Development and validation of an infant morbidity index using latent variable models

Area of Science:

• Pediatrics
• Biostatistics
• Public Health

Background:

• Four major infant morbidity outcomes include birth defects, abnormal newborn conditions, developmental delays, and low birth weight.
• Existing research often focuses on individual outcomes or their interrelationships, neglecting a comprehensive summary measure.
• A composite index for infant morbidity is needed to provide a holistic view of infant health status.

Purpose of the Study:

• To develop and validate a composite index for infant morbidity.
• To create a single variable summarizing multiple infant morbidity outcomes.
• To establish a measure representing an infant's propensity for morbidity.

Main Methods:

• Development of extended latent variable (LV) models for multiple multinomial morbidity outcomes.
• Utilized modified Gauss-Newton algorithms and estimated generalized nonlinear least-square methods.
• Modeled conditional probabilities of outcomes as nonlinear functions of a log-normal distributed LV.

Main Results:

• A novel Infant Morbidity Index (IMI) was developed, summarizing four key infant morbidity outcomes.
• The IMI demonstrated significant correlations with individual morbidity outcomes and infant mortality.
• The index showed strong correlation with a face-valid measure of overall infant morbidity.

Conclusions:

• The developed Infant Morbidity Index (IMI) serves as a valid summary measure of infant morbidity.
• The IMI can be effectively used in future research to assess an infant's propensity for morbidity.
• This index offers a valuable tool for a more comprehensive understanding of infant health outcomes.