HCLF: Maximizing decision support tools

CHICAGO—When it comes to clinical decision support (CDS), content is more generalizable but management of that content may have an even bigger impact, said David W. Bates, MD, MSc, senior vice president for quality and safety at Brigham and Women’s Hospital in Boston, speaking at the Healthcare Leadership Forum on Nov. 15.

Clinical decision support system functions include alerting, reminding, critiquing, interpreting, predicting, assisting, diagnosing and suggesting. Alerts, reminders and critiques are often simple if-then rules while the other functions are harder to program and require more data. For every 1,000 patients coming in for outpatient care, there appears to be 14 patients with life-threatening or serious adverse drug events. Also, for every 1,000 prescriptions written, 40 have medical errors and for every 1,000 women with a marginally abnormal mammogram, 360 will not receive appropriate follow-up care.

Bates cited the “10 commandments for effective CDS” from an article published in the Journal of the American Medical Informatics Association in 2003. One “commandment” is speed is everything. “If the system doesn’t deliver quickly, nothing else really matters. Providers won’t pay attention.”

CDS must fit into users’ workflows. “People want decision support that makes sense. If the system delivers something important or relevant at a time that doesn’t fit into the workflow, users will just ignore it.”

Physicians resist stopping, he said, but providing them with an option has been shown to have a higher chance of success. It’s also important to monitor feedback and respond to that. “Knowledge-based systems have to be managed and maintained and a lot of systems don’t have a robust approach to do so,” Bates said. “You can produce big improvements at multiple stages if the technologies are in place."

Most alerts for decision support for medication safety get overridden, said Bates, but it doesn’t have to be that way. His organization identified a highly selected set of drug alerts for the outpatient setting and made many of them noninterruptive so clinicians don’t have to do anything different. Of interruptive alerts, 67 percent were accepted. “I think it should be in that neighborhood,” he said. “If people are only paying attention to 5 to 10 percent of alerts, you need to switch things up.”

Research has identified ten grand challenges, according to an article published in 2007 in the Journal of Biomedical Informatics. These include the following:

1. Improve the human-computer interface.
2. Disseminate best practices in CDS design and development.
3. Summarize patient-level information.
4. Prioritize and filter recommendations to the user. Brigham and Women’s has several thousand rules in its system, Bates said. “Often, you will get a whole lot of suggestions and it’s hard to reconcile and figure out which are the most important.”
5. Create an architecture for sharing executable CDS modules and services.
6. Combine recommendations for patients with comorbidities.
7. Prioritize CDS content, development and implementation.
8. Create internet-accessible CDS repositories.
9. Use free-text information to drive CDS support.
10. Mine large clinical databases to create CDS.

In the future, “we should be able to implement more complex rules that take into account a wide array of factors. Leveraging analytics will be one of the most important things going forward. Eventually we should be able to be more directive in certain situations."

CDS can deliver great value, Bates said. “Most EHR benefits come from decision support.” To date, most implementations are very limited. “We’re only scraping the surface of this whole area.” He also pointed out that some key infrastructure is rarely present, such as an event monitor.

Bates also said he foresees a lot of decision support going to patients, which should help them follow appropriate care guidelines. For example, he said his hospital is working to 0 percent antibiotic prescribing for bronchitis. Currently, the rate is at 63 percent, he said, because “most patients are convinced an antibiotic will make them well.”

The Healthcare Leadership Forum was sponsored by ClinicalKey and presented by Clinical Innovation + Technology.