Participants were part of the Neighborhood Impact on Kids (NIK) Study, an NIH funded longitudinal, observational cohort study of children aged 6 to 11 and a parent in Seattle/King County, WA and San Diego County, CA [10, 11]. NIK was designed to evaluate the association of neighborhood and home environmental factors with children and parent’s weight status and weight-related behaviors. This study was approved by the Institutional Review Boards at Seattle Children’s Hospital and San Diego State University.
Participants were recruited September 2007 - September 2008 in San Diego and November 2007- January 2009 in King County. We attempted to contact a total of 8,616 households, of which 4,975 were screened for interest and eligibility, and 944 agreed to participate. Among families agreeing to participate, 730 consented and were enrolled. The final sample consisted of 713 child–parent pairs who completed the survey and had valid accelerometer data. Additional details regarding recruitment and inclusion/exclusion criteria have been previously published .
At a home or clinic visit, parents provided consent and children provided assent. The parent completed a survey (online or paper) that assessed, among other things, access to media and physical activity equipment at home, children’s sedentary behaviors, household rules and practices about physical activity and sedentary behavior, and sociodemographic information. The complete NIK survey is available at: http://www.seattlechildrens.org/research/child-health-behavior-and-development/saelens-lab/measures-and-protocols. Children and parents were instructed on having the child wear an Actigraph accelerometer for 7 days and were provided a log for recording when the accelerometer was worn. Study staff called participants several times within the week to answer questions and encourage daily wearing of the accelerometer.
The highest level of reported education of the parent(s) in the household and the household income were both used as SES indicators. The original 7 categories for educational attainment (ranging from < 7th grade to completed graduate/professional degree) and 11 categories for income (ranging from < $10,000 to > $100,000) on the survey were combined into 3 categories each for analyses according to the following a priori criteria: Education- low (≤completed high school), medium (completed college), high (completed graduate degree); income - low (≤$39,000), medium ($40,000-$89,000), high (≥$90,000). The Spearman’s rank correlation between household income and highest education in the household was 0.39. The 2008 median family income was $87,903 in Seattle/King County and $74,593 in San Diego County .
The physical home environment was assessed using survey items on the presence of electronic media in the child’s bedroom and access to fixed and portable equipment in and around the home that could be used for physical activity . A Bedroom Media Score was generated using 5 items from a reliable scale which asks if a TV, DVD/VCR, computer, video game system and/or hand held video game player are present in the child’s bedroom (prior test-retest reliability ICC = .51 - .96) . A Fixed Play Equipment Score was generated by summing yes/no items regarding presence of a basketball hoop, a swimming pool and/or a fixed swing set (prior test-retest reliability ICC = .53-.80) . A Portable Play Equipment Score was generated based on access to a bike, jump rope, sports equipment (balls, racquets) and/or roller skates(prior test-retest reliability ICC = .60 - .82) .
Other home environment measures included the presence of parental rules on outdoor play and on media use, and parent, sibling and friend participation in sedentary and physical activities with the child. The Safety Rules Score was the sum of “yes” responses by parents about whether they had the following rules: “Stay close/within sight of house/parent,” “do not go into street,” “do not ride bike on street.”(prior test-retest reliability ICC = .61-.74) . A Media Rules Score was generated by summing “yes” responses to the following 2 rules: “no TV before homework” and “<2 hours of TV per day” (prior test-retest ICCs of .57 and .73, respectively).
Screen time, sedentary time and physical activity
Parents reported their children’s “typical weekday time” spent watching TV/DVDS, playing video games and using the internet/other electronic media with response options of none, 15 min, 30 min, 1 hr, 2 hrs, 3 hrs, ≥4 hours per day (prior test-tetest reliability ICC = .66, .73, .72 respectively). Responses were summed to create a parent-reported child screen time value in average hours/day.
Child overall physical activity and sedentary behavior were measured by the GT1M Actigraph accelerometer (Pensacola, FL). The Actigraph has been validated and calibrated for use among children . Accelerometer data were collected in 30 second epochs. Participants were asked to wear the accelerometer for seven days during all waking hours. Upon return, the Actigraph was immediately downloaded and screened for completeness and irregularities/malfunction. A valid day was defined as having at least 10 valid hours of wearing time; and a valid hour contained no more than 20 minutes of consecutive zero counts. Data were included for children with at least 3 valid days. Data were converted to minutes engaged in sedentary behavior (≤ 100 counts per minute) and moderate-to-vigorous intensity physical activity (MVPA; ≥3 metabolic equivalents (METs)) using Freedson age-specific cut-points with the participant’s age rounded to half a year . Data were also examined using Evenson (4 METs) cut-points given a recent study which found that these gave the best classification accuracy for all four levels of physical activity intensity and performed well among children of all ages [17, 18]. Accelerometer data were cleaned and scored using MeterPlus version 4.0 (Santech, Inc., http://www.meterplussoftware.com).
As participants had been instructed to remove the accelerometers overnight, all data files were screened for non-zero counts between the hours of 11 pm and 6 am. In all, 93 participants were identified as having overnight activity during valid days. Participants' accelerometer log data were triangulated with the activity counts to determine sleep hours. In 92% of cases, the log-reported sleep start time corresponded with a significant drop in activity counts (below 1000), and the reported wake time corresponded to an increase in activity counts (above 1000) exactly or within 1/2-1 hour. In cases with the slight discrepancies between the log-reported times and activity counts, we relied more on the meter data and assigned the exact sleep and wake up times based on changes in activity counts. In 7 cases (including 3 sleepovers), there were discrepancies > 1 hour between the log and activity counts in the accelerometer data. In these cases, the data were reviewed by at least 2 individuals to arrive at an agreement about sleep hours based on activity counts (increase above 1000 or decrease below 1000). In 6 cases there was no log available and the following criteria were used: “asleep time” when counts dropped below 1000 for at least 3 consecutive 30 min blocks (1.5 hours) and “awake time” when counts increased above 1000 for at least 3 consecutive 30 min blocks (1.5 hours). Estimated sleep hours were converted to “non-wearing time” to prevent an overestimation of sedentary time due to the inclusion of overnight time.
Average accelerometer wear time for the whole sample was 5688 minutes. The differences in wear time across SES groups were not statistically significant. Wear times (in minutes) across income groups: Low: 5787, Med: 5750, High: 5640; across educational attainment groups: Low: 5772, Med:5675, High: 5622. Forty-seven percent of children’s accelerometer wearing time was spent at home.
Parents completed a place log of where their child went while wearing the accelerometer. Place categories were created to assess where children were while wearing the accelerometer. Accelerometer data were matched by day and time to the place log. From this, non-wear, sedentary, light, moderate, hard, and very hard accelerometer wear times were aggregated within the given timeframe of each location. For the purpose of the current study, the “Home” category included one single location for each participant (i.e. each child had only one address designated as home). If parent listed ‘front yard’ or ‘backyard’ in the place log, this was also considered home. “Home” did not include other parent/guardians’ homes or homes of relatives, friends or neighbors.
Children’s home physical activity and sedentary environments and their total and home-based activity levels were compared across different education and household income groups using chi-square test for categorical variables and linear regression for continuous variables. Parent’s age, marital status and ethnicity were included as covariates in the regression models as they differed across categories of income and education.
Home environment variables, which were found to vary in a statistically significant manner across SES, were selected for further analysis using the Sobel-Goodman test which tests if the indirect relationship between the independent and dependent variable through the mediator is significantly different from zero. Mediator analyses were conducted to examine the role of media in the bedroom, access to portable play equipment, rules around media and safety and parent screen time with their child as potential mediators in the statistically significant relationship between SES and screen time. For the mediation analyses, the original 11 categories of income and 7 categories of educational attainment were used (instead of the tertiles) in an effort to retain the most information. All analyses were conducted using STATA software version 10.1.