Development and internal validation of model predicting postoperative prolonged mechanical ventilation risk in pediatric patients following corrective operation of Tetralogy of Fallot

  AIM  To develop and internally validate a nomogram model predicting postoperative prolonged mechanical ventilation (PMV) risk in pediatric patients following corrective operation of Tetralogy of Fallot

  METHODS  Children aged from 6 months to 6 years old who underwent corrective operation of Tetralogy of Fallot in our hospital from June 2019 to December 2019 were selected and their clinical data were analyzed retrospectively. PMV was defined as the duration of postoperative mechanical ventilation exceeding 48 hours. Based on the selected children, a nomogram model predicting the risk of PMV was developed in the training set. The least absolute shrinkage and selection operator (LASSO) regression model was used to select the feature for multivariable logistic regression analysis applied to build a predicting model incorporating the feature selected in the LASSO regression model. The C-index, calibration plot and DCA were used to assess the discrimination, calibration and clinical utility of the prediction model. The bootstrapping validation was used to internally validate the model.

  RESULTS  The 109 children selected were divided into prolonged mechanical ventilation group (PMV group, 32 cases, 29.4%) and non-prolonged mechanical ventilation group (non-PMV group, 77 cases, 70.64%). The duration of postoperative mechanical ventilation in PMV group was longer than that in non-PMV group (P<0.01). Multivariate logistic regression analysis showed that the preoperative McGoon ratio < 1.5 (OR=3.564, 95% CI: 1.078-11.782, P<0.05) and the prolonged duration of cardiopulmonary bypass (CPB) (OR=1.020, 95% CI: 1.007-1.032, P<0.01) and lower postoperative left ventricular ejection fraction (OR=0.885, 95% CI: 0.792-0.988, P<0.05) were independent predictors of postoperative PMV. Moreover, the model displayed good calibration and good discrimination, with a C-index of 0.774. After the model was internally validated, the calibration plot performed well and high C-index value of 0.756 could still be reached. DCA showed that the nomogram was clinically useful when intervention was decided between the threshold probability of 2%−76%.

  CONCLUSION  We have developed and internally validated a highly accurate nomogram to assist ICU doctors in making clinical decisions regarding the postoperative prolonged mechanical ventilation. However, the model requires external validation before being recommended for clinical practice.