For almost 40 years, automated alerts from ground proximity warning systems have helped airline pilots avoid accidents. The healthcare field has a similar goal and is working toward creating a system of alerts that help clinicians avoid making errors. For instance, modules inserted in electronic health record (EHR) systems can check drug-to-drug interactions or warn a provider that a patient is allergic to a drug before the doctor orders it. But just as with pilots, one of the challenges to effective performance is being bombarded with too much information. How you filter that information to only the most critical items is crucial to success.
A consortium of some leading healthcare providers, software vendors and research organizations is making impressive strides at creating and sharing clinical decision support (CDS) tools and services across organizational boundaries. But the amount of information that has to be synthesized and then inserted in the electronic workflow presents a tremendous challenge.
"There is a huge amount of knowledge management in medicine," says Dean Sittig, a professor at the School of Biomedical Informatics in The University of Texas Health Science Center at Houston and a member of the UT Houston-Memorial Hermann Center for Healthcare Quality and Safety. "A doctor has to know something like a million concepts. We don't know how to represent the relationships between those very well. We can make a recommendation on one disease, but what if the person has complex comorbidities? Can the system respond to that? And the implications are life-critical. Being right 999 times out of 1,000 isn't good enough."
"Alerts in other industries are more binary," says Mark Van Kooy, M.D., director of informatics for Aspen Advisors, a Denver-based consulting firm. "You've entered the wrong information and are headed down the wrong path and it stops you. In healthcare, the kinds of alerts you create for patient safety deal with issues such as contra-indications about drug interactions, and the level of complexity is staggering."
Sittig co-authored a paper in 2007 called "Grand Challenges in Clinical Decision Support." The list included:
- improving the human/computer interface;
- disseminating best practices in CDS design, development and implementation;
- summarizing patient-level information;
- prioritizing and filtering recommendations to the user; and
- creating an architecture for sharing executable CDS modules and services.
Most of those challenges persist today, according to Sittig. "We are still closer to the beginning of the work than the fruition," he says. There is still a lot of KM work to be done, just in the naming of concepts. "In a sector like finance, there aren't these problems with labels and coding of concepts, but we have 20 different types of diabetes, and we can't agree across systems how to define and code them," Sittig says. "So, even at the lowest level of knowledge management, we struggle."
But the degree of difficulty does not make the goal less significant, and clinicians and medical informaticists are reporting slow progress. Virginia Mason Health System in Seattle did an extensive research study published in 2011 on clinical decision support built into ordering systems for high-volume imaging procedures involving back pain. The conclusion was that targeted use of imaging clinical decision support is associated with large decreases in inappropriate utilization of advanced imaging tests, says Craig Blackmore, M.D., scientific director for the Center for Health Care Solutions at Virginia Mason.
Gaining physician buy-in
"In imaging, there is a general acceptance that there is overutilization. Our goal was to put in a system that allows us to block imaging orders in certain scenarios," Blackmore explains. The research team defined cases for which the physicians would have to describe why they are ordering an imaging test, and they receive a yes or no based on clearly defined criteria. If they aren't able to identify those, they get a hard stop.
"For our system, it is not an alert, it is a wall," Blackmore notes. "We had to give them a backup. They can call a specialist in this area and seek an override, but I don't think that has happened even once."
Doctors don't tend to like change, he adds, so it was important to include them in the design process to get physician buy-in. The research project was part of a larger quality improvement effort that is now expanding to other areas of imaging at Virginia Mason.
But Blackmore says that although there is a lot of enthusiasm around clinical decision support, implementation has proven difficult. "It sounds great in theory, but once you try to apply these systems outside of the healthcare entities where they were developed, they aren't as successful," he says. The Center for Medicare & Medicaid Services (CMS) is piloting imaging alert systems around the country, Blackmore adds, but health systems have encountered pushback from physicians and IT infrastructure problems.