About ten years ago a global client asked why SIFs weren’t declining at the same rate that recordable injuries were. That led to a study drawing on the data of six large organizations. The answers were revealing in many ways. We found that SIFs were structurally different than smaller injuries, precursors were identifiable, and effective prevention required a new strategy.
As is often the case with original research, we confirmed an insight we already had, and we also found a new one.
We confirmed that studying incidents as a series yields understandings that are not apparent in single events. We also found that in order to study a series effectively you must have good incident reports. Understanding the first point leads to the importance of the second one.
1) Learning effectively from incidents requires looking at a series, at least thirty. The initial SIF study, (led by the author for BST and ORCHSE) reviewed 300 incidents. Each incident was analyzed independently and then overarching findings were extracted from the series.
Some things can be learned from individual events, but the range is very limited and the learning stops there. “Corrective action” after an event seems like the right thing to do, especially when it is a serious event. But “instruct employee on proper procedure” is not really corrective action. The real question, why is an employee not following the procedure, isn’t answered. And it is difficult to answer in a single case. Getting beyond the over-simple “behavior”, which is often incorrect, requires finding patterns in a series of events. This can be done if you know how to look.
2) In order to be useful, incident reports must explain what actually happened that resulted in the injury.
Working across organizations and studying many incident reports showed us that their data was often unusable. It was often necessary to interview employees in order to understand what had actually happened. Usually this was due to the way the incident report was written.
An incident report should say what happened, in straightforward clear language, so that the organization can learn from the event. It doesn’t have to be long and complex, but it should convey the real story. Political and legal considerations should not dominate the report because they distort critically important information.
For a variety of reasons root cause analysis of single events, doesn’t usually get to actionable data. If an employee falls and is killed, and the investigation finds that he wasn’t using fall protection, the root cause hasn’t been found. Was the fall protection system adequate, did employees generally use it, was the working environment supportive, was leadership influence sufficient, did the organization have a value for safety? It isn’t possible to find actionable answers to these questions from a single incident. But a good set of incident reports analyzed as a series can yield very useful information. Were there precursors that could have been identified prior to the incident? What kinds of decisions were made that set the stage for the incident? Were they influenced by cognitive biases and if so what kind?