A prediction model for childhood obesity in New Zealand

Area of Science:

• Pediatrics
• Public Health
• Epidemiology

Background:

• Early childhood obesity is a growing concern globally and in New Zealand.
• Existing obesity prediction models lack validation for New Zealand's diverse population.
• Accurate prediction is crucial for timely and effective early intervention strategies.

Purpose of the Study:

• To develop and validate a predictive model for obesity in 4-5-year-old children in New Zealand.
• To utilize parental and infant data from the Growing Up in New Zealand (GUiNZ) cohort.
• To assess the model's performance using internal and external validation cohorts.

Main Methods:

• Data from the GUiNZ cohort (n=1731 derivation, n=713 internal validation) were used.
• External validation was conducted using the Prevention of Overweight in Infancy (POI) and Pacific Islands Families (PIF) study cohorts.
• The model incorporated birth weight, maternal smoking, maternal/paternal BMI, and infant weight gain.

Main Results:

• The final prediction model demonstrated adequate discrimination accuracy (AUROC 0.74-0.80) across derivation and validation cohorts.
• Positive predictive values were generally low, indicating a high rate of false positives.
• The PIF cohort showed more consistent, though still variable, positive predictive values (52-61%).

Conclusions:

• The developed model can aid in early childhood obesity prediction for New Zealand children.
• High false positive rates necessitate careful consideration to avoid unwarranted parental anxiety.
• Further refinement may be needed to improve the model's positive predictive value for clinical application.