Objective To investigate the risk factors for bacterial pneumonia and the predictive value of early serum procalcitonin (PCT) level for bacterial pneumonia and sepsis classification in patients with acute stroke. Methods The patients with acute stroke in neurological intensive care unit were enroled retrospectively and divided into either a bacterial pneumonia group or a non-infection group according to whether they had bacterial pneumonia or not. The former was redivided into a non-severe sepsis subgroup and a severe sepsis subgroup according to the sepsis classification. The demographics, baseline clinical data, and PCT level (the bacterial pneumonia group was the PCT level when infection occurred, the non-infection group was the PCT level within 24 h of admission) were compared. Multivariate logistic regression analysis was used to identify the independent risk factors for bacterial pneumonia. Receiver operator characteristic (ROC) curve was used to analyze the predictive value of serum PCT level for bacterial pneumonia and sepsis
classification. Results A total of 164 patients with acute stroke were enroled in the study, including 114 in the bacterial pneumonia group (66 in the non-severe sepsis subgroup and 48 in the severe sepsis subgroup) and 50 in the non-infection group. There were significant differences in age, fasting blood glucose level, Glasgow coma scale (GCS) score, and PCT level between the bacterial pneumonia group and the non-infection group (P < 0. 05 ). Multivariate logistic regression analysis showed that fasting blood glucose level ≥7 mmol/L (odds ratio [ OR] 8. 488, 95% confidence interval [ CI] 2. 739 - 26. 300; P < 0. 01), GCS score ≤8 (OR 11. 361, 95% CI 2. 175 - 59. 352; P < 0. 01), and PCT level ≥0. 050 ng/ml (OR 16. 715, CI 5. 075 - 55. 049; P < 0. 01) were the independent risk factors for bacterial pneumonia. In the bacterial pneumonia group, the PCT level (median; interquartile range) in the severe sepsis subgroup was significantly higher than that in the non-severe sepsis subgroup (0. 835 [ 0. 164 - 1. 715 ] ng/ml vs. 0. 114 [0. 073 - 0. 275 ] ng/ml; Z = 4. 818, P < 0. 01 ). ROC curve analysis showed that PCT ≥0. 070 ng/ml could better predict the occurrence of bacterial pneumonia in patients with acute stroke, with sensitivity of 84. 2% , specificity of 74. 0% and the area under the ROC curve of 0. 865 (CI 0. 806 - 0. 924, P < 0. 01); PCT 0. 669 ng/mlcould better predict the occurrence of severe sepsis in acute stroke patients with bacterial pneumonia, with sensitivity of 56. 3% , specificity of 92. 4% and the area under the ROC curve of 0. 765 (CI 0. 672 - 0. 858; P < 0. 01). Conclusions The early PCT level ≥0. 050 ng/ml was an independent risk factor for occurring bacterial pneumonia in patients with acute stroke, its level had certaln predictive value for bacterial pneumonia and the severity of infection.