Sudden onset severe headache is usually caused by a primary headache disorder but occasionally is secondary to a more serious problem, such as subarachnoid hemorrhage (SAH). Guidelines recommend non-contrast brain computed tomography (CT) followed by lumbar puncture (LP) to exclude SAH. However, guidelines pre-date the introduction of more sensitive modern CT scanners. A systematic review was undertaken to assess the clinical effectiveness of different care pathways for the management of headache in the Emergency Department.

Eighteen databases (including MEDLINE and Embase) were searched to February 2020. Studies were quality assessed using criteria relevant to the study design; most studies were assessed using the QUADAS-2 tool for diagnostic accuracy studies. Where sufficient information was reported, diagnostic accuracy data were extracted into 2 × 2 tables to calculate sensitivity, specificity, false-positive and false-negative rates. Where possible, hierarchical bivariate meta-analysis was used to synthesize results, otherwise studies were synthesized narratively.

Fifty-one studies were included in the review. Eight studies assessing the accuracy of the Ottawa SAH clinical decision rule were pooled; sensitivity was 99.5 percent, specificity was 23.7 percent. The high false positive rate suggests that 76.3 percent SAH-negative patients would undergo further investigation unnecessarily. Four studies assessing the accuracy of CT within six hours of headache onset were pooled; sensitivity was 98.7 percent, specificity was 100 percent. CT sensitivity beyond six hours was considerably lower (≤90%; 2 studies). Three studies assessing LP following negative CT were pooled; sensitivity was 100 percent, specificity was 95.2 percent. LP-related adverse events were reported in 5.3–9.5 percent of patients.

The evidence suggests that the Ottawa SAH Rule is not sufficiently accurate for ruling out SAH and does little to aid clinical decision making. Modern CT within six hours of headache onset (with images assessed by a neuroradiologist) is highly accurate, but sensitivity reduces considerably over time. The CT-LP pathway is highly sensitive for detecting SAH, although LP resulted in some false-positives and adverse events.