Skip to content
ANDY

Quality systems guide

Deviations and CAPA that close properly

Most deviation systems fail in the same two places: root cause analysis that stops at the symptom, and closure without effectiveness review. Both are visible to an inspector reading the log.

Updated 7 August 20268 min read

What counts as a deviation

A deviation is a departure from an approved procedure, specification or standard. Under GDP that includes temperature excursions, picking and despatch errors, receipt of stock without required documentation, storage outside defined conditions, security breaches, and failures to complete a required check.

A frequent weakness is an unwritten threshold below which staff do not raise deviations. If minor events are not captured, the trend that would have shown the underlying problem never appears.

Immediate action versus corrective action

These are different steps and inspectors expect to see both. Immediate action contains the event: quarantine the stock, stop the process, retrieve the misdirected delivery. Corrective action addresses why it happened so it does not recur.

Records that show only immediate action read as incident logs rather than a quality system. Records that jump straight to corrective action without containment raise a different question about what happened to the affected product.

Root cause analysis that goes far enough

“Operator error” is almost never a root cause. It is a description of where the failure surfaced. The useful question is why the process allowed the error to occur and to go undetected — whether that is an ambiguous procedure, inadequate training, a system that permits an out-of-sequence action, or a workload pattern that makes the correct step impractical.

The corresponding test for a proposed corrective action: if the same person were placed in the same situation again, would the action prevent the outcome? Retraining alone rarely passes that test unless the investigation genuinely established a knowledge gap.

Classification and proportionality

Classifying deviations by impact lets investigation effort follow risk. A common three-tier scheme distinguishes events with potential patient or product impact, events affecting the quality system without direct product impact, and minor documentation issues.

Classification should be recorded with a reason. Consistently classifying events at the lowest tier is a pattern inspectors look for, particularly where the corrective actions are correspondingly light.

Closure and effectiveness review

Closing a CAPA means the action was completed and evidenced. Effectiveness review is a separate later step confirming the action worked, usually by looking at whether the event type recurred over a defined period.

Where effectiveness review is skipped, the log accumulates closed CAPAs alongside recurring events. That combination is one of the clearest signals that a quality system is being maintained on paper rather than in practice.

Common questions

Does every temperature excursion need a deviation record?
Departures from defined storage or transport conditions are deviations and should be recorded. Whether each requires a full investigation depends on documented risk-based criteria, but the event itself should be captured so trends are visible.
How long should deviations stay open?
There is no fixed period, but your procedure should define target timescales and overdue investigations should be visible and escalated. A log with long-overdue open items is a finding in itself.
Can AI tools write our deviation investigations?
Tools can structure an assessment, surface relevant procedure and guidance, and draft working material. The investigation conclusion and the CAPA decision remain human judgements, and the record should make clear who made them.

Sources

APPLY THIS TO YOUR OWN EVIDENCE

Need answers like this
against your own SOPs?

See how ANDY brings published guidance, controlled procedures and reviewable citations together while accountable decisions remain with your team.

Explore Pharmaceutical quality AI