DNP 825 Module 3 Clinical Decision Support and Alert Fatigue Example

Ninety-One Percent Ignored: Redesigning Drug Interaction Alerts for Nurse Practitioners in a Primary Care Network

Student Name

Doctor of Nursing Practice Program, Aspen University

DNP 825: Health Information Management and Informatics

Instructor Name

Month Day, Year

Ninety-One Percent Ignored: Redesigning Drug Interaction Alerts for Nurse Practitioners in a Primary Care Network

Clinical decision support exists to give the clinician who needs it the relevant fact at the moment a decision is made. Drug interaction alerts are among the most common forms, and among the most often ignored. When clinicians override nearly every alert, the few that matter are lost among the many that do not, and the alerts add time and frustration without improving safety. This paper analyzes drug interaction alerts in a composite primary care network of 14 clinics staffed mainly by nurse practitioners, reviews evidence on alert overrides and redesign, and proposes a redesign with measures to evaluate it.

The Problem in the Network

An extract of three months of alert logs showed that the network's prescribers received 11,420 drug interaction alerts, about 14 per prescriber per clinic day, and overrode 91% of them. The most frequent alerts involved combinations that are common and usually managed deliberately, such as an ACE inhibitor with a diuretic, or low-dose aspirin with an antidepressant. Nurse practitioners described the alerts as noise and admitted clicking through them without reading. A review of 50 randomly selected overrides found that most were clinically reasonable, but three involved combinations that warranted a change, including a strong interaction between a macrolide antibiotic and a statin.

What the Evidence Shows

High override rates are not unique to this network. A review of 17 studies found that clinicians overrode drug safety alerts in 49% to 96% of cases, that many overrides were justified, and that the alerting systems themselves invite error: they fire for too many harmless combinations, word their warnings vaguely and interrupt work that did not need interrupting, so clinicians learn to click past them (van der Sijs et al., 2006). In outpatient practice, a study of more than 157,000 alerts on two million medication orders found that 52.6% were overridden and that about half of overrides were appropriate, with appropriateness varying widely by alert type (Nanji et al., 2014).

Design changes can help. Comparing two academic hospitals that used the same drug interaction checking service, one of which presented alerts by severity with hard stops for the most severe while the other did not, researchers found that clinicians accepted 29% of alerts at the tiered site compared with 10% at the non-tiered site, and that all of the most severe alerts were accepted at the tiered site compared with only 34% at the other (Paterno et al., 2009).

The Redesign

The redesign has four elements. First, a multidisciplinary team of two nurse practitioners, a physician, a clinical pharmacist and an informatics nurse reviews the 100 most frequent interaction alerts, which account for most of the volume, and classifies each into one of three tiers. Tier 1 interactions, those with high risk of serious harm and a safer alternative, trigger an interruptive alert that requires an action or a documented reason. Tier 2 interactions, those that require monitoring or dose adjustment, appear as a passive notice in the prescribing screen without interrupting. Tier 3 interactions, those that are well known and routinely managed, are retired from alerting and listed in a reference view.

Second, Tier 1 alerts are rewritten so that the first line states the specific risk and the recommended action, for example suggesting azithromycin in place of clarithromycin for a patient on simvastatin, rather than a generic warning. Third, the system uses patient context to suppress alerts that are irrelevant, such as alerts for a potassium interaction when a potassium level was checked in the past month. Fourth, the team reviews override data quarterly and adjusts tiers.

Workflow and the Nurse Practitioner's Day

The redesign also considers when alerts appear. In the current system, interaction alerts fire at the moment of signing a prescription, after the clinician has already decided on the drug and often while the patient is waiting. Moving Tier 2 information into the medication search screen, where the prescriber sees it while choosing among options, turns a warning into guidance at the point of decision. For refills, which generate many alerts for combinations the patient has taken safely for years, alerts will fire only if a new interacting drug has been added since the last refill.

Nurse practitioners in the pilot clinics will also be asked to record, in one click, why they override any remaining Tier 1 alert, choosing from a short list such as benefit outweighs risk, patient already monitored, or alert not applicable. Those reasons will guide the quarterly review and show whether any Tier 1 alert should be moved to a lower tier or rewritten.

Safety Safeguards

Reducing alerts carries a risk of missing a harmful interaction. The team mitigates this by retiring alerts only after pharmacist and physician agreement, keeping all interactions visible in the medication profile, starting with a three-month pilot in four clinics, and monitoring for adverse drug events reported through the safety system and identified through a monthly pharmacist review of a sample of prescriptions.

Measures

Outcome and process measures include the number of interaction alerts per prescriber per day, the override rate for Tier 1 alerts, the proportion of Tier 1 alerts leading to an order change, and prescribers' ratings of alert usefulness. Balancing measures include adverse drug events related to interactions and pharmacist interventions for interactions not caught by alerts. The goal is to reduce interruptive alerts by at least 70% while increasing the acceptance rate for Tier 1 alerts, the pattern that tiering produced in the comparison study.

Results from the pilot clinics will be compared with the ten clinics not yet using the redesign before it spreads.

Conclusion

When nurse practitioners override 91% of drug interaction alerts, the alert system is failing both as a safety tool and as a use of clinicians' time. Evidence that override rates are high everywhere, that many overrides are appropriate, and that tiering by severity increases acceptance of the most serious alerts supports a redesign that interrupts only for high-risk interactions, gives specific recommendations and retires routine alerts. Careful safeguards and measurement ensure that making alerts fewer also makes them safer.

References

Nanji, K. C., Slight, S. P., Seger, D. L., Cho, I., Fiskio, J. M., Redden, L. M., Volk, L. A., & Bates, D. W. (2014). Overrides of medication-related clinical decision support alerts in outpatients. Journal of the American Medical Informatics Association, 21(3), 487-491. https://doi.org/10.1136/amiajnl-2013-001813

Paterno, M. D., Maviglia, S. M., Gorman, P. N., Seger, D. L., Yoshida, E., Seger, A. C., Bates, D. W., & Gandhi, T. K. (2009). Tiering drug-drug interaction alerts by severity increases compliance rates. Journal of the American Medical Informatics Association, 16(1), 40-46. https://doi.org/10.1197/jamia.M2808

van der Sijs, H., Aarts, J., Vulto, A., & Berg, M. (2006). Overriding of drug safety alerts in computerized physician order entry. Journal of the American Medical Informatics Association, 13(2), 138-147. https://doi.org/10.1197/jamia.M1809