Incident reporting and investigation

Despite good planning, incidents happen. When one does, the supervisor's first job is the injured worker, and the second is to make sure it cannot happen again. In between come duties to report, to protect the scene and to investigate. This lesson follows a real-feeling incident on Larkspur Commons from the moment it happened to the corrective actions that closed it, and covers the reporting rules, investigation methods and the mistakes that turn investigations into blame exercises.

The incident

At 10:40 on a Thursday in month 11, a first-year carpentry apprentice working for Thornbury, Kiran Dhaliwal, was carrying blocking along the level 4 corridor. He stepped on a sheet of plywood covering a duct opening about 2 by 3 ft. The cover shifted, his right leg went through the opening to the thigh, and he twisted his ankle and knee as he fell forward. A coworker helped him out. The first aid attendant suspected a fracture, and Kiran was taken to hospital, where an ankle fracture was confirmed. He would be off work for about eight weeks.

Nobody fell through to the floor below, but the opening was large enough for that, and the drop was a full story. This was a serious injury that could easily have been a fatal one, which made it a high-potential event as well as a lost-time injury.

The first hour

The first actions after an incident are the same everywhere.

• Care for the injured. First aid, emergency services if needed, and someone with the worker throughout. Kiran's foreman went to hospital with him and stayed until his family arrived.
• Make the area safe. The opening was barricaded immediately and every other duct opening on levels 3 to 6 was checked within the hour. Two more unsecured covers were found and fixed.
• Protect the scene. The cover and the area were left as they were, photographed and measured, except for the barricade.
• Report. The superintendent was told within minutes, and the regulatory reports were made as required (below).
• Start the investigation. Witnesses were identified and asked to write down what they saw before talking it over among themselves.

Checking the other openings in the first hour was the most important prevention step of the whole investigation. The cause of one incident is often present elsewhere on the site, waiting for the next person.

Reporting requirements

Under 29 CFR 1904.39, the employer must report a work-related fatality to OSHA within 8 hours, and any work-related in-patient hospitalization, amputation or loss of an eye within 24 hours. Kiran was treated in the emergency department and sent home with a cast that afternoon; he was not admitted, so the 24-hour report did not apply. The injury was, however, recordable on the OSHA 300 log because it involved medical treatment beyond first aid and days away from work, and it had to be entered within seven calendar days. Establishments with 100 or more employees in designated high-hazard industries, which include construction, must also submit their 300 and 301 data to OSHA electronically each year. The workers' compensation claim was filed with Thornbury's insurer under Idaho's rules.

Investigation duties and timelines

OSHA does not require employers to investigate incidents, but every serious safety program does, and an investigation is the best evidence of the reasonable care discussed in lesson 2. Thornbury's procedure follows the same pattern used by many companies and by the Canadian provinces: a preliminary review within 48 hours to fix immediate hazards and decide on interim controls, a full investigation within 30 days, and a corrective action plan with named people and dates. A worker representative from the site safety committee takes part, because the people who do the work know things the supervisors do not.

For Kiran's incident, the team was Omar as lead investigator, Priya as facilitator, the carpenter foreman Ben Kowalski, a worker member of the site safety committee, and Northgate Mechanical's foreman, because Northgate's crew had worked at that opening the day before.

Gathering the facts

An investigation gathers facts in four categories, sometimes remembered as people, position, parts and paper.

Interviews are conducted one at a time, in a quiet place, soon after the incident, by someone who explains that the purpose is to prevent a repeat, not to blame anyone. Open questions work best: 'Tell me what you were doing before it happened' rather than 'Didn't you see the cover?' The Northgate pipefitter's interview was the turning point. He explained that his crew had lifted the cover the previous afternoon to run a sleeve through the opening, put it back when they finished, and had not screwed it down because they had no screws with them and expected to return in the morning. Nobody had told anyone, and there was no rule that required them to.

Finding the causes

Investigations go wrong when they stop at the first cause they find, usually the action of the person who was hurt or the person nearest the hazard. 'Kiran didn't look where he was going' and 'the pipefitter didn't screw down the cover' are both true, and both are useless as conclusions, because they lead only to telling people to be more careful. Thornbury uses a layered approach, asking why at each level until it reaches causes that the company can change.

This approach resembles several formal methods used in industry, such as the 5 whys introduced in CON-345 and PMP-131, and incident cause analysis methods that look at absent or failed defenses, individual and team actions, task and environmental conditions, and organizational factors. HSE-403, Incident Investigation and Root Cause Analysis, in the Safety stream covers formal methods in depth. The point for a supervisor is to keep asking why past the person, to the conditions and systems that made the error possible.

The phrase human error deserves particular caution. People make errors all the time; a safe system is designed so that an ordinary error does not lead to serious harm. If an investigation's conclusion is 'human error', it has stopped too early.

Corrective actions

Corrective actions are chosen using the hierarchy of controls from lesson 5, and each one has a named owner, a date and a way to check that it worked.

The best action was the second one. Framing around duct openings with guardrails, instead of covering them, removed the cover from the walkway altogether on the upper floors. It cost a few hours of carpentry per opening and made the covers' weaknesses irrelevant.

Thirty days later, the corrective action report showed all seven actions complete. Priya checked effectiveness over the following two months: the opening log was being used, the daily walks found no unsecured covers, and near-miss reports involving covers fell to zero.

Measuring injury rates

Injury rates let a company compare its performance over time and against others. They are lagging indicators: they measure harm that has already happened. Lesson 11 covers why lagging indicators are not enough on their own.

# Injury rates for Thornbury's own workers on Larkspur Commons, year to date (illustrative)
hours = 186_400          # hours worked by Thornbury employees
recordable = 4           # OSHA recordable cases
dart_cases = 3           # days away, restricted or transferred
days_away = 2            # cases with days away from work
trir = recordable * 200_000 / hours
dart = dart_cases * 200_000 / hours
dafwii = days_away * 200_000 / hours
print(f"total recordable incident rate (per 200,000 h, i.e. 100 full-time workers): {trir:.2f}")
print(f"DART rate: {dart:.2f}")
print(f"days-away rate: {dafwii:.2f}")

total recordable incident rate (per 200,000 h, i.e. 100 full-time workers): 4.29
DART rate: 3.22
days-away rate: 2.15

A rate based on a few incidents moves sharply with each one, so a small site's rate says little about whether it is safe. Kiran's injury alone added about one point to the rates. That is one more reason to pay attention to high-potential events and leading indicators rather than to the injury rate alone.