Abstract Background and aims Distinguishing intracerebral hemorrhage from ischemic stroke in the prehospital setting would improve destination hospital and blood pressure strategy. The California Acute Stroke Subtype Prehospital (CASPR) Scales (CASPR-B and CASPR-ACS), leveraging information available to first responders, shows moderate performance. This study aimed to evaluate whether an Expanded-CASPR (E-CASPR) Scale, augmented with select NIH Stroke Scale (NIHSS) elements, would improve ICH identification. Methods In secondary analysis of data from the prehospital NIH FAST-MAG randomized clinical trial, we analyzed non-intubated patients with a final diagnosis of ischemic stroke or ICH. Prediction models included CASPR-B (n=998) and CASPR-ACS (n=102), enhanced using selection of 5 among 13 candidate NIHSS elements, via Akaike Information Criterion (AIC). Internal validation used a 70/30 training-testing split with ten-fold cross validation. Exploratory XGBoost models assessed feature importance. Results In derivation, CASPR-B ICH prediction improved with addition of NIHSS elements including lost consciousness (LOC), gaze deviation, leg weakness, dysarthria, and sensory, and CASPR-ACS with LOC, gaze deviation, facial palsy, visual field, and language. In validation analyses, CASPB-B performance improved AUC 0.707 to 0.743 (p0.05) and CASPR-ACS improved 0.647 to 0.787 (p0.05). NIHSS score alone performed near chance, and XGboost identified CASPR score as the most predictive variable for ICH. E-CASPR-B showed positive likelihood ratio (LR) 3.75 and negative LR 0.73 while E-CASPR-ACS demonstrated positive LR 5.15 and negative LR 0.62. Conclusions Incorporation of select NIHSS elements improves prehospital discrimination of ICH from ischemic stroke by CASPR. The selected NIHSS elements are feasible for prehospital implementation alongside CASPR, improving utility in stroke triage and treatment. Conflict of interest Alexander Fish nothing to disclose, Angshuman Saha nothing to disclose, Jeffrey Saver reports consulting fees for advising on rigorous and safe clinical trial design and conduct from Abbott, Aeromics, Bayer, Biogen, Boehringer Ingelheim, BrainQ, BrainsGate, CSL Behring, Medtronic USA, Roche, Stream Medical, Johnson & Johnson, MIVI Neuroscience and Occlutech and holds stock options from MindRhythm, Neuronics Medical, and Rapid Medical. Contracted hourly payments for advising on rigorous and safe clinical trial design and conduct.
Fish et al. (Fri,) studied this question.