This economic evaluation was conducted and reported in accordance with the Consolidated Health Economic Evaluation Reporting Standards (CHEERS 2022) guidelines [21] to ensure transparency, comparability, and completeness in reporting. A completed CHEERS checklist is provided in the Additional file 1.
Intervention
The DECIDE-Children intervention was a cluster RCT conducted among children aged 8 to 10 years in three geographically and culturally distinct regions in China—Beijing, Changzhi, and Urumqi. In each region, eight primary schools were selected, resulting in a total of 24 schools participating in the study. Within each school, one or two grade 4 classes were recruited to ensure that each school contributed approximately 50 students to the study population. Schools were randomized in a 1:1 ratio to either the intervention or the control group, stratified by district within each region. The intervention design was based on the social-ecological model [22], targeting multiple levels of influence, including the individual (student), family, and school environments. Activities for schools included implementing policies related to obesity and providing health education training for teachers. For families, the intervention included health education activities and encouraged caregivers (mainly parents) to support and supervise their child’s behavioral changes. Student-focused activities consisted of health education courses, enhancing physical activity, and regular monitoring of weight and height. Importantly, these student-focused components were embedded in the existing school curriculum, and no additional class periods were created. This design ensured that the program could be carried out without placing any extra academic or time burden on either teachers or students. A mobile application titled “Eat Wisely, Move Happily” was utilized to strengthen the connections among these intervention components, facilitating information diffusion, behavior monitoring, weight management, assessment, and feedback. To standardize intervention delivery, three manuals were developed for project staff, school teachers, and caregivers. Schools assigned to the control arm continued with their standard curriculum and routine physical education activities without any additional resources or obesity-specific components, thus incurring no intervention-related costs. The intervention was implemented for one school year, from late September 2018 to June 2019. Detailed information about the intervention has been reported previously [23, 24].
Costs of intervention
To estimate the economic impact of the intervention, we conducted a cost analysis from a societal perspective. The total intervention costs included both direct financial expenditures (e.g., materials, equipment) and labor components. To clarify cost attribution by stakeholder, we classified monetary costs (materials, equipment, and app-related operational expenses) as provider-borne costs, since they were directly funded by the research program. Labor costs reflected the opportunity cost of school staff time and were categorized under the school perspective. In the subsequent economic evaluation (cost-utility and cost–benefit analyses), we additionally considered medical costs and productivity losses associated with obesity, which correspond to the healthcare system and broader societal perspectives.
Monetary costs referred to direct expenditures incurred during the implementation period. These included spending on printed educational materials, sports supplies, and incentive gifts for children, all valued based on actual purchase prices. Importantly, the monetary costs also covered ongoing expenses associated with the mobile application, such as system maintenance, routine updates, user data processing, and data storage—costs that would be necessary for real-world replication. However, the initial app development costs were not included, as they were considered one-time sunk costs not expected to recur in future scale-up efforts.
Labor costs were estimated by translating time inputs from school personnel into monetary values. Specifically, we calculated labor costs by multiplying the average salary of school staff by the actual number of hours they dedicated to intervention-related activities, such as training sessions, health education, and coordination tasks. Time data were extracted from detailed process records maintained during the intervention. Four categories of staff were included: class teachers, principals, healthcare/health education teachers, and PE teachers. To reflect regional variation in wages, labor costs were calculated separately for the three intervention sites (Beijing, Changzhi, and Urumqi). We did not assign a monetary value to the time of caregivers (mainly parents). Although their participation was a crucial part of the intervention, it was facilitated entirely through a mobile application designed for flexible engagement. This design allowed parents to participate at their convenience outside of working hours, minimizing opportunity costs related to lost productivity. Similarly, we did not value children’s time, as the intervention was intentionally integrated into the existing school schedule. Health education components were delivered during classes already allocated to health or moral education, and physical activity promotion was embedded within regular breaks or physical education sessions. Consequently, no core academic instruction was displaced, and the opportunity cost of student time was effectively negligible.
To isolate the cost of implementing the intervention in real-world settings, research- and evaluation-related costs were excluded. The average cost per student was calculated by dividing the total intervention cost by the number of participants in the intervention group at baseline. No additional costs were assumed for the control group. All costs were converted to US dollars using the average exchange rate for 2018 (1 USD = 6.6322 CNY).
Obesity-related outcomes
Baseline and follow-up measurements on anthropometric measurements were collected in September 2018 and June 2019 by the trained outcome assessors using standardized procedures and instruments. The outcomes for the economic evaluation were the primary anthropometric indicators of the trial [23], including body mass index (BMI, calculated from height and weight), BMI Z-score (calculated according to the child growth standards released by the World Health Organization (WHO) [25]), body fat percentage (BF%), and waist circumference (WC). This approach, detailed a priori, ensured that the economic evaluation was not subject to post hoc outcome selection.
Cost-effectiveness analyses
We conducted the cost-effectiveness analyses based on the intention-to-treat principle, involving 705 participants in the intervention group and 687 participants in the control group. The incremental cost-effectiveness ratio (ICER) was calculated for primary anthropometric outcome indicator, representing the cost associated with achieving a one-unit change in the respective indicator. The ICER was calculated using the following formula [26]:
$$ICER= \frac{{Cost}_{IG} -{Cost}_{CG}}{{Effectiveness}_{IG} -{Effectiveness}_{CG}}$$
where:
-
\({Cost}_{IG}=\) Cost per student in the intervention group
-
\({Cost}_{CG}=\) Cost per student in the control group
-
\({Effectiveness}_{IG}=\)
Outcome value (e.g., change in BMI Z-score) in the intervention group -
\({Effectiveness}_{CG}=\) Outcome value (e.g., change in BMI Z-score) in the control group
For the anthropometric indicators, the ICER represented the expenditure per student in reducing one unit of BMI, 0.1 unit of BMI Z-score, 1% body fat percentage, and 1-cm waist circumference.
Cost-utility analyses and cost–benefit analyses
To estimate the long-term health and economic impact of childhood obesity prevention, we used a two-stage obesity-progression model to project the number of adult obesity cases prevented by the intervention [26, 27]. This approach was necessary due to the limited availability of longitudinal data linking childhood weight status directly to adult obesity incidence. Overweight and obesity were defined using both the Chinese BMI reference standards and WHO growth references to ensure broader applicability [25, 28]. In the first stage, we estimated the probability of transition between weight status categories (normal weight, overweight, or obese) from childhood/adolescence (ages 10–14) to young adulthood (ages 21–29), based on baseline weight status. The average age of children in our study was 9.6 years (SD 0.4). We referred to existing literature using the 10–14-year age band, which closely approximates the age of our participants [29]. In the second stage, we projected the transition from young adulthood (25–29 years) to adulthood (age 40). We assumed that BMI status remains stable after age 40, based on evidence from long-term cohort studies suggesting that adult weight trajectories are relatively stable or tend to increase further [30]. We assumed that the model was based on the observed reduction in childhood overweight and obesity prevalence after the 1-year intervention period. This reduction was applied to the initial weight distribution at baseline (ages 10–14), and the subsequent transitions to adulthood were projected using established progression probabilities [27, 29]. No additional long-term or sustained effects of the intervention were assumed beyond the childhood period, consistent with current evidence, which shows limited maintenance of intervention effects over time [31]. In other words, the intervention was assumed to reduce the number of children entering the progression model as overweight or obese, without modifying the transition probabilities themselves. The long-term number of adult obesity cases prevented was estimated by comparing adult obesity outcomes between the intervention and control scenarios. Transition probabilities were obtained from previous studies, considering gender [27, 29]. All model parameters are detailed in the Additional file 2 (Formula of calculation 1: The obesity progression model [23, 27, 29]).
The number of quality-adjusted life years (QALYs) gained per adult obesity case prevented was derived from a life-table approach, incorporating age- and sex-specific estimates of mortality risk, activity impairment, and life expectancy for obese vs. non-obese individuals [32]. We assumed that the health benefits from preventing adult obesity would begin at age 40 and persist through age 65. Based on literature-derived parameters [27, 29], we calculated the QALYs gained per obesity case prevented as the discounted difference in health utility over the expected lifetime between obese and non-obese adults, adjusted for survival probabilities. These included both quality-of-life (activity scale score) and life expectancy adjustments for those who die between 40 and 65 and those who survive. The average QALYs gained per case (Q) were then multiplied by the estimated number of adult obesity cases prevented (N prevent). A 5% annual discount rate was applied, consistent with pharmacoeconomic evaluation guidelines in China [33]. Formulas for calculating QALY saved were included in the Additional file 2 (Formula of calculation 2: QALY saved [27, 32, 33]).
Medical costs averted (C medical loss) per obese adult were estimated according to previous studies [26, 27]. The parameter was derived from a Chinese study that estimated the economic costs of obesity [34]. The study used the data from the National Behavioral Risk Factors Surveillance Survey and the National Health Service Survey to estimate the total medical costs of five major non-communicable diseases related to overweight and obesity [34]. To calculate the medical cost per obese adult, we further used the data from the Population Census of the People’s Republic of China to obtain the total number of obese adults [35]. This value was projected over 25 years (ages 40–65) and discounted at 5% annually to reflect lifetime medical cost savings per case prevented.
As for the cost of labor productivity loss (C productivity loss), it included averted costs from lost productivity due to morbidity or mortality per obese adult prevented. The parameters in the formula were based on Brown’s study [27]. Morbidity-related productivity loss was computed as the difference in annual absenteeism days between obese and non-obese adults, multiplied by sex-specific daily wages. Mortality-related loss was modeled as the difference in expected lifetime earnings between obese and non-obese adults from age 40 to 65, assuming premature mortality due to obesity. All estimates were adjusted for survival probability and discounted at 5%. Daily wage and annual wages in three cities (Beijing, Changzhi, and Urumchi) were calculated according to data from the National Bureau of Statistics of China [36]. The annual wage can be directly obtained from the statistics, while the daily wage is the annual income divided by the number of working days (365 days minus 104 statutory days off [37]). Formulas for averted medical costs and productivity loss costs were included in the Additional file 2 (Formula of calculation 3: Labor productivity loss [27, 32, 33, 36]; Formula of calculation 4: Medical costs averted [32, 33, 38,39,40]).
The incremental cost-utility ratio (ICUR) was calculated as the difference in total costs between the intervention and control groups divided by the difference in QALYs between the groups. The ICUR was calculated using the following formula:
$$ICUR = \frac{Net~\text{intervention}~cost}{N_{\text{Prevent}} \times Q}$$
where:
To assess the robustness of the long-term cost-utility analysis (CUA) results, we conducted probabilistic sensitivity analysis. We made the cost-effectiveness acceptability curve (CEAC) to show the relationship between the probability of the intervention being cost-effective and varying willingness-to-pay (WTP) thresholds per QALY gained. Probabilistic distributions were selected for each parameter based on its characteristics; the distribution of the parameters is detailed in the Additional file 3: Appendix Table 1 [41,42,43]. To account for parameter uncertainty, the range of variation for each parameter was defined as ± 20% of its base-case value [44]. We then performed 1000 independent Monte Carlo simulation trials using R 4.4.0 software.
For the cost–benefit analysis, we estimated the total economic benefit of the intervention as the sum of averted direct medical costs and indirect labor productivity loss, both per case of adult obesity prevented. The cost–benefit ratio (CBR) was calculated using the following formula:
$$CBR = \frac{\text{Net intervention cost}}{N_{\text{Prevent}} \times (C_{\text{medical loss}} + C_{\text{productivity loss}})}$$
where:
-
\({C}_{medical \;loss}=medical \;costs \;averted \;per \;case \;of \;obesity \;prevented\)
-
\({C}_{productivity \;loss}=the \;cost \;of \;labor \;productivity \;loss \;averted \;per \;case \;of \;obesity \;prevented\)
Projecting population-level impact and net economic benefits
To estimate the potential long-term population-level impact of scaling up the intervention nationally, we estimated net economic benefits through a five-step approach based on national demographic and educational statistics. First, we calculated the net effect of the intervention on childhood overweight and obesity prevalence by comparing changes between intervention and control groups. To adjust for potential attenuation in real-world implementation, we assumed a 10% reduction in intervention effectiveness due to implementation challenges during nationwide scale-up. Second, we projected the number of children affected using national enrollment statistics (Statistical Bulletin on Educational Development [45]) and sex-specific prevalence rates (the China Statistical Yearbook [46]). Third, we applied a two-stage obesity progression model to estimate the number of childhood cases likely to persist into adulthood. Fourth, we quantified the lifetime economic burden per adult with obesity—including direct medical costs and productivity loss—to derive total potential cost savings. Finally, we compared the projected economic benefit to the national implementation cost of the intervention to calculate the overall net economic return. The net benefits (NB) from national rollout were calculated using the following formula:
$$NB = N_{\text{national prevented}} \times (C_{\text{medical loss}} + C_{\text{productivity loss}}) – \text{Net intervention cost}$$
Sensitivity analyses
We adjusted the costs of the intervention and the parameters for the model estimates for the potential scenario to simulate nationwide rollout. We assumed that each school would have a full set of teaching aids, increasing the number of sets from 3 to 12 and adjusting the cost accordingly (12 × 188.47 USD per set), and that all 12 schools would require a full set of physical activity supplies, with the cost adjusted to 12 × 78.41 USD per set. To generalize productivity costs beyond the study regions, daily wages were updated to the national average (315.68 CNY) and annual wages to 82,413 CNY. Following international discounting practices, we tested a lower discount rate of 3%, and to reflect potential reductions in real-world implementation, the intervention effect on obesity prevention was reduced by 10% relative to the base-case estimate. These assumptions reflect plausible variations in real-world implementation and long-term projections (Additional file 3: Appendix Table 2 [47]).
link