Institutional Review Board from the University of California, Merced approved the study, which was conducted in compliance with ethical standards in the treatment of the participants between August 13, 2020 and August 21, 2020.
Participants
The study included 306 parents (51% fathers), satisfying common recommendations for sample size requirements [24], given 31 parameters and 8 variables in the structural equations modeling (SEM) of the proposed model. Eligibility requirements were participants being above 18 years of age and proficient in English, residing in the US, and having at least one child between 5 and 12 years of age (Mean child age = 8.71, SD = 1.97). When a participant had more than one child, he or she was asked to select one child in the age range to complete the questionnaire. Demographic information is presented in Table 1.
Table 1. Sample demographic information
Variable
|
%
|
Child's gender
|
|
Boy
|
67.40
|
Girl
|
32.60
|
Child's race
|
|
Asian
|
7.10
|
Black
|
16.45
|
Latinx
|
6.13
|
White
|
66.77
|
Other
|
3.55
|
Respondent's relationship to the child
|
|
Biological father
|
50.32
|
Biological mother
|
48.06
|
Adoptive father
|
0.97
|
Adoptive mother
|
0.65
|
Father's educational level
|
|
8th grade or less
|
0.63
|
High school graduate and GED
|
7.55
|
Some college, or 2-year degree
|
35.85
|
4-year college graduate
|
49.69
|
More than a 4-year college degree
|
6.29
|
Mother's educational level
|
|
8th grade or less
|
0.66
|
High school graduate and GED
|
4.64
|
Some college, or 2-year degree
|
32.19
|
4-year college graduate
|
48.58
|
More than a 4-year college degree
|
13.93
|
Father's feeding involvement a
|
|
Never
|
5.70
|
Seldom
|
25.30
|
About half the time
|
41.80
|
Most of the time
|
24.70
|
Always
|
2.50
|
Mother's feeding involvement a
|
|
Never
|
3.30
|
Seldom
|
30.70
|
About half the time
|
29.30
|
Most of the time
|
26.00
|
Always
|
10.70
|
a Response to the question “when your child is at home, how often are you responsible for feeding him or her?”
Procedures
Participants were recruited using a Facebook snowball sampling method, which was selected because it provided easy access to the target population at a lower cost than alternatives [25]. Moreover, the number of cases detected by Facebook has found to be higher than the traditional snowball sampling method [26]. Participants saw threads on Facebook and clicked the link for the study, which led to the Google Survey form through which the study was administered. To start, participants were presented with a description of the study and the informed consent. After indicating written consent, they answered questions online, using their smart phone or computer for approximately 20 – 30 minutes. The first 100 participants were compensated with a $10 Amazon e-gift card, and their remainder with a $5 card.
Measures
Fear of COVID-19
Participants were asked to indicate their level of agreement with seven items addressing fear of COVID-19, such as “I am most afraid of COVID-19” and “It makes me uncomfortable to think about COVID-19,” used in previous research [27]. Responses were recorded on a five-point Likert scale (1 = “strongly disagree” to 5 = “strongly agree”). A total sum across items was calculated, with higher scores indicating higher levels of fear of COVID-19. Internal consistency for this sample was α = .80.
Parent-child communication
Parent-child communication was evaluated with the Revised Family Communication Pattern Instrument – Parent Version (RFCP) [28], which has been developed to measure family communication according to the Family Communication Patterns Theory [29]. The RFCP consists of 26 items measuring two underlying dimensions of family communication pattern, using a Likert-scale (1 = “strongly disagree” to 5 = “strongly agree”): (1) conversation-oriented (CONV) family communication (15 items) such as “I often ask my child’s opinion when the family is talking about something” and “My child and I often have long, relaxed conversations about nothing in particular;” and (2) conformity-oriented (CONF) family communication (11 items) such as “When my child is at home, it is expected to obey the parents’ rules” and “In our home, the parents usually have the last word.” Scores were calculated as the mean item score separately for each communication dimension, with a higher score indicating stronger presence of that dimension. Internal consistency for this sample was α = .91 for CONV and .80 for CONF communication.
Parent’s controlling feeding practices
The parental controlling feeding practices with the child were assessed, using the Child Feeding Questionnaire [30], which is designed to assess seven dimensions. From these, three dimensions addressing controlling feeding practices were selected, including (1) monitor (3 items, e.g., “How much do you keep track of the sweets that your child eats?”), (2) restriction (5 items, e.g., “I have to be sure that my child does not eat too much of his or her favorite foods”), and (3) pressure to eat (5 items, e.g., “My child should always eat all of the food on his or her plate”). Responses were made on 5-point scales, with anchors from “never” to “always” for the monitoring items and from “disagree” to “agree” for the restriction and the pressure to eat items. Items were subjected to a confirmatory factor analysis (CFA) (see Results) to confirm the item structure of these three dimensions and that they could indicate a latent variable of controlling feeding practices. Internal consistency was α = .81.
Child’s dietary intake
Using items from the California Health Interview Survey Diet Screener [31], parents were asked about the child’s dietary intake of two food categories commonly used to mark healthy intake (fruits and vegetables) and four used to mark unhealthy intake (juice, soda, sweets, and fast-food consumption). These are food categories commonly used in survey research to mark healthy and unhealthy diet intake [32, 33, 34]. Questions ask about the number of servings the child consumed yesterday for each food category, with responses ranging from “0” to “more than 8 servings.” The exception was for fast-food consumption, for which the period was last week. Servings were self-defined by parent and considered to be child’s regular portion of the food. Items were subjected to CFA to confirm using them to measure the latent variables of healthy intake and unhealthy intake, respectively. Internal consistency for healthy food was α = .60 and unhealthy food α = .80.
Race/ethnicity
Participants indicated which one or more of eight racial/ethnic categories described the child from the following: American Indian or Alaska Native, Asian, Black/African American, Caucasian/White, Hispanic/Latino, Middle Eastern, Mixed and Multi-racial, or other race and ethnicity. Using the Census classification approach, the child was classified as Asian or Latino if so indicted, regardless of other racial/ethnic indication.
Statistical Analysis
IBM SPSS Statistics 20 was used for descriptive statistics and Mplus for SEM analysis. To assess the construct validity of the measurement models of the controlling feeding practices and child’s healthy and unhealthy dietary intake categories, a series of CFA were conducted to establish them as latent variables. After ensuring the adequate fit of the measurement models, three items from restriction, one item from pressure to eat, and two items from monitoring were included in the final model to measure the latent variable of controlling feeding practices.
A SEM path analysis was conducted first for the total sample. Subsequently, a multi-group SEM analysis was used with the father and mother subsamples. Child’s gender was entered as a control variable in the path analysis. Additionally, the structural models of indirect effects of family communication and parental controlling feeding practices between parental COVID-19 fear and children’s healthy and unhealthy food intake were tested using maximum likelihood estimation and a bias corrected bootstrapping procedure with 1000 iterations. The bootstrapping method can correct bias and thus give more accurate estimations [35]. The indirect effect is considered statistically significant if zero is not contained in the 95% confidence interval. Multi-group path analysis was completed also to test for fathers and mothers group differences in structural parameters of the indirect effect model.
To examine goodness of fit of the model, the comparative fit index (CFI), Tucker-Lewis index (TLI), and root mean square error of approximation (RMSEA) were assessed. Adequate fit for parsimonious SEM has been identified as CFI ≥ 0.90, TLI ≥ 0.90, and RMSEA ≤ 0.10 [36]. A value of α = 0.05 was set for statistical significance. Missing data was implemented under maximum likelihood estimation [37].