Introduction: A quarter of strokes occur in patients hospitalized for another reason. Inpatient stroke alerts are challenging due to complex medical situations. A stroke recognition instrument may be useful for general practitioners to discern neurologic deficits due to hemorrhagic or ischemic stroke from mimics such as seizures, delirium, and hypoglycemia.
Hypothesis: Clinical factors present at the initial evaluation of suspected inpatient stroke patients will distinguish strokes from mimics.
Methods: This was a retrospective review of inpatient stroke alerts from 9/1/2014 to 7/31/2016. Final diagnosis of stroke was based on stroke attending documentation and imaging confirmation if available. We evaluated predictors of stroke diagnosis including demographics, stroke risk factors, stroke alert reason, post-operative status and time from admission, NIHSS and CPSS scores abstracted from notes, vital signs, and laboratory values. Using the first 165 patients for derivation and the second 165 for validation, we used univariate and multivariate logistic regression and c-statistics to derive and validate a scoring system to predict stroke diagnosis.
Results: Among 394 alerts, 64 were excluded due to incomplete documentation. Of 330 remaining patients, 116 (35.2%) were diagnosed with stroke, 43 (13.0%) had a neurological mimic (e.g., seizure), and 171 (51.8%) had a non-neurologic mimic (e.g., sepsis, drug effect, or delirium). In multivariate analysis in the derivation cohort, we identified 4 independent predictors of inpatient stroke: abnormal CPSS, post-Cardiac procedure, history of Atrial fibrillation, and being a New (admitted within 24 hours) patient. The 2CAN score gives one point for each positive variable above, with additional 2 points for a CPSS ≥2 for a total score range from 0 to 6. The 2CAN score had c-statistic 0.93 (95% CI 0.88-0.97) in the derivation and 0.88 (95% CI 0.82-0.94) in the validation cohorts, respectively. A 2CAN score ≥2 had 92.2% sensitivity, 69.6% specificity, 62.2% positive predictive value, and 94.3% negative predictive value for identifying stroke.
Conclusion: The 2CAN score is a sensitive screening tool for identifying inpatient stroke. Further external and prospective validation of this score is warranted.