HCD 602 Module 1 Article Analysis Example

Too Many, Too Often: An Analysis of a Study on Alert Fatigue in Primary Care

Student Name

MS in the Science of Health Care Delivery, Arizona State University

HCD 602: Introduction to Health Informatics for Future Health Professionals

Instructor Name

Month Day, Year

Too Many, Too Often: An Analysis of a Study on Alert Fatigue in Primary Care

Article Selected and Why

I chose Ancker et al. (2017), published in BMC Medical Informatics and Decision Making, for three reasons. It is original research in a peer-reviewed medical informatics journal. It addresses a central problem in health informatics, the gap between the promise of clinical decision support and clinicians' frequent overriding of alerts. And it tests competing explanations rather than describing the problem, which allows a real analysis of methods.

Research Question and Background

Clinicians override most drug safety alerts; a review found override rates from 49% to 96% (van der Sijs et al., 2006). The usual explanation is alert fatigue, but the term is loosely defined. Ancker et al. (2017) tested two mechanisms. Cognitive overload predicts that clinicians accept fewer alerts when they have more work, more complex work or more uninformative alerts. Desensitization predicts that acceptance of a new alert declines over time with repeated exposure.

Data and Methods

The study is a retrospective cohort using electronic health record data from January 2010 through June 2013 for 112 primary care clinicians in a network of community health centers. It analyzed drug alerts and clinical practice reminders. Workload was measured by numbers of encounters and patients, complexity by patient comorbidity and alerts per encounter and informational value by the share of alerts that repeated for the same patient within a year. For desensitization, the authors tracked acceptance of newly deployed reminders over time. Regression models estimated the association of each factor with alert acceptance.

Findings

Repetition was common: roughly one drug alert in four, and one reminder in three, duplicated a warning that same patient had triggered within the year. Acceptance was associated with work complexity and repeated alerts but not with the amount of work. Acceptance dropped sharply as reminders piled up in a visit, by roughly 30% per additional one, and fell a further 10% or so with each five-point increase in the proportion that were repeats. Newly deployed reminders did not show declining acceptance, so the desensitization hypothesis was not supported. Nurse practitioners were four times as likely as physicians to accept drug alerts (Ancker et al., 2017).

Critical Analysis

Strengths. The study uses real-world data from many clinicians over several years, avoiding the bias of self-reported attitudes. It turns a vague term into testable hypotheses, and the finding that one hypothesis was not supported adds credibility. Measures are drawn directly from system logs.

Limitations. Acceptance is an imperfect outcome: some overrides are appropriate, so lower acceptance does not always mean worse care. The data come from one network and one electronic health record, which limits generalizability. As an observational study, it shows associations, not causes; clinicians with complex patients may differ in ways that also affect acceptance. The desensitization test relied on a small number of new reminders.

Contribution. The study shifts attention from clinicians' workload, which organizations cannot easily change, to alert design, which they can.

Implications for Practice

The findings point to two design changes: stop re-firing an alert a clinician has already seen for that patient, and cap how many alerts can appear in a single visit. This is consistent with an expert panel that identified 33 classes of drug interactions that should not interrupt clinicians, which accounted for over a third of displayed interaction alerts at one institution (Phansalkar et al., 2013). Health systems should measure repeats and alerts per encounter as part of alert governance.

Conclusion

Ancker et al. (2017) provide evidence that alert fatigue in primary care is driven by the number and repetition of alerts rather than workload or habituation over time. Despite limits in outcome measurement and generalizability, the study gives informatics teams a practical target.

References

Ancker, J. S., Edwards, A., Nosal, S., Hauser, D., Mauer, E., & Kaushal, R. (2017). Effects of workload, work complexity, and repeated alerts on alert fatigue in a clinical decision support system. BMC Medical Informatics and Decision Making, 17(1), 36. https://doi.org/10.1186/s12911-017-0430-8

Phansalkar, S., van der Sijs, H., Tucker, A. D., Desai, A. A., Bell, D. S., Teich, J. M., Middleton, B., & Bates, D. W. (2013). Drug-drug interactions that should be non-interruptive in order to reduce alert fatigue in electronic health records. Journal of the American Medical Informatics Association, 20(3), 489-493. https://doi.org/10.1136/amiajnl-2012-001089

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