Cost‐effectiveness of scaling up a whole‐of‐community intervention: The Romp & Chomp early childhood obesity prevention intervention

Summary Background Given the high prevalence of early childhood overweight and obesity, more evidence is required to better understand the cost‐effectiveness of community‐wide interventions targeting obesity prevention in children aged 0–5 years. Objectives To assess the cost‐effectiveness of the Romp & Chomp community‐wide early childhood obesity prevention intervention if delivered across Australia in 2018 from a funder perspective, against a no‐intervention comparator. Methods Intervention costs were estimated in 2018 Australian dollars. The annual Early Prevention of Obesity in Childhood micro‐simulation model estimated body mass index (BMI) trajectories to age 15 years, based on end of trial data at age 3.5 years. Results from modelled cost‐effectiveness analyses were presented as incremental cost‐effectiveness ratios (ICERs): cost per BMI unit avoided, and cost per quality‐adjusted life year (QALY) gained at age 15 years. Results All Australian children aged 0–5 years (n = 1 906 075) would receive the intervention. Total estimated intervention cost and annual cost per participant were AUD178 million and AUD93, respectively, if implemented nationally. The ICERs were AUD1 126 per BMI unit avoided and AUD26 399 per QALY gained (64% probability of being cost‐effective measured against a AUD50 000 per QALY threshold). Conclusions Romp & Chomp has a fair probability of being cost‐effective if delivered at scale.


| INTRODUCTION
Early childhood obesity is a significant public health issue, with approximately 40 million (5.9%) children aged under five classified as having overweight or obesity worldwide. 1 Childhood obesity negatively affects psychosocial, respiratory, orthopaedic, endocrine and reproductive health. 2 The condition is also associated with economic consequences, including increased healthcare costs and indirect costs such as school performance or lost productivity. 3 Evidence also suggests that children and adolescents with overweight and obesity are at increased risk of overweight in adulthood. 2,3 The prevention of overweight and obesity in the early years of life (i.e., in the first 5 years of life) is of increasing importance internationally, with recognition that the behavioural and biological responses of a child to obesogenic environments can be shaped from a young age. 1 The World Health Organization (WHO) Commission on Ending Childhood Obesity 1 outlined a comprehensive package of recommendations to address childhood obesity, including multi-sectoral, multi-faceted approaches to prevention. As part of this comprehensive approach, community-wide interventions, involving community capacity building and engagement with population-level obesity prevention strategies, have been recognized as promising in achieving modest reductions in population weight gain among children. 4,5 Evidence also suggests that community-wide interventions may represent an equitable approach to obesity prevention intervention, which may be of particular importance considering the socioeconomic patterning of obesity starts from a very young age. 6 Given scarce societal resources, it is important that interventions to reduce childhood obesity represent good value for money. To have an impact at the population level, effective and cost-effective obesity prevention interventions need to be scaled-up and widely available. 7 Limited evidence currently exists on the cost-effectiveness of interventions aiming to prevent overweight and obesity in the preschool-age population. 8 Published literature primarily focuses on school-based strategies and the impacts of community-wide interventions on schoolage children. 4,5,9 In addition, much of the published economic evaluation literature has modelled the costs and effects of community-wide interventions in primary-school aged children over their lifetime, to estimate the longer-term health benefits and healthcare cost-savings from the prevention of chronic diseases into adulthood (i.e., 5,10 ). No published studies have estimated the cost-effectiveness of community-wide interventions in early childhood populations across shorter time horizons, accounting for the health benefits and healthcare cost-savings that might accrue throughout childhood and adolescence. 11 Yet such evidence is important because it could assist decision making in a shorter-term, policy-relevant timeframe. 2 This paper aims to estimate the cost-effectiveness of a community-wide early childhood obesity prevention intervention-the Romp & Chomp (R&C) intervention-assuming it was hypothetically scaled-up and nationally delivered to all Australian children from 0 to 5 years of age. R&C was a quasi-experimental trial of a multi-setting, multi-strategy, community-wide obesity prevention intervention conducted in Geelong, Victoria, Australia from 2004 to 2008 that targeted children aged from 0 to 5 years through community capacity building (i.e., professional training, support provision to early childhood environments to favour obesity prevention) and environmental changes in early childhood education and care (ECEC) settings. 12 These ECEC settings comprised long day care centres, family day care services, and preschools. 12 The intervention was designed in collaboration with key stakeholders, including the regional health service, local government and state government departments of health and education. Intervention strategies involved health promotion and activities designed to develop sustainable policy, and sociocultural and environmental changes in early childhood settings. R&C demonstrated a significant difference in body mass index (BMI) (À0.06 kg/m 2 [95% CI: À0.10; À0.01, p <0.01]) among 3.5-year-olds between the intervention and comparator groups, 12 however no economic evaluation was undertaken at the time of the trial.
We undertook modelled cost-effectiveness analysis (CEA) and cost-utility analysis (CUA) over a 10-year time horizon, from 5 to 15 years of age. The research question was: From a funder perspective, would the R&C intervention delivered nationally to the population of all Australian children aged from 0 to 5 years be cost-effective by age 15 years (in terms of cost per BMI unit avoided at age 15 years and cost per quality-adjusted life year (QALY) gained by age 15 years), compared to a "no intervention" comparator?

| METHODS
Modelled CEA and CUA were undertaken, using published estimates of effectiveness of the R&C intervention 12 applied to a nationally representative cohort and using intervention cost estimates from a detailed retrospective costing analysis. CEA compares the cost of an intervention relative to control with the intervention outcome measured in natural units (i.e., cost per BMI unit avoided). CUA compares the cost of an intervention relative to control with the intervention outcome measured in a metric incorporating the impact on both quality and quantity of life (i.e., cost per QALY gained).
To conduct our modelled economic evaluation, we extrapolated the within-trial intervention costs and effects to a nationally representative cohort of the Australian children aged 0 to 5 years. A previously published health economic model, the Early Prevention of Obesity in CHildhood (EPOCH) model, 11 was used to estimate the health benefits, healthcare cost-savings and incremental cost-effectiveness of the R&C intervention beyond the duration of the R&C intervention efficacy trial.
The economic evaluation followed the recommendations of the Second Panel on Cost-Effectiveness in Health and Medicine. 13 Analyses were reported following the Consolidated Health Economics Evaluation Reporting Standards 14 (Appendix S1, Supporting Information).

| The intervention and comparator
R&C was a community-wide obesity prevention intervention that targeted children aged 0 to 5 years across the City of Greater Geelong and Borough of Queenscliffe in Victoria, Australia (n $ 12 000), their families and the organizational caregivers at ECEC settings. 12 The intervention emphasized community capacity building and sustainable changes in policy, sociocultural and physical environments in ECEC using a socioecological framework to encourage healthy eating, active play, reduced screen time and attainment of healthy weight. 12 Through professional training, policy development and messaging materials, the intervention focused on four key messages: (1) daily active play, (2) daily water and fewer sweet drinks, (3) daily fruit and vegetables and (4) less screen time. 12 During the intervention phase of R&C, there were two other health promotion programs available in ECEC in the Geelong and  16 Given the obvious synergies, these programs were delivered with the R&C intervention as an integrated intervention package in the intervention region. 12 Further details of the R&C, S4M and KGFYL intervention have been reported elsewhere. 12,15,16 We defined the comparator for both the CEA and the CUA as a nationally representative cohort of Australian children that did not receive the R&C intervention. Given the lack of available data on implementation of S4M and KGFYL we assumed that the comparison group received no intervention (i.e., this group did not receive R&C, S4M or KGFYL interventions). This may result in an overestimation of intervention cost as compared to the cost of the control, but was deemed to be the most conservative approach to estimating costeffectiveness.

| Measurement of effectiveness
Intervention effect was estimated using a repeat cross-sectional quasi-experimental design to measure the differences in outcomes between the population exposed to the R&C intervention (the intervention sample) compared to the comparison population drawn from other LGAs across Victoria. 12 Height and weight data were collected by trained Maternal Child Health nurses as part of routine Key Ages and Stages health checks. 12 The survey collected data pre-and post-intervention in the intervention and non-intervention LGAs; therefore, the intervention effect sizes were the average of all the children in the intervention communities compared to those of the non-intervention communities. 12 Data comprised children who had attended their 2-and 3.5-year-old health checks in 2004 and 2007. 12 In the 3.5-year-old intervention sample, the R&C intervention demonstrated a statistically significant reduction in BMI of À0.06 kg/m 2 (95% CI: À0.10; À0.01, p <0.01) relative to the control group. 12 Our base case analysis assumed effect maintenance until age 15 years, meaning that the children remained in and moved on a BMI trajectory informed by the EPOCH model 17 based on national data 18 and according to their BMI at age 3.5 years.

| Target population and setting
To conduct our economic evaluation, we assumed that the R&C intervention was scaled up and delivered nationally to all Australian children aged from 0 to 5 years (n = 1 906 075) 19 given the evaluation sample was a cross-section of all children in the intervention age group using Maternal and Child Health Key Age and Stage health checks. 12 The intervention settings for the national modelling included all ECEC settings (n = 12 463 20,21 ): family day care (FDC) (n = 906 20,21 ), centre-based long day care (LDC) and preschools (n = 11 557 20 ). FDC is a type of formal care provided by a registered early childhood educator and carer in a home setting for a small group of children. 22 Centre-based LDC is delivered by trained educators and carers, and may involve an integrated preschool program. 22 Preschool programs (called "Kindergartens" in Victoria) are noncompulsory, government-funded learning programs, normally delivered by early childhood educators to children within 1 or 2 years before starting formal education (typically, when children are aged between 3 and 5 years). 20

| Resource use and costs
Intervention costs from a funder perspective were estimated retrospectively, using trial records and micro-costing techniques. All assumptions on how the intervention would be implemented at scale were based on the existing literature on community-wide obesity prevention intervention, 23 the management structure reported in trial records (unpublished documents) and in consultation with members of the R&C research team.
Costs were categorized as: time costs, travel costs and intervention material costs. Time costs were estimated using published wage rates including salary on-costs (i.e., overhead costs, superannuation, employer taxes, compensation, and leave loading). 24 We assumed that an average of one state-level full-time equivalent (FTE) Project Administrator would be required in each Australian state and territory (n = 8), to develop policy, manage the intervention implementation within each jurisdiction and provide online training to an average of one 0.5FTE Health Promotion Officer located within each Australian LGA (n = 562 25 ). While some smaller LGAs may share a Project Administrator and Health Promotion Officer fraction, some larger LGAs may need more capacity. The Health Promotion Officer would, consistent with the R&C implementation, assist the localized interven-

Costs would be incurred by each LGA-level Health Promotion
Officer (n = 562 25 ) for travel to local events to promote intervention messages, and to ECEC settings to provide intervention training to early childhood educators and carers. Travel costs were estimated using published guidelines. 27 The travel distance for each one-way trip was assumed to be 16 km, which was based on the average commute distance of Australian residents. 28 This assumption was tested in sensitivity analyses (see Section 2.6).
Intervention material costs consisted of marketing and promotional materials provided to ECEC settings and participants. Each ECEC setting received a program print, early childhood services toolkit documents, display posters, fact sheets, and stickers to encourage healthy eating and active play. 12 An annual sweet drink demonstration was presented to parents by early childhood educators. 12 While the base case analysis assumed that participants received most intervention materials electronically (i.e., newsletters, information postcards, tip sheets), intervention participants received a water bottle and lunch bag. 12,15,16 We assumed the intervention was promoted at four local events per LGA per year (n = 2 248 presences nationally), based on trial records (unpublished documents).
This incurred a stall fee and promotional materials (i.e., a trestle table, show bags, lunch box brochures). Unit costs for intervention materials and the cost of acquiring and annually maintaining the intervention website domain and hosting were estimated using market prices (Table S3).
The intervention was costed assuming it was in steady state, running at its full effectiveness potential (i.e., excluding costs associated with research and intervention planning and development). In the base case analysis, intervention costs were assumed to be borne by all children aged 0 to 5 years. All costs were estimated in 2018 Australian dollars (AUD1 = USD0.70 29 ), and if required, unit costs were adjusted to 2018 values using the Consumer Price Index. 30 All future costs and benefits were discounted at 5% annually. 31 Intervention costs by major cost category and a detailed summary of intervention costs are presented in Tables 1 and S3 Data from the LSAC used as the representative national level input population for our modelled economic evaluation. 18 The LSAC is a national, comprehensive, and multi-disciplinary Australian dataset of children from two cohorts, the "baby" ("B") cohort and the "kindergarten" ("K") cohort, followed from aged 0-1 to 4-5 years, respectively. 18 The B cohort was selected as our input population for our analyses given the data collection of the B cohort corresponds to the roll out of R&C. Intervention effect size (À0.06 kg/m 2 ) was applied to child BMI at age 4/5 years to estimate BMI trajectories of the intervention group to age 14/15 years and compared to the trajectories of the same cohort without any intervention effects applied. QALY weights associated with child weight status to inform the estimation of QALYs were obtained from a recent systematic review and meta-analysis. 34 The QALY weights of children with healthy weight, overweight and obesity were 0.85, 0.83 and 0.82, respectively. 34 The classification of weight status was based on WHO growth standards. 35 Healthcare costs of participants to age 15 years were modelled following a 'top down' method, using administrative records of annual hospital, 36

| Sensitivity analysis
A series of univariate and multivariate sensitivity analyses were performed to evaluate the impact of assumptions made (Table 2). In sensitivity analysis 1, intervention costs were varied, assuming that they were borne only by the population in which the intervention effect was modelled (i.e., children aged 4-5 years, n = 642 178 19 ). In sensitivity analysis 2, higher intervention costs were assumed, based on: (i) intervention resources for participants being paper-based; (ii) the Health Promotion Officer at 562 LGAs being employed at 1 FTE; (iii) the allocation of 2 h each year for early childhood carers and educators to attend training and perform the sweet drink demonstration; (iv) the allocation of 1.5 h each year for dentists to engage with parents and staff training and (v) a longer distance of 31.2 km was travelled by Health Promotion Officers within LGAs to deliver training and to attend festivals. 41 In sensitivity analysis 3, a "worst case" scenario was also examined, using the low confidence interval (CI) of the intervention effect on BMI (i.e., À0.01 kg/m 2 ) 12 and the higher intervention cost. In sensitivity analysis 4, the discount rate was reduced from 5% to 3%. 13

| RESULTS
Under the base case assumptions, the intervention costs totalled AUD177 536 705 per year, with the annual cost per participant averaging AUD93 (Table 3). By age 15 years, the modelled healthcare cost-saving was AUD15 per participant as compared to the nointervention comparator ( Table 3).
The intervention led to a 0.07 reduction in BMI units at age 15 years compared to the no intervention comparator, resulting in an ICER of AUD1 126 per BMI unit avoided at age 15 years (Table 3 and Figure 1).
At age 15 years, 0.003 QALYs were gained per child, as compared to the no intervention comparator ( Table 3). The national delivery of the R&C intervention, with mean ICER of AUD26 399/QALY gained, has a 64% probability of being cost-effective compared to the commonly accepted AUD50 000/QALY gained threshold ( Figure 2).
The intervention is not cost-effective when intervention costs are borne only by children aged 4 to 5 years (sensitivity analysis 1), as a result of the higher intervention cost per participant (AUD276; approximately three times higher than the base case cost). The intervention is also not cost-effective under worst case assumptions (sensitivity analysis 3), as a result of the intervention effect being six times lower and the mean intervention cost per participant more than five times higher than in the base case analysis (Table 3).  given the intervention was more resource-intensive.
Our cost-effectiveness results suggest that R&C should be con-

| CONCLUSIONS
Excess BMI affects Australian children from a very young age, which negatively impacts their quality of life in the short term and causes increased risks of developing chronic diseases in the long term. 3 R&C was a community-wide obesity prevention intervention that effec- Tran wrote the first draft of the paper. All authors reviewed and commented on the paper.