Frontiers in Pediatrics (Apr 2023)

Predictive parameters and model for extubation outcome in pediatric patients

  • Kan Charernjiratragul,
  • Kantara Saelim,
  • Kanokpan Ruangnapa,
  • Kantisa Sirianansopa,
  • Pharsai Prasertsan,
  • Wanaporn Anuntaseree

DOI
https://doi.org/10.3389/fped.2023.1151068
Journal volume & issue
Vol. 11

Abstract

Read online

BackgroundProlonged mechanical ventilation is associated with significant morbidity in critically ill pediatric patients. In addition, extubation failure and deteriorating respiratory status after extubation contribute to increased morbidity. Well-prepared weaning procedures and accurate identification of at-risk patients using multimodal ventilator parameters are warranted to improve patient outcomes. This study aimed to identify and assess the diagnostic accuracy of single parameters and to develop a model that can help predict extubation outcomes.Materials and methodsThis prospective observational study was conducted at a university hospital between January 2021 and April 2022. Patients aged 1 month to 15 years who were intubated for more than 12 h and deemed clinically ready for extubation were enrolled. A weaning process with a spontaneous breathing trial (SBT), with or without minimal setting, was employed. The ventilator and patient parameters during the weaning period at 0, 30, and 120 min and right before extubation were recorded and analyzed.ResultsA total of 188 eligible patients were extubated during the study. Of them, 45 (23.9%) patients required respiratory support escalation within 48 h. Of 45, 13 (6.9%) were reintubated. The predictors of respiratory support escalation consisted of a nonminimal-setting SBT [odds ratio (OR) 2.2 (1.1, 4.6), P = 0.03], >3 ventilator days [OR 2.4 (1.2, 4.9), P = 0.02], occlusion pressure (P0.1) at 30 min ≥0.9 cmH2O [OR 2.3 (1.1, 4.9), P = 0.03], and exhaled tidal volume per kg at 120 min ≤8 ml/kg [OR 2.2 (1.1, 4.6), P = 0.03]; all of these predictors had an area under the curve (AUC) of 0.72. A predictive scoring system to determine the probability of respiratory support escalation was developed using a nomogram.ConclusionThe proposed predictive model, which integrated both patient and ventilator parameters, showed a modest performance level (AUC 0.72); however, it could facilitate the process of patient care.

Keywords