This is a ready-to-use field checklist. Replace every <<FILL: ...>> placeholder with your own specifics and route the completed copy through your normal audit reporting and CAPA process. A worked filled specimen follows the template. This checklist follows the five-layer method in Data Integrity Self-Audit: A Compliance Checklist; read that article first if the layered method is new to you.
Audit header
| Field | Entry |
|---|---|
| System(s) / area audited | <<FILL>> |
| Audit period covered | <<FILL: from>> to <<FILL: to>> |
| Sample size and selection method | <<FILL>> |
| Lead auditor (independent of area) | <<FILL>> |
| Audit dates | <<FILL>> |
Every row below requires a named evidence artifact, not an impression. Use the Evidence column to record the specific record, log export, or observation examined, its identifier, and the date reviewed.
Layer 1: Infrastructure controls
| Item | Pass/Fail/NA | Evidence reviewed |
|---|---|---|
| All GxP systems synchronized to one authoritative time source; users cannot change local time | <<FILL>> | <<FILL>> |
| Concurrent events across two systems agree in timestamp within documented tolerance | <<FILL>> | <<FILL>> |
| No shared or generic accounts in use on GxP systems | <<FILL>> | <<FILL>> |
| Departed-staff accounts disabled within the SOP window, with evidence | <<FILL>> | <<FILL>> |
| Backup schedule documented and followed; at least one dated restore test per critical system | <<FILL>> | <<FILL>> |
| Oldest archived records confirmed still readable | <<FILL>> | <<FILL>> |
| No unattended, logged-in GxP session reachable by an unauthorized person | <<FILL>> | <<FILL>> |
Layer 2: System configuration
| Item | Pass/Fail/NA | Evidence reviewed |
|---|---|---|
| Audit trail enabled, captures create/modify/delete/approve with old and new values | <<FILL>> | <<FILL>> |
| Audit trail unalterable, including by system administrators | <<FILL>> | <<FILL>> |
| Documented audit trail review procedure exists, with defined frequency and scope | <<FILL>> | <<FILL>> |
| Audit trail reviews actually performed and documented on schedule | <<FILL>> | <<FILL>> |
| Electronic signatures carry two components, explicit meaning, and are bound to the record | <<FILL>> | <<FILL>> |
| No single role both generates and independently approves the same record (segregation of duties) | <<FILL>> | <<FILL>> |
| A documented baseline configuration exists and current state matches it, or every difference traces to an approved change | <<FILL>> | <<FILL>> |
Layer 3: Procedural controls
| Item | Pass/Fail/NA | Evidence reviewed |
|---|---|---|
| SOPs require data recorded directly to the GxP record, not scratch paper or temporary notes | <<FILL>> | <<FILL>> |
| OOS procedure requires investigation before retest and prohibits invalidation without documented justification | <<FILL>> | <<FILL>> |
| Data review SOP requires reviewers to examine the audit trail, not just the final result | <<FILL>> | <<FILL>> |
| DI training is role-specific and documented for everyone who generates GxP records | <<FILL>> | <<FILL>> |
Layer 4: Work practice verification
| Item | Pass/Fail/NA | Evidence reviewed |
|---|---|---|
| Sampled analytical runs reconcile across LIMS and CDS with no undisposed injections | <<FILL>> | <<FILL>> |
| Instrument timestamps consistent with staffed hours per access logs | <<FILL>> | <<FILL>> |
| Sampled batch records show GMP activities recorded when they occurred | <<FILL>> | <<FILL>> |
| No sign of blank forms filled out in advance | <<FILL>> | <<FILL>> |
| Access-log review shows no unexplained after-hours, concurrent-location, or bulk-modification activity | <<FILL>> | <<FILL>> |
Layer 5: Culture indicators
| Item | Pass/Fail/NA | Evidence reviewed |
|---|---|---|
| Staff interviews indicate analysts feel safe reporting an unexpected result | <<FILL>> | <<FILL>> |
| No pattern of pressure to keep re-testing until an assignable lab error is found | <<FILL>> | <<FILL>> |
| Deviations and OOS investigations described as learning, not blame, exercises | <<FILL>> | <<FILL>> |
| DI issues tracked, trended, and reviewed at the quality system level | <<FILL>> | <<FILL>> |
| Leading indicators (OOS invalidation rate, batch amendment rate, after-hours access, re-injection rate) reviewed and trended | <<FILL>> | <<FILL>> |
Summary
| Field | Entry |
|---|---|
| Total items assessed | <<FILL>> |
| Pass / Fail / NA counts | <<FILL>> |
| Critical findings (list) | <<FILL>> |
| Major findings (list) | <<FILL>> |
| Minor findings (list) | <<FILL>> |
| Overall conclusion | <<FILL>> |
| Lead auditor signature / date | <<FILL>> |
| QA sponsor signature / date | <<FILL>> |
References
FDA Data Integrity and Compliance With Drug CGMP guidance (December 2018). MHRA GxP Data Integrity Guidance and Definitions (March 2018). 21 CFR Part 11; EU GMP Annex 11.
Confirm the current version of each reference before issue.
Filled specimen (excerpt)
| Item | Pass/Fail/NA | Evidence reviewed |
|---|---|---|
| Audit trail enabled, captures create/modify/delete/approve with old and new values | Pass | CDS-HPLC-07 audit trail export, 01-07 June 2026, 14 entries reviewed, all show old/new values |
| Audit trail reviews actually performed and documented on schedule | Fail | Batch disposition packages BR-2026-0441 through 0452 (n=12), none contain a CDS audit trail review record |
| No single role both generates and independently approves the same record | Pass | Role matrix export for CDS-HPLC-07, 06 June 2026; Analyst role has no Approve right |
| Sampled analytical runs reconcile across LIMS and CDS with no undisposed injections | Fail | Sequence HPLC-07-2206-031, LIMS shows one result (99.2%), CDS shows 3 undisposed injections preceding it |
Overall conclusion (excerpt): Two Major findings raised (audit trail review not performed for six months; one sequence showing undisposed injections preceding a reported result, escalated per the triage path as a potential testing-into-compliance pattern and referred for record forensics). CAPA opened for both; report closed pending CAPA completion.
Common inspection findings this checklist prevents
- A self-audit that reads as a documentation review, with no named evidence artifact behind any answer.
- Findings recorded as pass/fail with no supporting record identifier, which does not survive a challenge.
- A clean result from an audit that never actually pulled a sample of real records.
How to adapt this checklist
- Add or remove items to match your actual system inventory; the five-layer structure should stay intact even if the specific checks change.
- Set the sample size and selection method before the audit starts, and record it in the header, not after the fact.
- Pair this checklist with the DI self-audit finding classification severity scheme so every fail gets a defensible severity.