<p>Purpose:　The aim of this study was to develop an algorithm to assess the risk of developing aspiration pneumonia.</p><p>Subjects:　Among those who were urgently admitted to the study hospital between April and August 2013 within 5 days after the onset of cerebral infarction or cerebral hemorrhage and who received conservative treatment without endotracheal intubation, we obtained consent for 160 patients to enter the study either directly from them or their families.</p><p>Methods:　We performed this study with the approval of the study ethical review board of Aichi Prefectural University and the study hospital. We collected basic data on subjects as well as relevant data regarding aspiration pneumonia from the medical records. Physical examination, intraoral observation, bacterial count, and tongue fluid volume measurements were carried out every other day from the 2nd through 14th hospital day. All subjects were grouped into either a pneumonia group or a non-pneumonia group based on definite. Subsequently, based on the algorithm, we determined the aspiration pneumonia risk and calculated the sensitivity and specificity, defining the pneumonia group as the gold standard.</p><p>Results:　One hundred and sixty subjects were classified into the pneumonia group (23 subjects; 16 developed aspiration pneumonia and 7 developed suspected aspiration pneumonia) and the non-pneumonia group (137 subjects). Fourteen subjects (87.5%) developed aspiration pneumonia by the 5th hospital day, for which we took particular note of the 2nd and 4th hospital days. Those determined as aspiration pneumonia-risk individuals by physical examination of the algorithm were 57 subjects on the 2nd hospital day and 60 subjects on the 4th hospital day. While the sensitivity and the specificity were 0.86 and 0.71 on the 2nd hospital day, they were 0.75 and 0.67 on the 4th hospital day, respectively. When there was tongue coating and secretion in the oral cavity on the 2nd hospital day, both the sensitivity and the specificity were 0.75. The inter-rater reliability on assessing each physical examination data was 82.0 to 95.3%, thus verifying reliability.</p><p>Conclusion:　We developed an algorithm for assessing the risk of aspiration pneumonia, and showed that the algorithm appears to be valid. When subjects with aspiration risk were screened using the algorithm on the 2nd hospital day, sensitivity was 0.86 and specificity was 0.71. Furthermore, when tongue coating and secretion in the oral cavity were observed, both the sensitivity and specificity were 0.75.</p>