The Problem with Accident Investigations

By Thomas A. Smith

 They can hinder improvement

One question I’m frequently asked by safety directors is, “What can I do to improve accident investigations?” I get an incredulous reaction when I respond, “Stop doing them.” For some safety professionals, accident investigations are The Holy Grail. Without them, many wonder, how could we tell a worker about his mistake that led to the accident? If we want the employee to understand that poor safety performance will not be tolerated, we must be able to tell him what he did wrong so he can avoid repeating his mistake, the conventional safety wisdom goes.

Unfortunately, accident investigations never lead to continual improvement when they are used to affix blame for accidents. A better goal for safety is to fix the system. Although many managers say that’s what they are trying to do with accident investigations, in reality, they are not. This article will explain why.

Missing the mark

Traditional safety management postulates that accidents are caused by unsafe actions and unsafe conditions coupled with the element of time. This formula also suggests that unsafe actions are the major culprit when it comes to finding a root cause of an accident.  Using this axiom passed down from Heinrich, we have developed a system of superstitious learning when it comes to understanding why accidents occur. The accident investigation provides us with neat, simple and easy solutions based on deductive reasoning. Unfortunately, it often misses the point.

Ultimately, the goal of an accident investigation is to find out what safety rule or regulation was disobeyed and then take corrective action. In the past this meant finding the guilty party and making him pay. We are more delicate about how corrective action is administered. It might start with a verbal warning by the supervisor. If the safety violation is more serious, or the accident had a higher degree of severity, the disciplinary action is enhanced appropriately. Maybe a written reprimand is placed in the employee=s record file.  Further, if an employee has frequent accidents, he can be dismissed. This is all done in the name of progressive discipline for misbehavior documented by the accident investigation.

The administration of corrective action is where superstitious learning comes into play. Supervisor=s warnings to some employees seem to work. Some employees never have another accident of the same type: the remedy succeeds. However, when the same remedy administered to a different employee fails to prevent that employee from having another accident, the supervisor assumes there must be something wrong with the employee. He has a safety behavior problem.

Naïve view

The problem with accident investigations is they take a naïve view of a system. They are based on the assumption that if only employees would do as they are told no accidents would occur. (Too which I would respond B sweet dreams.)

Our superstitious management system assumes that variation has been eliminated from the workplace through the edicts and regulation laid down by management. This misunderstanding is perpetuated by the use of accident investigations as a method to control variation. When in fact, the only thing they are good for is to maintain the status quo.

To understand accidents at work we should view them from a systems perspective. Each element of the work system has variation.  Dr. W. Edwards Deming taught us that variation stems from common causes or special causes. Variation will always exist. Accidents and/or near hits will always occur. The key to understanding accidents is to determine if they are being influenced by the common causes or special causes in the system.

When an accident occurs, the typical reaction is to attribute it to someone=s carelessness. So we try to eliminate these employees by describing them as being accident prone or having a poor safety behavior.  What we have failed to understand is that the system guarantees and average frequency of accidents to occur at unpredictable times and places.

We would be much better off if the time we spent on accident investigations were allotted to trying to improve the underlying causes in the system that creates the accidents. Mangers fail to review safety training and procedures to see if they meet the needs of their safety customers, i.e. the employees that are actually doing the work. They are prone to look for something special and take action against a particular person when the common causes would have led to the accident regardless of who was performing the task.

Who is in control?

The basic principle here is that no one should be blamed or penalized for an accident over which he/she has little or no control. The traditional definition of what causes an accident leaves employees without a chance to defend themselves: it assumes that all employees have control over there actions at all times. Therefore they are primarily responsible for any accident that he/she is involved.

The question is: How much control does the employee have over the system? Does he/she decide how much safety training he/she will receive? Does he/she determine the quality of the training? Does the employee decide the speed of production?  Can he/she improve the lighting? Can he/she install better guards or redesign his/her workspace so it will be safer than what is provided? Can he/she influence the attitude of their immediate supervisor about safety?

It is apparent we need a better definition of what causes accidents if we are to understand and reduce accident frequencies. In his book, “Normal Accidents” Charles Perrow describes a “normal accident” as the interaction of multiple failures that are not in direct operational sequence. And if a system is complex, there is almost no way to understand beforehand that an accident is inevitable, he says.

When variations merge

I would like to suggest a definition that takes it further, one that addresses what I call “system accidents.”  A system accident is the misalignment of variation in the component parts of the system that prevent proper completion of a task.

For example: Suppose you have a number of employees attend a safety training session. Everyone will not obtain the same amount of information, knowledge or understanding from your training. This is the first form of variation. Some may understand and retain up to ninety-five per cent of what was presented. Others in the same group may retain less then fifty percent of the subject matter.  So right from the start you’ve got variation in a very important part of your system. However as long as they’ve attended the training and you’ve documented it, they can go to work.

Now let=s put another element of variation into the equation. Your training is based on production at a particular speed with no unusual problems. But as everyone knows, production speed varies over time for all kinds of reasons: customer demands, breakdowns, etc.

Now add a third element of variation. A supervisor’s management style varies. Some days a supervisor is very cheerful and understanding (On these days production has usually gone according to plan.) Other days he/she can be the boss from hell. (When he could be described as an ogre.)

For the most part, employees will be able to complete their work tasks and get through the day accident free. However, there will be days when these common causes will line up and create conditions that prevent this. When common cause variation lines with within certain parameters, accidents will occur. The system all but guarantees it.

For example, when the supervisor who is having a bad day is told to speed up production of the employee who only retained fifty per cent of safety training the system is pushed to the brink. (And remember human beings often retain less knowledge over time so the fifty per cent figure could be much lower depending on how long it=s been since the safety training tool place.)  The problem is, even the best accident investigations cannot identify the variation of common causes built into the production or service systems.

Re-tool the system

System accidents are best understood using a simple control chart. Their purpose is twofold. First they show if the safety system is “stable.” That is statistically under control. Second, they prevent an improper reaction by everyone regarding how to respond to accidents in the system.

The key is to know if accidents stem from common causes or special causes. Management often assumes that accidents are a result of special causes. Therefore they react to them by blaming the person that was injured.  Control charts are the best tools for making this determination. You react differently to common causes and special causes.

Common causes require action by management on the system if they are to be eliminated. An individual can act on special causes with no system change required to prevent them in the future. We have learned that common causes are responsible for approximately eighty-five to ninety-nine percent of accidents in a work process!

Control charts provide the signals of any special causes in your safety/work system. This means managers and employees can avoid the mistake of blaming a particular person for an accident when in truth it stems from the variation built into the system.

Even simple systems are complex. We need a whole new set of tools to study the system so we can improve it and make it do what we want.  The same tools we’ve learned to use to improve quality, such as process flow charts, pareto charts, cause and effect diagrams and the control chart should be applied to safety management.

The next time you start your accident investigation ask yourself, Will this time and energy result in a change in the system? Or will you merely be trying to fix the blame before you carry on as usual, trying to put out the next fire?

This article was published in Industrial Safety and Hygiene Magazine, November 1995. Mr. Smith can be reached at 248-391-1818 or at www.mocalinc.com.