This is a ready-to-use trend log for the tier mix of a quality event population, the signal the triage and classification article describes as distinct from reviewer calibration: calibration measures whether reviewers agree with each other on the same event, this measures whether the whole population of real events is drifting, together, toward a lighter tier over time. A team can pass every calibration session and still be quietly under-classifying if the drift is shared rather than individual. Replace every <<FILL: ...>> placeholder, maintain it in a queryable system rather than a narrative log, and route the periodic review through your normal QA oversight. Verify each cited regulation against the current source before you rely on it.
Part 1: Classification register (one row per event)
| Field | Format | Required | Notes |
|---|---|---|---|
| Event ID | Text | Yes | Links to the full triage and classification record |
| Event type | Deviation / OOS / OOT / complaint / audit finding / excursion / DI event | Yes | For per-type distribution |
| Date opened | Date | Yes | For time-based trending |
| Tier at open | Critical / Major / Minor | Yes | The provisional tier |
| Tier at close | Critical / Major / Minor | Yes | The final tier |
| Reclassified? | Yes / No | Yes | If Yes, direction matters (section 3) |
| Reclassification direction | Up / Down / N/A | Conditional | Up means the event proved more serious than first read; Down means less |
| Notification met the timeline for its tier | Yes / No | Yes | Per the escalation matrix |
| Investigation closed on time for its tier | Yes / No | Yes | Per the SOP-defined target |
| Reportability screen completed at classification, not at closure | Yes / No | Yes | The clock-start discipline the article calls out |
Register table
| Event ID | Type | Date opened | Tier at open | Tier at close | Reclassified | Direction | Notified on time | Closed on time | Screen at classification |
|---|---|---|---|---|---|---|---|---|---|
<<FILL>> | <<FILL>> | <<FILL>> | <<FILL>> | <<FILL>> | <<FILL>> | <<FILL>> | <<FILL>> | <<FILL>> | <<FILL>> |
Part 2: Trending metrics
Compute on a defined cadence (commonly quarterly, with a rolling 12-month view for management review). Compare against a stated baseline or expected range rather than a bare number; a mix with no comparison point cannot show drift.
| Metric | Definition | Signal to watch for |
|---|---|---|
| Tier distribution | Share of events at close that are Critical, Major, and Minor, over the period | A falling Critical share and a rising Minor share with no process improvement to explain it |
| Reclassification rate, net direction | Count of Up reclassifications minus count of Down reclassifications, and each as a share of all reclassified events | A rate heavily skewed Down with few or no Up reclassifications, suggesting the initial call is being set high on paper and quietly settled low, rather than genuinely refined on evidence |
| On-time notification rate by tier | Events notified within the tier’s timeline, divided by total events at that tier | A falling rate specifically for Critical events, since that is where a late notification carries the most regulatory exposure |
| On-time closure rate by tier | Events closed within the tier’s target, divided by total events at that tier | Chronic lateness that is never flagged as an extension with a documented rationale |
| Reportability screen timeliness | Screens completed at classification (day zero) divided by total events requiring a screen | A screen habitually completed only at investigation closure, missing the regulatory clock the article warns about |
| Recurring-minor escalation count | Minor events sharing a root cause, counted across a rolling window, that were not escalated to a trend review or a CAPA | Any count above zero without a documented reason it was not escalated |
Part 3: Thresholds and signals for concern
Set these before the data arrives, against your own baseline distribution, not a generic industry figure. A distribution is a fingerprint of the product type and process maturity as much as of the classification discipline behind it.
| ID | Signal | Definition | Basis for the level chosen |
|---|---|---|---|
| S1 | Critical-share decline | Critical share of closed events falls below <<FILL: e.g. 80%>> of the trailing 4-quarter average, with no documented process improvement in the period | <<FILL: set from your own historical variation; a decline larger than ordinary period-to-period noise>> |
| S2 | Minor-share growth | Minor share rises above <<FILL: e.g. 120%>> of the trailing 4-quarter average | <<FILL: mirrors S1 from the other direction>> |
| S3 | Down-classification skew | Down reclassifications exceed <<FILL: e.g. 3>> times Up reclassifications in a quarter | <<FILL: "classify high, refine on evidence" should produce some Up movement over time, not none>> |
| S4 | Critical notification lateness | Any Critical event notified outside its timeline | <<FILL: zero-tolerance signal given the regulatory exposure>> |
| S5 | Recurring minor unescalated | <<FILL: e.g. 3>> or more minors sharing a root cause within <<FILL: e.g. 90 days>>, with no trend review or CAPA raised | <<FILL: recurrence is itself a severity driver per the classification article>> |
| S6 | <<FILL: your own>> | <<FILL>> | <<FILL>> |
Part 4: Periodic review record
| Field | Entry |
|---|---|
| Review reference | <<FILL: e.g. TIER-TREND-YYYY-QN>> |
| Review period | <<FILL: from>> to <<FILL: to>> |
| Review date | <<FILL>> |
| Performed by | <<FILL: name, role>> |
| Total events in period, by tier | <<FILL: Critical n / Major n / Minor n>> |
| Trailing 4-quarter baseline distribution | <<FILL: Critical % / Major % / Minor %>> |
| Signals crossed this period | <<FILL: list, or "none">> |
| Companion calibration session referenced | <<FILL: session ID from the classification calibration session log, if one occurred in the period>> |
| Reviewed by (name, role, date) | <<FILL>> |
| QA approval (name, role, date) | <<FILL>> |
Part 5: Signal-crossing escalation
A crossing obliges investigation, not an automatic conclusion. Record for each:
- Verify the data. Confirm no coding change moved events between tiers or types mid-period, and no denominator change (a change in event volume, a new event type brought into scope) explains the shift before treating it as a classification signal.
- Sample and re-score. Pull a sample of the period’s events, run them through the risk-ranking worksheet, and compare the re-score to the tier recorded. A mismatch here is the direct evidence of drift.
- Trigger a calibration session per the classification calibration session log if the sample re-score disagrees with the recorded tiers on more than an isolated event.
- Assess whether any individual event needs re-opening on its own evidence; a population-level signal is a prompt to look, not itself grounds to overturn a specific closed classification without that event’s own documented basis.
- Report the finding and the action to management review, whether or not any individual event changes.
| Field | Entry |
|---|---|
| Signal crossed | <<FILL: ID and value>> |
| Data verified, result | <<FILL>> |
| Sample re-scored, size and outcome | <<FILL>> |
| Calibration session triggered, reference | <<FILL: or N/A with reason>> |
| Individual events re-opened | <<FILL: list, or none, with reasoning>> |
| Reported to management review | <<FILL: date>> |
Acceptance criteria
The tier-distribution trend program is defensible when a defined cadence exists and reviews actually occur on it; the distribution, reclassification direction, and timeliness metrics are computed against a stated baseline every period, including periods with no signal; every signal crossing is verified, sampled, and, where warranted, linked to a calibration session; and the trend is visibly carried into management review alongside, not instead of, the reviewer-agreement trend from the calibration log.
Retention
Retain trend review records per <<FILL: retention period>>, per the records retention schedule.
References
ICH Q9(R1), Quality Risk Management (2023 revision), reducing subjectivity in risk assessment, the basis for treating population-level drift as a program signal. ICH Q10, Pharmaceutical Quality System, management review and continual improvement. 21 CFR 211.192 (investigation of discrepancies and failures), the basis for consistent classification across the event population.
Confirm the current version of each reference before issue.
Filled specimen
The following is a worked quarterly review showing a genuine drift signal. Company, event counts, and names are illustrative; replace them with your own.
Review header
| Field | Entry |
|---|---|
| Review reference | TIER-TREND-2026-Q3 |
| Review period | 1 July 2026 to 30 September 2026 |
| Review date | 10 October 2026 |
| Performed by | S. Nwosu, Manufacturing QA |
| Total events in period, by tier | Critical 6 (7%), Major 31 (36%), Minor 49 (57%); total 86 |
| Trailing 4-quarter baseline distribution | Critical 12%, Major 38%, Minor 50% (average of Q4 2025 through Q2 2026) |
| Signals crossed this period | S1 (Critical-share decline) and S2 (Minor-share growth) |
| Companion calibration session referenced | CAL-2026-Q3-01, 15 September 2026 |
Reclassification summary
| Direction | Count | Share of reclassified events |
|---|---|---|
| Up | 2 | 22% |
| Down | 7 | 78% |
| Total reclassified | 9 | of 86 events, 10% |
S3 (down-classification skew) crossed: 7 Down against 2 Up is well over the 3-to-1 trigger.
Timeliness
| Tier | On-time notification | On-time closure |
|---|---|---|
| Critical | 6 of 6, 100% | 5 of 6, 83% (1 justified extension on record) |
| Major | 29 of 31, 94% | 27 of 31, 87% |
| Minor | 49 of 49, 100% | 46 of 49, 94% |
No S4 crossing (no late Critical notification). Reportability screen completed at classification for 6 of 6 Critical events and 30 of 31 Major events; one Major event’s screen was completed 3 days after classification, flagged separately for correction, not a population signal on its own.
Signal-crossing escalation
| Field | Entry |
|---|---|
| Signal crossed | S1: Critical share 7% against a 12% baseline, a decline to 58% of baseline (trigger: below 80% of baseline). S2: Minor share 57% against a 50% baseline, 114% of baseline (trigger: above 120%; did not itself cross, but recorded as contributing to S1’s read). S3: Down-classification skew, 7 Down versus 2 Up. |
| Data verified | No coding change in the period; the OOS event type was newly brought fully into this register in Q3 rather than tracked separately before, adding 14 events not present in the Q2 baseline. Re-computed the baseline including a back-filled OOS count for Q2 for comparability: revised baseline Critical 11%, Minor 51%. Signal persists after the correction. |
| Sample re-scored | 15 of the 31 Major events and all 7 Down-reclassified events re-scored using the risk-ranking worksheet by a reviewer not involved in the original classification. 4 of the 7 Down-reclassified events re-scored to the same lower tier on the evidence recorded at re-classification, consistent with a genuine, evidence-based downgrade. 3 of the 7 re-scored to the tier they held before the downgrade, meaning the re-classification was not adequately supported by the evidence on file. |
| Calibration session triggered | CAL-2026-Q3-01 was already scheduled for this period; its results (section on divergence analysis in that log) independently identified the same failure pattern, reviewers not applying the detectability bump and anchoring on eventual outcome rather than potential at open, which is consistent with, and likely the mechanism behind, this quarter’s down-classification skew. |
| Individual events re-opened | The 3 events whose Down-reclassification did not hold up under re-score were referred to QA for a documented re-classification decision on their own evidence; none were changed unilaterally from this trend review alone. |
| Reported to management review | 17 October 2026, with the calibration session findings presented together |
Review conclusions
The Critical-share decline and the down-classification skew are not independent findings this quarter; they trace to the same root cause the calibration session found on the same date, classifiers anchoring on eventual outcome and under-applying the detectability bump when reclassifying downward. The corrective actions from the calibration log (anchored examples for the detectability bump, for confirmed DI events, and for classifying at open rather than on outcome) are the corrective action for this trend signal as well; a second trend review is scheduled for Q4 2026 specifically to confirm the Critical share and the reclassification balance move back toward baseline once those anchored examples are in use, rather than assuming the fix worked.
Common inspection findings this log prevents
- A tier distribution that has drifted toward Minor over several periods with no trend review ever performed.
- Reclassifications trending heavily downward with no population-level review, only individual event rationales that look reasonable in isolation.
- A notification or closure timeliness failure visible only in individual event records, never rolled up by tier so the pattern is invisible.
- Reportability screens habitually completed late in the investigation rather than at classification, with no metric surfacing the pattern.
- A calibration session and a tier-distribution signal pointing at the same root cause, but never connected because they were never reviewed together.
How to adapt this log
- Set your document number, owner, cadence, and data source, and name the report you extract from.
- Calculate your own baseline distribution from your own event history; do not adopt the numbers in the specimen.
- Set the signal thresholds in Part 3 against your own observed period-to-period variation, the same discipline used for complaint and OOS trending elsewhere on this site.
- Align event types and tier definitions with the classification SOP and the triage and classification record so the three documents cannot disagree about what a tier or an event type is.
- Schedule this review and the classification calibration session on a cadence that lets you present both together to management review, the way the specimen does, since a shared root cause is easiest to see when the two signals sit side by side.
- Confirm every regulation referenced against the current published version before issue.