Defect Class Coverage Parity Index (DCCP)

GMI-PH-029Pharmaceutical OperationsIntegration-Required

Definition

Proportion of required defect classes where computer vision sensitivity meets or exceeds the human baseline or regulatory minimum. Prevents a strong overall detection rate from masking a defect class the system barely sees.

The Measurement Gap

A CV inspection system for prefilled syringes demonstrates an overall defect detection rate of 96.8%. That number goes into management reviews and regulatory submissions. But the overall rate averages across defect types: 99.2% for particulate matter, 97.5% for fill volume deviations, 91.0% for plunger position defects, and 68.3% for hairline cracks. The 68.3% is below the validated human inspection sensitivity of 82% for that class.

The regulatory filing describes the system as "performing equivalently to or better than qualified manual inspection." It does not, for one defect class with the most serious patient safety implications. The high detection rates on common defects mathematically overwhelm the low rate on rare defects in an average.

Decision enabled:

Any DCCP below 1.0 means the inspection system has gaps requiring compensating controls (supplemental manual inspection for the failing defect class) until remediation and revalidation. DCCP identifies exactly which defect classes need attention, enabling targeted rather than blanket remediation.

Formula

DCCP = defect_classes_meeting_standard / total_required_defect_classes

Where defect_classes_meeting_standard counts categories where CV sensitivity meets or exceeds the human baseline or regulatory minimum.

Worked Example:

A sterile injectable line requires inspection for 8 defect classes per USP <790>/<1790>:

Particulate (visible, >50 μm): CV 99.2% vs human 89.0%. Meets standard.
Particulate (subvisible, 25–50 μm): CV 94.1% vs human 78.5%. Meets standard.
Fill volume high: CV 97.5% vs human 92.3%. Meets standard.
Fill volume low: CV 98.0% vs human 93.1%. Meets standard.
Container closure integrity: CV 96.7% vs human 88.4%. Meets standard.
Plunger position: CV 91.0% vs human 85.6%. Meets standard.
Cosmetic (labeling): CV 93.4% vs human 91.2%. Meets standard.
Hairline cracks: CV 68.3% vs human 82.0%. Does NOT meet standard.

DCCP = 7 / 8 = 0.875

Below 1.0, a GMP compliance gap requiring compensating controls for crack detection until the CV system is remediated.

Sign Convention:

1.0 is the minimum acceptable value, not a target. Below 1.0 means the inspection system has regulatory gaps. This is a binary coverage metric.

Proposed formula, pending validation pilots.

At a Glance

Unit
Percentage
Trend Direction
Higher
Reporting Frequency
Quarterly; per defect-class audit
Category
Predictive Manufacturing Quality
← Batch Release Independence Score (BRIS)Inspection Performance Throughput Decay (IPTD) →

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