clicks; it should be applied only when necessary, minimize redundant data entry and enable users to act on the decision support recommendation.

In addition to prioritizing efficiency, physicians also tend to be competitive, said Hentel. “Provide feedback on performance and show them how they compare to peers.”

Another challenge in system design is that one size does not fit all. Neurosurgeons and internal medicine physicians need different support, so recommendations should be tailored to various physician users.

Hentel concluded by encouraging his colleagues to consider decision support and recommended those that do leverage the experiences of early adopters to set the stage for a smooth implementation.