External Validation of a Tool to Predict Neurosurgery in Patients with Isolated Subdural Hematoma

Abstract

Subdural hematoma (SDH) is the most common form of traumatic intracranial hemorrhage. Orlando and colleagues derived a prediction tool for neurosurgical intervention, the "Orlando Tool," consisting of (a) maximum thickness of hematoma, and (b) presence of acute-on-chronic (AOC) hematoma. This study externally validated the Orlando Tool. We performed a retrospective chart review of consecutive patients aged ≥16 years with a Glasgow Coma Scale score ≥13, and a computed tomography-documented isolated, traumatic SDH, who presented to a university-affiliated, urban, 100,000-annual-visit emergency department from 2009-2015. The primary outcome was neurosurgical intervention. Thickness of hematoma and presence of AOC hematoma were abstracted from cranial computed tomography scan reports by 2 trained physician abstractors. A total of 607 patients with isolated SDH were included in the validation dataset. Median hematoma thickness was 6 mm. AOC hematoma was noted in 13% of patients. Mortality was 2.5%, and 15.7% of patients underwent neurosurgery. The Orlando Tool had an area under the curve of 0.93 in the validation, comparable to 0.94 reported in their derivation set. At the prespecified cutoff of 9.96% risk, the tool had a 88% (95% CI, 80-94) sensitivity in the validation cohort compared with 94% in the derivation cohort. The specificity of 82% (95% CI, 78-85) was comparable with 84% in the derivation group. Negative likelihood ratio was 0.14 (95% CI, 0.08-0.25), compared with 0.09 in derivation. The Orlando Tool accurately predicts neurosurgical intervention in patients with isolated, traumatic SDH and preserved consciousness.