Form: Instrument Out-of-Tolerance Investigation and Impact Assessment

This is a ready-to-use form for an instrument out-of-tolerance (OOT) event, meaning a calibration as-found error that exceeds the tolerance. An OOT on a critical instrument is a quality event, not a maintenance ticket, because a measuring device was potentially giving wrong values to a GMP process. This form drives the confirm, quantify, assess-impact, root-cause, disposition sequence an inspector expects. Replace every <<FILL: ...>> placeholder. A filled specimen follows. This content is general educational reference, not legal or regulatory advice. Note: an instrument OOT (out-of-tolerance) is a different event from an out-of-trend (OOT) result on a stability or analytical dataset; this form covers the instrument case.

Field	Entry
OOT record number	`<<FILL: OOT-YYYY-####>>`
Linked deviation number	`<<FILL: DEV-YYYY-####>>`
Instrument ID and description	`<<FILL>>`
Classification	Critical / Non-critical
Date OOT identified	`<<FILL>>`
Identified by	`<<FILL>>`
Calibration record reference	`<<FILL>>`

Section 1: Confirm the OOT is real

Check	Result
Instrument tagged out of service	Yes / No
As-found fully captured before any adjustment	Yes / No
Re-measured to rule out setup error (wrong standard, bad connection, not stabilized)	Yes / No
OOT confirmed real (not an artifact)	Yes / No

Section 2: Quantify the error

Point	Tolerance	As-found error	Magnitude out	Direction (reads high/low)
`<<FILL>>`	`<<FILL>>`	`<<FILL>>`	`<<FILL>>`	High / Low

Worst-case error used for impact: <<FILL: value and direction>>
Why direction matters here: <<FILL: e.g. reads high -> cycle runs cold -> sterility risk>>

Section 3: Define the affected window

Field	Entry
Last known-good calibration date	`<<FILL>>`
OOT identification date	`<<FILL>>`
Affected window	`<<FILL: from last good to now>>`
Records / batches / datasets in the window	`<<FILL: list or query reference>>`

Section 4: Impact assessment (per affected record)

For each affected batch, lot, or result, apply the worst-case corrected value against the process or specification limit.

Batch / result	Recorded value	Corrected value (apply error)	Limit	Could it have crossed the limit?	Impacted?
`<<FILL>>`	`<<FILL>>`	`<<FILL>>`	`<<FILL>>`	Yes / No	Yes / No

Screening rule applied: <<FILL: e.g. any recorded value within [error magnitude] of the near limit is impacted>>
Records requiring formal evaluation / OOS: <<FILL: list or none>>

Section 5: Root cause

Field	Entry
Root cause	`<<FILL: wear, damage, environment, interval too long, mishandling, prior bad repair>>`
Method used	`<<FILL: e.g. 5-why, fishbone>>`
Is the cause systemic (model, duty, interval)?	Yes / No

Section 6: Corrective and preventive action

Action	Type (C/P)	Owner	Due
Adjust/repair and re-verify the instrument	C	`<<FILL>>`	`<<FILL>>`
Shorten interval / change device / add interim check (if systemic)	P	`<<FILL>>`	`<<FILL>>`

Section 7: Disposition and closure

Field	Entry
Product disposition (with QA)	`<<FILL: release / additional testing / reject / further investigation per batch>>`
OOS / recall assessment raised?	Yes / No, ref `<<FILL>>`
Conclusion (defensible on its own)	`<<FILL: what was affected, corrected numbers, why product/data are or are not impacted>>`
QA approval	`<<FILL: name, signature, date>>`

Acceptance criteria

The investigation is acceptable when: the OOT was confirmed and as-found captured before adjustment; the error is quantified with magnitude and direction; the affected window runs from the last known-good calibration; impact is assessed with the actual worst-case corrected values against limits, not asserted; affected product is dispositioned with QA; and the root cause feeds back into the interval or device selection.

Filled specimen

The following shows the form completed for an in-process balance found reading high. Numbers are illustrative.

Instrument: BAL-1102 analytical balance, critical. OOT-2026-0088, DEV-2026-0201.

Confirm: tagged out of service; as-found captured before adjustment; re-measured against OIML E2 weights, confirmed real.
Quantify: at the 100 g point, tolerance ±0.1 g, as-found reads 0.5 g high. Worst-case error +0.5 g. Reads high, so true weights were 0.5 g lower than recorded.
Affected window: last good calibration 4 months ago to now; all in-process weight checks on BAL-1102 in that window (query WQ-118).
Impact: in-process limit 95.0 to 105.0 g. A recorded 100.0 g was truly 99.5 g, still in limit. But any recorded value within 0.5 g of the lower limit was truly below it. Screening rule: any recorded value at or below 95.5 g is impacted. Three batches flagged (recorded 95.2, 95.4, 95.1 g), sent to formal evaluation; the remainder justified as not impacted with corrected numbers shown.
Root cause: worn load cell; systemic for this model under heavy daily use.
CAPA: repair and re-verify (C); shorten interval from 12 to 6 months and add a daily control-weight check for this model (P).
Disposition: three flagged batches to additional evaluation with QA; OOS assessment OOS-2026-0044 raised for one that failed re-evaluation; rest released.

The arithmetic is the assessment. “We looked and found no impact” without the corrected numbers, the screening rule, and the batch list would not survive an inspection.

Common inspection findings this form prevents

An out-of-tolerance result handled as a maintenance event, so the quality system never sees it and no impact is assessed.
An impact assessment asserted (“no impact”) with no arithmetic, no direction, and no affected-record list.
The affected window taken as “since we noticed” instead of “since the last known-good calibration.”
No feedback loop, so the same instrument goes out of tolerance again at the next cycle.

How to adapt this form

Point the linked deviation and OOS references to your real quality-event procedures.
Set the screening rule to your process; the principle is that any record within the error magnitude of a limit is potentially impacted.
Where a fleet of identical instruments shares a failure mode, escalate to a device-level CAPA rather than repeating per-instrument investigations.
File the completed form with the triggering calibration record so the two are always retrieved together.

Header