Work

Exploring Novel Data-Driven Clustering Methods for Uncovering Patterns in Longitudinal Neonatal Postoperative Temperature Measurements

Year: 2025

Type: article

Abstract: Objective To identify distinct postoperative temperature trajectories in neonates with congenital heart defects (CHD) after cardiopulmonary bypass (CPB), using advanced unsupervised machine learning (ML) clustering methods, corroborate findings, and evaluate their prognostic value on outcomes. Patients and Methods Secondary cohort analysis of prospective data collected from a single pediatric referral center's CardioAccess data registry, consistent of neonates who underwent CPB between January 1, 2015, and January 1, 2019. Postoperative temperatures were extracted from medical records (48 hours). Group-based trajectory modeling (GBTM) performance was compared to self-organizing maps (SOM) and k-means clustering. Cluster membership and model fit were optimized for three temperature clusters per method. The primary outcome was a composite of postoperative complications. Clustering techniques were compared and associated with outcomes using adjusted multivariable binary logistic regression. Results Neonates ≥ 34 weeks gestation undergoing CPB (N= 450). GBTM, SOM, and k-means identified membership for three groups: 1) persistent hypothermia (n= 38, 9% vs. n=49, 11% vs. n=40, 9%), 2) resolving hypothermia (n= 233, 51% vs. n=227, 50% vs. n=147, 33%), and 3) normothermia (n= 179, 40% vs. n=174, 39% vs. n=263, 58%). Concordance between techniques showed strong agreement between GBTM and SOM (kappa= 0.92), and weak agreement between GBTM and k-means (kappa=.41). After adjustment, persistently hypothermic compared to normothermic neonates were associated with higher odds of the complication composite outcome in the GBTM (OR: 2.8 [1.0-7.3], p .04 ) and SOM (OR: 2.3 [1.0-5.4], p.04 ) models, but not in the k-means model (OR: 1.4 [0.7-3.1], p.38 ). Conclusion Exploring concordance between different ML techniques demonstrates that temperature in neonates after CPB follows unique trajectories. Those exhibiting persistent hypothermia trends are at higher risk of adverse outcomes.

Source: Mayo Clinic Proceedings Digital Health

DOI: 10.1016/j.mcpdig.2025.100270