Oral Presentation ANZOS Annual Scientific Meeting 2021

i-PATHWAY: Development and validation of a clinical prediction model for childhood obesity in Australia (#95)

Oliver J Canfell 1 2 3 , Robyn Littlewood 4 5 , Olivia Wright 6 , Jacqueline L Walker 6
  1. Digital Health Cooperative Research Centre, Australian Government, Sydney, NSW, Australia
  2. UQ Business School, The University of Queensland, St Lucia, QLD, Australia
  3. Centre for Health Services Research, The University of Queensland, Herston, QLD, Australia
  4. Health and Wellbeing Queensland, Australian Government, Brisbane, QLD, Australia
  5. Faculty of Health and Behavioural Sciences, The University of Queensland, St Lucia, QLD, Australia
  6. School of Human Movement and Nutrition Sciences, The University of Queensland, St Lucia, QLD, Australia

Introduction: Childhood overweight/obesity prevention requires prioritisation, yet clinical practice in Australia focuses on identification and treatment. A model that can accurately predict childhood overweight/obesity from the first 1,000 days may be a clinically useful preventive tool. This study aimed to develop and validate i‐PATHWAY–a model to predict childhood (age 8–9 years) overweight/obesity from infancy (age 12 months) using an Australian prospective birth cohort. 

Methods: The Transparent Reporting of a multivariable Prediction model for individual Prognosis or Diagnosis (TRIPOD) checklist was followed. Participants were n=1947 children (aged 8–9 years) from the Raine Study Gen2 – an Australian prospective birth cohort – who had complete anthropometric measurement data available at follow up. The primary outcome was childhood overweight or obesity (age 8–9 years), defined by age‐ and gender‐specific cut‐offs. Multiple imputation was performed to handle missing data. Predictors were selected using 2000 unique backward stepwise logistic regression models. Predictive performance was assessed via: calibration, discrimination and decision‐threshold analysis. Internal validation of i‐PATHWAY was conducted using bootstrapping (1000 repetitions) to adjust for optimism and improve reliability. A clinical model was developed to support relevance to practice.  

Results: At age 8–9 years, 18.9% (n=367) of children were classified with overweight or obesity. i‐PATHWAY predictors included: weight change (0–1 year); maternal pre‐pregnancy body mass index (BMI); paternal BMI; maternal smoking during pregnancy; premature birth; infant sleep patterns; and sex. After validation, predictive accuracy was acceptable: calibration slope=0.956 (0.952–0.960), intercept=−0.052 (−0.063, −0.048), area under the curve=0.737 (0.736–0.738), optimised sensitivity=0.703(0.568–0.790), optimised specificity=0.646 (0.571–0.986). The clinical model retained acceptable predictive accuracy without paternal BMI. 

Conclusion: i‐PATHWAY is a simple, valid and clinically relevant prediction model for childhood overweight/obesity - the first in Australia. After further validation, this model can influence state and national health policy for overweight/obesity screening in the early years.