Abstract: Objective To investigate the risk factors for post-stroke pneumonia and assess the value of A2DS2 score in predicting
post-stroke pneumonia in elderly stroke patients. Methods The clinical data were retrospectively collected from elderly stroke
patients from January, 2007 to December, 2012. A2DS2 score was then assigned using the clinical information from the medical
record. The ability of the score to discriminate between patients with post-stroke pneumonia and those without was quantified
using ROC analysis. The calibration of the score was analyzed using Hosmer-Lemeshow goodness-of-fit test. Results A total of
131 elderly male stroke patients were enrolled in this study, among whom the incidence of post-stroke pneumonia was 29.01%.
The independent risk factors for post-stroke pneumonia identified included moderate (P=0.0081, OR: 5.6089; 95% CI:
1.5663-20.0854) and severe (P=0.0048, OR: 44.4827; 95%CI: 3.1847-621.3126) neurological impairment, dysphagia (P=0.0005, OR:
7.5265; 95%CI: 2.4282-23.3292), and atrial fibrillation (P=0.0226, OR: 4.1778; 95%CI: 1.2221-14.2825). The incidence of post-stroke
pneumonia ranged from 2.2% in patients with a A2DS2 score less than 3 to 75% in those with a score higher than 8. The
C-statistic of A2DS2 score for predicting post-stroke pneumonia was 0.86 (95%CI: 0.784-0.911) by the ROC analysis. The A2DS2
score was well calibrated to predict post-stroke pneumonia in elderly patients by Hosmer-Lemeshow test (7.083, P=0.528).
Conclusion The A2DS2 score can be useful for predicting post-stroke pneumonia and for routine monitoring of high-risk
elderly stroke patients in the clinical setting.