Bivariate poisson-poisson model of zero-inflated absenteeism data

Area of Science:

• Biostatistics
• Epidemiology
• Pediatric Health Research

Background:

• Bimodal count distributions, particularly with a zero-inflation mode, are common in medical and health research.
• Childhood illness data, such as days missed from activities and days spent in bed, often exhibit zero inflation.
• The existence and impact of a critical window for growth and nutrition in early life on subsequent health is debated.

Purpose of the Study:

• To analyze bivariate count data from a health survey on childhood illness (days missed, days in bed).
• To investigate the relationship between early life growth (infancy weight-for-age, weight gain) and childhood health status.
• To evaluate the hypothesis of a critical window in infancy versus the impact of childhood weight gain on health.

Main Methods:

• Development and application of a bivariate Poisson-Poisson regression model to handle zero-inflated count data.
• Extension of the model to incorporate a constraint where days missed activities (Y(1)) are greater than or equal to days in bed (Y(2)).
• Analysis of health survey data in relation to infant weight-for-age and weight gain from infancy to age 7.

Main Results:

• The bivariate Poisson-Poisson model effectively addressed the zero-inflated nature of the illness count data.
• Analysis of the data did not provide support for a critical window of growth and nutrition in infancy affecting later health.
• A potential association was observed between weight gain during childhood and subsequent health status.

Conclusions:

• The findings challenge the notion of a critical window in infancy for long-term health outcomes.
• Childhood weight gain appears to be a more relevant factor influencing children's health status than previously emphasized critical infancy periods.
• The developed bivariate Poisson-Poisson models offer a robust statistical framework for analyzing complex, zero-inflated health-related count data.