A Predictive Model for Pediatric Postoperative Respiratory Failure: A National Inpatient Sample Study

To identify risk factors for pediatric postoperative respiratory failure and develop a predictive model.

This retrospective case-control study utilized the US National Inpatient Sample (NIS) from 2012 to 2014. Significant predictors were selected, and the predicted probability of pediatric postoperative respiratory failure was calculated. Sensitivity, specificity, and accuracy were then calculated, and receiver-operator curves were drawn.

National Inpatient Sample data sets from years 2012, 2013, and 2014 were used.

Patients aged 17 and younger in the 2012, 2013, and 2014 NIS data sets.

Candidate predictors included demographic variables, type of surgical procedure, a modified pediatric comorbidity score, presence of substance abuse diagnosis, and presence/absence of kyphoscoliosis.

The primary outcome measure was the pediatric quality indicator (PDI 09), which is defined by the Agency for Healthcare Research Quality, and identifies pediatric patients with postoperative respiratory failure.

The incidence of pediatric postoperative respiratory failure in each year's data set varied from 1.31% in 2012 to 1.41% in 2014. Significant risk factors for the development of postoperative respiratory failure included abdominal surgery ([OR] = 1.92 in 2012 data set, 1.79 in 2013 data set), spine surgery (OR = 7.10 in 2012 data set, 6.41 in 2013 data set), and an elevated pediatric comorbidity score (score of 3 or greater: OR = 32.58 in 2012 data set, 22.74 in 2013 data set). A predictive model utilizing these risk factors achieved a C statistic of 0.82.

Risk factors associated with postoperative respiratory failure in pediatric patients undergoing noncardiac surgery include type of surgery (abdominal and spine) and higher pediatric comorbidity scores. A prediction model based on the identified factors had good predictive ability.