Accident investigation process: from first response to verified change

A workplace accident investigation should establish what happened, why the controls failed, what will prevent recurrence, and whether those changes worked. Start by making the scene safe and caring for people. Then gather facts promptly, analyse immediate and underlying causes without stopping at blame, choose measures that control the risk, and verify that someone implemented them. A report that ends with “worker reminded” but leaves the same hazard in place is not a completed investigation.

This is a process guide for Great Britain. It follows the four stages in the Health and Safety Executive's HSG245 accident and incident investigation workbook: gathering information, analysing information, identifying risk control measures, and the action plan and its implementation. Northern Ireland has a separate health and safety regulator, and the exact reporting and legal regime should be checked before using this page there. This article does not determine liability or replace competent advice after a serious incident.

First: protect people and preserve the scene

The first response is safety and medical care. Stop the relevant activity if it remains unsafe, summon first aid and emergency services as needed, prevent another exposure, and account for anyone else at risk. Do not delay rescue to protect evidence. Once the immediate danger is controlled, preserve the scene so far as it is safe and lawful to do so: equipment settings, guards, permits, work instructions, photographs and witness contacts may be lost quickly. Avoid moving equipment or clearing debris simply to resume production before the relevant facts are recorded.

Assign an investigation lead and consider whether the event is serious enough to need a team, specialist engineer, occupational health professional, worker representative, insurer or legal advice. HSG245 links investigation depth to the potential and actual risk; a near miss with the capacity to kill can warrant a deeper investigation than a minor injury caused by a familiar low-risk event. The lead should have authority to secure information and recommend change. If the regulator or police may investigate, coordinate evidence handling and any internal interviews appropriately; do not obstruct an external investigation.

At the same time, assess any separate notification and record duty. HSE's RIDDOR guide explains what is reportable, who the responsible person is and how to submit. A RIDDOR report is a notification to the regulator; it is not the whole causal investigation. Conversely, many events worth investigating are not RIDDOR-reportable. Do not wait for an investigation conclusion to decide whether a time-sensitive report is due. HSE defines the responsible person as the employer, certain self-employed people or the person in control of premises depending on the event and relationship. Record who made that determination.

Stage 1: gather facts before memories or conditions change

Create a basic timeline from sources that can be checked. Record the date, time, place, task, people present, equipment, materials, environmental conditions, instruction and supervision, permits, maintenance and any change from the normal method. Speak to witnesses separately, with support where appropriate, and ask open questions such as “What did you see before the event?” rather than “Why did you ignore the procedure?” Record what a person actually observed, what they inferred, and what is still unknown. Do not make a worker repeat an account unnecessarily after a traumatic event.

Collect the relevant records, not every file the organisation owns. The risk assessment or RAMS, safe system of work, training, induction, inspection and maintenance records may show what control was intended. Machine logs, photographs, measurements or CCTV may show what actually happened. Note who captured each item, when, the original location and any change to it. If evidence is missing, record that gap rather than filling it with a confident guess.

HSG245 advises starting as soon as practicable because evidence and recollections degrade. However, do not interview an injured person before they are medically able, or pressure a witness into agreement. Privacy and employment-law obligations still apply. Keep medical detail in an appropriately restricted record and include only what the causal analysis needs in the wider report.

A useful investigation record structure

The form should make uncertainty visible. A polished narrative can conceal that an important machine setting was never checked. If an external authority has taken possession of evidence, record that and use an agreed route to obtain information.

Stage 2: analyse causes without using blame as the conclusion

Begin with the event sequence. What condition allowed a person to be exposed to the hazard? Which control was expected to prevent it? Was the control absent, not suitable for the task, bypassed, unavailable, poorly maintained or not understood? Ask why each failure was possible in the work system. Immediate causes may include a slipped foot, moving vehicle or exposed blade. Underlying causes may include route design, a missing guard, unrealistic production time, ineffective supervision or a risk assessment that did not cover the actual task. A root or organisational cause may be a procurement, maintenance or change-control decision that let the underlying condition persist.

Avoid the single line “human error”. People act within equipment, work organisation, information and incentives. If a worker took a shortcut, ask whether the safe method was feasible, supplied with the right equipment, modelled by supervisors and checked in practice. The finding can still record a deliberate unsafe act where evidence supports it, but it should also ask what control could prevent the next exposure. HSG245's analysis stage is designed to identify immediate, underlying and root causes so action reaches the system.

Check the counterfactual: if the proposed explanation were fixed, would the event probably have been prevented? If the answer is unclear, gather more evidence or state the limitation. A near miss is particularly valuable because the absence of injury can be luck rather than proof that controls were adequate. Consider credible worst-case consequences when deciding investigation depth and urgency.

Stage 3: choose measures that change the risk

Choose controls against the identified failure, then assign an interim measure while a permanent fix is developed. Prefer eliminating or reducing the exposure and engineering or work-design measures where practicable over relying solely on reminders, signs or retraining. Review whether a similar hazard exists at other sites, shifts or equipment models. An action limited to the individual involved may leave the same failure everywhere else.

For each proposed control, write the hazard it addresses, why it should work, who owns it, the date it must be in place, resources required, and how the result will be checked. If a guard must be redesigned, an engineer or competent supplier may need to approve the design. If a procedure changes, affected workers need access and instruction before work resumes. If the root cause is an unrealistic schedule, a revised training slide will not resolve it.

Example. A worker slips while carrying materials down an external stair in rain. A weak investigation says “take care.” A fuller one finds poor drainage and a worn anti-slip surface, a delivery method requiring hand carrying, and an inspection that never covers the stair during wet shifts. Interim action diverts deliveries. Longer-term actions repair the drain/surface, change the material route and add a wet-weather inspection. The lead checks the stair after rain and asks users whether the route now works. The example illustrates the method; the actual control decision requires a site assessment.

Stage 4: implement, verify and share the lesson

The investigation is not closed when a manager accepts recommendations. Maintain an action queue with owners and due dates. Verify physical controls on the job, not only in a meeting. A new guard can be installed but prevent maintenance access, encouraging another bypass. A rewritten method may be technically sound but impossible on night shift. Test the control with the people who use it, close deficiencies, and decide whether the risk assessment, RAMS, training or contractor instructions need updating.

Brief relevant workers and managers on the *change*, not on unnecessary personal details of the people involved. If other sites use the same equipment or process, check whether they need the control. Monitor leading signals after release: inspections, near misses, defects, worker feedback and any recurrence. HSG245's final stage includes the action plan and implementation; a dated verification is the evidence that it happened.

Keep investigation records secure and appropriately restricted. Separate medical, HR or legal material from a general safety-learning summary. HSE specifies record requirements for RIDDOR-reportable events, but that is not a universal retention period for every investigation file. Apply the organisation's legal retention, privacy, insurance and litigation-hold requirements to the actual event.

How investigation and RIDDOR fit together

An internal investigation answers why the incident occurred and what must change. RIDDOR answers whether a prescribed work-related death, injury, disease or dangerous occurrence must be reported by the responsible person. HSE's reportable-incident summary distinguishes categories and includes the over-seven-day worker injury rule, which has a reporting deadline; over-three-day incapacity may require recording even when it is not reportable. Check the current HSE guidance and facts rather than assuming that every hospital visit or near miss is reportable. Record the decision and source.

For a serious event, other routes may matter too: police, emergency responders, the enforcing authority, insurer, client, principal contractor or professional regulator. Do not assume one form notifies them all. If the event involves a specialist sector, gas, rail, aviation, offshore or healthcare rules may add specific duties. Give those cases to the appropriate competent lead.

Questions employers ask

Should every near miss be investigated?

Record and screen it. The investigation depth should follow potential severity and likelihood, not simply whether an injury occurred. A dropped object that narrowly missed a person may deserve urgent review because the next event could be fatal. HSG245 provides a proportionate method.

Can the supervisor investigate their own team's event?

For a low-risk event, a competent supervisor may lead. More serious or complex events need suitable independence, technical expertise, worker representation and senior oversight. A person whose decision is central to the failure should not be the sole judge of that decision.

Does a RIDDOR report complete the investigation?

No. It notifies the regulator of a reportable event. The employer still needs to understand causes and act on risk. An event may also be worth investigating even if it is not RIDDOR-reportable.

Next step

Take a recent incident or near miss and test its file against four questions: are the facts traceable, are underlying causes evidenced, are controls assigned, and has their effect been verified? If any answer is missing, reopen the action rather than polishing the narrative. For a product discussion, use the verified Complys home page and ask how its current implementation supports the real incident-record and action workflow. The manifest's /incident-reporting-software route and any claim of investigation automation require route and feature verification before publication.

Source, owner, link, product and writer-side QA

Terminal writer-side disposition: READY.