Journal article
Comparing machine learning techniques for neonatal mortality prediction: insights from a modeling competition
Pediatric research, Vol.98(2), pp.405-411
08/2025
DOI: 10.1038/s41390-024-03773-5
PMCID: PMC12454155
PMID: 39681666
Abstract
Predicting mortality risk in neonatal intensive care units (NICUs) is challenging due to complex, variable clinical and physiological data. Machine learning (ML) offers potential for more accurate risk stratification.BACKGROUNDPredicting mortality risk in neonatal intensive care units (NICUs) is challenging due to complex, variable clinical and physiological data. Machine learning (ML) offers potential for more accurate risk stratification.To compare the performance of various ML models in predicting NICU mortality using a team-based modeling competition.OBJECTIVETo compare the performance of various ML models in predicting NICU mortality using a team-based modeling competition.We conducted a modeling competition with five neonatologist-led teams applying ML techniques-logistic regression, CatBoost, neural networks, random forest, and XGBoost-to a shared dataset from over 6,000 NICU admissions. The dataset included static demographic and clinical variables, alongside daily samples of heart rate and oxygen saturation. Each team developed models to predict mortality risk at baseline and within 7 days. Models were evaluated using the area under the receiver operator characteristic curve (AUC). Results were presented at a national meeting, where an audience poll ranked models before AUC results were revealed.METHODSWe conducted a modeling competition with five neonatologist-led teams applying ML techniques-logistic regression, CatBoost, neural networks, random forest, and XGBoost-to a shared dataset from over 6,000 NICU admissions. The dataset included static demographic and clinical variables, alongside daily samples of heart rate and oxygen saturation. Each team developed models to predict mortality risk at baseline and within 7 days. Models were evaluated using the area under the receiver operator characteristic curve (AUC). Results were presented at a national meeting, where an audience poll ranked models before AUC results were revealed.The audience favored the most complex model (CNN) for real-world application, though logistic regression achieved the highest AUC on test data. Teams employed varied feature selection, tuning, and evaluation strategies.RESULTSThe audience favored the most complex model (CNN) for real-world application, though logistic regression achieved the highest AUC on test data. Teams employed varied feature selection, tuning, and evaluation strategies.Logistic regression outperformed more complex models, highlighting the importance of selecting modeling methods based on data characteristics, interpretability, and expertise rather than model complexity alone.CONCLUSIONSLogistic regression outperformed more complex models, highlighting the importance of selecting modeling methods based on data characteristics, interpretability, and expertise rather than model complexity alone.By demonstrating that model complexity does not necessarily equate to better predictive performance, this research encourages the careful selection of modeling approaches.IMPACTBy demonstrating that model complexity does not necessarily equate to better predictive performance, this research encourages the careful selection of modeling approaches.
Details
- Title: Subtitle
- Comparing machine learning techniques for neonatal mortality prediction: insights from a modeling competition
- Creators
- Brynne A Sullivan - University of VirginiaAlvaro G Moreira - The University of Texas at San AntonioRyan M McAdams - University of Wisconsin–MadisonLindsey A Knake - University of IowaAmeena Husain - University of UtahJiaxing Qiu - University of VirginiaAvinash Mudireddy - University of IowaAbrar Majeedi - University of Wisconsin–MadisonWissam Shalish - Montreal Children's HospitalDouglas E Lake - University of VirginiaZachary A Vesoulis - Washington University in St. Louis
- Resource Type
- Journal article
- Publication Details
- Pediatric research, Vol.98(2), pp.405-411
- DOI
- 10.1038/s41390-024-03773-5
- PMID
- 39681666
- PMCID
- PMC12454155
- NLM abbreviation
- Pediatr Res
- ISSN
- 1530-0447
- eISSN
- 1530-0447
- Publisher
- SPRINGERNATURE
- Grant note
- National Institutes of Health (NIH) Eunice Kennedy Shriver National Institute of Child Health and Human Development: K23 HD101701, K23 HD097254 NIH National Institute of Neurologic Disorders and Stroke: K23 NS111086
We would like to thank the Pediatric Academic Societies Meeting for hosting our modeling competition presentation and the audience for their participation. National Institutes of Health (NIH) Eunice Kennedy Shriver National Institute of Child Health and Human Development K23 HD101701 (PI Moreira) and K23 HD097254 (PI Sullivan); NIH National Institute of Neurologic Disorders and Stroke K23 NS111086 (PI Vesoulis).
- Language
- English
- Electronic publication date
- 12/16/2024
- Date published
- 08/2025
- Academic Unit
- Stead Family Department of Pediatrics; Engineering Administration; Neonatology
- Record Identifier
- 9984758267602771
Metrics
5 Record Views