Establishment and evaluation of prognostic prediction models for patients with severe pneumonia complicated with ARDS in emergency department with different scores / 中华急诊医学杂志

ABSTRACT

Objective:To establish a predictive model of acute physiological and chronic health status score (APACHEⅡ) and the British Thoracic Society modified pneumonia score (CURB-65) score on the prognosis of patients with emergency severe pneumonia complicated with acute respiratory distress syndrome (ARDS) and to evaluate the predictive effect.Methods:The relevant clinical data of patients with severe pneumonia combined with ARDS admitted to the Emergency Intensive Care Unit (EICU) of General Hospital of Ningxia Medical University from January 2017 to December 2021 were retrospectively collected, and different logistic regression models were established. On this basis, three prediction models (model 1: APACHE Ⅱ score, model 2: CURB-65 score, Model 3: APACHE Ⅱ score combined with CURB-65 score) were established and the accuracy of the prediction model was evaluated by repeating 50 times of 10-fold cross-validations. The efficacy of the prediction model was evaluated by C statistics, Kendall's tau-a rank correlation coefficient, R2, Brier score, calibration curve, net reclassification index (NRI), composite discriminant improvement index (IDI) and decision curve (DCA).Results:The study eventually included 108 patients, including 81 males and 27 females, with mean age (57.92 ± 16.56) years. Forty-eight patients survived and 60 patients died. The age of the death group was older, and APACHEⅡ score and CURB-65 score of the death group were all greater than those in the survival group, and the differences were statistically significant ( P<0.05). Different logistic regression models showed that the OR value of model 1 was 1.12 (95% CI: 1.06 -1.20), that of model 2 was 2.21 (95% CI: 1.43 - 3.40), and that of model 3 was 1.10 (95% CI: 1.03 - 1.18) and 1.95 (95% CI: 1.24 - 3.07). The average accuracy of model 1, model 2, and model 3 were 0.68±0.14, 0.66±0.11, and 0.72±0.13, respectively. The C statistic, Kendall's Tau-a rank correlation coefficient, R2 and Bril score of model 3 were better than those of model 1 and model 2, and the different models fit well ( P<0.05). The calibration curve results of 500 resampling showed that the calibration degree of model 2 was better than that of model 1 and model 3, and the predictive ability of model 3 was improved compared with model 1, and the IDI was increased by 0.08 ( P<0.01). Compared with model 2, the reclassification ability of cases and the comprehensive discrimination ability of model 3 were improved ( P<0.01). The decision curves of different models showed that the net benefit of model 3 was higher than that of single model 1 and model 2 when the prediction probability was about 25% to 55%, while the benefits of model 1, model 2 and model 3 in other probability prediction intervals were basically equal. Conclusions:Both APACHE Ⅱ score and CURB-65 score have certain predictive power for prognosis of patients with emergency severe pneumonia and ARDS, and their combination has the best prediction effect. CURB-65 score has fewer parameters, and its prognostic benefit in emergency patients with severe pneumonia complicated with ARDS is basically equivalent to APACHE Ⅱ score, which may be more suitable for the prognosis evaluation of emergency patients with severe pneumonia complicated with ARDS.