Abstract: Objective　To explore and establish a clinical predictive model for patients with severe pneumonia. Methods　A total of 185 patients with severe pneumonia treated in our hospital were retrospectively analyzed. According to whether the patient received mechanical ventilation within 24 hours, they were divided into mechanical ventilation group (n=123) and non-mechanical ventilation group (n=62). Data of patient sex, age, blood gas analysis indicators such as arterial partial pressure of oxygen [p(O2)], arterial partial pressure of carbon dioxide [p(CO2)], alveolar-arterial partial pressure of oxygen [p(A-a)O2], oxygenation index (OI), and some laboratory findings at admission were analyzed. Multivariate Logistic regression was used to screen the risk factors affecting the need for mechanical ventilation in patients with severe pneumonia. A predictive model based on these selected indicators was constructed, and a nomogram was plotted. The predictive value of the model was evaluated through the receiver operating characteristic (ROC) curve and calibration curve. Results　Compared with the non-mechanical ventilation group, data of age, p(A-a)O2, APACHEⅡ score and p(CO2) were significantly higher in the mechanical ventilation group, while procalcitonin (PCT), central venous oxygen saturation (ScvO2), p(O2) and OI were significantly lower (P＜0.05). Logistic regression analysis showed that age, p(O2), p(CO2), p(A-a)O2 and OI were independent factors affecting whether the patients needed mechanical ventilation. The nomogram model constructed by the above five indicators showed a good discrimination (AUC=0.827, 95%CI: 0.785-0.898) and accuracy, which was better than the traditional p(O2)+ p(CO2)+OI model and OI model. Conclusion The nomogram model established based on age, p (O2), p (CO2), OI and p (A-a)O2 can accurately predict whether mechanical ventilation is required in the early stage of severe pneumonia patient.

Key words: pneumonia, respiration, artificial, blood gas analysis, Logistic models, severe pneumonia, nomograms