Development of a risk prediction model for respiratory failure in elderly patients with severe pneumonia based on DynNom dynamic scoring

OBJECTIVE To develop a risk prediction model for respiratory failure in elderly patients with severe pneumonia based on DynNom dynamic scoring. METHODS A total of 205 elderly patients with severe pneumonia admitted to Shanghai Seventh People's Hospital Affiliated to Shanghai University of Traditional Chinese Medicine from Jan. 2021 to Jan. 2024 were selected as the study subjects. Patients were divided into a respiratory failure group (n=91) and a non-respiratory failure group (n=114) based on the occurrence of respiratory failure. Lasso regression analysis was used to screen variables, and multivariate logistic regression analysis was employed to identify risk factors for respiratory failure in elderly patients with severe pneumonia and establish a DynNom dynamic scoring prediction model. RESULTS The incidence rate of respiratory failure in elderly patients with severe pneumonia was 44.39%. Multivariate logistic regression analysis revealed that oxygenation index (OR: 0.911, 95%CI: 0.880-0.944), serum prealbumin (PAB) (OR: 0.986, 95%CI: 0.976-0.995), red blood cell volume distribution width (RDW) (OR: 1.515, 95%CI: 1.232-1.863), C-reactive protein (CRP) (OR: 1.061, 95%CI: 1.021-1.102) and interleukin-6 (IL-6) (OR: 1.011, 95%CI: 1.005-1.017) were influencing factors for respiratory failure in elderly patients with severe pneumonia (P<0.05). Validation results of the DynNom model: The C-index was 0.853 (95%CI: 0.825-0.881). The calibration curve showed that the model's calibration curve approached the ideal curve, with a Hosmer-Lemeshow goodness-of-fit test result of (χ²=10.829, P=0.212). The area under the receiver operating characteristic (ROC) curve (AUC) was 0.857 (95%CI: 0.814-0.887), and the AUC for internal validation through 1 000 Bootstrap-based repeated samplings was 0.855 (95%CI: 0.809-0.901). The net benefit curve indicated that when the predicted value ranged from 8% to 98%, the model's net benefit rate was >0. CONCLUSION The DynNom dynamic scoring prediction model developed based on the aforementioned factors has certain prediction value for the risk of respiratory failure in elderly patients with severe pneumonia.