The number of alerts generated by clinical decision support systems can be overwhelming for clinicians. It’s little wonder that the Joint Commission has long identified alert fatigue as a critical safety issue for providers, particularly given how many turn out to be unimportant or even irrelevant.
Unfortunately, however, there’s a flipside to this issue. Sometimes, CDS alerts can actually prevent care problems, clearly suggesting that clinicians shouldn’t dismiss them out of hand either. In fact, recently-published research found that at least in an ICU setting, overriding alerts might be associated with patient harm.
The study, which appeared in BMJ Quality & Safety, focused on the nature and impact of medication-related CDS overrides in the ICU. To conduct the analysis, the authors gathered data on adults admitted to any of six ICUs between July 2016 and April 2017.
The research team looked at a total of 2,448 overridden alerts from 712 unique patient encounters. The studies looked at patients with provider-overridden CDS alerts for dose, drug allergies, drug-drug interaction, geriatric and renal alerts. They also looked at how frequently patients suffered adverse drug events following alert overrides and the risk of adverse drug events given the appropriateness of the overrides.
A team of two independent reviewers concluded that while 81.6% of the overrides were appropriate, the roughly 19% remaining were inappropriate.
Researchers found that inappropriate overrides were associated with a greater risk of adverse drug events. In addition, they concluded that they could find more potential and definite adverse drug events following inappropriate overrides than appropriate overrides. They also found that inappropriate overrides were associated with an increased risk of adverse drug events.
Overall, inappropriate overrides were six times as likely to be associated with potential and definite adverse drug events. That’s too big a correlation to ignore.
One thing the study doesn’t comment on is how the alerts were presented. Given that they may have been presented through multiple interfaces, the question arises of how big a difference those interfaces make in how clinicians respond to alerts. It could be that these interfaces have more impact than the clinical content of the alerts.
Bottom line, this problem may very well fall under the larger umbrella of usability problems. Just one more reason why the industry needs to keep a laser focus on improving usability in HIT across the board.