Abstract:

Objective To investigate the roles of carbohydrate antigen 125（CA125）,neutrophil/lymphocyte ratio（NLR）,platelet/lymphocyte ratio（PLR）and high-sensitivity C-reactive protein（hs-CRP） combined determination for community-acquired pneumonia with pleural effusion. Methods A total of 284 patients with community-acquired pneumonia were enrolled. There were 180 patients with pneumonia without pleural effusion and 104 patients with pneumonia with pleural effusion. The levels of CA125,NLR,PLR and hs-CRP in the 2 groups were determined. The multilayer perceptron（MLP） neural network model,Chi-square automatic interaction detection（CHAID） decision tree model and Logistic regression model were established,and the diagnostic efficiency of CA125,NLR,PLR and hs-CRP combined determination was analyzed by receiver operating characteristic（ROC） curve. Results There was statistical significance in CA125,NLR,PLR and hs-CRP between pneumonia groups without and with pleural effusion（P<0.01）. In pneumonia group with pleural effusion,there was statistical significance in CA125,NLR,PLR and hs-CRP in small,medium and large amounts of pleural effusion（P<0.01）,and CA125,NLR,PLR and hs-CRP were positively correlated with the volume of pleural effusion（P<0.01）. The accuracy of the MLP neural network model was 78.1%,and the area under curve（AUC） was 0.803,that of the CHAID decision tree model was 75.0%（AUC=0.775）,and that of the Logistic regression model was 74.3%（AUC=0.785）. Conclusions MLP neural network model combined with CA125,NLR,PLR,hs-CRP determinations for community-acquired pneumonia with pleural effusion has certain auxiliary diagnostic value.

Key words: Carbohydrate antigen 125, Neutrophil/lymphocyte ratio, Platelet/lymphocyte ratio, High-sensitivity C-reactive protein, Datum mining, Community-acquired pneumonia, Pleural effusion