Impacts of sharing research findings with study participants: Evidence from a field experiment

Participants will be randomly assigned to one of three groups:
1. Control group involves those who receive general public information on weight and nutrition;
2. Treatment group 1 involves those who receive research findings from the broadly defined scientific literature and from the baseline survey data to which they contributed; and
3. Treatment group 2 involves those who receive research findings from the broadly defined scientific literature and from the baseline survey data to which they contributed, as well as child-specific comparisons with baseline survey summary findings.

Participants will be randomly assigned to one of three groups and will participate in two surveys, namely the baseline survey and the follow-up survey. Specifically, treatment groups 1 and 2 will receive the following information:
Treatment group 1 receives general research findings + baseline survey summary findings.
Treatment group 2 receives general research findings + baseline survey summary findings + child-specific comparisons with baseline survey summary findings.
The information will be delivered to parents through scripted virtual workshops, virtual workshop recordings, and written reports. The control group will participate in a workshop where a standard, publicly available nutrition video is shown without making specific reference to research findings.
General research findings include 3 main pieces of information: the prevalence of overweight and obesity, the perception of parents about the healthiness of their child’s weight and diet, and researchers’ efforts to reduce the prevalence of overweight and obesity in primary schoolchildren. In particular, participants will be told that Ho Chi Minh City (HCMC), Vietnam’s largest city, has witnessed increases in youth overweight and obesity rates. A study conducted in 2016 shows that nearly 1 in 2 primary schoolchildren aged 7-9 years in HCMC was overweight or obese (Nam et al., 2020). Overweight and obesity in turn can lead to health problems such as cardiovascular disease, diabetes, and cancer later in life. Despite the potential negative consequences of being overweight or obese, many parents have a misperception about the healthiness of their child’s weight and diet. Previous studies have found that a high proportion of parents do not accurately classify their child’s weight status, especially their overweight status (Chen et al., 2014; Hong et al., 2017). Also, parents’ weight perception can be associated with the healthfulness of the meals prepared for children. If mothers perceive their child to be heavy, they are more likely to adopt strategies to promote a healthy weight for their child. For example, Do et al. (2015) found that a higher maternal perception of their child’s weight positively influenced monitoring, which in turn helped to reduce their child’s consumption of snacks and fatty foods. Given the negative effects and the prevalence of overweight and obesity in young children, researchers have endeavored to find measures to prevent childhood overweight and obesity. For example, Nam et al. (2022) conducted a study with parents of primary schoolchildren in HCMC. They found that receiving nutrition counseling along with goal setting and mobile text reminders led to a decrease in body fat, waist circumference, and the risk of overweight or obesity among children. The study found that the information-based intervention activities were successful in improving anthropometric status because they caused improvements in child diets and in the perceptions of parents about whether their child’s diet was healthy. Also, the authors found that girls had better anthropometric improvements than boys and improvements in girls’ anthropometric status still persisted even after two years. (Nam et al., 2022).

2024-03-15

2024-04-30

Please see details in the attached file.

Please see details in the attached file.

Please see details in the attached file.

Please see details in the attached file.

Participants will be randomly assigned to one of three groups and will complete two surveys at home (the baseline survey and the follow-up survey):
- In the baseline survey, all participants will answer the same questions.
- After about 2-3 weeks after the intervention, they will conduct a follow-up survey.

We will also qualitatively interview a small subset of parents from all three groups about 2 months after the follow-up survey to explore their experience with the study and better understand potential effects.

Following the study by Nam et al. (2022), we will communicate the research findings to parents of primary schoolchildren and examine whether and how the delivery of such information can affect health- and diet-related parental perceptions. There are three dimensions that concern us: health perception, diet perception, and research acceptance.
Model approach:
We examine whether the delivery of research findings can impact parents’ health perceptions, diet perceptions, and research acceptance. We accomplish this objective by regressing health perception, diet perception, and research acceptance on group indicators in different regression specifications. In particular, we specify the following conditional model:
Yi = α + β1TreatmentGroup1i + β2TreatmentGroup2i + β3Yi,0 +β4Xi + εi
where:
Yi denotes the outcome variable at follow-up.
TreatmentGroup1 equals 1 if subjects are in treatment group 1, and 0 otherwise.
TreatmentGroup2 equals 1 if subjects are in treatment group 2, and 0 otherwise.
The omitted category is the control group.
Yi,0 is the lagged value of the dependent variable (if collected at baseline).
β1 and β2 are the coefficients of interest, and X is a set of covariates at baseline. These covariates include child traits (age, gender), respondent traits (relationship to child, age, gender, education), and household characteristics (number of members, monthly expenditure). Missing covariate values are replaced with the sample mean and a dummy indicating missingness is included in the model.
ε is the error term. Robust standard errors will be computed.
Define health perception, diet perception, and research acceptance:
(1) Health perception
Health perception is expressed through two dimensions. First, we examine parents’ perception of their child’s current weight. Second, we examine parents’ perception of their child’s overweight/obesity status. Particularly, whether parents correctly determine their child’s weight (based on school anthropometric records), and whether they correctly identify their child’s weight status as underweight, normal weight, overweight, or obese. Health perception questions will be asked in both the baseline survey and the follow-up survey.
(2) Diet and physical activity perception
In both the baseline survey and the follow-up survey, participants will be asked about their child’s daily consumption of foods and drinks. These foods and drinks include vegetables, fruits, salty snacks, sweet snacks, foods prepared away from home, and sugary drinks. We will also ask about parents’ perception of their child’s physical activities.
(3) Research acceptance
Do participants find the results from the wider literature or the baseline survey of health and dietary perceptions helpful? Or are they not really interested in such academic research findings? Can sharing research findings improve subsequent research participation and research perceptions? We attempt to answer these questions by considering their acceptance of the existing literature. This is done through different ways. First, participants will use a Likert scale to rate the usefulness of the study. Second, we will ask if they are interested to participate in a qualitative study to be conducted after 2-3 months. Third, we will use the dictator game to explore their preferences for such research. Specifically, participants will be able choose to donate to a future study (scenario 1) or an anonymous nutrition center to disseminate research results (scenario 2). Questions regarding research acceptance will be asked in the follow-up survey only.
Missing data and attrition:
We will check if missing outcome data and attrition are balanced by experimental condition.
Subgroup analysis:
Nam et al. (2022) indicate that the average effects of the intervention on anthropometric status are stronger among girls than among boys. Therefore, we will qualitatively explore subgroup effects by estimating linear regressions by gender to examine whether there are any differences between female and male children.
Multiple hypothesis testing:
Following Anderson (2008), we will calculate q-values by family of indicators for multiple hypothesis testing (see general content groups in attached file).
Average standardized effects:
As the outcome variables can include a number of specific variables, we will calculate the average effect by outcome family coded to move in the same direction based on the approach of Kling, Liebman, and Katz (2007) (see general content groups in attached file).
Spillovers:
We will explore potential spillover effects due to connections between treatment and control participants by using data on friendship networks among parents and children and controlling for connections or dropping highly connected control units.
Social Desirability Bias:
We will explore social desirability bias by controlling for an index collected at baseline and also interacting it with the treatment variables to see if treatment effects vary by the desire to please the experimenters.

Subjects will be randomly assigned to three groups. Randomization is conducted using STATA.
We will check randomization by testing for balance on outcomes at baseline as well as basic socio-demographics.

Individual

No

6 schools with treated and control groups in each.

We will survey a total of 1500 primary schoolchildren in control and treatment units from 6 schools in HCMC at baseline and then again after about 2-3 weeks after the intervention.

There are three experimental groups, each group includes 500 parents/children from 6 schools.

We aim to examine the effects of sharing findings with research participants with a power of 80%. We build on Nam et al. (2022) who investigated the effect of an information-based treatment package on childhood overweight/obesity. We perform a power calculation based on effects on parents’ dietary perceptions of their children in by Nam et al. (2022) (parents’ perception of their child’s diet being excellent or very good). Setting parents’ perception in the control group to 0.34, the effect size to -0.083, power to 0.8, and the significance level to 0.05, we find that each group should include at least 476 subjects. In Stata, this is calculated as power twoproportions 0.34 0.257. We round up to 500 observations.