BACKGROUND AND OBJECTIVES: To describe the clinical features of anti-NMDAR encephalitis, emphasizing on late-onset patients and antibody test characteristics in serum and CSF. METHODS: Nationwide observational Dutch cohort study, in patients diagnosed with anti-NMDAR encephalitis between 2007 and 2019. RESULTS: One hundred twenty-six patients with anti-NMDAR encephalitis were included with a median age of 24 years (range 1-86 years). The mean annual incidence was 1.00/million (95% CI 0.62-1.59). Patients ≥45 years of age at onset (19%) had fewer seizures (46% vs 71%, p = 0.021), fewer symptoms during disease course (3 vs 6 symptoms, p = 0.020), and more often undetectable serum antibodies compared with younger patients (p = 0.031). In the late-onset group, outcome was worse, and all tumors were carcinomas (both p < 0.0001). CSF was more accurate than serum to detect anti-NMDAR encephalitis (sensitivity 99% vs 68%, p < 0.0001). Using cell-based assay (CBA), CSF provided an unconfirmed positive test result in 11/2,600 patients (0.4%); 6/11 had a neuroinflammatory disease (other than anti-NMDAR encephalitis). Patients with anti-NMDAR encephalitis, who tested positive in CSF only, had lower CSF antibody titers (p = 0.003), but appeared to have an equally severe disease course. DISCUSSION: Anti-NMDAR encephalitis occurs at all ages and is less rare in the elderly patients than initially anticipated. In older patients, the clinical phenotype is less outspoken, has different tumor association, and a less favorable recovery. Detection of antibodies in CSF is the gold standard, and although the CBA has very good validity, it is not perfect. The clinical phenotype should be leading, and confirmation in a research laboratory is recommended, when in doubt.