External validation and comparison of three prediction tools to identify low risk patients with isolated subdural hematomas

Abstract

Introduction: Three prediction tools have been developed to identify low risk patients with isolated subdural hematoma (SDH) and preserved consciousness: the Orlando Tool, the SafeSDH Tool, and the Brain Injury Guidelines (BIG) Criteria. We externally validated and compared the performance of these tools.

Methods: We performed a secondary analysis of a retrospective chart review of patients age > 16 with a Glascow Coma Scale (GCS) ≥ 13 and isolated SDH. The primary outcome was composite of neurologic deterioration, neurosurgical intervention, discharge on hospice, and/or death. For each tool, we calculated sensitivity, specificity, negative likelihood ratio, and AUC (if possible) along with 95 % confidence intervals.

Results: The dataset included 753 patients; 230 (30.5 %) met the composite outcome. Of those, 195 (25.9 %) underwent neurosurgical intervention, 36 (4.8 %) had neurologic decline, and 31 (4 %) died or were discharged on hospice. Sensitivity for detecting the composite outcome was 100 %[95 % CI 98-100] for BIG1, 98 %[95-99] for BIG1 + 2, 99 %[97-100] for SafeSDH and 97 %[94-99] for Orlando. Specificity was 25 %[21-29] for BIG1, 40 %[36-44] for BIG1 + 2, 31 %[27-35] for SafeSDH and 54 %[50-58] for Orlando.

Conclusion: This is the first known study to externally validate and compare the performance of the Orlando Tool, SafeSDH Tool, and BIG Criteria to identify low risk patients with isolated SDH in a multicenter cohort. All three prediction tools had high sensitivity but limited specificity. These tools may, if validated prospectively, identify patients who do not require neurosurgical consultation or transfer to tertiary care centers. It remains to be seen which tool has the highest sensitivity and specificity when prospectively applied.

Keywords: Neurosurgery; Prediction tool; Subdural hematoma.