Did EP Decide Not to Follow Recommendation of Computer Decision Aid?
By Stacey Kusterbeck
Emergency physicians (EPs) are using clinical decision aids more often, but the computer-generated recommendations sometimes are not appropriate. “Clinicians should never blindly follow any computer-generated recommendation,” says Dean F. Sittig, PhD, professor at UT Health School of Biomedical Informatics in Houston.
Typically, the EP knows additional information about the patient the computer does not. “EPs should think carefully about what the computer is suggesting,” Sittig stresses.
EPs may know something that makes the clinical decision support recommendation irrelevant for a particular patient. In most cases, clinical decision aid alerts are configured to increase sensitivity. “They don’t want to miss anything, so they alert more often than needed,” Sittig explains. “The override rate on most clinical decision support is over 90%.”
This means clinicians ignore most decision support recommendations. Sittig says this is with good reason. “A major problem with computer-generated recommendations is that they can be totally wrong, whereas humans are often close to the right answer even when they are wrong,” he notes.
Clinical decision aids can be wrong for many reasons.1,2 For example, the data the tool uses might be wrong. The tool’s logic can include an error (e.g., not including the route of administration of the medication in the logic for drug-drug interactions). An ED patient might be prescribed a topical medication, flagged by the clinical decision aid because the patient is taking another orally administered medication in which the ingredients normally would pose an interaction with the first medication. However, since the first medication was administered topically, there is no need to worry about the two drugs interacting. “Most alerts do not account for route of administration,” Sittig observes.
The patient might tell the provider they are no longer taking a medication. The computer may not know this, so it produces an alert. Clinicians can help by taking the time to clean up the patient’s current medication list. “However, this may not be realistic in the ED, since other physicians may have ordered and be managing the meds,” Sittig admits.
Emergency providers can document the reasoning for following or not following the computer’s suggestion. “Even if after the fact, it may be clear that the clinical decision support was right; at the time of the event, it wouldn’t be so clear,” Sittig explains.
The medical record should demonstrate the clinician saw the recommendation, thought about it, and decided what to do. “The clinician may still be wrong. But it is now more of a judgment error than simple carelessness,” Sittig says.
EPs can document their reason for overriding alerts at the time of the event. “In general, ED providers are actually better positioned in terms of malpractice when clinical decision aids are present than if they are absent,” says David W. Bates, MD, a professor of medicine at Harvard Medical School.
It is easier to defend malpractice cases when a guideline or clinical decision support is in place, compared to decisions made when no guideline or decision aid was involved, according to Bates. However, there are many recommendations ED providers need to override, often numerous times in a single shift. This is especially common for medication warnings. “The best safe practice is to document why you are not following a recommendation in that specific patient,” advises Bates, a professor of health policy and management at the Harvard T.H. Chan School of Public Health.
It may be related to the patient’s specific condition (e.g., if the risk is uterine cancer, but the patient had undergone a hysterectomy). The patient already may have tolerated the issue that represents a concern (e.g., the patient took two medications together without an issue).
EPs might document that a recommendation is not appropriate because the patient is at the end of life, because he or she presented with a Medical Orders for Life-Sustaining Treatment form, or because the recommendation does not factor in prior surgery (e.g., a recommendation for a pregnancy test in a patient who had undergone a hysterectomy). “It is always reasonable to document why you think a recommendation is not appropriate, especially if there is a high-stakes situation like a new potential stroke or possible cardiac ischemia,” Bates offers.
1. Wright A, Nelson S, Rubins D, et al. Clinical decision support malfunctions related to medication routes: A case series. J Am Med Inform Assoc 2022;29:1972-1975.
2. Wright A, Aaron S, McCoy AB, et al. Algorithmic detection of Boolean logic errors in clinical decision support statements. Appl Clin Inform 2021;12:182-189.
The medical record should demonstrate the clinician saw the recommendation, thought about it, and decided what to do. The clinician still may be wrong. But now, it is more of a judgment error than simple carelessness.
Subscribe Now for Access
You have reached your article limit for the month. We hope you found our articles both enjoyable and insightful. For information on new subscriptions, product trials, alternative billing arrangements or group and site discounts please call 800-688-2421. We look forward to having you as a long-term member of the Relias Media community.