Deviation Root Cause Verification Rate (DRCVR)

GMI-PH-009Pharmaceutical OperationsPilot-Ready

Definition

Percentage of AI-suggested root causes that an investigator independently verified through evidence beyond accepting the suggestion, such as additional data analysis, operator interviews, or physical inspection.

The Measurement Gap

The Gap: CAPA Closure Rate does not catch this because CAPAs are closed on time. Deviation Rate tracks count, not investigation quality. Repeat deviation rate is a lagging indicator: it takes 6–12 months for wrong root causes to manifest as repeats. Investigation cycle time improved 35% with the AI tool, and the quality metrics review shows green across the board. But the 20% of deviations that are genuinely novel get the AI's best historical match instead of a thorough investigation.

The Decision:

DRCVR below 80% for critical deviations triggers mandatory independent verification for all critical deviation investigations. DRCVR trends inform training investments in investigation methodology. The lagging companion (correlation between unverified AI suggestions and repeat deviations) provides definitive confirmation of whether rubber-stamping is producing real harm.

Formula

Notation:

DRCVR = (Σ Independently_Verified / Σ AI_Root_Cause_Suggested) × 100

Variables:

  • Independently_Verified = deviations where the investigation record contains documented evidence of verification activity beyond accepting the AI suggestion (additional data analysis, operator interviews, equipment inspection, cross-functional review)
  • AI_Root_Cause_Suggested = deviations where the AI system provided a root cause recommendation

Worked Example: Last quarter: 120 deviations, AI provided suggestions for 98. Of 98: 31 showed documented independent verification (14 additional data analysis, 8 operator interviews, 6 equipment inspection, 3 cross-functional review). 67 showed concurrence without additional investigation. DRCVR = 31/98 = 31.6%. By criticality: Critical (12 events) = 58.3%, Major (34) = 41.2%, Minor (52) = 19.2%.

Edge Cases:

  • "Independent verification" must be defined in SOP: writing "agree with AI assessment" is not verification. The investigator must document what they did to verify.
  • Risk-stratify: critical deviations should target 100% verification. Minor deviations may accept 50%.
  • Track the lagging companion: correlation between unverified root causes and repeat deviations

Sign convention: Higher DRCVR is better. 100% means every AI suggestion was independently verified.

Proposed formula, pending validation pilots.

At a Glance

Unit
Percentage
Trend Direction
Higher
Reporting Frequency
Quarterly (6-12 month repeat-deviation lag)
Category
Human-AI Trust & Override Governance
← Batch Record AI Reliance Gap (BRARG)Diversion Alert Investigation Quality (DAIQ) →

Assess Your Organization

The AI-GPS Assessment evaluates your AI governance maturity and recommends KPIs like this one for your implementation roadmap.

Take the Free AI-GPS Assessment