Clinical Alert Override Rate

Clinical alert override rate measures the share of clinical decision support (CDS) alerts, such as drug-drug interaction warnings, allergy alerts, and dose-range checks generated by a computerized provider order entry (CPOE) system, that a clinician dismisses or overrides rather than acts on. It is the standard quantitative signal used to detect and manage "alert fatigue": the well-documented tendency for clinicians to become desensitized to alerts once the volume of low-value warnings becomes overwhelming.

Why it matters

Published override rates for drug interaction alerts commonly range from roughly half to over ninety percent, and a high rate is not automatically a safety failure: many interruptive alerts fire for interactions that are clinically insignificant in context, or repeat an alert the clinician has already acted on earlier in the same order set, so a well-tuned system deliberately fires fewer, higher-value alerts rather than trying to drive the override rate to zero. What genuinely matters for safety is the trend over time, the distribution across severity tiers, and whether clinicians document a reason when they override a high-severity alert; a rising override rate on high-severity, well-evidenced interactions is a real governance concern even when the average across all alerts looks stable.

How it's calculated

Override rate = alerts overridden / total alerts fired × 100

Segment by:
  - severity tier (e.g. contraindicated, major, moderate)
  - alert type (drug-drug interaction, allergy, duplicate therapy, dose range)
  - whether an override reason was documented

A "documented override rate" tracks the share of overrides that carry a
recorded justification, which is a governance measure in its own right.

Worked example

A hospital's CPOE system fires 10,000 drug-drug interaction alerts in a month, of which 8,700 are overridden, giving an overall override rate of 87%. Segmenting by severity shows that of 500 "contraindicated" alerts, 60 are overridden (12%), while of 6,000 "moderate" alerts, 5,700 are overridden (95%). The moderate-tier figure is broadly consistent with published benchmarks and is not, by itself, cause for concern; the contraindicated-tier figure warrants individual case review, and the fact that only 340 of the 500 overrides at that tier carry a documented reason is the more actionable governance finding.

Data sources and caveats

The electronic health record's audit log, or the CDS vendor's own alerting module, records each alert-fired and alert-response event, including whether the clinician entered free-text or structured justification. Comparing override rates between organizations, or even between departments in the same organization, requires checking that the underlying alert rule sets and severity tiering are the same; a hospital with an aggressively tuned rule set will show a lower override rate for reasons that have nothing to do with clinician behaviour.

Pitfalls

• Treating the raw override rate as a single safety score: it conflates well-justified overrides of low-value alerts with unsafe overrides of genuinely dangerous interactions; always segment by severity.
• No override-reason capture: without a documented reason, it is impossible to distinguish "this alert was wrong" from "this alert was right and the clinician made an unsafe judgement call", which is the actual distinction that matters for patient safety.
• Alert-rule inflation over time: adding more alerts to "be safe" without pruning low-value ones is the direct cause of rising override rates and alert fatigue; alert governance should include regular review and retirement of poorly performing rules, not only monitoring.
• Comparing rates across systems with different interruption design: an interruptive, hard-stop alert produces different override behaviour from a passive, non-blocking one, so the two are not directly comparable metrics.

Sources

• Peer-reviewed literature on clinical decision support alert fatigue, widely published in journals including JAMIA and npj Digital Medicine
• ONC / HealthIT.gov, health IT safety guidance on clinical decision support
• Institute for Safe Medication Practices (ISMP), guidance on CDS alert design and governance