Why a Closed CAPA Doesn't Always Mean a Resolved Problem

Most CAPA dashboards lead with the same number: how many CAPAs are open, how many are overdue, how many were closed this quarter. Those numbers measure workflow. They say very little about whether the underlying problems were understood and removed — and that is the question an auditor, an investigator or a management review is really asking.

What the requirement says

ISO 13485:2016 clause 8.5.2 does not stop at taking action. It asks the organisation to review nonconformities (including complaints), determine their causes, evaluate the need for action so they do not recur, plan and implement that action, verify that the action does not adversely affect the ability to meet regulatory requirements or the safety and performance of the device, and review the effectiveness of the corrective action taken.^[1] Since 2 February 2026, FDA's Quality Management System Regulation incorporates ISO 13485:2016 by reference, so for manufacturers marketing devices in the United States this is a regulatory expectation, not only a certification one.^[2]

The GHTF guidance on CAPA, still widely used as a reference, frames corrective action as part of a wider measurement, analysis and improvement process fed by data from across the quality system — and treats verifying effectiveness as part of the process rather than an administrative step at the end.^[3]

Three measures that are easy to confuse

• Closure — the record has completed its workflow. Necessary, but it proves only that steps were signed off.
• Root-cause evidence — the stated cause is supported by evidence, and the evidence that would contradict it has been considered.
• Demonstrated effectiveness — after the action, the problem did not come back, measured by criteria defined before the check rather than chosen afterwards.

A register can score well on the first while being weak on the other two. That is not a hypothetical failure mode; it is what happens whenever closure is the metric that gets reported upward.

Our interpretation: what closure hides

Reading a CAPA register as a population, rather than record by record, tends to surface the same patterns. None of them is a finding on its own; each is a reason to look closer:

• Closed without an effectiveness check, or with the check deferred indefinitely.
• Closed although the effectiveness check returned an ineffective result.
• Closed while one of its actions was still open.
• A chronology that does not hold together — an effectiveness check dated before the action it verifies.
• The same problem recurring after a CAPA was recorded as effective.
• Root causes that recur across unrelated events ("operator error", "training") without evidence specific to each event.

Effectiveness is the claim. Recurrence is the evidence that tests it.

Practical questions for the next CAPA review

• For each closed CAPA in the period: what evidence shows the problem did not recur, and when was it measured?
• Were effectiveness criteria defined before the check, or written to fit the result?
• Which root causes appear most often, and how many of those are supported by event-specific evidence?
• Has anything been recorded as effective that later recurred in a complaint, nonconformance or audit finding?
• Is the management review seeing closure rates, or effectiveness outcomes?

None of these questions needs new software to ask. They do need the register to be read as a whole, and someone accountable for the answer.

Where QMSROOT fits

This section describes QMSROOT, not a regulatory requirement. QMSROOT's CAPA register analysis reads the register as a population and flags overdue actions, overdue effectiveness checks, CAPAs closed without an effectiveness check, CAPAs closed with an ineffective result or an open action, inconsistent chronology, incomplete records, repeat signals and effectiveness contradicted by recurrence. RootAí brings evidence-grounded intelligence to that review; it does not create, close or accept a CAPA. The decision stays with the accountable Quality professional.