Techniques for preventing accidents
For every serious injury there are hundreds of incidents that hurt no one. They are the cheapest information for prevention, and the one that is almost never recorded.
01Accidents have causes
Prevention starts from a simple idea: an accident is not bad luck but the result of causes that can be identified and eliminated. Three models shaped the way we think about it:
| Model | Central idea | Consequence for prevention |
|---|---|---|
| Heinrich’s domino (1931) | A chain of factors ends in the injury; unsafe acts and conditions are the central domino | Removing one domino stops the chain |
| Bird’s pyramid (1969) | For every serious injury there are many minor ones, plus property damage and hundreds of incidents without consequences | Incidents are warnings: recording them makes it possible to act sooner |
| Reason’s Swiss cheese (1990) | Every barrier has holes; the accident occurs when they line up | Several independent barriers, and a search for the latent failures of the organization |
Reason’s model changed the focus: the operator’s error is usually the last failure in a series that began in design, maintenance or organizational decisions. Blaming the person does not eliminate those latent failures.
02Analytical and operational techniques
Prevention techniques are divided into analytical ones, which look for the causes, and operational ones, which act on them:
- Safety inspections: planned walk-throughs with checklists to detect hazardous conditions before they cause an accident.
- Accident and incident investigation: the causal tree, developed in France in the 1970s, starts from the event and asks, for each antecedent, what was necessary and sufficient for it to occur, until it reaches facts that can be corrected. It looks for facts, not culprits.
- Job analysis: dividing the task into steps and listing the hazards and controls for each one.
- Accident statistics: measuring in order to compare periods and sectors, and to see whether the measures work.
- Training, signage, standards and procedures: the operational techniques covered in industrial hygiene and safety and in the unit on safe work procedures.
03Accident rates
Counting accidents is not enough: a large company has more accidents than a small one even if it is safer. The rates relate them to the exposure:
One year of data from a plant. On the right, the rates; in the chart, Bird’s pyramid scaled to the recorded injuries, using the 1 : 10 : 30 : 600 ratio of the original study, which is only indicative: each activity has its own.
The Superintendency of Occupational Risks publishes incidence rates by economic activity, which make it possible to compare a plant with its sector. A low rate with many unrecorded incidents does not indicate safety, but a lack of information.
04In the laboratory
Take an accident described in a public report and build its causal tree as a group. Identify which of the facts could have been corrected and with what measure.
Design a checklist for an electronics laboratory, apply it and classify the findings by severity.
Propose a short form for recording no-injury incidents in the workshop and discuss how to prevent it from being used to look for culprits.
05Common mistakes
- Ending the investigation at human error. It is the starting point for asking why it was possible.
- Comparing numbers of accidents between companies of different sizes. Rates are what get compared.
- Ignoring incidents. They are the base of the pyramid and the cheapest information for prevention.
- Taking Bird’s ratio as a law. It is an average from one study; what matters is the shape, not the numbers.
06Self-assessment
A company of 400 workers had 10 lost-time accidents. What is its incidence rate?
25 per thousand workers.
If those workers each worked 1800 hours, what is the frequency rate?
\( 10/720\,000 \times 10^6 \approx 13.9 \).
What does the Swiss cheese model propose for prevention?
Several independent barriers, and correcting the latent failures of the organization, not just the last error.
What question guides the construction of a causal tree?
For each fact: what was necessary and sufficient for it to occur.
07Further reading
- James Reason. Human Error. Cambridge University Press, 1990. The origin of the Swiss cheese model and the distinction between active and latent failures.
- Sidney Dekker. The Field Guide to Understanding 'Human Error'. 3rd ed., CRC Press, 2014. How to investigate accidents without stopping at individual blame.
- Frank E. Bird and George L. Germain. Practical Loss Control Leadership. International Loss Control Institute, 1985. The accident pyramid and loss control.