Prioritizing Disease Diagnosis in Neonatal Cohorts through Multivariate Survival Analysis: A Nonparametric Bayesian...

Area of Science:

• Biostatistics
• Epidemiology
• Health Informatics

Background:

• Accurate analysis of disease relationships is crucial for effective healthcare.
• Existing methods struggle with censored multivariate time-to-event data, limiting analytical precision.
• Understanding disease precedence aids in prevention and recovery strategies.

Purpose of the Study:

• Introduce Censored Event Precedence Analysis (CEPA), a novel nonparametric Bayesian approach.
• Develop a robust methodology for exploring precedence relationships in censored multivariate events.
• Enhance the prediction of subsequent disease occurrences.

Main Methods:

• Developed CEPA, a nonparametric Bayesian statistical approach.
• Applied CEPA to neonatal health insurance data, analyzing International Classification of Diseases (ICD) codes.
• Conducted simulation studies to compare CEPA with traditional models on censored multivariate datasets.

Main Results:

• Identified a typical neonatal disease diagnostic sequence: respiratory, skin, infectious, digestive, ear, eye, and injury-related diseases.
• CEPA demonstrated superior performance in simulation studies, with accuracy reaching 76% (uniform distribution) and 65% (exponential distribution).
• The method proved effective across four tested environments for censored multivariate data.

Conclusions:

• CEPA significantly enhances the understanding of disease interrelationships compared to existing methodologies.
• Identifying disease precedence through CEPA enables proactive intervention against subsequent conditions.
• The findings support the development of a disease-sequence-informed healthcare system for improved patient outcomes.