Canadian Head CT Rules: Accuracy and Key Limitations

The Canadian CT Head Rule is a clinical decision tool that helps emergency physicians determine which patients with minor head injuries actually need a CT scan and which can safely go home without one. Developed from a study of more than 3,000 patients across ten Canadian hospitals, the rule uses seven clinical findings, split into high-risk and medium-risk categories, to identify people at risk for brain injuries that matter clinically. In the original derivation, the high-risk criteria alone caught every single patient who needed brain surgery, while potentially cutting CT use by nearly 70 percent. The rule has since been validated internationally and compared against rival decision tools, though its real-world track record is more complicated than its promising origins suggest.

What the Rule Actually Checks For

The Canadian CT Head Rule applies to adults who come to the emergency department after a head injury with a Glasgow Coma Scale score between 13 and 15, meaning they are conscious and generally alert. The rule divides its criteria into two tiers. The five high-risk factors flag patients who may need neurosurgical intervention:

GCS below 15 at two hours: if the patient hasn’t returned to a perfect alertness score within two hours of injury
Suspected open skull fracture: signs that the skull bone has been broken with an overlying wound
Signs of basal skull fracture: bruising behind the ears, raccoon eyes, or fluid leaking from the nose or ears
Vomiting two or more times: repeated vomiting after the injury
Age 65 or older: older adults are at higher risk for clinically significant bleeding

If any one of those five is present, the patient gets a CT scan. If none of those apply, two medium-risk factors are checked to decide whether imaging is still warranted for a clinically important brain injury that might not need surgery but still matters:

Amnesia before impact longer than 30 minutes: the patient can’t remember events in the half hour leading up to the injury
Dangerous mechanism of injury: being hit by a car as a pedestrian, being ejected from a vehicle, or falling from a height of more than three feet (or five stairs)

If neither medium-risk factor is present either, the rule says CT can be safely skipped. In the original derivation study, the five high-risk criteria were 100 percent sensitive for predicting who would need neurosurgical intervention, while requiring only about a third of patients to undergo a scan. Adding the medium-risk criteria raised sensitivity for clinically important brain injury to roughly 98 percent while still allowing almost half of patients to avoid imaging altogether.¹

How Well It Performs Outside the Original Study

Derivation studies tend to look better than what follows, because the rule is being tested on the same kind of patients it was built from. External validation is where the real test happens. A meta-analysis pooling multiple validation studies found that the rule’s sensitivity for predicting clinically important traumatic brain injury was about 93 percent, with specificity around 40 percent.² That sensitivity number is strong but noticeably lower than the near-perfect figure from the derivation cohort. For predicting who needs neurosurgical intervention specifically, the rule has performed at 100 percent sensitivity in several validation studies, including one in the Netherlands that tested it on more than 3,000 patients.³

The specificity numbers deserve attention too, because they determine how many unnecessary scans you avoid. A specificity of 40 percent means the rule still sends a lot of people to the CT scanner who turn out to have nothing wrong. That is by design: a decision rule for something as serious as brain bleeding should err heavily on the side of catching everyone who is hurt, even if it means scanning some people who are fine. But it does limit how much the rule can trim imaging costs in practice.

How It Compares to the New Orleans Criteria

The Canadian CT Head Rule’s main rival in emergency medicine is the New Orleans Criteria, which uses a different set of findings (headache, vomiting, age over 60, drug or alcohol intoxication, persistent memory problems, evidence of injury above the collarbone, and seizure). The two tools represent different philosophies. The New Orleans Criteria cast a wider net, catching slightly more patients with any CT abnormality, but at the cost of sending almost everyone to the scanner. In the same meta-analysis, the New Orleans Criteria had about 97 percent sensitivity for positive CT findings but only around 12 percent specificity, compared to the Canadian rule’s roughly 90 percent sensitivity and 38 percent specificity.⁴

In practical terms, the Dutch validation study found that applying the Canadian rule could have reduced emergency CT scans by about 37 percent, while the New Orleans Criteria would have cut scans by only 3 percent.⁵ So the trade-off is clear: the New Orleans Criteria miss fewer injuries, but they barely reduce imaging at all. The Canadian rule misses a small number of CT-positive cases that wouldn’t have needed surgery anyway, while sparing far more patients from unnecessary radiation and emergency department delays. Which approach a hospital prefers depends on local risk tolerance, litigation culture, and how strained the CT scanner schedule is.

The Problem with Older Adults

Age 65 or older is one of the five high-risk criteria in the Canadian CT Head Rule, which means every older adult who comes in after a head injury automatically qualifies for a scan under the rule. That made sense in the derivation population, where older adults had meaningfully higher rates of dangerous findings. But in practice it creates a blunt instrument. Emergency departments that serve aging populations, particularly those near residential care facilities, end up scanning a huge proportion of their head-injury patients simply because of the age cutoff.

A study of older adults presenting after falls found that the prevalence of clinically important intracranial bleeding varied considerably depending on the circumstances. Among those who clearly met the rule’s application criteria, about 8 percent had significant bleeding. But among those who fell and did not clearly hit their heads, the rate dropped below 1 percent.⁶ That suggests the rule may need refinement for older adults, particularly when the history of whether the head was actually struck is murky, which is common in unwitnessed falls from care homes.

One research group tested what would happen if the age cutoff were raised from 65 to 75. Sensitivity stayed at 100 percent for clinically important brain injury, and the change would have reduced CT scans by about a quarter among patients aged 65 to 74.⁷ Pushing the threshold to 80 dropped sensitivity to about 91 percent, which most clinicians would consider too risky. The 75-year mark seems to offer a sweet spot, but it hasn’t been widely adopted yet.

An Australian audit of a residential aged-care population showed that formally implementing the Canadian rule with its original inclusion criteria cut CT scanning by about 20 percent without missing any patients who needed neurosurgical intervention.⁸ That reduction came largely from properly applying the rule’s inclusion criteria, since many of these patients didn’t actually have a witnessed head strike and therefore fell outside the rule’s scope entirely. The finding highlights an underappreciated feature of the Canadian rule: its exclusion criteria are as important as its risk factors, and misapplying the rule to patients it was never designed for inflates CT rates.

Patients on Blood Thinners

The original Canadian CT Head Rule explicitly excludes patients taking anticoagulant or antiplatelet medications. This means anyone on warfarin, a direct oral anticoagulant, or even daily aspirin is supposed to get a CT scan regardless of whether any other risk factor is present.⁹ The rationale is straightforward: blood thinners can turn a minor bleed that would otherwise seal itself into an expanding hemorrhage, and the rule wasn’t validated in this population so its reassurance doesn’t apply.

This exclusion has become a growing practical issue because anticoagulant use has surged, particularly among older adults with atrial fibrillation. An emergency department serving an elderly population may find that a substantial fraction of their head-injury patients are automatically excluded from the rule and sent to the scanner by default. Whether a more nuanced approach could safely be developed for lower-risk anticoagulated patients is an active area of research, but for now the guidance is clear: if you’re on blood thinners and you hit your head, you’re getting a scan.

Why It Doesn’t Work for Children

The Canadian CT Head Rule was derived from adult patients and should not be applied to children. Pediatric head injuries follow different patterns: children’s skulls are thinner, their brains are more susceptible to certain types of injury, and the threshold for worrying about radiation exposure is lower because of their longer remaining lifespan. When researchers tested the adult Canadian rule in a pediatric emergency department, its sensitivity dropped to about 65 percent, meaning it would have missed roughly a third of children with positive CT findings.¹⁰ That is nowhere near acceptable.

Separate pediatric decision rules exist for children. The best known is the PECARN rule from the United States, developed specifically for children under 18. Canada developed its own pediatric counterpart called CATCH (Canadian Assessment of Tomography for Childhood Head Injury), later updated as CATCH2. In a validation study, CATCH2 achieved 100 percent sensitivity for detecting intracranial pathology in children, compared to 80 percent for the PECARN rule, though neither difference was statistically significant given the small number of positive cases.¹¹ Both pediatric rules help reduce CT rates in children, where minimizing radiation exposure is especially important. A decision aid for parents of children with minor head trauma has shown that involving parents in the imaging decision increased their knowledge and reduced decisional conflict without increasing the rate of missed injuries.¹²

The Implementation Paradox

Here is the uncomfortable truth about the Canadian CT Head Rule: introducing it in hospitals doesn’t always reduce CT scanning. In a large cluster-randomized trial across Canadian emergency departments, the hospitals that actively implemented the rule actually saw their CT imaging rate go up, from about 63 percent to 76 percent. The control hospitals, which received no special training, also saw rates climb, from about 68 percent to 74 percent. There was no significant difference between the two groups.¹³

This result puzzled the researchers who had developed the rule precisely to reduce unnecessary scanning. The likely explanation lies in how the rule interacts with physician behavior. Before the rule existed, many doctors were already applying informal clinical judgment that happened to skip CT in low-risk patients. Formalizing the criteria paradoxically made some physicians more aware of potential risk factors they had previously been comfortable ignoring. A doctor who previously would have sent a healthy 67-year-old with a minor bump home without imaging might now feel obligated to scan because the rule says “age 65 or older” is high-risk.

Interviews with physicians at intervention sites identified several barriers to the rule reducing scans as intended. Despite knowing the rule was highly sensitive and believing in its value, doctors cited concerns about consequences of missing an injury, difficulty changing established scanning habits, and environmental pressures within the emergency department.¹⁴ A separate study at a Nova Scotian trauma center found that physicians followed the rule’s recommendation in about 88 percent of cases, meaning roughly one in eight patients was managed in a way that contradicted the rule’s guidance.¹⁵

Fear of Lawsuits and Patient Expectations

A recurring theme in studies on CT ordering patterns is that clinical decision rules compete with powerful nonclinical motivations. In one survey, about a fifth of physicians admitted that fear of being sued influenced their decision to order a head CT, and 8 percent cited pressure from the patient or their family.¹⁶ A separate study looking specifically at patients with minimal head injuries, people so mildly hurt that they arguably fell below the Canadian rule’s threshold for even being evaluated, found that doctors listed their own reassurance and patient reassurance as the top reasons for ordering scans. Reducing legal liability accounted for about 12 percent of CT orders in that group.¹⁷

These findings help explain the implementation paradox. A clinical rule can be perfectly designed on paper, but if doctors are also managing their own anxiety, their patients’ expectations, and their perceived legal exposure, the rule becomes one input among many rather than the definitive decision-maker. The 12 percent figure for legal liability may sound modest, but in a busy emergency department handling thousands of head injuries a year, that translates to hundreds of scans driven not by clinical need but by defensive medicine.

Validation in Resource-Limited Settings

The Canadian CT Head Rule was developed in well-resourced hospitals with rapid access to CT scanners. Its potential value is arguably even greater in settings where scanners are scarce or expensive to operate, because every scan saved has a larger impact. Validation studies in lower-resource countries have generally been encouraging. In Ethiopia, the rule achieved 100 percent sensitivity for identifying patients who needed neurosurgical intervention, with a specificity of about 42 percent.¹⁸ A study in a developing-nation setting similarly found 100 percent sensitivity and about 46 percent specificity when emergency physicians applied the rule, and reported that neurosurgeons at the same institution chose to scan every single patient based on clinical judgment alone.¹⁹ That contrast between a rule-based approach and a scan-everyone approach is exactly the kind of gap the rule was designed to close.

An Austrian study struck a more cautionary note, finding that the rule’s sensitivity for detecting who needed neurosurgery was 80 percent in their elderly patient cohort, compared to 90 percent for a locally developed set of predictors.²⁰ That study specifically focused on older patients, a population where, as noted above, the rule’s blunt age cutoff may not be fine-grained enough. The broader picture from international validations is that the rule travels well for detecting surgical emergencies, but its performance varies more for the full spectrum of clinically important brain injuries depending on the patient mix.

Blood Biomarkers as a Possible Upgrade

The most promising evolution of the Canadian CT Head Rule may involve combining it with blood tests. Several proteins that leak from damaged brain cells into the bloodstream can be measured within hours of an injury. Glial fibrillary acidic protein (GFAP) and ubiquitin C-terminal hydrolase L1 (UCH-L1) are the two that have attracted the most attention. A composite blood test using these two markers achieved a negative predictive value of 100 percent for abnormal CT findings in one study, meaning that when the blood test came back negative, no patients had anything on their scan.²¹

More interesting is what happens when you pair the biomarkers with the clinical rule rather than using either alone. Research combining the Canadian CT Head Rule with GFAP measurement found that the two together outperformed either tool individually, and that GFAP added meaningful predictive power on top of the clinical criteria. A separate approach using the biomarker S100B alongside clinical variables suggested the combination could reduce unnecessary brain CT scans by about a third.²² The idea is appealing: a quick blood draw could help the emergency physician decide whether a borderline patient really needs the scanner, combining the speed and simplicity of the clinical rule with the biological specificity of a lab test. Point-of-care blood biomarker assays already exist and have received regulatory approval in some countries, though they haven’t yet been widely integrated into standard emergency protocols for head injury triage.