HIM 500 Module 6 Decision Support Short Paper Example

Ninety-One Percent Overridden: Making Decision Support Work at Bramble Bay Medical Center

[Student Name]

Southern New Hampshire University

HIM 500: Healthcare Informatics

Module Six Short Paper

[Instructor Name]

[Date]

The organization, setting and figures below are a composite written as a model document. No real employer, client, colleague or patient is described.

Ninety-One Percent Overridden: Making Decision Support Work at Bramble Bay Medical Center

The hospital's pharmacy reports that clinicians override 91% of the medication alerts the record displays. The vendor's answer is a larger decision support package that adds predictive models for deterioration and readmission. Before adding more, the hospital needs to understand why the alerts it already has are ignored. This paper reviews what research says about effective decision support, explains the override problem, reports a local review of overridden alerts and proposes a plan.

What Makes Decision Support Work

Decision support has a solid evidence base when designed well. Kawamoto et al. (2005) analyzed randomized trials and found that systems improved clinical practice in most of them, and identified the features most closely tied to success: support that appears on its own inside the clinician's normal work, a recommended action rather than a bare assessment, timing that coincides with the decision itself and delivery by computer. Bright et al. (2012) reviewed a larger set of trials and found consistent improvements in process measures, such as ordering recommended tests and preventive care, but thinner evidence for effects on clinical outcomes, costs and efficiency. Together, these reviews suggest that decision support reliably changes what clinicians do when it fits their work, while its effect on patients depends on whether the recommended actions matter.

Why Alerts Are Overridden

Bramble Bay's override rate is high but not unusual. A review of drug safety alerting in order entry systems reported that clinicians overrode anywhere from 49% to 96% of alerts, mostly because the warnings seemed irrelevant, repetitive or trivial (van der Sijs et al., 2006). Ancker et al. (2017) tied acceptance to circumstance: busier clinicians and alerts seen many times before were both linked to more dismissals, which is what alert fatigue looks like in practice. Overrides are not always errors. Nanji et al. (2018) reviewed medication alert overrides among inpatients and judged some alert categories far more likely than others to be dismissed for good reason, and that inappropriate overrides were associated with preventable adverse drug events. The goal, therefore, is not a low override rate for its own sake but alerts that deserve attention.

A Local Review of Overridden Alerts

A pharmacist and the informatics manager reviewed a random sample of 300 overridden alerts from one month and judged whether each override was clinically appropriate. Table 1 summarizes the results. Duplicate therapy and low-severity interaction alerts made up most of the volume and were almost always overridden appropriately, while a small number of high-severity interaction and renal dosing alerts were overridden inappropriately. Those few inappropriate overrides are the safety problem; the many appropriate ones are the noise that hides them.

Table 1. Review of 300 Overridden Medication Alerts

Note. Review by a clinical pharmacist and the author; composite data.

An Alert Rationalization Plan

The plan has four parts. First, retire or silence alert types with high volume and near-universal appropriate override, starting with duplicate therapy alerts for scheduled and as-needed forms of the same drug and low-severity interactions, which would remove about 70% of displayed alerts. Second, tier the remaining alerts so that only high-severity interactions and dose alerts interrupt ordering, while others appear passively. Third, redesign the renal dosing alert to show the patient's current kidney function and a recommended dose, following the finding by Kawamoto et al. (2005) that recommendations work better than warnings. Fourth, clean up allergy lists so intolerances are recorded as intolerances, a health information task that removes false allergy alerts at their source.

Governance will prevent the list from growing again. A decision support committee with pharmacy, nursing, physicians and informatics will approve every new interruptive alert, require a stated target and review each alert's override rate and appropriateness each year.

The Health Information Contribution

Several causes of low-value alerts sit in data that health information staff manage. The allergy list is the clearest example: 41 of the 300 overridden alerts in the review fired on allergies, and most of those were recorded intolerances such as nausea from an opioid. The problem list matters too, because dose and interaction rules that depend on diagnoses cannot work if conditions are missing or outdated. Medication lists that carry discontinued drugs create duplicate therapy alerts that clinicians rightly ignore. A rationalization plan that changes only the alert logic will leave these sources of noise in place, so the health information team will run a quarterly audit of allergy entries and coordinate with pharmacy on medication list reconciliation at admission.

Measures

Four measures will track the plan: interruptive alerts per 100 orders, the override rate for high-severity alerts, the share of overrides judged inappropriate in a quarterly sample and pharmacist interventions for renal dosing errors. A falling override rate alone would not prove success; the key measure is whether the alerts that remain lead to changed orders when they should.

What This Means for Predictive Models

The vendor's predictive models for deterioration and readmission would add new alerts to a system clinicians already tune out. Until the medication alerts are rationalized and the governance process exists, adding models risks deepening alert fatigue and burying their signals. The Module Five legal review also noted that the hospital should see the developer's information on how each model was tested before relying on it. The recommendation is to complete rationalization first and then evaluate models one at a time through the full process.

Conclusion

Decision support works when it delivers relevant recommendations within workflow and fails when it floods clinicians with warnings they have learned to ignore. Bramble Bay's review shows that most overrides are reasonable responses to low-value alerts, while a few dangerous ones hide among them. Retiring, tiering and redesigning alerts under active governance should make the remaining alerts worth reading, which is the precondition for any expansion.

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), Article 765. https://doi.org/10.1136/bmj.38398.500764.8F

Nanji, K. C., Seger, D. L., Slight, S. P., Amato, M. G., Beeler, P. E., Her, Q. L., Dalleur, O., Eguale, T., Wong, A., Silvers, E. R., Swerdloff, M., Hussain, S. T., Maniam, N., Fiskio, J. M., Dykes, P. C., & Bates, D. W. (2018). Medication-related clinical decision support alert overrides in inpatients. Journal of the American Medical Informatics Association, 25(5), 476-481. https://doi.org/10.1093/jamia/ocx115

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