The range of outcomes and outcome measurement instruments collected in multisectoral community‐based obesity prevention interventions in children: A systematic review

Multicomponent and multisectoral community‐based interventions (CBIs) have proven potential in preventing overweight and obesity in children. Synthesizing evidence on the outcomes collected and reported in such CBIs is critical for the evidence of effectiveness and cost‐effectiveness. This systematic review aimed to identify the range of outcomes and outcome measurement instruments collected and reported in multisectoral and multicomponent CBIs for obesity prevention in children. A systematic search updated an existing review and extended the search to 11 academic databases (2017–2023) and gray literature. Outcomes were classified into outcome domains, and common measurement instruments were summarized. Seventeen outcome domains from 140 unique outcomes were identified from 45 included interventions reported in 120 studies. The most frequently collected outcome domains included anthropometry and body composition (91% of included interventions), physical activity (84%), dietary intake (71%), environmental (71%), and sedentary behavior (62%). The most frequently collected outcomes from each of these domains included body mass index (89%), physical activity (73%), fruit and vegetable intake (58%), school environment (42%), and screen time (58%). Outcome measurement instruments varied, particularly for behavioral outcomes. Standardization of reported outcomes and measurement instruments is recommended to facilitate data harmonization and support quantifying broader benefits of CBIs for obesity prevention.


| INTRODUCTION
Obesity in children is a global public health crisis, causing significant health, social, and economic consequences. 1Although child and adolescent body mass index (BMI) may have plateaued at unacceptably high levels in some high-income countries (HICs), it is still increasing in other areas of the world, including low-and middleincome countries (LMICs). 2,3Overweight and obesity are now recognized as a product of the interplay of multiple complex systemic and institutional drivers. 3,4For example, food and agriculture, transportation, urban design, and land use systems have each been recognized as interacting to create environments that promote unhealthy weight gain. 3Interventions that work collaboratively with community stakeholders to address systemic drivers are urgently required, with the potential for synergistic action that could positively impact not only the very real challenge of obesity but also other challenges with common drivers, such as undernutrition and climate change. 3mmunity-based interventions (CBIs) for obesity prevention have been recognized for their potential in preventing overweight and obesity through participatory methods, including systems science, 4 to address the complex drivers of unhealthy weight. 4,5CBIs designed in partnership with community stakeholders result in multifactorial, multilevel, and multistrategy interventions that foster wholeof-community engagement, and intentionally target multiple drivers of childhood obesity, across a variety of settings and involving a variety of actors. 4CBIs have demonstrated promising intervention effectiveness in preventing unhealthy weight gain in children, compared with single-sector interventions.For instance, comprehensive multisectoral CBIs like Shape-Up Somerville in Massachusetts, USA, and the Romp & Chomp intervention in Victoria, Australia, have demonstrated success in reducing BMI among children by engaging sectors such as households, schools, and communities through environmental changes and community capacity building initiatives. 6,7There is also promising evidence of intervention effectiveness on a range of secondary outcomes, including diet and physical activity-related behaviors, and health-related quality of life (HRQoL). 8,9CBIs have also demonstrated evidence of cost-effectiveness, 10 but there is a need for more comprehensive inclusion of the broader benefits of CBIs into health economic evaluation.To date, published economic evaluations have included a relatively narrow subset of intervention benefits mostly pertaining to changes in body weight. 11This may be undervaluing CBIs for obesity prevention, if there are other substantial and tangible benefits that are not being accounted for in the relatively limited economic evidence that exists.
The increasing number of CBIs for obesity prevention undertaken in recent years 4 presents a significant opportunity to combine insights from these studies to better understand how these interventions may work, for whom and at what cost.Combining data across trials presents a highly efficient method that can result in more powerful and in-depth analyses and is being undertaken in some areas of child obesity (e.g., early childhood obesity prevention 12 ).A challenge of evidence synthesis, however, is achieving data harmonization between studies that have collected a range of different outcomes, using a variety of outcome measurement instruments. 13e first step towards efficient and successful data harmonization is a comprehensive understanding of the data collected, and the methods for data collection, within CBI studies.To date, a comprehensive systematic review of the outcomes collected and reported in CBIs for obesity prevention has not been undertaken.A 2018 study by Karacabeyli et al. 4 aimed to systematically review the methodological approaches for outcome evaluation of CBIs for obesity prevention in children; however, it only reported outcomes across four broad categories: (i) anthropometry (BMI, waist circumference, and percentage body fat); (ii) self-reported outcomes (behaviors and psychosocial outcomes); (iii) observed outcomes (physical activity and fitness); and, (iv) system change (including process evaluation).Bleich et al.'s 14 systematic review summarized single-sector and multisectoral obesity prevention CBIs in the United States and other HICs, but also only summarized four broad categories of outcomes.The categories were (i) anthropometry (BMI, obesity prevalence, and percentage of body fat); (ii) dietary intake (energy intake, fruit and vegetable intake, and sugar-sweetened beverage intake); (iii) physical activity; and (iv) sedentary behaviors.The broad categorization of outcomes in published systematic reviews to date limits our ability to harmonize finer grained data, both on outcomes and processes, across studies.

Scoping reviews of obesity prevention interventions specifically in
early childhood have identified wide heterogeneity in outcomes collected and reported 15,16 and argued for the development of core outcome sets (COS) to standardize data collection and support data synthesis. 16,17is review aimed to identify the full range of outcomes and outcome measurement instruments currently collected and reported in CBIs for obesity prevention in children.This information is critical for intervention trialists designing and evaluating CBIs for obesity prevention, in order to better inform outcome measurement and reporting.
Consideration of the range of different outcomes, along with the instruments used to collect data, is also important when designing studies interested in maximizing data utility through data synthesis.In addition, a key step in development of any COS is identifying the existing knowledge about outcomes in an area.Results from systematic reviews of outcomes, such as this one, can inform further steps in COS development (i.e., by informing Delphi or other consensus approaches). 18Findings will also be useful to health economists, in considering the availability of data to support the inclusion of broader benefits into the economic evaluation of CBIs for obesity prevention.
As such, this review addresses the research question: What is the current range of outcomes collected and reported in multisetting and multisector CBIs for obesity prevention in children, and how were the outcomes measured?

| METHOD
This review was prospectively registered with PROSPERO (CRD 42022351496) and was undertaken following Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) guidelines. 19

| Information sources and search strategies
We set out to build on the previous systematic review of complex childhood obesity prevention CBIs by Karacabeyli et al.,4 (in MEDLINE, CINAHL, and PsycINFO); however, the current review Further, a gray literature search using Google Scholar was conducted in May 2022 and February 2023 by one reviewer (M.S.) to identify any additional intervention(s), and/or any additional studies of already identified interventions (i.e., those identified from the previous review 4 and academic database searches (Table S1.1).Specifically, to identify additional intervention(s), an advanced Google Scholar search was conducted using predefined search terms (Table S1.2).Following that, another search was conducted using the title of each intervention identified from the academic database search, together with the predefined search terms, to identify studies/reports of already identified interventions.These searches were limited to the first five pages of the results.
Comprehensive details of the search strategies are presented in Supporting information (Tables S1.1 and S1.2).
Studies identified from the updated academic databases search were imported into EndNote.After removing duplicates, records were exported to Covidence (https://www.covidence.org) and screened independently by two reviewers (M. S. and J. J.).Conflicts were resolved through discussion, and through consultation with a third reviewer (V.B.).Reference lists of included studies were screened for additional study inclusions by one reviewer (M.S.).
To increase the comprehensiveness of the data about each of the • focused on the prevention of obesity in children or adolescents; • multisectoral and multicomponent CBIs targeting the whole community and engaging at least two sectors such as schools, households/families, community, recreation centers, childcare centers, businesses, health services, media, and local governments through a participatory approach 4,11 ; • reported the primary or secondary outcomes and/or outcome measurement instruments used; • were full-text studies/reports published in the English language.
Interventions were excluded if they were • reported outcomes only in adult populations (aged >18 years); • conducted within a single sector (e.g., interventions only conducted within the school sector); • were only reported methodological studies (e.g., examining intervention design) without stating anticipated outcomes.Following extraction, data were analyzed using a deductive iterative approach to sort data and structure the conceptualization of outcomes and outcome domains. 18Outcomes collected for each intervention were carefully checked and duplicates (i.e., where an outcome was reported across multiple peer-reviewed publications, or in an academic paper and a trial registry record) were removed to avoid double-counting.We did not analyze outcomes according to whether they were primary or secondary outcomes, as there were multiple instances of conflicting reporting (i.e., where one academic paper stated the outcome was a primary outcome, but the same outcome was categorized as a secondary outcome in a subsequent paper or trial registry record).Categorization of outcomes into outcome domains was adapted from domain frameworks from recently published reviews that were conducted to inform a core set of outcomes for early childhood obesity prevention interventions 15,17 and lifestyle interventions of older populations. 20r instance, anthropometric outcomes such as body composition (e.g., body fat percentages), BMI, height, weight, and circumference were grouped under the "anthropometry and body composition" outcome domain. 20Outcomes with similar definitions were merged by one reviewer (M.S.), and checked by a second reviewer (V.B.).

| Data extraction and synthesis
Outcomes were categorized into outcome domains by one reviewer (V.B.) and checked by a second reviewer (M.S.).Outcome domains and definitions were reviewed by all co-authors.Following domain categorization, outcome domains was categorized by age "up to 5 years," "5 to 12 years," "13 to 18 years," and "other."All frequencies were summarized using descriptive statistics and presented in outcome matrices.
Given the large number of outcome domains and outcomes, the extensive variety in outcome measurement instruments used, and the limited detail reported about outcome measurement instruments in some studies, it was not feasible to present frequencies of use for outcome measurement instruments.Instead, a list of the outcome measurement instruments was compiled for all outcomes identified.In addition, the outcome measurement instruments used for the most frequently reported outcomes in the five most frequently reported outcome domains were narratively summarized.Broad groups of outcome measurement instruments were coded where appropriate (e.g., "Tanita BWB-800 scale" was coded as "scale"; "ActiGraph GT1M" was coded as "accelerometer"; "self-reported behaviors, attitudes and knowledge questionnaire comprised of 83 questions" was coded as "83-item survey questionnaire").Assessment of the validity and/or reliability of the outcome measurement instruments used was outside the scope of this review.Given that the aim of the review was to identify the range of outcomes and measurement instruments, rather than the methodology used for the study, risk of bias assessment was not undertaken.
The least frequently collected outcome domains were oral health, neuropsychological, and motor function (each identified in only one intervention).The community outcome domain was collected in 12 interventions (27%) (Table 2).Economic and eating behavior outcome domains were each collected in 36% of the included interventions (n = 16).The quality-of-life outcome domain was collected in less than half of the included interventions (n = 15, 33%, Table 2).
All outcome domains were most frequently collected in studies of children aged 5-12 years, except oral health, which was only collected in one study in children aged up to 5 years (Tables 2, S2, and S4).
Table 3 provides more information on the outcome domains collected in each intervention.The "Texas Childhood Obesity Research Demonstration (TX CORD)" intervention 41,91,116,119,134,135  Glucose Study) collected outcomes from a single outcome domain (anthropometry and body composition). 56,87The outcome domains were collected in various populations and settings (e.g., child, parent, school, and community), detailed in Tables S3, S4, and S5.Further, detailed analysis of the population group in which the outcome domain was collected, as well as the method for measurement (e.g., self-report and measured), is presented in Supporting information 6-14.

| Outcomes
From the 17 outcome domains, a total of 140 unique outcomes were identified (Supporting information S3; Table S3).Frequencies of outcomes within each outcome domain are presented in Supporting    S30, Figure S30).Self-or parent-reported outcomes    The potential for spillover effects has been recognized in the literature 64,[144][145][146] ; however, limited and sometimes conflicting evidence of these spillovers means this is a significant area for future work.This means that evidence on an important dimension of possible effects of CBIs is underexplored, and this may also be a missed opportunity for confirmation that CBIs do not result in potential unintended harms (e.g., stigma).Lack of reporting community domain could be due to the relative lack of validated instruments as well as lack of scope or funding to collect such outcomes.For HRQoL, the Pediatric Quality of Life Inventory 148 was most frequently used for collecting QoL data (n = 6), but it is a non-preference-based instrument, focusing on health domain functions without providing patient-preferred utility values.In contrast, preference-based or utility measures assess how patients value experiencing a defined health state based on functioning and well-being in those domains. 149The limited evidence on preference-based HRQoL poses a challenge to economic evaluation using cost-utility analysis.Cost-utility analyses are commonly undertaken in health economic analysis to assess costs and benefits, incorporating an intervention's impact on both quality and quantity of life using preference-based instruments.The limited data collection related to these outcomes may be attributed to the interventions' primary focus on preventing obesity in children.In addition, collecting such information might be challenging given the lack of validated instruments as well as insufficient funding.

| Outcome collection population and measurement instruments
The variance in the methods for selection and measurement of behavioral and other outcomes, and even the variance in ways in which anthropometric data are collected and reported, hampers the ability to compare across studies.Considerable inconsistencies in reporting anthropometric and other behavioral outcomes across studies for obesity prevention interventions are also highlighted by available literature. 14,15,17,152This is hampering evidence synthesis, which is of critical importance for understanding what works and how, and designing new interventions accordingly.One potential direction for future research is the standardization of a set of outcomes recommended for collection and reporting in CBIs for obesity prevention (i.e., a COS). 152Development of a COS, with accompanying core outcome measurement instrument recommendations, would better inform evaluation design and data collection and encourage maximum scientific gain from evidence synthesis and cross-study comparison of intervention effects. 17,153is review is an advancement to the literature in multiple ways.
The review was conducted systematically using a range of academic databases including gray literature and was registered in PROSPERO.
Another major strength includes comprehensive cataloguing of outcomes collected in CBIs for obesity prevention, including details on the target populations and instruments used.Despite these strengths, this review has several limitations.Although the clinical trial registries were searched for identified interventions, no detailed search was conducted using clinical trial registries to identify additional interventions.In addition, only studies published in the English language were considered.Given the large volume of interventions and outcomes included in our review, and the lack of a fully detailed, and up-to-date database of all outcomes and measurement instruments used in intervention studies, it is possible that some outcomes and outcome measurement instruments may have been missed.Our comprehensive search strategy across multiple databases, trial registries, and the gray literature was designed to minimize this risk to the best of our ability.
Finally, investigating reliability/validity of the reported measurement instruments was not considered within the scope of this study, but is an area for future work.

| CONCLUSIONS
This review summarized a broad range of outcomes and measurement instruments collected in multicomponent and multisectoral CBIs for obesity prevention in children.Seventeen outcome domains were identified, and six domains were reported by over 50% of the included interventions.The outcomes and measurement instruments are considerably heterogeneous.Summaries of commonly collected outcomes and measurements instruments can be used as a key reference for intervention trialists when planning obesity prevention interventions and is an important step in the creation of a COS for obesity prevention CBIs.Findings will also support health economists to consider the broader range of outcomes that may be used in any economic evaluation aiming to incorporate potential broader benefits into cost-effectiveness evidence.
intended to capture more comprehensive outcomes including details of instruments used for outcome measurements.Therefore, all interventions (n = 33) included in that study were included and updated with additional interventions published on and after July 2017.The updated search expanded the search strategy utilized in Karacabeyli et al. 4 by adding additional search terms, to ensure all relevant inclusions were identified and to reflect more recent advances in the literature on CBIs that have been published since July 2017.The search was designed in conjunction with a subject specialist librarian and included a broader range of academic databases (n = 11; Academic Search Complete, CINAHL Complete, EconLit, Global Health, MED-LINE Complete, APA PsycINFO, SPORTDiscus, Embase, Scopus, PubMed, and Cochrane Library).The search was run in September 2021 and updated in February 2023.
included interventions, a targeted search of publicly available clinical trial registries (www.clinicaltrials.gov;and the World Health Organization International Clinical Trials Registry Platform [WHO ICTRP], www.who.int/clinical-trials-registry-platform) was undertaken by one reviewer (M.S.) using the title/trial identification number of all included interventions from Karacabeyli et al. 4 and those identified in the updated search.Interventions met the inclusion criteria if they were • conducted in the general child or adolescent population (aged 0-18 years) and/or parents of children or adolescent aged 0-18 years;

A
data extraction template was developed in Microsoft Excel based on the template used in a recently published scoping review of outcomes collected and reported in early childhood obesity prevention interventions. 15Data extracted included intervention name, study design, setting, recruitment country, study aim, target population, recruitment process, sample size, description of the intervention and comparator, outcomes, outcome measurement instruments, time-points for measurement, measured population (e.g., children, parents, and teachers), sector (e.g., school, community), evaluation methods, and details of statistical analyses, summary results, generalizability, limitations, source of funding, and conflicts of interest.Data extraction was completed by one reviewer (M.S.), and all data extraction was checked by another reviewer (V.B.) to ensure accuracy.
All included studies from 45 interventions were published between 2004 and 2023.Forty interventions were conducted in HICs (the United States [n = 18], Europe [n = 9], Australia [n = 7], the United Kingdom [n = 4], New Zealand [n = 1], and South Korea [n = 1]), three interventions were conducted in upper middle-income countries (UMICs; one each in China, Tonga, and Fiji), one conducted in both HIC and UMIC (n = 1), and, one intervention was conducted in a LMIC (Iran) (Table 1).Thirty-one interventions were quasi-experimental in design (n = 20 with control groups), and 14 interventions were either randomized controlled trials (RCTs) or cluster RCTs.The primary target populations were children (n = 40), parent-child dyads (n = 3), and parents with the outcomes measured in children (n = 2) (Table S2).The majority of included interventions focused on children aged between 5 and 12 years (n = 23).Eight interventions targeted children up to 5 years, four interventions targeted adolescents (aged 13-18 years), and the rest (n = 10) targeted a range of ages (e.g., 0-18 years).
utilized to collect the 140 outcomes identified in our review.Supporting information S29-S33 summarize the outcome measurement instruments used to measure the most frequently collected outcome in each of the five most frequently reported outcome domains (i.e., anthropometry and body composition, physical activity, dietary intake, environmental, sedentary behavior).Thirty-six interventions measured BMIz, and four interventions measured BMI in children with specific measurement tools.Eight interventions collected self-reported BMIz among parents/caregivers (Supporting information S29).When calculating BMI percentiles or classifications for children, nine interventions used the International Obesity Task Force reference, 140,141 eight used the US Centers for Disease Control and Prevention reference, 142 and six used World Health Organization child growth standards or reference 143 (Figure S29).Physical activity (e.g., daily steps and time spent in moderate to vigorous physical activity) was the most frequently reported outcome in the physical activity outcome domain (collected in 87% of interventions that collected physical activity domain outcomes; 73% of all interventions); and all interventions collected the outcome data among children, mostly for the children aged 5-12 years (n = 17).Of these interventions, 48% (n = 16) measured the outcome objectively, such as through the use of an accelerometer (33%), pedometer (9%), or other methods, while the remaining 52% (n = 17) utilized selfreport (n = 7, 21%), parent-report (n = 7, 21%), or other methods (n = 3) (Table

•
Body mass index • Body composition • Height and weight, reference (e.g., CDC and WHO) CBI participants (number of studies in age group collecting outcome from domain; % of all CBIs included in review) be attributed to the inclusion of a broad range of age-groups (e.g., 0-18 years) or specifically reporting BMI for adults.For instance, a few interventions (n = 8) collected and reported anthropometric outcomes in parents or caregivers.In addition, few other outcomes were collected and reported in broader target populations than the child (e.g., teachers), despite broader impacts community members feasibly resulting from CBIs that involve systems-level community change.
Environmental outcomes were collected by 71% of included interventions in our review and the school environment outcome was most frequently collected.Relatively high collection of outcomes in this domain may be a reflection of the design of CBIs, where most CBIs incorporate strategies aimed at promoting healthy environments for communities.The school environment outcome was the most frequently collected, possibly due to the fact that school environments are relatively contained and amenable to intervention.Evidence also suggests that the school environment may be influential for physical activity related behaviors among children.147In contrast, outcome domains such as quality of life (QoL) related outcomes in children and community were collected in only 33%(n = 15) and 27% (n = 12) of the included interventions, respectively.
Table S3 presents a comprehensive list of all collected outcomes, target population with age groups and measurements instruments of the study selection process.Study characteristics (country/region).Outcome domains, age categories, examples of outcomes, and measurement instruments.
15A B L E 129,37,106,123Communities and Schools Together (CAST)43Children's Healthy Living (CHL) a Program69,93,99,112,117,118,124,125,129,138Healthy Eating, Active Communities (HEAC) Among the32interventions that collected the dietary intake outcome domain, fruit and vegetable intake in children was the most frequently collected outcome (n = 25, 78% of 32 interventions, Supporting information 20) and was mostly measured either via self-report (n = 12, 48%) or parent report (n = 11, 44%) (Supporting information 30).Three interventions also collected the outcome among parents.Measurement instruments for fruit and vegetable intake varied across interventions, and included a range of survey tools (TableS31).4 | DISCUSSIONThis systematic review categorized 140 unique outcomes into 17 outcome domains and summarized measurement instruments from 45 multicomponent, multisectoral obesity prevention CBIs in children, as collected in 120 included studies.The list is extensive and demonstrates considerable heterogeneity that currently exists in the choice of outcomes and outcome measurement instruments between CBI studies.Decisions about what to measure in multisectoral and multicomponent obesity prevention CBIs in children are influenced by many factors, both pragmatic and theoretical, including the specific objectives and priorities of the CBI, the program logic of the intervention, project resources (funding, personnel, and time available), ethical constraints, researchers'perceptions of an acceptable level of participant burden, and expectations about likely size of effect, and whether the effect may be detectable in the available sample population.Decisions about what outcomes to report in peer-reviewed or other publications may further be influenced by the participation rates achieved, and actual results observed; for instance, outcomes with significant changes may be more likely to be reported than those on which the intervention had no effect.15Giventhe primary aim of these interventions is obesity prevention, the frequency of collection of anthropometric outcomes is expected.Although reporting BMI as an outcome, age and sex standardized BMI was not measured for four interventions.This might T A B L E 1 (Continued) Interventions Country/region implemented in Studies reporting outcomes included in our review Healthy Habits, Happy Homes Scotland (4HS) 76 Optimising Family Engagement in HENRY (OFTEN) a Intervention conducted in both HIC and UMIC.T A B L E 2 Cambridge Neuropsychological Test Automated Battery; CDC, Centers for Disease Control and Prevention, FFQ, Food frequency questionnaire; FGD, Focus group discussion; KII, Key informant interviews; KIDMED, Mediterranean Diet Quality Index; SFFQ, Short food frequency questionnaire; WHO, World Health Organization.Outcome domains collected by each of the included interventions and by age (n = 45).
a Wide/various range of age groups.T A B L E 3