Machine learning for prediction of bronchopulmonary dysplasia-free survival among very preterm infants

Abstract

Background Bronchopulmonary dysplasia (BPD) is one of the most common and serious sequelae of prematurity. Prompt diagnosis using prediction tools is crucial for early intervention and prevention of further adverse effects. This study aims to develop a BPD-free survival prediction tool based on the concept of the developmental origin of BPD with machine learning. Methods Datasets comprising perinatal factors and early postnatal respiratory support were used for initial model development, followed by combining the two models into a final ensemble model using logistic regression. Simulation of clinical scenarios was performed. Results Data from 689 infants were included in the study. We randomly selected data from 80% of infants for model development and used the remaining 20% for validation. The performance of the final model was assessed by receiver operating characteristics which showed 0.921 (95% CI: 0.899–0.943) and 0.899 (95% CI: 0.848–0.949) for the training and the validation datasets, respectively. Simulation data suggests that extubating to CPAP is superior to NIPPV in BPD-free survival. Additionally, successful extubation may be defined as no reintubation for 9 days following initial extubation. Conclusions Machine learning-based BPD prediction based on perinatal features and respiratory data may have clinical applicability to promote early targeted intervention in high-risk infants.

Comparative effectiveness of explainable machine learning approaches for extrauterine growth restriction classification in preterm infants using longitudinal data

“…When comparing with other related studies, Han et al ( 31 ) achieved an AUROC of 74%, Leigh et al ( 32 ) achieved a receiver operating characteristics performance of 92.10%, Wu et al ( 33 ) achieved an AUROC of 88.10%, and Podda et al ( 34 ) achieved an accuracy of 91.49%, our study achieved an AUC of 83 and 78% for the weight-based dataset using LR and SVM, respectively. For the length-based dataset, our study achieved an AUC of 75 and 71% using LR and SVM, respectively.…”

“…Given that explicitly describing a black-box model remains a niche ( 74 ), our study employed a global interpretable ML models to construct a decision support system, especially for making critical medical decisions. Although preterm infant prediction models that utilize ML have been previously reported ( 32 , 34 , 75 ), longitudinal studies involving the comparative evaluation of ML methods in an interpretable approach have been limited. Due to dynamic nature of infant growth, infants undergo rapid growth and development within their first few weeks of life.…”

“…The authors concluded that machine learning algorithms, particularly XGB, could help neonatologists identify high-risk infants for PGF and enable early intervention. Leigh et al ( 32 ) applied machine learning to predict bronchopulmonary dysplasia (BPD)-free survival among very preterm infants using data from 689 infants. The final model demonstrated 92.10% receiver operating characteristics performance in both the training and validation datasets.…”

Early prediction of pulmonary outcomes in preterm infants using electrical impedance tomography

“…We speculate that the combination of lung ultrasound and EIT may allow a good prediction of BPD even as early as 30 min after birth but prospective clinical studies with a larger sample size are needed to evaluate this hypothesis. As BPD is a multifactorial disease, an all-encompassing approach may be needed which includes clinical and visualization (EIT, lung ultrasound) parameters as well as biomarkers ( 29 ) and machine-learning ( 30 ) to accurately predict this important outcome for patients and their families.…”

A clinical informatics approach to bronchopulmonary dysplasia: current barriers and future possibilities

“…Additionally, existing models do not account for the infant's clinical trajectory over time, which could provide a dynamic approach for better personalized preventive treatment and targeted trial recruitment. Nonetheless, a recent machine learning (ML)-based prediction model demonstrated the predictive power of postnatal respiratory support for 14 consecutive days ( 84 ). While this model was developed at a single center without validation, it demonstrates how prediction models and tools are evolving to better meet clinical needs.…”