Diagnosis of subarachnoid haemorrhage following a negative computed tomography for acute headache: a Bayesian analysis

Objectives As access to computed tomography scanning has become easier, patients with a lower probability of serious pathology are being scanned. We asked how many lumbar punctures need to be performed to detect each subarachnoid haemorrhage in these lower-risk patients. Methods Literature review and Bayesian analysis of the application of the data to clinical practice. Results A computed tomography scan for acute headache has a negative likelihood ratio of 0.02 if the computed tomography is performed at < 12 h, 0.07 at < 24 h and 0.18 at > 24 h. A low pre-test probability, for example 1 in 20 (5%), and a negative computed tomography at < 12 h means that more than 1000 lumbar punctures would be required to detect each subarachnoid haemorrhage. Conclusions In patients who have a low pre-computed tomography probability of subarachnoid haemorrhage and undergo an early computed tomography scan, the risk/benefit ratio of lumbar puncture is unclear. A decision rule (risk stratification system) might improve our ability to help the patient make an informed choice. (C) 2006 Lippincott Williams & Wilkins.