Delirium occurs frequently in patients with stroke and neurocritical illness but is often underrecognized. We developed a novel delirium screening tool designed specifically for neurocritical care patients called the fluctuating mental status evaluation (FMSE) and aimed to test its usability and accuracy in a representative cohort of patients with intracerebral hemorrhage (ICH). We performed a single-center prospective study in a pilot cohort of patients with ICH who had daily delirium assessments throughout their admission. Reference-standard expert ratings were performed each afternoon using criteria from the Diagnostic and Statistical Manual of Mental Disorders, Fifth Edition, and were derived from bedside assessments and clinical data from the preceding 24 h. Paired FMSE assessments were performed by patients' clinical nurses after receiving brief one-on-one training from research staff. Nursing assessments were aggregated over 24-h periods (including day and night shifts), and accuracy of the FMSE was analyzed in patients who were not comatose to determine optimal scoring thresholds. We enrolled 40 patients with ICH (mean age 71.1 ± 12.2, 55% male, median National Institutes of Health Stroke Scale score 16.5 [interquartile range 12-20]), of whom 85% (n = 34) experienced delirium during their hospitalization. Of 308 total coma-free days with paired assessments, 208 (68%) were rated by experts as days with delirium. Compared with expert ratings, FMSE scores ≥ 1 had 86% sensitivity and 73% specificity on a per-day basis, whereas FMSE scores ≥ 2 had 68% sensitivity and 82% specificity. Accuracy remained high in patients with aphasia (FMSE scores ≥ 1: 83% sensitivity, 77% specificity; FMSE scores ≥ 2: 68% sensitivity, 85% specificity) and decreased arousal (FMSE scores ≥ 1: 80% sensitivity, 100% specificity; FMSE scores ≥ 2: 73% sensitivity, 100% specificity). In this pilot study, the FMSE achieved a high sensitivity and specificity in detecting delirium. Follow-up validation studies in a larger more diverse cohort of neurocritical care patients will use score cutoffs of ≥ 1 as "possible" delirium and ≥ 2 as "probable" delirium.