What it is and what it represents
A cause tree is a representation of the events leading up to a specific accident. It connects facts established during the investigation and shows how they contributed to the sequence of events that caused the harm. The method described by the INSST (National Institute for Safety and Health at Work) starts with the outcome and works backward to its antecedents. This approach allows for the consideration of multiple contributing factors, rather than relying on a single explanation.
The diagram is not the end goal. It serves to organize the reasoning and make it easier for others to examine the proposed relationships. A clear presentation should allow one to distinguish what happened, what evidence supports it, and what information is still missing. The appearance of a complete diagram does not demonstrate that the research is complete, especially if its connections have been built on assumptions.
Difference compared to other methods
The tree is applied retrospectively to the event under investigation. It should not be confused with a fault tree used to study possible combinations that lead to an event. Nor does it serve exactly the same function as the Ishikawa diagram, which is useful for categorizing hypotheses before testing them. Choosing a tool involves knowing what it can and cannot prove.
The five whys method can help delve deeper into a situation, but a single sequence of questions can mask different aspects. For example, a fall might require separate analyses of the loss of stability, exposure to a hazardous area, and the effectiveness of the fall protection system. Each aspect requires its own data and may lead to different measures.
Prepare the information before drawing
After attending to those involved and bringing the situation under control, the investigation must preserve available information and respect the actions of the relevant authorities. When working with the cause tree, it is advisable to prepare a record of events with its source: observations of the scene, documentation, relevant photographs, or testimonies. Discrepancies should not be erased when drafting a seemingly uniform account.
A work table can contain an identifier, a specific description, evidence, and the degree of confirmation. “The platform had an unprotected opening at the access point” is a verifiable statement; “it was poorly maintained” is an imprecise judgment. The second statement does not allow for reconstructing the sequence of events or deciding what technical or organizational changes should be implemented to prevent it.
Building relationships between facts
The analysis begins with the final event, identifying its immediate antecedents. This process is then repeated for each antecedent, determining whether only one was involved or if several were necessary. The method distinguishes between chaining relationships, concurrent antecedents, and branching consequences. Independent events should not be linked simply because they occurred close together.
During the construction process, it’s helpful to ask another team member to explain each connection in words. If the explanation requires adding information not included in the investigation, there’s a gap that needs to be addressed or acknowledged. When a precedent is unknown, this limitation should be made clear. Filling it with a common cause of other accidents can lead to an irrelevant measure.
Practical example
In a hypothetical example, a person slips in a hallway after stepping on liquid from a faulty connection. The investigation confirms the leak, its extent in the passageway, and that traffic was still permitted. The tree diagram allows investigators to separate the appearance of the liquid from the presence of people exposed to it. The investigation then examines why the connection wasn’t repaired and why the affected section wasn’t isolated.
The records show prior notification with no assigned responsible party and an incident procedure that did not define temporary measures. These conclusions must be supported by the case data, not automatically deduced from any leak. Proposed actions may include repair, containment, isolation rules, and assigned follow-up. Each addresses a specific part of the investigated sequence.
Turn analysis into measures
Corrective measures must be linked to the facts and conditions that need to be changed. If the investigation reveals a design flaw, simply reiterating an instruction may leave the problem unresolved. If notification management failures are identified, it’s advisable to review how notifications are assigned, prioritized, and closed, in addition to addressing the underlying material defect that caused the specific incident.
For each measure, you can specify the responsible party, deadline, and evidence of expected effectiveness. In the example, closing the repair order does not, in itself, demonstrate that other leaks will be managed better. A subsequent check can review a sample of reports, their response times, and their resolution times. Thus, the tree becomes a starting point for improving decisions and controls.
How to check tree quality
A useful review checks that the chart maintains a logical order, that each event is accurately described, and that the connections are supported by evidence. It also asks whether the technical conditions, organization, and context of the decisions have been examined. Focusing on a person’s last action can obscure why the working conditions made the harm possible.
The team must distinguish between confirmed results and open questions. This may require revisiting the site, verifying dates, or requesting additional technical information. The review should not be about finding a version that is convenient for the organization. Nor does it require a specific level of graphical complexity: a short, well-supported tree diagram can be more useful than a lengthy one with speculative relationships.
Regulatory scope and limits
In Spain, the Occupational Risk Prevention Law requires the investigation of certain injuries to identify their causes. Cause tree analysis is a technical tool to support this obligation, but it is not the only possible method. The chosen method should be appropriate to the complexity of the incident and the competence of the investigating team, and should allow for the adoption of well-founded preventive measures.
The method does not calculate probabilities on its own, nor does it guarantee that all causes have been identified. It also does not replace specialized investigation when complex technical failures are involved. Its contribution is to make reasoning about facts explicit, so that it can be reviewed and used for prevention. Learning requires supplementing the analysis with decisions, resources, and verification of their effects.
