Factors associated with significant MRI findings in medical walk-in patients with acute headache

Abstract

Among walk-in patients who underwent MRI for acute headache with low suspicion for intracranial bleeding, 20% had a significant MRI finding. A simple clinical score identified all patients with significant findings that explained the headache. If prospectively validated, this might be a useful tool in selecting those walk-in headache patients requiring urgent cranial MRI.

Diagnostic yield of emergency MRI in non-traumatic headache

“…Regarding MRI findings in patients with headache in general, a fairly recent meta-analysis by Jang et al [5] found potentially significant abnormalities assessed by MRI in 5.7% (95% CI: 1.6-20%) of all patients suspected of primary headache. Budweg et al found that ~22% (18/82) of their walk-in patients had findings that could at least potentially explain their acute headache, of whom 10% (8 patients) had findings that were considered significant [9]. In both studies, only patients with a provisional diagnosis of a primary headache were included.…”

“…Several clinical risk scores have been developed for nontraumatic headache to reduce unnecessary imaging [6,8,9,11]. In these studies, the most frequently presented predictors of intracranial pathology have been age > 50 years, focal neurological deficit, nausea/vomiting, and altered mental status.…”

“…In these studies, the most frequently presented predictors of intracranial pathology have been age > 50 years, focal neurological deficit, nausea/vomiting, and altered mental status. However, most of these prediction models have been developed for cranial CT. Budweg et al presented a clinical score for MRI, but so far it has not been validated in a prospective study setting [9].…”

“…Studies using computed tomography (CT) have found secondary causes in 13-15% of emergency patients who had undergone cranial CT for headache, which were mostly intracranial hemorrhages or ischemia [6][7][8]. Magnetic resonance imaging (MRI) is a suitable alternative, with superior soft tissue characterization and no ionizing radiation, but only a few studies have investigated its yield in an emergency setting [9,10]. Budweg et al reported that ~22% (18/82) of their walk-in outpatients had at least potentially significant findings that explained acute…”

ROC and AUC with a Binary Predictor: a Potentially Misleading Metric

“…Therefore the ROC curve of a binary predictor is not really appropriate, but should be represented by a single point on the curve. But alas, ROC and AUC analysis is done on binary predictors and used to inform if one variable is more predictive than the other (E et al 2018;TV et al 2017;Glaveckaite et al 2011;Blumberg et al 2016;Budwega et al 2016;Mwipatayi et al 2016;Xiong et al 2018, Shterev et al (2018; Kushnir et al 2018;Snarr et al 2017;Veltri et al 2018). For example, these cases show that researchers use ROC curves and AUC to evaluate predictors, even when the predictors are categorical or binary.…”