Background: For the early detection of post-stroke dysphagia (PSD), valid screening parameters are crucial as part of a step-wise diagnostic procedure. This study examines the role of the National Institute of Health Stroke Scale (NIH-SS) as a potential low-threshold screening parameter. Methods: During a ten-year period, 687 newly admitted patients at University Hospital Muenster were included in a retrospective analysis, if they had ischemic or haemorrhagic stroke confirmed by neuroimaging and had received NIH-SS scoring and endoscopic swallowing evaluation upon admission. The NIH-SS score was correlated with dysphagia severity as measured by the validated 6-point fiberoptic endoscopic dysphagia severity score (FEDSS), and the ideal cut-off score to predict PSD, defined as FEDSS > 1, was calculated. Supra- and infratentorial strokes were analysed separately due to their differing role in the pathophysiology of neurogenic dysphagia. Results: NIH-SS and dysphagia severity show a significant positive correlation in the whole study population (R2 = 0.745) as well as in both analysed subgroups (R2 = 0.494 for supra- and R2 = 0.646 for infratentorial strokes, p < 0.0005, respectively). For supratentorial strokes, the ideal NIH-SS cut-off is > 9 (sensitivity 68.3%, specificity 61.5%, positive predictive value 89.7%, negative predictive value 28.4%). For infratentorial strokes, a lower ideal cut-off > 5 was calculated (sensitivity 67.4%, specificity 85.0%, positive predictive value 95.1%, negative predictive value 37.8%). Conclusions: NIH-SS may be used as an adjunct to predict dysphagia in acute stroke patients with moderate sensitivity and specificity. Differentiation between supra- and infratentorial regions is essential not to miss dysphagia in infratentorial stroke.