A chain of causation is a sequence of connected events where each one leads to the next, ultimately producing a final outcome. The concept appears most often in law, medicine, and safety science, where professionals need to trace backward from a result (an injury, a death, an accident) to identify what set the whole sequence in motion. In legal cases, proving this chain is essential: without a clear link between someone’s action and the harm that followed, there’s no liability.
How It Works in Law
In tort law and criminal law, the chain of causation connects a defendant’s conduct to a plaintiff’s injury. Courts break this into two separate questions. The first is factual causation: did the defendant’s action actually contribute to the harm? The standard test here is the “but-for” test, which simply asks, “But for the defendant’s action, would the harm have occurred?” If the answer is no, factual causation exists.
The second question is proximate causation, which asks whether the connection is close enough to justify legal responsibility. This matters because technically, any event has an infinite chain of prior causes. A defendant’s mother gave birth to them, but that obviously doesn’t make her liable for their actions decades later. Proximate cause draws a practical boundary: at some point in the causal chain, a defendant’s conduct is too remotely connected to the injury to count as a legal cause. To qualify as a proximate cause, the action must be a substantial factor in bringing about the injury, and the general type of harm must have been reasonably foreseeable to someone exercising ordinary care.
Both elements must be present. Factual causation without proximate causation, or the reverse, is not enough to establish liability.
Breaking the Chain
A chain of causation can be legally severed by what’s called a “novus actus interveniens,” a new intervening act that disrupts the link between the original wrongdoing and the final outcome. If an independent event enters the sequence and is strong enough to redirect the outcome, it can reduce or eliminate the original wrongdoer’s liability.
Medical negligence cases illustrate this well. Say a surgeon makes an error during a procedure, but the patient’s harm ultimately results from a separate, unrelated mistake by an anesthesiologist. In that scenario, the anesthesiologist’s independent negligence broke the causal chain. If the patient’s family filed suit, the damages would be linked to the anesthesiologist’s care rather than the surgeon’s. The intervening act has to be genuinely independent and significant, though. A minor, foreseeable complication wouldn’t qualify.
The Eggshell Skull Rule
One important exception to the foreseeability requirement is the eggshell skull rule. This doctrine holds that a defendant is liable for the full extent of a plaintiff’s injuries, even if the plaintiff was unusually vulnerable. If someone with an extraordinarily fragile skull suffers a severe head injury in an accident that wouldn’t have seriously hurt most people, the person who caused the accident is still fully responsible. You take your victim as you find them. The chain of causation isn’t broken just because the outcome was worse than expected.
How Medicine Uses the Chain
Doctors and medical examiners use the same concept when documenting cause of death. The cause-of-death section on a U.S. death certificate, as outlined by the CDC, is structured as an explicit causal chain. It has multiple lines, each connected by the phrase “due to (or as a consequence of).” The immediate cause of death goes on the top line: the final disease, injury, or complication that directly killed the person. Each subsequent line records the condition that led to the one above it. The lowest used line captures the underlying cause of death, the disease or injury that set the entire chain of events in motion.
For example, a certificate might read: Line a, pulmonary embolism (the immediate cause), due to Line b, deep vein thrombosis, due to Line c, immobilization from a hip fracture, due to Line d, a fall. Medical examiners are specifically instructed not to list terminal events like cardiac arrest or respiratory arrest without documenting the underlying chain that led to them, because those events are final common pathways that tell you nothing about what actually caused the death.
The international coding system for health records, ICD-11, formalizes this further. It provides a structured mechanism for linking conditions in causal relationships using clusters of codes. A three-part model connects harm, cause, and mode (the mechanism linking cause to harm). When documentation doesn’t clearly support a causal relationship between two conditions, the guideline is to code them separately rather than assume a link.
Establishing Causation in Science
In epidemiology and public health, proving that one thing causes another requires more than showing they occur together. In 1965, Sir Austin Bradford Hill outlined nine criteria for evaluating causal relationships, originally used to demonstrate the link between smoking and lung cancer. These criteria remain the standard framework for assessing whether an observed association is genuinely causal.
The most universally accepted criterion is temporality: the cause must come before the effect. Beyond that, scientists look for strength of association (a stronger statistical relationship suggests causation), consistency (the same relationship appears across different studies and settings), and a dose-response relationship (more exposure leads to more of the effect). The proposed mechanism should also be biologically plausible given current scientific knowledge, and new evidence should align with what’s already established.
Two additional criteria round out the framework. Experiment asks whether reducing or removing the exposure changes the outcome, though scientists now recognize that diseases often involve multiple interacting causes, so removing just one factor doesn’t always reverse the condition. Analogy considers whether a known causal relationship between one exposure and outcome suggests a similar relationship for a comparable exposure. No single criterion is definitive on its own. Researchers weigh all nine together to judge whether a causal chain is supported by the evidence.
Chains of Causation in Accident Investigation
Safety science applies the concept differently. Rather than a single linear chain, the Swiss Cheese Model, developed by James Reason, visualizes accidents as the result of multiple failures aligning across layers of a system. Each layer represents a level of defense: organizational policies, supervisory decisions, preconditions, and individual actions. Every layer has weaknesses, represented as holes in slices of Swiss cheese. When holes across multiple layers line up at the same moment, a hazard passes through every barrier and an accident occurs.
The model distinguishes between two types of failures. Active failures happen at the point of contact, during surgery, while dispensing medication, or while operating equipment. They occur in close proximity to the harm event and constantly shift as people make errors, catch them, and correct them. Latent failures sit higher in the system, at the organizational or supervisory level. They can go undetected for months or years, lying dormant until they combine with other failures to produce a harmful outcome. Most serious accidents involve multiple active and latent failures interacting, not a single point of failure.
This model changed how investigators think about causation in healthcare, aviation, and industrial safety. Instead of tracing a single chain to one root cause, investigators map the full landscape of contributing factors, recognizing that removing any one of them might have been enough to prevent the outcome.