A mixed-effects two-part model for twin-data and an application on identifying important factors associated with...

Area of Science:

• Biostatistics
• Pediatric Health Research
• Statistical Modeling

Background:

• Analyzing the Positive Child Health Index (PCHI) in preterm infants presents challenges due to distinct data processes for healthy versus affected children.
• Correlations within multiple births (twins, triplets) require specialized statistical approaches.
• Existing methods dichotomize health disorder data, potentially reducing statistical power.

Purpose of the Study:

• To develop and evaluate a novel mixed-effects two-part model for analyzing PCHI data.
• To jointly model the two distinct data processes (healthy vs. disordered children) within a unified framework.
• To account for correlations in health outcomes among multiple births.

Main Methods:

• Development of a mixed-effects two-part model to analyze correlated semi-continuous PCHI data.
• Joint modeling of two data processes: one for children with no disorders and another for those with disorders.
• Simulation studies to compare the proposed method with existing logistic regression approaches.

Main Results:

• The proposed joint-test procedure demonstrates increased power in detecting factors associated with health disorders.
• The new method consistently outperforms existing approaches in simulations when controlling for Type I error.
• The model is robust to misspecifications and applicable to correlated semi-continuous data.

Conclusions:

• The proposed mixed-effects two-part model provides a more powerful and robust framework for analyzing PCHI data.
• This approach enhances the ability to identify factors influencing child health outcomes in preterm populations.
• The method is suitable for complex health index data, particularly in the presence of multiple births and varying health statuses.