The derivation and vulidafion of a prediction rule for hypoxemia in infants with Pierre Robin sequence after weaning

Objective: To set up a prediction scoring system for the hypoxemia in infants with Pierre Robin sequence after weaning and evaluate its clinical value. Methods: Data of consecutive patients from November 2016 to June 2019, who underwent mandibular distraction osteogenesis in Guangzhou Women and Children's Medical Center, were retrospectively analyzed (n=148). All the cases were divided into two groups according to the appearance of hypoxemia after weaning. They were randomly divided into the derivation cohorc (2/3,n=100) and the validation cohort (1/3,n=48). Single factor and multiple logistic regression analysis were used to select the independent risk factors related to hypoxemia and establish a prediction model. A prediction scoring system was developed in accordance with assigning of the value of each variable beta in the model. Internal verification of scoring system by validation population. Data of consecutive patients from July 2019 to November 2019, who underwent mandibular distraction osteogenesis, were prospectively analyzed (n=26). The diagnostic accuracy were conducted to evaluate the clinical value of the scoring system. Results: The logistic regression demonstrated that age at operation, pulmonary infection and the length of distraction less than 5 mm at weaning were the independent risk factors for hypoxemia. The P value of logistic regression model in Hosmer and Lemeshow goodness of fit test was 0.848, and a prediction scoring system was established accordingly. The area under the ROC curve of the scoring system was 0.890, and the optimum critical value was 53. The sensitivity, specificity, accuracy of the model were 78.6%(11/14),86.1%(74/86), 85.0%(85/100) respectively. The predictive effectiveness of the scoring system in the retrospective validation population was similar to that in the modeling population. 26 patients were included in the prospective analysis. The area under ROC curve of the scoring system was 0.870. The sensitivity, specificity and accuracy were 80.0%(5/6),95.0%(20/21), 96.1%(25/26) respectively. Conclusion: The prediction scoring system established in the study are efficacious for the hypoxemia in infants with Pierre Robin sequence after weaning.