65 External Validation and Comparison of Three Predictive Tools to Identify Low-Risk Patients With Isolated Subdural Hematomas

Abstract

Three predictive scoring tools have been developed to identify low-risk patients with isolated subdural hematoma and preserved consciousness: the Orlando criteria, the SafeSDH Tool, and the Brain Injury Guidelines (BIG) criteria. We aim to validate and compare these three predictive tools in a single cohort of patients with isolated subdural hematomas. We performed a retrospective chart review of patients age > 16 with GCS >= 13 with CT-confirmed isolated subdural hematomas who presented to 1 academic and 3 community EDs. The Orlando criteria, SafeSDH Tool, and BIG criteria were applied to this dataset with a primary composite outcome (CO) of neurologic deterioration (severe headache, altered mental status, seizure, intubation), neurosurgical intervention, or death. Predictors used in the Orlando tool were: hematoma size and the presence of acute-on-chronic hematoma. The Orlando criteria was applied using the maximum sensitivity cutoff of 0.0235 from the derivation study. The Orlando Tool was also tested on both neurosurgical intervention (NI) (as in the derivation study), as well as the primary CO. The SafeSDH Tool includes the following predictors: use of warfarin, use of clopidogrel, number of hematomas, hematoma thickness, GCS, and midline shift. Cutoff used for this model was 0.0432, as in derivation. BIG criteria predictors included: intoxication, hematoma size, number of hematomas, and presence of skull fracture. Patients meeting both BIG1 and BIG2 criteria were considered low risk as these patients were deemed safe for management without neurosurgical consultation in derivation. The BIG criteria were not derived via regression, so no threshold cutoff was used and area under the receiver-operator curve (AUROC) was not calculated. The validation dataset consisted of 599 patients. 173 patients (28.9%) met the composite outcome. 157 (26.2%) underwent neurosurgical intervention, 22 patients (3.7%) had clinical deterioration, and 10 patients (1.7%) died. The performance of these tools in identifying low-risk patients, their sensitivity, specificity, and AUROC are summarized in Table 1 below. This is the first study to externally validate and compare performance of the Orlando criteria, SafeSDH Tool, and BIG Criteria in identifying low-risk patients with isolated subdural hematomas using a multicenter cohort. All rules had a similarly high sensitivity in predicting need for neurosurgical procedure, neurologic decline, or death. Although specificity is limited, with prospective validation all three tools could be used to identify patients who do not require neurosurgical consultation or transfer to tertiary care centers.