Home Environment Factors and Health Behaviors of Low-income, Overweight, and Obese Youth

Abstract

Objectives

Home environment may influence children’s health behaviors associated with obesity. In this study, we examined home environment factors associated with diet and physical activity behaviors of overweight or obese youth.

Methods

We analyzed baseline data from child and parent/caregiver dyads enrolled in an urban family weight management program. Multivariable logistic regression examined how home environment (parenting practices, family meal habits, and home availability of fruits/vegetables, sugar-sweetened beverages (SSBs), screen media, and physical activity resources) are related to children’s intake of fruit, vegetables, and SSBs, and moderate-vigorous physical activity and sedentary time (ST) after adjusting for potential confounders.

Results

Children were more likely to consume fruit if their families frequently ate meals together and infrequently watched TV during meals, and more likely to consume vegetables with high fruit/vegetable availability and low SSB availability. Children were more likely to engage in ST if parents practiced monitoring and frequently watched TV during meals.

Conclusions

Overweight or obese children appear to have healthier habits if their families eat meals together without watching TV and if healthy food choices are available in the home. Encouraging parents to focus these practices may promote healthier body weight in children.

Childhood obesity is a major health concern in the United States (US) with significant disparities among racial/ethnic minority subgroups¹ and children with lower socioeconomic status.² National Health and Nutrition Examination Survey data from 2013–2014 indicate that the prevalence of obesity is higher and occurs at a younger age among black and Hispanic youth with the highest prevalence (25%) found in elementary school-age Hispanic youth.³ Standard of care⁴ to address pediatric obesity indicate that pediatric primary care providers (PCPs) should provide family-based counseling that includes culturally appropriate dietary and physical activity recommendations for behavioral change. With almost 80% of US children receiving one or more well-child care visits annually,⁵ PCPs are well positioned to address parenting behaviors related to obesity. Prior research suggests that parents view PCPs as highly valued weight management advisors^6,7 relying on the PCP to initiate weight management counsel.⁸ PCPs need to guide parents to become positive behavior-change agents in the prevention and treatment of childhood obesity. This includes empowering parents to create healthier home food environments and develop effective parenting practices related to food and physical activity.^4,9

However, there are important gaps in our understanding of how PCPs can leverage parenting and home environment factors to promote health behavior change,⁴ particularly among high risk racial/ethnic minority populations. Limited but growing evidence among Hispanic/Latino youth suggests that targeting parenting practices/skills and the home environment among preschool and early elementary-aged (kindergarten-2^nd grade) children is effective for addressing pediatric obesity disparities.^10–13 However, whether parenting and home environment variables are applicable to older children, who face increased exposure to external social and environmental determinants of obesity, needs elucidation. Findings from our pediatric ambulatory care-based family intervention trial suggest that semi-structured PCP counseling, which includes several targets of parenting behavior, can achieve a modest improvement in body mass index (BMI) Z-scores among overweight and obese elementary school-age youth (ages 7–12 years old).¹⁴

The present study focuses on a more in-depth examination of baseline data from this cohort of predominantly Hispanic/Latino and underserved children and their parent or primary caregiver (hereafter referred to as parent). The aim was to evaluate cross-sectional associations between children’s diet and physical activity behaviors and components of the home environment, including parenting practices, family meal habits, home availability of fruits/vegetables, home availability of sugar-sweetened beverages, and home availability of screen-media and physical activity resource. Results can contribute to clinical practice research gaps and inform the development of primary care-based obesity prevention or treatment interventions that aim to target home environments of families to promote weight management, particularly among high-risk Hispanic/Latino youth.

METHODS

Study Design, Setting, Participants, and Procedures

The present study is a cross-sectional analysis of baseline data of the Family Weight Management (FWM) study (registered at www.clinicaltrials.gov, NCT00851201; also known as the Fun Healthy Families Study).¹⁴ The FWM study¹⁴ was a 12-month randomized controlled trial of a pediatric weight management intervention conducted through a Bronx municipal hospital’s safety-net (with predominantly Medicaid coverage) primary care pediatric practice. Bronx County, New York, is a predominantly ethnic minority (53.9% Hispanic/Latino; 30% non-Hispanic/Latino black or African American),^15,16 low-income, urban community with an estimated childhood overweight and obesity prevalence of 40%,^17,18 which exceeds state and national levels.¹ Participants were child-parent dyads. Children (ages 7 to 12 years old) were overweight or obese (BMI ≥ 85^th for age and sex). We previously reported that our participants had high cardiometabolic risk.¹⁹ Recruitment and enrollment occurred from July 2009 through December 2011. Inclusion/exclusion criteria have been published elsewhere.^19,20 One parent (or primary caregiver) per child were enrolled in the study. Two to 3 visits were used to obtain baseline data. Families could begin completing questionnaires after providing consent/assess, complete additional questionnaires during pre-randomization visit for obtaining fasting blood specimens, and could finish completing questionnaires during a third visit in which eligibility was verified prior to randomization.

Independent survey data were not available for siblings, limiting this study to 301 parent-child (index) dyads that had complete anthropometric and survey measurements assessed in our analyses. As previously described,¹⁴ 536 children were assessed for eligibility, of which 170 were excluded for the following reasons, which were not mutually exclusive; overall, 23 did not meet inclusion criteria, 80 declined to participate, and 76 did not complete baseline assessments. A total of 360 eligible children assented. In the case that the child recruited into the study (called the index child) had sibling(s) that met eligibility criteria, the sibling(s) was invited to participate (N = 42); however, surveys used in this analyses were for the index child only.

Dependent Variables

Children’s dietary intake

Children’s fruit, vegetable, and sugar-sweetened beverage (SSB) intake, defined in Table 1, were assessed using the Block Kids 2004 Food Frequency Questionnaire (Block Kids FFQ) Spanish version (NutritionQuest, Berkeley, CA). The Spanish version contains food items of typical Mexican-Americans diets that were selected using nationally representative surveys.²¹ The Block Kids FFQ is a widely accepted reliable instrument²² with a third-grade reading level. It has been validated for use among racial/ethnic minority children ages 8 to 17 years of age.²³ In addition, it is a valid estimate of beverage intake among children ages 7 to 9 years,²⁴ and has been used to measure sugar-sweetened beverage intake in children as young as 3 to 5 years old.²⁸

Table 1.

Definitions and Selected Cut-points of Children’s Diet and Physical Activity Behaviors (Dependent Variables) Assessed in Bivariate and Logistic Regression Analyses

Children’s Weight-related Behaviors (Dependent Variables)	Instrument Unit of Analysis	Selected Cut-point for Analyses	Definition Used in Our Study Sample
Fruit Intake	Block 2004 Food Frequency Questionnaire, Spanish Version (NutritionQuest; Berkeley, CA) Cup equivalents per 1000 kcal per day (USDA MyPyramid Equivalents Database version 2.0)	< 1 cup equivalent per 1000 kcal/day versus ≥ 1 cup equivalent per 1000 kcal/day	Bananas; apples or pears; oranges or tangerines (excludes juices); strawberries or other berries; applesauce, fruit cocktail or pineapple slices; jelly or jam; any other fruit like grapes, peaches, watermelon, cantaloupe, or fruit roll-ups; real orange juice (excludes Sunkist or other orange sodas); any other real fruit juices like apple juice or grape juice (includes juice boxes), additional mixed foodsa
Vegetable Intake	Block 2004 Food Frequency Questionnaire, Spanish Version (NutritionQuest; Berkeley, CA) Cup equivalents per 1000 kcal per day (USDA MyPyramid Equivalents Database version 2.0)a	< 1 cup equivalent per 1000 kcal/day versus ≥ 1 cup equivalent per 1000 kcal/day	Salad with lettuce, green salad; avocado, guacamole; green beans, string beans, or peas; pinto, chili with beans, or bean burrito; refried beans; spaghetti, ravioli, or lasagna with tomato sauce; vegetable soup, vegetable beef soup, or tomato soup; any other soup or stew; greens like collards, mustard greens or spinach; broccoli; carrots, carrot sticks, or cooked carrots; French fries, Tater Tots, hash browns or home fries; any other kind of potatoes, like mashed, baked, or boiled; sweet potatoes, sweet potato, or pumpkin pie; any other vegetables like squash, cauliflower, asparagus, nopales; ketchup, salsa, or barbecue sauce, additional mixed foodsb Spanish added foods: tamales or tamale pie; cooked green chile peppers; cooked green peppers, chile rellenos, or green chile stew; other starchy vegetables like yucca or plantains
Sugar-sweetened Beverage (SSB) Intake	Block 2004 Food Frequency Questionnaire, Spanish Version (NutritionQuest; Berkeley, CA) Percentage of total energy intake, kcal/day, from SSBc	SSB intake < 5% kcal/day versus ≥ 5% kcal/day	Sodas like Coke, Dr. Pepper, 7-Up, Sprite, Sunkist, Orange Crush (excludes diet); slurpees, snow cones, popsicles (not ice cream); Hawaiian Punch™, Kool-Aid™, Sunny Delight™, Gatorade™, iced tea, Snapple™; Hi-C™, Tang, Tampico™, Mr. Juicy™, Ssips punch ™
Sedentary Time	Actigraph GT3X Accelerometer (Pensacola, FL) minutes/day	< 8 hours/day versus ≥ 8 hours/day	The amount of time spent in periods of activity with counts that corresponded to <1.00 METs (or <100 counts/minute; also equivalent to <1 MET in our sample). Sedentary time measures all daily activities that do not increase energy above the resting level, such as sitting, laying down, and standing still. Time spent sleeping was not measured. Children were instructed to remove the monitor when they went to bed at night, bathed, or swam.
Moderate-to-Vigorous Physical Activity (MVPA)	Actigraph GT3X Accelerometer (Pensacola, FL) minutes/day	< 60 minutes/day versus ≥ 60 minutes/day	The amount of time spent in periods of activity with counts that corresponded to ≥4.00 METs. Examples of MVPA include brisk walking, bicycling, aerobics, and playing sports.

Note.

Abbreviations: kcal, kilocalorie; METs, metabolic equivalents; USDA, United States Department of Agriculture

^a:The USDA MyPyramid Equivalents Database²⁶ version 2.0 defines 1 cup equivalent as 1 cup of raw, cooked, or canned fruit or vegetables; 2 cups of raw leafy green vegetables; 1 cup of 100% fruit vegetable juice; ½ cup of dried fruit or vegetables; or 4 ounces of dry beans and peas.

^b:Mixed foods are disaggregated into individual ingredients and assigned into appropriate MyPyramid food groups.²⁷ For example, apple pie is disaggregated before food group assignments are made. Thus, certain mixed foods are included in the fruit and vegetable groups but contribute negligible amounts to the calculations and are not considered major food group contributors.

^c:Sugar-sweetened beverage energy intake is estimated using the following standard serving sizes: 12 ounces of soda; 5 ounces of Slurpee, snow cones, popsicles (not ice cream); 8 ounces of Hawaiian Punch™, Kool-Aid™, Sunny Delight™, Gatorade™, iced tea, Snapple™; Hi-C™, Tang, Tampico™, Mr. Juicy™, Ssips punch™

The Block Kids FFQ,²⁵ available in English or Spanish languages, was obtained during an in-person interview at baseline conducted by trained, bilingual (English/Spanish) research staff with children and parental assistance, as needed for clarification. The Block Kids FFQ²⁵ Spanish version consists of 84 food/beverage line items and takes approximately 25 minutes to complete. Instructions on the Block Kids FFQ asked children (and assisting parents), “How many days last week?” did they eat a specific food item(s), and to include foods eaten at home, at school, from fast food, or from a restaurant. Responses were recorded using a 6-point frequency of intake scale: none, 1 day, 2 days, 3–4 days, 5–6 days, or every day. Portion sizes for each item were determined using a handout with pictures of serving sizes to enhance the accuracy of quantification.

Block Kids FFQs²⁵ were mailed to Nutrition-Quest (Berkeley, CA) for data processing and analyses. Data output provided daily average intake of food groups based upon the USDA MyPyramid Equivalents Database (MPED) version 2.0.²⁶ Table 1 provides definitions of children’s dietary intake measures and selected cut-points for data coding in bivariate and logistic regression analyses. We assessed fruit and vegetable intake, respectively, in cup equivalents, which is consistent with the 2015 Dietary Guidelines for Americans (DGAs).²⁷ Both fruit and vegetable intake were dichotomized for analyses as < 1 cup equivalent per 1000 kcal/day versus ≥ 1 cup equivalent per 1000 kcal/day, based on the 2015 DGAs²⁷ intake recommendations for children. We assessed SSB intake by the total number of kilocalories (kcal) per day from SSB, calculated as (kcal/day from SSB)/(total kcal/day). This method adjusts for differences in total energy intake to facilitate comparisons across sex and age groups, and has been used to estimate national trends in SSB intake.²⁸ SSB intake was dichotomized for analyses as < 5% kcal/day versus ≥ 5% kcal/day. This cut-point was selected based on DGA²⁷ recommendations to limit added sugar and solids fats be to 5%−15% of total daily calories.²⁷ Most added sugars in the US diet are from sugar-sweetened beverages (soda, energy drinks, sports drinks, and fruit drinks).²⁸ Children who did not complete the Block Kids FFQ (N = 8) and children with implausible reported energy intakes of ≤ 500 kcal/day were excluded (N = 16) from analyses, limiting the sample size for diet-related outcomes to 277 parent-child dyads.

Children’s physical activity

Children’s sedentary time and moderate-vigorous physical activity (MVPA) were objectively measured using the Actigraph GT3X accelerometer (ActiGraph, Pensacola, FL). Accelerometers are activity monitors that capture the intensity of an individual’s daily physical activities in units of measurement called counts. ActiGraph accelerometers have established validity for use among children ages 6 to 16 years old.^29–32

According to standardized procedures,³³ children were instructed to wear the accelerometer on their hips at waistline for 5 days, excluding during bedtime, bathing, and swimming. To be included in this analysis, children must have had 10 hours of valid wear time for 3 or more of the 5 days. Non-wear periods were defined as 60 minutes of consecutive zero counts with an allowance for up to 2 minutes of nonzero counts.^34,35 Data from the accelerometer were downloaded, processed and screened for wear time.

Epoch lengths (sampling intervals) of 60 seconds were used to determine activity cut-point values.³⁶ The GT3X device has an inbuilt inclinometer that detects standing, lying, sitting and “off.” The ActiGraph algorithm classifies counts >100 counts min−1 as standing. If the counts are below 100, the data from the axes is used to classify sitting or lying or “off.” Thus, this analysis defined sedentary time as <100 counts min−1.

Whereas several definitions of sedentary time have been proposed, there is no standard definition of the term. It is recommended to use cut-points that are consistent with unit of analysis (eg, epoch length), instrumentation.³⁷ Our analysis defined sedentary time using < 100 counts min-1 because threshold has been calibrated³⁰ and validated^29,38 to measure sedentary time and physical activity with the ActiGraph accelerometer in young children (ages ≥ 3 years old), and is consistent with published literature using the ActiGraph GT3X.^19,34

A MET is an estimate of relative intensity such that 1 MET represents the energy expenditure for an individual at rest; whereas a 5 MET activity requires 5 times that amount. Average time (minutes/day) spent in MVPA was calculated by using an age- and sex- specific energy expenditure prediction equation, developed by Freedson et al³⁹ and used by others^31,33,40 for count cut-points that correspond to levels of PA measured in metabolic equivalents (METS): 2.757 + [0.00153 × counts/minute] – [0.0896 × age (years)] – [0.000038 × counts/minute × age (years)]. Table 1 provides definitions of children’s sedentary time and MVPA, and selected cut-points for data coding in bivariate and logistic regression analyses. Sedentary time was dichotomized for analyses as < 8 hours/day versus ≥ 8 hours/day, based on the average time that American youth spend engaged in sedentary time, which includes hours during and outside of school. Specific recommendations for total time spent in sedentary time are not available.^41,42 MVPA was dichotomized for analyses as < 60 minutes/day versus ≥ 60 minutes/day, based on the DGAs²⁷ and American Academy of Pediatric recommendations^4,41,42 for children to perform 60 minutes or more of physical activity per day, of which most of the minutes should be either moderate- or vigorous-intensity.

The biggest barrier to accelerometer data collection was children forgetting to put on the accelerometer, resulting in non-wear periods. We excluded 88 (29.2%) children from this analysis due to too few (< 3) valid accelerometer wear days. Compared to children with ≥ 3 valid wear days (data not shown), these children had a greater proportion of parents that obtained more than a high school education (34.1% vs 23.0%; p = .05). No other statistically significant differences in child/parent characteristics, HE, or anthropometric measures assessed in this study were observed between the 2 groups. This limited the sample size when sedentary time or MVPA were dependent variables to 213 parent-child dyads.

Independent Variables

Sociodemographic and home environment surveys (available in English and Spanish) were administered face-to-face by trained, bilingual (English/Spanish) research staff. Parental and child sociodemographic and home environment questionnaires were answered by the parent. Additional details of the HE measures assessed in this analysis have been published elsewhere.²⁰

Anthropometric

Parents’ and children’s standing heights and weights were objectively measured by trained research staff in a private patient examination room using a stadiometer and digital scale, following standardized procedures.^43,44 Participants wore light clothing without shoes and emptied their pockets. For children, BMI percentiles and z-scores were calculated using the SAS Program for the 2000 CDC Growth Charts (ages 0 to < 20 years),⁴⁵ and categorized as overweight (BMI 85^th to 94.9^th % for age and sex) or obese (≥95^th % for age and sex). Parents’ BMI was calculated using the formula BMI = (Weight, kilograms)/(Height, meters²) and categorized as normal/overweight (BMI 18.5 < BMI < 30 ) or obese (BMI ≥ 30.0).⁴⁶ One parent was underweight (BMI = 16.9) and categorized as normal/overweight for analyses.

Sociodemographics

Due to a large proportion of Hispanics/Latino children (74.8%) in our population, race and ethnicity were collapsed: Hispanic/Latino, Non-Hispanic black, and other. Child characteristics included age, sex, BMI percentile, weight category, and race/ethnicity. Familial characteristics included parent age, sex, BMI, weight category, and proportion of years spent living in the US (calculated as [(years living in the US)/(age in years]).⁴⁷ Additional sociodemographic details of the study population have been published elsewhere.²⁰

Parenting practices

The Parenting Practices for Eating and Activity Scale (PEAS) is a 26-item instrument that was developed and validated by Larios et al^48,49 among Latina/Mexican-American mothers of elementary-aged children to assess parenting strategies related to children’s dietary and activity-related behaviors. The PEAS has demonstrated good internal consistency in additional populations.^50–52 To examine the factor structure of the PEAS instrument in our population, we conducted exploratory principal component analysis. Methods and results of our PCA analysis have been published previously.²⁰ Six factors were revealed: monitoring, discipline, limit setting of soda/snacks, limit setting of sedentary behavior, pressure to eat, and reinforcement. Parents rated each item (eg, “My child should always eat all the food on his/her plate”) using a 5-point Likert-type scale with responses ranging from 1 = never to 5 = always or 1 = strongly disagree to 5 = strongly, as appropriate. Mean construct scores (possible range of 1 to 5) were calculated for each factor, demonstrating good internal consistency in our sample with Cronbach’s alpha ranging from 0.67 to 0.87. Higher mean construct scores indicate higher engagement in the parenting practice.

Family meal habits

The frequency of family meals and frequency of watching TV during family meals, respectively, were assessed using 2-items:^53,54 “How many times does your family sit down together for dinner?” and “How many times does your family have meals in front of the TV?” Response scales were never, once per month, 2–3 times per month, 1–3 times per week, and 4+ times per week. Response options were collapsed into more frequent ≥ 1–3 times per week) and less frequent (≤ 2–3 times/month).

Home fruit/vegetable and SSB availability

Parents were asked to answer the following items to indicate the frequency of fruit and vegetable availability in the home and at dinner: (1) How often would you say fruits and vegetables are available in your home? (2) How often are vegetables served at dinner? (3) How often is fruit served at dinner?”^55,56 Parents were asked to answer the following items to indicate the frequency of sugar-sweetened beverage (SSB) availability in the home and at meals: (1) How often is juice (like apple or orange) served at meals in your home? (2) How often are other drinks (like iced tea, lemonade, fruit punch, Kool-Aid™, Capri Sun™, Sunny Delight™, Snapple™, Gatorade™, Vitamin Water™) served at meals in your home? (3) How often is regular soda served at meals in your home?” These items have been previously reported in the literature.^55–59 Response scales were always, usually, sometimes, or never. Response options were collapsed into always/usually and sometimes/never.

Home screen-media availability

The number of TV’s in the home and the presence of a TV in the child’s bedroom were assessed by parent-response to 2-items:^58,60,61 “How many TVs do you have at home?” (numeric response) and (2) “Is there a TV in your child’s bedroom?” (yes/no).

Home physical activity resource availability

Availability of active video games and sports equipment in the home, respectively, were assessed by parent-response to 2 items:^58,62,63 “Now, we would like to know if your child or somebody else at home has: (1) Active video games (like Dance Dance Revolution™, Wii™ etc)?” and (2) “Sports equipment (balls, rackets, bats, sticks)?” Response options were yes/no.

Data Analysis

Descriptive statistics are presented as N (%) for categorical variables, mean ± standard deviation (SD) for normally distributed continuous variables, or median (interquartile range, IQR) for skewed continuous variables (Supplementary Tables 1 and 2). The distributions of variables were assessed for normality visually by histograms and quantitatively with tests of skewness and kurtosis. The distributions of the 5 dependent variables of interest (children’s weight-related behaviors: fruit intake, vegetable intake, sugar-sweetened beverage intake, sedentary time, and MVPA) were non-normally distributed. Because this was an exploratory analysis, we first conducted non-parametric univariate analysis using Wilcoxon Rank-Sum test or chi-square test to identify independent variables (home environment components) that were associated with children’s weight-related behaviors (dependent variables). Any variable with a p < .25 in bivariate analysis was subsequently assessed in a multiple logistic regression model.

Logistic regression was used because relationships between independent and dependent variables were not linear, as determined by analysis of standardized residuals from linear regression analyses (data not shown). Respective multiple logistic regression models were constructed for each of the 5 children’s weight-related behaviors (dependent variables). Based on clinical judgement and previously published data from the FWM study,^19,20 it was decided a priori to adjust models for the following potential confounders: child sex (10–12 vs 7–9 years old), gender (male vs female), ethnicity (Hispanic/Latino vs non-Hispanic/Latino), and BMI z-scores; and parent age, sex, BMI, and proportion of parent years spent living in the US (calculated as = (parent time spent living in the US in years)/(parent age in years). Using backward stepwise elimination procedures, the least statistically significant variables (p > .05) were removed, one-by-one, until there were no additional variables, other than a priori adjustment factors, that were non-significant. Each removed variable was added back to the model, one-by-one, to assess for confounding. The importance of each home environment component in the model was assessed using the likelihood ratio test. Assumptions of the models were tested using methods consistent with logistic regression (Hosmer and Lemeshow Good-ness-of-Fit Test, influential covariate patterns, and multicollinearity). A p-value ≤ .05 was considered statistically significant. Data were analyzed using STATA (version 13.1, 2014, StataCorp LP, College Station, TX).

RESULTS

Characteristics of the Sample and Children’s Diet and Physical Activity Behaviors

Median child age was 10 (8.4, 11.4) years. Children were predominantly Hispanic/Latino (74.8%) and obese (76.4%) (Table 2). Median (IQR) reported energy intake on the Block Kids FFQ was 1040.7 (768, 1452) kcal/day (Table 3). About one-half of children consumed ≥ 1 cup fruit equivalents/1000 kcal/day, whereas less than one-third met the recommended intake of ≥ 1 cup vegetable equivalents/1000 kcal/day. The majority (61%) of children consumed < 5% of total energy intake from SSBs. Less than one-half (45.5%) achieved ≥ 60 minutes/day of MVPA. Most (56.8%) of children engaged in < 8 hours/day of sedentary time. Median parent age was 36.0 (31.0, 42.0) years (Table 2), and most were female (92.4%), obese (62.5%), and lived in the US for ≤ 67% of their lives. Additional characteristics of the sample population and descriptive statistics of home environment measures are described in detail elsewhere.^14,19,20

Table 2.

Characteristics of the Sample Population (N = 301 Parent/Child Dyads)

Variable	Summary Statistic
Child Characteristics
Age (years), median (IQR)	10 (8.4, 11.4)
Age category (years),a N(%)
7–9	153 (50.8)
10–12	148 (49.2)
Sex, N(%)
Female	162 (53.8)
Male	139 (46.2)
BMI percentile,b median(IQR)	97.8 (95.3, 98.9)
Weight Category, N(%)
Overweight (BMI ≥85th to <95th %)	71 (23.6)
Obese (BMI ≥95th percentile)	230 (76.4)
Race/ethnicity, N(%)
Hispanic or Latino	225 (74.8)
Non-Hispanic Black	55 (18.2)
Othera	21 (7.0)
Parent Characteristics
Age (years),b median(IQR)	36.0 (31.0, 42.0)
Age category (years),b N(%)
22–36	153 (51.3)
37–67	145 (48.7)
Sex, N(%)
Female	278 (92.4)
Male	23 (7.6)
BMIb median (IQR)	32.7 (7.0)
Weight Category,c N(%)
Normal/Overweight (18.5 < BMI < 30)	112 (29.1)
Obese (BMI ≥ 30.0)	187 (62.5)
Proportion years living in the USb,c)
≤ 0.33	86 (28.9)
0.34 – 0.67	127 (42.6)
> 0.67	85 (28.5)

Note.

Abbreviations: BMI, body mass index; IQR, interquartile range; US, United States of America

^a:Child other race/ethnicity (N) = Asian (N = 6); Caucasian or white (N = 5); Pakistani (N = 2); Sri Lankan (N = 1); Bangladeshi (N = 1); East Indian (N = 1); Bengali (N = 1); Egyptian (N = 1); Arabic, unspecified (N = 1); Dominican/Puerto Rican, not self-identified as Hispanic/Latino or black (N = 1); Afro-Caribbean (N = 1)

^b:For parent age, N = 298 parents/caregivers (3 parents/caregivers did not report birth dates).

^c:Proportion of parent years living in the United States = (parent time spent living in the US in years)/(parent age in years)

Table 3.

Summary Statistics of Children’s Weight-related Behaviors

Weight-related Behaviors	Summary Statistic
Dietary Intake (N = 277)	Median (IQR) or N(%)
Total kilocalories/day	1040.7 (768, 1452)
Fruits, cup equivalents/1000 kcal/day	1 (0.6, 1.7)
≥ 1 cup fruit equivalents/1000 kcal/day	144 (52.0%)
Vegetables, cup equivalents/1000 kcal/day	0.7 (0.4, 1.0)
≥ 1 cup vegetable equivalents/1000 kcal/day	75 (27%)
Sugar-sweetened beverages, % kcal/day	3.2 (0, 7.5)
< 5% kcal/day	169 (61%)
Physical Activity (N = 213)
Sedentary time	458.1 (376.5, 538.2)
< 8 hours/day	121 (56.8%)
Moderate-vigorous physical activity	56 (34.9, 88.1)
≥ 60 minutes/day	97 (45.5%)

Note.

Abbreviations: IQR, interquartile range; kcal, kilocalorie

Associations between the Home Environment and Children’s Dietary and Physical Activity Behaviors

Children’s fruit intake

In bivariate analysis (Supplementary Table 1), children’s fruit intake was associated at the p < .25 level with the variables limit setting of sedentary behavior (p = .13), frequency of family meals (p = .003), frequency of TV watching during family meals (p = .02), and TV in child’s bedroom (p = .14). These variables were evaluated in the backward stepwise logistic regression procedure. Results of the backward stepwise logistic regression (Table 4, Outcome 1) indicated that family meal frequency and TV watching during family meals were independently associated with children’s fruit intake in a final multivariable model (Hosmer and Lemeshow Goodness-of-Fit test p = .84). Children were more likely to have fruit intake ≥ 1 cup equivalent/1000 kcal/day versus < 1 cup equivalent/1000 kcal/day when the family reported frequent family meals (OR = 1.84; 95% CI 1.0, 3.39; p = .05) compared to a family who does not have frequent family meals. Children were less likely to have fruit intake ≥ 1 cup equivalent/1000 kcal/day versus < 1 cup equivalent/1000 kcal/day when parents reported more frequent TV watching during family meals (OR = 0.59; 95% CI 0.36, 0.98; p = .04) compared to parents who reported less frequent TV watching during family meals.

Table 4.

Multivariable Logistic Regression Analysis of Associations between Home Environment Measures (Independent Variables) and Children’s Health Behaviors (Dependent Variables)^a

	Children’s Health Behaviors
	Dietary Intake			Physical Activity
	Outcome 1	Outcome 2	Outcome 3	Outcome 4	Outcome 5
Home Environment Measure	Fruit Intakeb	Vegetable Intakec	SSB Intaked	Sedentary Timee	Moderate-Vigorous PAf
	(N = 271)	(N = 271)	(N = 271)	(N = 209)	(N = 209)
	Odds Ratio (95% Confidence Interval)
Parenting Practices
Monitoring	--	--	0.75 (0.55, 1.04)ǂ	1.95 (1.36, 2.77)**	--
Limit setting soda/snacks	--	--	0.78 (0.58, 1.06)	--	--
Family Meal Habits
Family meals (more frequent)e	1.84 (1.0, 3.39)*	--	--	--	--
TV during meals (more frequent)e	0.59 (0.36, 0.98)*	--	--	2.17 (1.16, 4.0)*	--
Food Availability
Availability FVs (high)g	--	1.90 (1.0, 3.6)*	--	--	--
Availability SSBs (high)g	--	0.44 (0.20, 0.94)*		--	2.42 (0.96, 6.0)ǂ
PA Resource Availability
Own active video games (yes)h	--	--	--	--	0.39 (0.18, 0.84)*

^ǂp < .10

^*p ≤ .05

^**p < .001

Note.

Abbreviations: FVs, fruits and vegetables; SSB, sugar-sweetened beverages; SA, sedentary activity; PA, physical activity; TV, television

-- Variable not included in final model because it did not contribute to overall significance or confound associations between the independent and dependent variables.

^a:Each column represents one of the 5 dependent variables, respectively. All models included the following covariates: child age (10–12 vs 7–9 years old), gender (male vs female), race/ethnicity (other and non-Hispanic black vs Hispanic/Latino), and BMI percentile category (≥ 95^th vs < 95^th); and parent age (> 36 years vs ≤ 36 years), sex (male vs female), BMI category (≥ 30 vs < 30), and proportion of years spent living in the US (> 0.67 and 0.34–0.67 vs ≤ 0.33)

^b:Categorized as ≥ 1 cup equivalent per 1000 kcal/day versus <1 cup equivalent per 1000 kcal/day

^c:Categorized as ≥ 5.0% of total kilocalories from sugar-sweetened beverages/day versus < 5.0% of total kilocalories from sugar-sweetened beverages/day

^d:Categorized as ≥ 8 hours of sedentary time/day versus < 8 hours of sedentary time/day

^e:Categorized as ≥ 60 minutes of moderate-vigorous PA/day versus < 60 minutes of moderate-vigorous PA/day

^f:Categorized as More frequent (≥ 1–3 times per week) versus less frequent (≤ 2–3 times/month)

^g:Categorized as High (always/usually) versus low (sometimes/never)

^h:Categorized as Yes versus No

Children’s vegetable intake

In bivariate analysis (Supplementary Table 1), children’s vegetable intake was associated at the p < .25 level with the variables frequency of TV watching during family meals (p = .13), availability of FV in the home (p = .01), availability of SSB in the home (p = .02), TV in child’s bedroom (p = .15), and owning active video games (p = .22). These variables were evaluated in the backward stepwise logistic regression procedure. Results of backward stepwise logistic regression (Table 4, Outcome 2) indicated that FV availability and SSB availability were independently associated with children’s vegetable intake (Hosmer and Lemeshow Goodness-of-Fit test p = .35). Children were more likely to have vegetable intake ≥ 1 cup equivalent/1000 kcal/day versus < 1 cup equivalent/1000 kcal/day when parents reported high home FV availability (OR = 1.90; 95% CI 1.0, 3.6; p = .04) compared to parents who reported low home FV availability. Compared to parents with low home SSB availability, parents with high home SSB availability had children who were more likely to consume < 1 cup equivalent vegetables/1000 kcal/day compared to ≥ 1 cup equivalent vegetables/1000 kcal/day (OR = 0.44; 95% CI 0.20, 0.94; p = .03)

Children’s SSB intake

In bivariate analysis (Supplementary Table I), children’s SSB intake was associated at the p < .25 level with the variables monitoring (p = .003), discipline (p = .03), limit setting soda/snacks (p = .03), limit setting of sedentary behavior (p = .21), reinforcement (p = .21), frequency of TV watching during family meals (p = .02), availability of FV in the home (p = .02), availability of SSB in the home (p = .07) and owning active video games (p = .09). These variables were evaluated in the backward stepwise logistic regression procedure. Results of backward stepwise logistic regression (Table 4, Outcome 3) indicated that monitoring (OR = 0.75; 95% CI 0.55, 1.04; p = .08) and limit setting of soda/snacks (OR = 0.78; 95% CI 0.58, 1.06; p = .11) were not independently associated with children’s SSB Intake, but the relationship between children’s SSB intake and monitoring approached statistical significance (Hosmer and Lemeshow Goodness-of-Fit test p = .11) Table 4, Outcome 3). Monitoring confounded the relationship between SSB Intake and limit setting of soda/snacks (percent change in Beta coefficient > 10%), and therefore, was included in the final model. The interaction term for (monitoring) X (limit sitting of soda/snacks) was not statistically significant (data not shown).

Children’s sedentary time

In bivariate analysis (Supplementary Table 2), children’s sedentary time was associated at the p < .25 level with the variables monitoring (p = .003), limit setting soda/snacks (p = .11), control (p = .11), reinforcement (p = .15), frequency of TV watching during family meals (p = .03), availability of FV in the home (p = .14) and owning active video games (p = .05). These variables were evaluated in the backward stepwise logistic regression procedure. Results of backward stepwise logistic regression (Table 4, Outcome 4) indicated that parent monitoring and the frequency of watching TV during meals were independently associated with children’s sedentary time (Hosmer and Lemeshow Goodness of-Fit test p = .22). Children were more likely to engage in ≥ 8 hours/day versus < 8 hours/day of sedentary time when parents reported higher engagement in monitoring (OR = 1.95, 95% CI 1.36, 2.77; p = .001), and when parents reported frequently TV watching during family meals (OR = 2.17; 95% CI 1.16, 4.0; p = .02) compared to parents who reported less frequent TV watching during family meals.

Children’s moderate-vigorous activity

In bivariate analysis (Supplementary Table 2), children’s MVPA was associated at the p < .25 level with the variables limit setting of sedentary behaviors (p = .09), control (p = .03), availability of SSB in the home (p = .04), number of TVs at home (p = .14), and owning active video games (p = .04). These variables were evaluated in the backward stepwise logistic regression procedure. Results of the backward stepwise logistic regression (Table 4, Outcome 5) indicated that owning active video games was independently associated with children’s MVPA (Hosmer and Lemeshow Goodness-of-Fit test p = .84). Children were more likely to engage in ≥60 mins/day versus <60 mins/day of MVPA when parents reported owning active video games (OR = 0.39; 95% CI 0.18, 0.84; p = .01) compared to parents who did not report owning active video games. Additionally, the relationship between SSB availability and MVPA approached statistical significance (p = .06). SSB availability did not confound the relationship between MVPA and owning active video games, but it was kept in the final model because the likelihood ratio test indicated that that the model fit better with both SSB availability and owning active video games as independent variables (LR = 3.66; pr > LR = .05).

DISCUSSION

Our study identified several measures of the home environment that were independently associated with children’s diet and physical activity behaviors, including the parenting practice monitoring, frequency of family meals, frequency of watching TV during meals, availability of fruits and vegetables in the home, availability of sugar-sweetened beverages in the home, and owning active video games. The standard care recommendations for pediatric weight management encourage pediatricians in primary care settings to provide family-focused weight management counsel.^41,42 However, low-income familes commonly experience barriers to lifestyle changes, including time, financial costs, and physical environmental challenges (eg, lack of safe places to play). Despite these barriers, Cason-Wilkerson et al⁶⁴ found that engaging low-income children in a family-oriented obesity treatment program that teaches specific skills/strategies, such as increasing family mealtimes, can attenuate barriers to change and promote the adoption of healthful diet and physical activity behaviors. Thus, weight management counseling by PCPs should extend beyond BMI to address specific learning skills to enhance efficacy of the PCP-patient interaction.⁶⁵

In this analysis, higher frequency of family meals increased the odds of children consuming ≥ 1 cup equivalent fruit/1000 kcal/day, whereas higher frequency of watching TV during meals decreased these odds. In a cross-sectional analysis of Latino children and caregivers at baseline, Andaya et al⁵⁴ similarly observed positive effects of an increased frequency of family meals and decreased frequency of watching TV during family meals on the fruit and vegetable intakes of elementary-aged Latino children. However, we assessed fruit and vegetable intake separately, and found that food availability, rather than family meals, was associated with children’s vegetable intake. Parents who reported higher FV availability in the home, and lower availability of SSBs in the home, had children with increased odds of consuming ≥ 1 cup vegetables/1000kcal/day. In a cohort study among a predominantly white, middle class population in the Neighborhood Impact on Kids Study, Couch et al⁶⁶ similarly reported that children’s diet quality increased as healthful home food availability increased. Although slightly more than one-half of our sample achieved the recommended³⁰ fruit intake of ≥ 1 cup fruit equivalents/1000 kcal/day, few (27.0%) met this recommendation for vegetables. Corroborating myriad evidence that fruit and vegetable intake and availability/accessibility is an important target for pediatric weight management among low-income families.

In addition, we observed that the higher report of parent monitoring and limit setting of soda/snacks was associated, although not statistically significantly, with reduced odds of children consuming > 5% total energy from SSBs. The majority of our sample (61.0%) consumed within the recommended target of 5% or less of total energy intake from added sugars.⁶⁷ Given that sugar-sweetened beverage reduction is a major public health recommendation, this observation may be due to social desirability bias or the fact that our participants had enrolled in a family family weight management program and already changed their SSB intake.

We also found that parent monitoring was associated with increased odds of children engaging in ≥ 8 hours/day sedentary time. This is in contrast to the findings of Arredondo et al⁶⁸ who reported that monitoring was positively associated with physical activity levels in cross-sectional analysis of the Aventuras para Niños study. It is possible that our data reflects parent concern about their child’s weight^55,69 if they engaged in frequent sedentary behaviors, resulting in a greater need for monitoring. Whereas we did not measure concern in this study, future investigations should consider exploring this association. Watching TV during meals was also associated with increased odds of children’s sedentary time. Watching TV during meals has been previously associated with obesity,^70,71 but data relating it to sedentary time, as measured by accelerometry, is limited. Pediatricians’ promotion of more frequent family meals, with an emphasis on turning off the TV, may help children achieve recommendations of one hour of physical activity per day, with limits of ≤2 hours of screen time per day.^41,42

Interestingly, owning active video games (eg, Nintendo Wii™, Dance Dance Revolution™, etc) was associated with decreased odds of children engaging in 60 minutes/day of moderate-vigorous physical activity. Although several studies support positive benefits of active video games on children’s physical activity,^72,73 other studies have shown no effects.^74,75 One explanation may be that active video games are not played intensely enough to achieve moderate -vigorous levels of physical activity. Future studies should assess this association to help clarify the relationship.

Whereas our study has many strengths, several limitations should be noted. Our sample consisted of overweight/obese, predominantly Hispanic/Latino children in a safety-net primary care setting,¹⁴ and may not be representative of other populations. This was a cross-sectional analysis, and temporal relationships cannot be established. Recall and social desirability bias may affect parent-report of the home environment and child-report of the Block Kids FFQ. The more accurate 24-hour multiple pass recall method⁷⁶ was not used in our study due to costs and time restraints. Although the Block Kids FFQ is a widely used and validated instrument,^21,23,24 FFQs are susceptible to underreporting and social desirability.⁷⁷ To minimize underreporting, we excluded participants with ≤ 500 total reported kcals/day from this analysis. Whereas we used the Spanish version of the Block Kids FFQ, food items may not reflect the specific food preferences of Bronx families. For example, aguas frescas, a popular Mexican-American sugar-sweetened beverage, was not measured. Lastly, FFQs rely on children to remember foods from a previous time, introducing recall bias.

Moreover, our use of accelerometry to measure children’s physical activity is susceptible to underestimating activity spent engaged in upper body movements, cycling, and swimming.⁷⁸ A large number of children (about 30% our sample) did not adhere to accelerometry protocols due to children forgetting to put on the accelerometer, resulting in non-wear periods. This reduced the sample size for analysis. Future investigations to improve accelerometry adherence in our study population would be beneficial.

CONCLUSIONS

The home environment is an important component of pediatric weight management. Our findings elucidated several home environment components for PCPs to encourage parents to focus that may facilitate increasing fruit and vegetable intake and decreasing sedentary time to promote healthier body weight in children. More specifically, overweight or obese and predominantly Hispanic/Latino children appear to have healthier habits if their families eat meals together without watching TV and if healthy food choices are available in the home. Future studies are needed to investigate how changes in the home environment influence changes in children’s weight-related behaviors over time, and implications for long-term weight management.

Human Subjects Approval Statement

The Institutional Review Boards of Albert Einstein College of Medicine (FWA #00023382) and Tufts University approved all study procedures (IRB # 2005–582). Each participating parent/caregiver provided written informed consent, and children provided assent.