Skip navigation

Skull X-ray for mild head injury?

Head injury is not uncommon. For most (80-90%) people who sustain a head injury it is mild, and they need neither admission to hospital or complex health care. The worry is the development of intracranial haemorrhage. A meta-analysis [1] of radiological diagnosis indicates that a plain skull X-ray is of little value in initial assessment.

Mind you, after reading this necessarily complex paper, one feels as if a mild head injury would come as a welcome relief. That is not any adverse comment on the authors, who have done a fantastic job, but on the innate difficulty in sorting out diagnostic tests. This is another example of where the archaeology of past attempts to sort out the value of a test shows them to crumble to dust when exposed to the bright lights of contemporary reasoning.


A number of databases were used to find studies that were relevant on two counts. First were studies that informed about the prevalence of intracranial haemorrhage in patients with mild head injury. Second were studies that informed on the diagnostic value of a finding of skull fracture. Prospective and retrospective studies were included.

Mild head injury was defined as trauma to the head with a Glasgow coma score of 13 to 15 on initial presentation. The diagnosis of intracranial haemorrhage was sensible, ideally requiring a CT scan, though uneventful recovery in the absence of a CT scan was considered to indicate the absence of intracranial haemorrhage. Studies with fewer than 50 patients were excluded, as were those on children and older people.


The mean prevalence of intracranial haemorrhage after mild head injury was 8% (95% confidence interval 3% to 13%) in 13 studies with 12,750 patients. Loss of consciousness or post traumatic amnesia occurred in 61% to 100% of patients in individual studies (most commonly 100%).

Differences in patient selection and the percentage of patients receiving a CT scan to verify skull fracture were important sources of variation between studies. This variation was large. Sensitivity varied between 13% and 75%, and specificity between 91% and 99.5%. The mean sensitivity was 50% at a specificity of 97%. For studies that had less bias where more than 50% of patients had loss of consciousness or post trauma amnesia and more than 50% had a CT scan, the mean sensitivity was 38% at a mean specificity of 95%.

What does this mean in practice? For a hypothetical 1000 patients with mild head injury, 83 would have an intracranial haemorrhage and 917 would not (Table). The likelihood ratio for a positive test for skull fracture was 7.7, raising the post-test probability to about 35%. The likelihood ratio for a negative test for skull fracture was 0.7, with a post-test probability of about 5%.

Table: Findings for a hypothetical population of 1000 patients with mild head injury

  Intracranial haemorrhage
  Present Absent
Fracture present 32 46
Fracture absent 51 871
Total 83 917
Sensitivity of positive test 0.39
Specificity of negative test test 0.95
Likelihood ratio for a positive test 7.7
Likelihood ratio for a negative test test 0.6


Like many, even most, reviews of diagnostic tests this one is hard to get to grips with. The simple take home message from the authors was that plain skull X-ray has no place in the assessment of mild head injury in adult patients. If an intracranial haemorrhage is not seen on a plain skull X-ray, then intracranial haemorrhage can still not be ruled out. We are reminded as well that it depends on who reads the X-ray: experienced physicians miss up to 10% of skull fractures. The authors of the review further concluded that patients with a Glasgow coma score of 15 with loss of consciousness or post trauma amnesia, and patients with a score of 13 or 14, require observation, a CT scan, or both.

This is useful information. The archaeology of the extant literature helps explain differences between studies of diagnostic tests or strategies. Even more it informs on how to design studies that might in future help better identify patients at risk of developing an intracranial haemorrhage after mild head injury. The lesson of this review is that when it comes to making effective and efficient diagnosis we need much better information than we have now. More archaeology won't help. We need new studies constructed to a better design, and providing outputs that will actually help in everyday situations. It may be expensive, and it may take time, but without better information we can't do better.


  1. PA Hofman, P Nelemans, GJ Kemerink, JT Wilmink. Value of radiological diagnosis of skull fracture in the management of mild head injury: meta-analysis. J Neurol Neurosurg Psychiatry 2000 68: 416-422.
previous or next story in this issue