DNP/715 Week 4: Clinical Decision Support Design and Evaluation, sample paper

The Dosing Alert Everyone Overrides: Redesigning Weight-Based Medication Alerts in a Pediatric Emergency Department Using the Features That Make Decision Support Work

[Student Name]

University of Phoenix

DNP/715: Information Systems and Health Care Delivery Technology

Week 4 Assignment

[Instructor Name]

[Date]

The emergency department and figures are composites written for a model paper.

Our pediatric emergency department sees about 32,000 children a year. When a clinician orders a weight-based medication, the record checks the dose against a range and fires an interruptive alert if the dose falls outside it. A pharmacy report showed that these alerts fired 11,400 times last year and were overridden 96% of the time. Two months ago, a tenfold acetaminophen overdose order was overridden and caught only by the pharmacist. This paper evaluates the alert and redesigns it using evidence on decision support.

What Decision Support Can Achieve

Bright et al. (2012) systematically reviewed 148 randomized trials of clinical decision support systems and found that both commercial and locally developed systems improved health care processes, including preventive services, ordering of studies and prescribing, with odds ratios around 1.4 to 1.7. Few studies measured clinical outcomes, costs or unintended consequences. Decision support can work, but its effects depend heavily on design.

Features That Predict Success

Kawamoto et al. (2005) reviewed 70 trials of decision support and found that systems improved practice in 68% of trials and that four features independently predicted success: automatic provision as part of clinician workflow, provision of recommendations rather than just assessments, delivery when and where the choice is made and computer-based delivery. Of 32 systems with all four features, 94% improved practice.

Evaluating Our Alert Against the Features

Our alert is automatic, computer-based and appears at the time of ordering, meeting three of four features. It fails the second: it tells the clinician the dose is "outside range" without recommending a correct dose or showing the calculation. It also fires on doses that are clinically appropriate, such as rounded doses to the nearest available tablet.

When nearly every alert is false, clinicians learn to dismiss them all, including the one that would have stopped a tenfold dose.

Why Clinicians Override

Ancker et al. (2017) studied alert acceptance among primary care clinicians and found that about a quarter of drug alerts and a third of reminders were repeats for the same patient within a year, and that acceptance fell as clinicians received more alerts, especially repeated ones, with reminder acceptance dropping by 30% for each additional reminder per encounter. They found no evidence that workload itself explained declining acceptance. Our override rate fits this pattern: most of our alerts are low value, and clinicians have learned to dismiss them.

Analyzing Our Alerts

I reviewed a random sample of 200 overridden alerts. In 71%, the dose was appropriate but slightly outside the range because of rounding to available formulations. In 18%, the weight was entered in pounds but interpreted as kilograms or was outdated. In 7%, the range itself was wrong for the indication. Only 4% were true dosing errors, and 3 of those 8 reached the pharmacist.

Redesign

First, adjust ranges to allow rounding to available dose sizes within 10%, removing most false alerts. Second, require weight entry in kilograms with a plausibility check against age, displayed on the order screen. Third, when an alert fires, show the weight, the calculated correct range and a suggested dose that can be accepted with one click, meeting Kawamoto et al.'s (2005) recommendation feature. Fourth, make alerts for doses more than twice the maximum hard stops requiring pharmacist override, so the most dangerous errors cannot be dismissed.

Pharmacist Partnership

Pharmacists already review every pediatric order in our department. The redesign gives them a report of all overridden alerts each shift, so a second check focuses on the orders most likely to be wrong. Decision support and human review work best as layers, each catching what the other misses.

Tiering

Not all alerts need to interrupt. Minor deviations can appear as passive information on the order screen, while major deviations interrupt. Tiering reduces alert volume and restores meaning to interruptive alerts.

Learning From the Near Miss

The tenfold acetaminophen order was reviewed with the clinician, who said the alert looked like all the others and was dismissed by habit. That account, offered without blame, was the most persuasive evidence for redesign presented to the review committee.

Evaluation Plan

Process measures include alert volume, override rate and the proportion of overridden alerts that were true errors. The key safety measure is dosing errors reaching the patient or intercepted by the pharmacist, per 1,000 weight-based orders. Balancing measures include time to medication administration and clinician satisfaction. Bright et al. (2012) noted that few decision support studies measured unintended consequences; I will track whether hard stops delay urgent medications.

Human Factors

The alert's layout also matters. The current alert appears as a gray box with small text and an "Override" button in the position where "OK" usually sits. The redesign places the suggested dose first in large type, puts "Accept suggested dose" as the default button and requires a reason from a short list to override. Small design choices shape what busy clinicians do.

Governance

Changes will go to the hospital's medication safety and informatics committees for approval, with a named owner responsible for reviewing alert performance quarterly.

Testing Before Go-Live

Before activation, the redesigned alerts will be run silently for four weeks, logging when they would have fired without showing them to clinicians. Comparing silent-mode alerts with pharmacist interventions will show whether the new ranges catch true errors and how many alerts clinicians would see. Silent testing avoids inflicting a flawed alert on busy clinicians.

Weight as the Foundation

Weight-based dosing is only as accurate as the weight. In our sample, nearly a fifth of overridden alerts involved weight problems. Nurses now weigh every child in kilograms at triage, the scale is set to kilograms only, and the order screen shows the weight and when it was recorded. Fixing the input is as important as fixing the alert.

Stakeholder Input

Emergency physicians, nurse practitioners, nurses and pharmacists reviewed the redesign in a simulation session. Their feedback added a display of the last recorded weight date.

Expected Effect

If ranges are adjusted for rounding, most false alerts should disappear, reducing volume by roughly two thirds. With fewer and more accurate alerts, acceptance should rise, consistent with the finding that acceptance falls as repeated alerts accumulate (Ancker et al., 2017). The hard stop should prevent any tenfold error from reaching a patient.

The DNP Role

My role was to quantify the alert's performance, apply evidence on decision support design, lead the interprofessional redesign and plan its evaluation.

Conclusion

Our weight-based dosing alert was overridden 96% of the time because most alerts were false, it gave no recommendation and repeated low-value alerts eroded attention. Evidence on decision support features and alert fatigue guided a redesign that corrects ranges, fixes weight entry, recommends doses, tiers alerts and adds hard stops for dangerous doses. Evaluation will measure both safety and alert burden, including unintended consequences.

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, Article 36. https://doi.org/10.1186/s12911-017-0430-8

Bright, T. J., Wong, A., Dhurjati, R., Bristow, E., Bastian, L., Coeytaux, R. R., Samsa, G., Hasselblad, V., Williams, J. W., Musty, M. D., Wing, L., Kendrick, A. S., Sanders, G. D., & Lobach, D. (2012). Effect of clinical decision-support systems: A systematic review. Annals of Internal Medicine, 157(1), 29-43. https://doi.org/10.7326/0003-4819-157-1-201207030-00450

Kawamoto, K., Houlihan, C. A., Balas, E. A., & Lobach, D. F. (2005). Improving clinical practice using clinical decision support systems: A systematic review of trials to identify features critical to success. BMJ, 330(7494), 765. https://doi.org/10.1136/bmj.38398.500764.8F