Clinical Decision Support Sensitivity vs. Alert Fatigue

Business Context

Clinical decision support systems, including medication interaction alerts and abnormal result flags, are configured to be highly sensitive in order to minimize the risk of a clinician missing a genuinely dangerous interaction or finding. However, high sensitivity, applied without adequate specificity, generates a large volume of low-value alerts, and clinicians exposed to frequent low-value alerts develop alert fatigue, a documented phenomenon in which clinicians begin reflexively dismissing alerts, including, occasionally, the rare high-value alert that would have prevented genuine harm.

Healthcare TRIZ Resolution

Rather than uniformly reducing alert sensitivity, which risks missing genuine safety signals, or uniformly maintaining high sensitivity, which perpetuates fatigue, the resolution applies tiered alerting based on clinical severity and actionability: high-severity, high-actionability alerts are preserved as hard, interruptive alerts, while low-severity or low-actionability alerts are downgraded to passive, non-interruptive indicators that a clinician can review without workflow disruption. Alert performance is also reviewed on a recurring basis using override-rate data to continuously recalibrate which alerts belong in which tier.

Applicable TRIZ Principles

Principle 3 – Local Quality Differentiate alert presentation by clinical severity and actionability rather than a single uniform alert style for all triggers.

Principle 23 – Feedback Use ongoing override-rate data as feedback to continuously recalibrate alert tiering rather than setting it once and leaving it static.

Principle 21 – Skipping Allow low-value alert review to happen quickly and passively, skipping the interruption of an active clinical workflow.

Expected Outcome

Reduced overall alert volume

Preserved high-severity alerting

Lower measured alert fatigue

Improved clinician trust in alerts

Decision Indicators

Early indicators that this contradiction is limiting organizational performance include:

Override rates for interruptive alerts exceeding 90 percent, suggesting most alerts are not perceived as clinically valuable

No differentiation between high-severity and low-severity alert presentation

No recurring review process to recalibrate alert configuration based on override data

Clinicians self-reporting that they routinely dismiss alerts without reading them

A known missed high-severity alert incident where investigation reveals the clinician had become desensitized to interruptive alerts generally

Monitoring these indicators helps clinical informatics leadership identify when alert configuration itself, rather than clinician diligence, is the root cause of missed safety signals.