This is a ready-to-use report template. Replace every <<FILL: ...>> placeholder with your own data and route it into management review. A worked filled specimen follows the template. It is an educational aid to adapt and verify; the numeric thresholds shown are illustrative and must be set from your own process history, not copied.
Report control
| Field | Entry |
|---|---|
| Report title | Data Integrity Behavioral Metrics Review |
| Report number | <<FILL: report ID>> |
| Period covered | <<FILL: from>> to <<FILL: to>> |
| Prepared by | <<FILL: name, role>> |
| Reviewed by | <<FILL: QA lead>> |
| Scope | <<FILL: site / area / product family>> |
1. Purpose
This report tests whether the leading indicators of a healthy data integrity culture are behaving plausibly for the period. It does not ask whether targets were met; it asks whether the numbers tell a believable story, because the metrics that suppression distorts look good precisely when something is wrong. A falling out-of-specification rate, a low deviation count, and a quiet reporting channel can each be genuine health or genuine concealment, and only a plausibility check tells them apart.
2. Indicators reviewed
Each indicator is reported with its value, its recent trend, and a plausibility verdict (plausible, watch, or investigate), with the reasoning recorded.
2.1 OOS rate versus production volume
| Field | Entry |
|---|---|
| OOS rate this period | <<FILL: value>> |
| OOS rate prior periods | <<FILL: trend>> |
| Production volume trend | <<FILL: up / flat / down>> |
| Verdict | <<FILL: plausible / watch / investigate>> |
Reasoning to apply: a healthy process finds some out-of-specification results. A rate that falls toward zero while volume rises is not automatically good news; statistically, more volume should surface at least as many problems. Treat a falling rate as a question until you confirm investigations are still being opened when they should be.
2.2 Deviation severity mix
| Field | Entry |
|---|---|
| Minor / Major / Critical counts | <<FILL: n / n / n>> |
| Share that is Minor | <<FILL: percent>> |
| Verdict | <<FILL: plausible / watch / investigate>> |
Reasoning to apply: a realistic distribution includes some major and occasionally critical events. A mix that is almost entirely minor over a long period is a flag for downgrading, not a trophy.
2.3 Result clustering near specification limits
| Field | Entry |
|---|---|
| Distribution shape near the limit | <<FILL: describe>> |
| Results just inside versus just outside | <<FILL: counts or ratio>> |
| Verdict | <<FILL: plausible / watch / investigate>> |
Reasoning to apply: results piling up just inside a limit, with very few just outside, suggest borderline values are being nudged to pass rather than investigated. A natural process spread does not have a wall at the specification.
2.4 Self-reported data integrity concerns
| Field | Entry |
|---|---|
| Concerns raised this period | <<FILL: count>> |
| Trend since channel launch | <<FILL: trend>> |
| Outcomes recorded and fed back | <<FILL: yes / no>> |
| Verdict | <<FILL: plausible / watch / investigate>> |
Reasoning to apply: a rise in self-reported concerns after a confidential channel launches is a sign of trust, not decay. A channel that receives nothing over a long period is more likely untrusted than unnecessary.
2.5 Audit trail review yield
| Field | Entry |
|---|---|
| Reviews performed versus due | <<FILL: n / n>> |
| Discussion items or exceptions raised | <<FILL: count>> |
| Verdict | <<FILL: plausible / watch / investigate>> |
Reasoning to apply: audit trail reviews that are always signed and never raise anything suggest review by rote. A real review of real systems produces occasional questions.
2.6 Repeat and downgraded events
| Field | Entry |
|---|---|
| Repeat deviations on the same step | <<FILL: count>> |
| ”Retrain the analyst” as sole CAPA | <<FILL: count>> |
| Verdict | <<FILL: plausible / watch / investigate>> |
Reasoning to apply: repeat deviations each closed as a one-off, and deliberate acts closed with training, both point to root causes that were never addressed.
3. Overall assessment
<<FILL: a short narrative that names the indicators on watch or investigate and what they collectively suggest about the culture, not an average score>>
4. Actions
| Action | Owner | Due date |
|---|---|---|
<<FILL>> | <<FILL>> | <<FILL>> |
5. References
FDA guidance, Data Integrity and Compliance With Drug CGMP: Questions and Answers (December 2018). ICH Q10, Pharmaceutical Quality System, on management review and continual improvement. Statistical process control concepts (control charts, capability, run rules) for judging plausibility. See statistics in quality.
Confirm the current version of each reference before you rely on it.
Revision history
| Version | Date | Author | Summary of change |
|---|---|---|---|
<<FILL: 1.0>> | <<FILL: date>> | <<FILL: author>> | Initial issue. |
Approvals
| Role | Name | Signature | Date |
|---|---|---|---|
| Prepared by | <<FILL>> | ||
| Reviewed by (QA) | <<FILL>> |
Filled specimen
An illustrative quarter, worked:
| Indicator | Value | Verdict | Reasoning |
|---|---|---|---|
| OOS rate versus volume | 0.4 percent, down from 1.9 percent; volume up 30 percent | Investigate | Rate halved while volume rose; not credible without confirmation that investigations still open |
| Deviation severity mix | 96 minor, 2 major, 0 critical | Watch | Very heavy minor share; check for downgrading over the last year |
| Clustering near limits | 41 results within 1 percent inside the lower limit, 2 just outside | Investigate | A wall just inside the limit suggests borderline results are being passed rather than investigated |
| Self-reported concerns | 9, up from 3 | Plausible | Rise after the channel launched; each has a recorded outcome |
| Audit trail review yield | 24 of 24 done, 0 exceptions | Watch | Full completion but zero findings across 24 reviews suggests review by rote |
| Repeat / downgraded events | 3 repeats on one step; 2 “retrain” CAPAs | Investigate | Same-step repeats and training-only closures point to unaddressed root causes |
Overall assessment for the specimen: the headline metrics look excellent (low OOS, few majors, all reviews done), which is exactly why the plausibility lens matters. Three indicators on Investigate describe a plausible pattern of borderline results being absorbed before they reach the system. The action is to pull the underlying distributions and the recent invalidations, not to celebrate the low numbers.
Common inspection findings this report surfaces early
- A falling OOS rate presented as an achievement without a plausibility check.
- A deviation profile that is almost entirely minor, indicating downgrading.
- Audit trail reviews that are always clean, indicating review by rote.
- Repeat deviations and training-only CAPAs that show root causes are not being found.
How to adapt this report
- Set each threshold from your own process history; the specimen numbers are illustrative only.
- Add indicators that fit your operation (invalidation rate, aborted-run rate, shared-login incidents).
- Run it on a fixed cadence and carry the Investigate and Watch items into management review with owners and dates.
- Confirm each reference in section 5 against its current published version before issue.