Guide · Root cause and CAPA

Root cause analysis and CAPA: a practical guide to strong actions

Short answer

Root cause analysis finds the system conditions behind an event. CAPA, corrective and preventive action, changes them and checks the change worked. Stronger actions redesign the process, while training alone rarely lasts. Every action needs one owner, a date and a measure.

Root cause analysis finds causes; CAPA fixes them

Root cause analysis (RCA) finds the system reasons an event happened. An OSHA and EPA fact sheet calls a root cause a fundamental system reason that points to correctable failures. Fixing only the immediate cause removes a symptom, not the problem. See root cause analysis.

Corrective and preventive action (CAPA) is what you do with the findings. A corrective action removes the cause of this event. A preventive action stops the same weakness elsewhere. See corrective and preventive action.

CMS gives a surgery center example. Confusing emergency drug storage in one operating room is corrected in that room, and prevented by reviewing the others.

The patient safety field calls the process RCA2 (root cause analyses and actions), because analysis without action prevents nothing. The CAPA label comes from manufacturing. FDA's device rule is now the quality management system regulation, which incorporates ISO 13485.

The old section 820.100 is gone from the current eCFR. The logic carries over unchanged.

Match investigation depth to risk

Triage every event. A full team analysis of every near miss would collapse under its weight. A five-minute review of a serious event would miss the point.

A common triage tool is the VA National Center for Patient Safety's Safety Assessment Code matrix, reproduced in RCA2. It pairs severity with probability and returns 3 (highest risk), 2 or 1. It covers actual events and close calls.

Safety Assessment Code matrix (VA National Center for Patient Safety, as reproduced in RCA2)
ProbabilityCatastrophicMajorModerateMinor
Frequent3321
Occasional3211
Uncommon3211
Remote3211

Some analyses are required. The Joint Commission expects a comprehensive analysis of every sentinel event. For events reported to it, the analysis and action plan are due within 45 business days of the event or of becoming aware of it.

RCA2 recommends starting within 72 hours, using 4 to 6 people not involved in the event, and finishing in 30 to 45 days. OSHA's process safety standard requires covered employers to investigate any incident that did, or could reasonably have, caused a catastrophic release.

Start within 48 hours. A near miss can trigger this duty.

Suggested tiers. These are recommendations, not regulatory requirements.
LevelWhenMethodTiming
Quick reviewScore 1, one-offManager-led five whysSame week
Focused reviewScore 2, or a repeating patternSmall group, timeline, cause-and-effect diagramWithin 30 days
Full analysisScore 3, sentinel event, or a near miss that could have been catastrophicRCA2-style team, interviews, flow diagram, action hierarchyStart in 72 hours; finish in 30 to 45 days

Use the simplest method that finds the cause

Combine methods for complex events. OSHA and EPA say their listed tools are ideally combined. Brainstorming and checklists may do for simple incidents. Complicated ones need logic trees backed by timelines and causal factor analysis.

MethodBest forWatch out for
Five whysSimple events with one clear threadStops at the first plausible answer, often a person. Misses parallel causes.
Fishbone diagramMany contributing factors; group brainstormingLists causes without ranking them. Pair with a timeline.
Timeline and flow diagramRebuilding what happened and where the process driftedNeeds interviews and records. Build it before debating causes.
Fault tree analysisHigh-consequence events where several failures combinedNeeds skill and time. Overkill for routine events.
Barrier analysisWhich defenses failed or were missingEasy to stop at the barrier and skip why it failed.
Failure mode and effects analysisReviewing a high-risk process before harm occursJoint Commission hospital standards require a proactive risk assessment of one high-risk process every 18 months.

OSHA and EPA build every analysis on four questions: what happened, how, why, and what needs correcting. See five whys vs fishbone vs fault tree.

Write causes that point to systems, not people

Write each cause as a chain: something led to something, which made the event more likely. RCA2 and the VA use five rules, adapted from an FAA study of maintenance errors.

RuleWeak statementStronger statement
1. Show cause and effectFatigue.Back-to-back double shifts raised fatigue, which made a skipped double check more likely.
2. Use specific words, not poor, bad, careless or failedThe handoff was poor.Handoffs had no standard format and happened in a hallway, so allergy status was missed.
3. A human error needs a preceding causeWrong vial selected.Look-alike vials were stored side by side.
4. A violation needs a preceding causeTime-out not followed.The schedule left no pause between cases, so time-outs started after draping.
5. Failure to act is causal only if there was a duty to actAlarm log not checked.No role was assigned to check the alarm log, so overnight alarms went unreviewed.

Human error is not an acceptable root cause, says RCA2. If a well-trained person erred in a typical setting, system factors helped. Retraining one person does nothing for the next.

Define blameworthy events, such as deliberately unsafe acts, in advance and route them to HR, not RCA.

Stronger actions change the system, not memory

Rank every action by how little it relies on memory. Stronger actions change the system. Weaker ones ask people to try harder. The ranking comes from the VA National Center for Patient Safety (2001) and follows NIOSH's hierarchy of controls. See hierarchy of controls.

Action hierarchy as shown in RCA2, Figure 3
StrengthAction categoriesExamples from RCA2
StrongerPhysical plant changes; usability-tested devices; forcing functions; simpler processes; standardized equipment or process; visible leadership involvementSliding doors instead of revolving doors to cut falls; tubing that connects only one way; bar-coded medication administration
IntermediateRedundancy; staffing or workload changes; software alerts; fewer distractions; simulation training with refreshers; checklists; no look-alike or sound-alike items; read-back; enhanced documentationTwo nurses independently calculating high-risk doses; quiet rooms for programming infusion pumps; pre-incision checklists
WeakerDouble checks; warnings; new procedure or policy; trainingCaution labels; a reminder to check IV sites every two hours; demonstrating hard-to-use equipment

Two points surprise people. Double checks are weaker, because they rely on someone remembering to look. Training and policy are weaker too, though they are the most common fixes. RCA2 says they are often needed to set expectations but unlikely to be enough alone.

IHI's action hierarchy tool shows one case. A patient slid off a shower chair. The stronger action: a chair with secure straps. The intermediate action: find patients at risk of falling and add staff for showering. The weaker action: retrain staff.

Seven steps to a CAPA that works

  1. Tie each action to one causeAn action that answers no stated cause is a good idea, not a corrective action.
  2. Choose the strongest action you can doLog a weaker action as interim if the stronger one needs time or capital.
  3. Name one owner and a dateRCA2 says assign a person, not a committee, with authority and resources.
  4. Write the measure nowState what, who, target and date. Use a process measure (was it done) and, where possible, an outcome measure (did it work).
  5. Get leadership approvalRCA2 recommends the CEO or another top leader approve or reject each action. Record why if rejected.
  6. Pilot before rolloutCMS advises testing in one area first. Some changes have unintended effects.
  7. Extend the fixAsk where else the weakness exists: other rooms, units, shifts, sites. That is the preventive half of CAPA.

The Joint Commission says an acceptable plan names the responsible title, the start date including any pilot, how effectiveness will be evaluated and sustained, and when alternatives apply if targets are missed. The corrective action plan template and root cause analysis worksheet follow that structure.

Prove the action worked with two measures

An effectiveness check is a planned re-measure that shows the problem is smaller and stays smaller. Closing the task is not closing the loop. See effectiveness review.

RCA2 gives a worked pair for new hand-hygiene technology. Process measure: watch 100 staff-patient encounters over seven days, expecting 95 percent compliance. Outcome measure: a 20 percent drop in hospital-acquired infections spread by staff contact. Process data comes fast. Outcome data takes longer. Use both.

Illustrative effectiveness checks
ActionProcess measureOutcome measure
Bar-code scanning at the bedsideScan rate across 100 observed administrations at 30 daysWrong-drug reports at 90 days
Strapped shower chairs in every shower roomAudit confirms every room is equippedFalls during bathing per 1,000 resident days at 90 days

Sustainment matters as much as the first result. CMS tells surgery centers to keep collecting data after a fix. AAAHC's model says plan more than one re-measure and return to the cause if the goal is missed.

IncidentKit does not close a corrective action until it is verified.

Six signs of a weak analysis

RCA2 lists conditions that mean a review needs redoing. The Joint Commission reviews each sentinel event analysis for thoroughness, credibility and acceptability.

  • No contributing factors, or factors with no supporting data.
  • Individuals named as the cause, or causes that point to blame.
  • No stronger or intermediate actions.
  • Causal statements that break the five rules.
  • No actions, or actions that miss the stated vulnerabilities.
  • Follow-up assigned to a committee, not a person.

The Joint Commission also expects a credible analysis to be precise, complete, systematic, and deep and broad enough to cover systemic factors. It should include a process owner, a patient or family member when fitting, and people close to the process.

A sentinel event alone does not affect the accreditation decision. Willful failure to respond appropriately could.

Industrial and OSHA context

The same logic works on the plant floor. In OSHA and EPA's example, a worker slips on oil. A traditional investigation says oil was spilled, cleans it up and tells the worker to be careful.

Root cause analysis asks why the oil was there, where it came from, why it stayed, and whether anyone reported it. It may find a missing mechanical integrity program, a system cause.

OSHA's process safety standard (29 CFR 1910.119(m)) also requires covered employers to:

  • Use a team that includes someone who knows the process.
  • Report the date, description, contributing factors and recommendations.
  • Resolve and document findings, and review them with affected staff.
  • Keep the report for five years.

See process safety incident investigation. NIOSH's hierarchy of controls runs from elimination and substitution through engineering and administrative controls to personal protective equipment. RCA2 modeled its action hierarchy on it, so safety and risk managers can share the same logic.

How IncidentKit supports RCA and CAPA

Lauren drafts the investigation from what reporters and others said, with a timeline and a first pass at contributing factors and five whys. Every drafted field shows "Lauren · draft" until a person reviews, edits and signs. The team decides the root cause.

Human-authored RCA templates are rolling out.

Investigations hold contributing factors, five whys and the disposition. Corrective actions carry an owner, due date, evidence and effectiveness check, and nothing closes until verified. See the root cause analysis and close corrective actions use cases.

Frequently asked questions

What is the difference between root cause analysis and CAPA?

Root cause analysis finds the system conditions behind an event. CAPA is the follow-through: corrective actions that remove those causes, preventive actions that stop similar problems elsewhere, and a check that they worked. RCA without CAPA changes nothing, which is why RCA2 adds "actions."

How long should a root cause analysis take?

RCA2 recommends starting within 72 hours and finishing in 30 to 45 days. The Joint Commission expects the analysis and action plan for a reported sentinel event within 45 business days. OSHA's process safety standard requires starting an incident investigation within 48 hours.

Is training a good corrective action?

Training is a weaker action in the VA and RCA2 hierarchy. It sets expectations and builds skill, but alone it relies on memory and rarely lasts. Pair it with at least one stronger or intermediate action that changes the process or equipment.

What are stronger, intermediate and weaker actions?

Stronger actions need little human memory: physical changes, forcing functions, simpler or standard processes. Intermediate actions give people tools: checklists, alerts, redundancy, read-back. Weaker actions rely on memory: double checks, warnings, policies, training. Plans need at least one stronger or intermediate action.

How do you measure whether a corrective action worked?

Set a process measure and an outcome measure before the action starts. The process measure shows it was done, such as an observed compliance rate. The outcome measure shows the problem shrank, such as fewer events. Name one person, set a date, and plan a later re-measure.

Who should be on a root cause analysis team?

RCA2 recommends 4 to 6 people: process experts plus others from different levels, with a patient representative considered. Interview the people directly involved, but keep them off the team. Give staff protected time, and have leadership review the result.

Sources

Reviewed against the sources above on Oct 5, 2026. Rules change: confirm current requirements with the issuing body or your counsel before relying on any summary.

Start free

Start with one incident.

Create your kit in about ten minutes and report the first incident the same day. Free to start, no card.