Incident management guide

How to check whether an incident corrective action is effective

An incident action is complete only when the intended control has been put in place, checked in the real work setting, and shown to address the cause it was meant to fix. HSE’s investigation workbook calls for actions that deal with immediate, underlying and root causes, a plan with specific and timed objectives, communication to those affected, and arrangements to monitor implementation.

That creates two separate checks. Implementation asks whether the promised action happened, for example a guard was installed or training delivered. Effectiveness asks whether it reduces the risk in actual use, so that the guard stays in place and prevents access and workers can perform the revised method. A record marked “done” after a purchase order or training attendance does not answer the second question. Keep RIDDOR reporting separate from the decision about what actions prevent recurrence. HSE: investigating accidents and incidents, HSG245; HSE: managing health and safety.

Begin with what the investigation actually found

A useful action should be traceable to an evidence-based finding. If a worker slipped, “remind staff to take care” leaves the leaking pipe or damaged floor unchanged. HSE advises identifying controls that were missing, inadequate or unused and measures that address immediate, underlying and root causes, and it criticises vague recommendations that fail to reach the causes. Do not force every investigation into one favourite “root cause” formula. The depth should match potential consequences and uncertainty. For every action ask: “which finding and causal pathway does this address?”

Deal with immediate risk, then write an action with a measurable outcome

An investigation takes time and people may keep facing the hazard meanwhile, so make the area safe, isolate defective equipment or change the method within competent arrangements, and record interim controls and their owner. Do not confuse immediate correction, such as cleaning a spill or giving first aid, with prevention. Then write the action as a risk control to be achieved, not an activity purchased: “order new guarding” is an activity; “install and verify guarding that prevents access to the danger point during normal operation and foreseeable cleaning” is an outcome. HSE calls for specific, measurable, agreed, realistic and timescaled objectives. Set a verification method before the action starts, or the team will close on whatever evidence arrives.

Prefer controls that change the work

A reminder, toolbox talk or poster can help when workers need new information, but may not be enough if the hazard could be removed, guarded, segregated or controlled by changing the process. Ask why a control was unused. Was it available, did it fit the task, was the equipment broken, or did the schedule make the safe method impracticable? Retraining a worker without fixing a missing resource produces a signed attendance record and no improvement. If training is part of the action, define what competence will be demonstrated: attendance proves presence; a practical demonstration or later observation shows whether the method is applied.

Check implementation first, then test effectiveness in real conditions

Confirm the promised measure exists on the right machine, at all entrances and on every shift, and watch for partial implementation. A company may fix the incident site but leave identical equipment elsewhere. Then check whether the control performs under the conditions that caused or could cause harm: observe normal work, maintenance, shift changes and foreseeable deviations; ask workers whether they can use the measure and whether it introduces a new problem. Do not use “no repeat accidents” as the sole proof. Harmful events can be rare, and a weak control may appear successful because exposure has been low. Look for leading evidence: whether the control is present, used and maintained.

Reopen or strengthen an action that fails

If the check shows the measure is absent, bypassed, ineffective or producing another hazard, keep the action open or reopen it, control immediate risk, and investigate why the remedy failed. Avoid gaming a dashboard. An overdue action is an operational signal, not a reason to relabel it “monitoring” and remove it from view. Feed the learning into the relevant risk assessment and similar work elsewhere, and share findings without turning the process into blame, which discourages reporting.

FAQs

What is the difference between corrective action and a task closure?

A task closure records that an assigned step was completed. A corrective action aims to control the causal weakness found in the incident. A completed task may be one piece of evidence, but the risk control should be checked for effectiveness before the finding is considered resolved.

Does every near miss need a full root-cause investigation?

The general HSE guidance supports a proportionate approach, and potential consequences, recurrence and uncertainty should guide depth. A simple event can still reveal a serious weakness, so do not judge importance only by whether injury occurred. HSE does not endorse one analysis technique for all cases.

Should the action be closed when the new procedure is issued?

Only if issuing it is enough to address the identified risk, which often needs checking in actual use. Confirm people received and understood it, can follow it, and that the relevant controls work. HSE advises monitoring action-plan implementation and checking management controls.

Can software verify effectiveness automatically?

Complys describes assigning incident corrective actions with owners, dates, status and an audit trail. It does not physically observe a guard, route or work method, so use it to keep the action and evidence together, then have a competent person perform and record the effectiveness check.

Where Complys fits

Complys can log the incident and assign corrective actions with owners, dates, status and an audit trail. It does not physically observe a guard, route or work method. Record a separate, person-led check that the control works before you treat the finding as resolved. A status change does not prove risk reduction.

Incident reporting software →