EPHM7320 Chap.5 Accident Causation Models
Accident Causation Models
Models direct attention to different evidence
Domino, human-factors, accident or incident, epidemiological, systems, combination and fault-tree approaches organise different questions about an event. A model earns its place by changing what the investigator seeks and which intervention becomes visible.
This chapter compares sequential and interactive explanations, shows why explaining human action is not excusing it, and requires combined accounts to connect findings rather than stack labels. Sequential models clarify order, barrier models expose defence relationships, and systems approaches reveal feedback and adaptation. Each also has blind spots.
Choosing a lens should therefore follow the investigation question and preserved facts. Strong application states what evidence could challenge the proposed explanation and shows how the model changes prevention rather than merely changing vocabulary.
What this chapter covers
- 01
Domino sequences and interruption
- 02
Human factors and performance conditions
- 03
Accident and incident interaction
- 04
Epidemiological combinations
- 05
Systems feedback and interfaces
- 06
Fault trees and combination theory
Worked example · free
Choose a model for alarm-response failure
- 3Identify evidence of habituation and expectation.
- 3Map alarm design and feedback relationships.
- 3Connect the combined account to controls.
Key terms
- Domino Theory
- A sequential model that treats an accident as linked events and seeks a preventable link.
- Human Factors
- The conditions shaping human capability, perception, workload, expectation and response during work.
- Systems Theory
- An approach examining control and feedback among people, equipment, organisation and environment.
- Epidemiological Theory
- An approach considering how predispositions and environmental conditions combine to produce an outcome.
- Combination Theory
- The use of connected explanatory lenses when one model leaves material evidence outside its account.
- Fault Tree
- A backward logical analysis that traces combinations of contributing conditions from a defined unwanted event.
Accident Causation Models FAQ
Which accident causation model is best?
No model is universally best. Choose the one that fits the relationships in the evidence and directs useful investigation. A sequence model suits ordered dependencies, while systems reasoning is stronger where feedback, adaptation and interacting components shape the event.
Does a human-factors explanation excuse unsafe action?
No. It explains how capability, workload, cues, expectations and work design shaped performance, which creates additional intervention points. Accountability questions remain, but prevention improves when the organisation understands why an action made sense under the conditions.
When is combination theory justified?
Use it when distinct models explain connected parts of the same evidence and the combined account changes control selection. Simply listing several theory names is not combination analysis; the answer must show how the relationships interact.
Exam move
Create a comparison card for each model with one diagnostic question, required evidence, blind spot and likely intervention. Apply two models to the same original event and explain which facts each reveals, then decide whether a connected combined account is justified. For a sequential model, reconstruct the event order and identify where a barrier or decision could interrupt progression.
For an epidemiological or barrier-oriented account, separate hazards, defences, local conditions and organisational influences. For a systems model, examine interacting decisions, feedback, adaptations and constraints across levels. Do not treat the diagrams as competing decorations; use each model to generate particular evidence questions.
Work with one stable scenario, such as maintenance on energised equipment, and redraw it through several lenses. Record what enters or disappears each time. A chain may clarify temporal dependence but understate parallel influences. A defence model may reveal latent weaknesses but tempt you to list holes without explaining interaction.
A broad systems model may capture adaptation yet become vague unless actors, constraints and evidence are named. Practise selecting a model by the question and available evidence, not by familiarity. If the problem asks how the event unfolded, sequence matters. If it asks why safeguards failed together, barrier relations matter.
If it asks why local behaviour was rational within organisational pressure, a wider systems account may be necessary. Test causal claims with counterfactual reasoning: if the proposed factor had been different, is there a supported mechanism by which the event trajectory would likely change? Avoid calling every antecedent a root cause.
Prioritise factors that explain the event, can be influenced and connect to durable prevention. In written answers, briefly state the model’s logic, map specific facts to its elements, identify a limitation and derive interventions at more than one level. Compare the interventions produced by two models and explain whether they complement or contradict each other.
Finish by building a combined causal map only when the connections are explicit. It should remain readable enough to guide action. Add effectiveness evidence beside every intervention so the causal account continues into organisational learning rather than ending at explanation.
