On Monday, I reviewed a study by Kaveh Shojania and colleagues which showed that computer-based alerts generated a 4.2% improvement in the process of care, a result they deemed to be “below thresholds for clinically significant improvement.”
It was merely the latest in a series of recent, negative studies concerning clinical decision support (CDS), yet I continue to believe CDS can improve the quality of care. We just haven’t figured out how to do it yet. But it’s early. Fewer than 10% of US physicians in outpatient settings even use EHRs, after all.
What we need is a research agenda that can inform design efforts for CDS tools. So I’m proposing one in my next 2 posts. I’d love to hear how you think I can make it better.
The way I see it, researchable questions for CDS fall into 2 categories: system design issues and user-related issues. Below, I list some key questions in each category.
CDS System Design Issues
What kind of CDS tools work best? Shojania’s study focused on alerts, which interrupt clinical workflows (like e-prescribing and charting) to provide information that helps physicians correct errors and make better decisions. It makes sense to deliver real-time decision support like this, but the problem is that alerts interrupt normal cognitive processing by physicians. As such, they can frustrate physicians and paradoxically, create new errors as I discussed here.
By contrast, reminders—another form of CDS—are presented before or after clinical processing, at a time when physicians may be relatively clear-headed and receptive to input. A third form of CDS, standing orders, attempts to prevent errors before they occur by codifying best practices.
Is there an optimal number of alerts? Shojania’s study did not look at the overall quantity of alerts presented to physicians. Did physicians see, on average, one alert for every 5 patients, or was it more like 5 alerts for every patient? Too many alerts will likely overwhelm physicians, no matter how clinically relevant they might be. Amid a deluge, physicians might choose to ignore many of them. With CDS, more is not necessarily better.
Can performance assessment report cards increase physician responsiveness to CDS? As a rule, physicians respond to meaningful feedback about their performance, especially if it is compared with their peers. This is likely to be true in cases where physicians receive data regarding their response rates to CDS, although the idea has not been studied so far as I know.
Can incentive payments increase physician responsiveness to CDS? Similarly, pay-for-performance incentive programs can influence physician behavior. Perhaps such programs could have a similar impact on responsiveness to CDS tools.
In what instances should prohibitive and permissive alerts be used? A prohibitive alert prevents the physician from carrying out something she intended to do, such as ordering a medication that is contraindicated for her patient. Permissive alerts require physicians to acknowledge they are doing something unusual, and then permit them to carry-on as they wish. Physicians routinely weigh dozens of factors before initiating treatment plans for their patients. The complex decision making environment in which they function creates challenges for CDS designers in deciding when, if ever, to say “no.”
How long should decision support messages persist in the EHR and where should they appear? Anything that tends to pile-up in a a physician’s in-box can discourage action. Yet many physicians prefer to handle at least some aspects of patient care at a time of their choosing. Perhaps CDS should remain visible for a relatively short period, and then scheduled to re-appear as a “second notice” at a later time. Once again, CDS design trade-offs will likely affect performance of the CDS system as a whole.
What is the best way to display decision support messages? The visual display might take the form of a pop-up screen, flashing lights around key data entry boxes or something else. Certain background colors might prove more likely to generate a response, as might certain text colors, fonts, point sizes, or the use of italicized, underlined, bolded or capitalized words. Things like these have been worked out for jet pilots. Perhaps some quality time with the Air Force could help us out here?
I’ll pick-up this discussion on Friday.
Glenn Laffel, MD, PhD
Sr. VP Clinical Affairs, Practice Fusion
