I-PATHWAY: a clinical prediction model to prevent childhood obesity

Abstract

BackgroundIn Australia, almost one-in-four (24.9%) children (aged 5-17 years) live with overweight or obesity (OW/OB). Prevention requires prioritisation, yet clinical practice in Australia focuses on identification, treatment and management. There are limited clinical, health system-integrated strategies to facilitate childhood OW/OB prevention. The first 1,000 days (conception-2 years) is an optimal time period for childhood OW/OB prevention. Prediction modelling is an emerging field of preventive medicine with ability to identify patients at-risk of disease, enhance clinical decision-making and reduce disease risk. A prediction model that can identify risk for childhood OW/OB from the first 1,000 days may promote preventive clinical decision-making by health professionals as part of routine care. This thesis aimed to develop and validate a prediction model (i-PATHWAY, infant-Practitioner Assessment Tool for Healthy Weight in pre-Adolescent Years) for childhood OW/OB that is translatable to clinical practice in Australia.MethodologyThis thesis employed mixed-methodology across five progressive yet unique studies:(1) Study 1 (systematic literature review) investigated the global availability of prediction models for OW/OB across the lifespan.(2) Study 2 (literature review) narratively assessed the predictive strength and clinical utility of predictors for childhood OW/OB to inform the selection of candidate predictors for i-PATHWAY.(3) Study 3 (predictive analytics) used a Western Australian prospective birth cohort – the Raine Study (n 1,947) – and considered 14 candidate predictors to develop and internally validate (via bootstrapping) i-PATHWAY to predict childhood OW/OB at age 8-9 years. Missing data were handled using the multiple imputation by chained equations (mice) method. Predictive accuracy and clinical relevance were assessed via calibration, discrimination and decision-threshold analysis.(4) Study 4 (qualitative semi-structured interviews) investigated the attitudes of i-PATHWAY end-users – health professionals (n 18) and caregivers of infants (aged 0-2 years) (n 13) towards predicting childhood OW/OB in clinical practice and explored ways to optimise the language and phrasing of i-PATHWAY to promote sensitivity and acceptability.(5) Study 5 (digital survey) determined enablers, barriers and individual and organisational readiness to implement i-PATHWAY into routine clinical practice in Queensland with a sample of health professionals (n 78) and health system and organisational decision-makers (n 12).The methodology of this thesis was underpinned by the seven steps for clinical prediction model development and validation (study 2 and 3) and implementation science theory, including: Diffusion of Innovations (study 4), integrated-Promoting Action on Research in Health Services (i-PARIHS) (study 5) and Organisational Readiness for Evidence-based Interventions (OREBI) (study 5).ResultsThis thesis presents five major findings from each of the five studies:(1) Twelve prediction tools for childhood OW/OB have been developed in eight international countries; none were developed in Australia. Predictive accuracy of the tools was mostly adequate; however, there was no evidence of attempted or actual translation into clinical practice.(2) Thirteen candidate predictors encompassing 5 domains (infant anthropometric, parental anthropometric, behavioural, socioeconomic and demographic) were appropriate for consideration in i-PATHWAY.(3) After internal validation, i-PATHWAY can predict childhood OW/OB (age 8-9 years) with strong calibration (slope = 0.956 [0.952 - 0.960], intercept = -0.052 [-0.063- -0.048]) and acceptable discrimination (AUC = 0.737 [0.736 – 0.738]); optimised sensitivity (0.703 [0.568 – 0.790]); and optimised specificity (0.646 [0.571 – 0.986]) using seven simple predictors collected in infancy (age 1 year). i-PATHWAY predictors included: weight change (0-1 year), maternal pre-pregnancy BMI, paternal BMI, maternal smoking during pregnancy, premature birth, infant sleep patterns and sex.(4) Health professionals and caregivers expressed optimism and acceptance towards predicting childhood OW/OB in clinical practice and identified that clinical training and supportive resources are necessary to ensure i-PATHWAY is sensitive and effective.(5) Health professionals expressed readiness to use i-PATHWAY in practice; it was considered an advantage to standard care and compatible with personal and clinical values. Enablers to implementation included integration with clinical guidelines and the digital health system. In Queensland, one health system and one statutory health promotion agency conveyed organisational readiness for implementing i-PATHWAY.Implications and conclusionsThis thesis presents a significant advance in the clinical prevention of childhood OW/OB in Australia. i-PATHWAY is the first prediction model for childhood OW/OB to be developed, validated and co-designed for use in an Australian population. i-PATHWAY is simple, accurate and acceptable to its end-users. Optimising translatability to clinical practice will require external validation and decision- and impact-analysis. Integrated into digital health systems alongside supportive training and resources, i-PATHWAY can substantially improve clinical decision-making and patient care by encouraging informed, evidence-based and targeted preventive intervention for infants at-risk of childhood OW/OB and their families. Stakeholders influenced by this thesis include: health professionals, caregivers, policymakers, health organisation leaders and hospital and health service decision-makers. i-PATHWAY can be used to influence state and national health policy to encourage universal childhood OW/OB risk identification in the first 1,000 days. In the long-term, health system-integration will encourage sustainable use of i-PATHWAY and may contribute to a reduction in childhood OW/OB prevalence in Australia.