Leisure Time Computer Use and Overweight Development in Young Adults – A Prospective Study

Sara Thomée; Lauren Lissner; Mats Hagberg; Anna Grimby-Ekman

Disclosures

BMC Public Health. 2015;15(839) 

In This Article

Methods

The study design was a prospective cohort study of young adults, with measurements at baseline and after 1 and 5 years. The study was approved by the Regional Ethics Review Board in Gothenburg, Sweden (Registry numbers 191–05 and 876–11).

Study Population and Data Collection

A cohort of Swedish young adults aged 20–24 years was recruited in 2007. Twenty thousand persons who were born in the years 1983–1987, 50 % of each sex, were randomly selected from the registry of the total Swedish population (held by the Swedish Tax Agency), and were sent a survey questionnaire containing questions about health, work- and leisure-related exposure factors, demographic factors and lifestyle factors.[22] The questionnaire could be returned by post or completed online, if desired. As an incentive to respond, a lottery ticket (value approx. 1 Euro) was attached to the cover letter and could be used regardless of participation in the study. After two reminders, there were 7125 respondents (36 %). Only those with data on self-reported height and weight, i.e. with data to calculate body mass index (BMI), were included in the current study. An additional four individuals were excluded due to implausible values in height, leaving 2662 men and 4073 women at baseline (Fig. 1). After twelve months, those who had agreed to be invited in future studies (n = 5433) were asked by post to respond to an identical web-questionnaire. The data collection process was similar to that at baseline, including the initial lottery ticket, but with the addition of a third reminder offering a paper version of the questionnaire and two cinema tickets for participating. After 3 reminders, the response rate was 73 %, and after exclusion of 21 persons with missing BMI data, n = 3928 (1403 men and 2525 women). In 2012, a 5-year follow-up was conducted with an almost identical web-questionnaire. Of 4788 invited (by post and with the lottery ticket incentive), 55 % had responded after three reminders. After excluding three individuals with implausible values for weight and 18 with missing data on BMI, 2593 (957 men and 1636 women) remained in the study (Fig. 1). Of these, 488 were missing at the 1-year follow-up.

Figure 1.

Study participation process. m = men, w = women

Leisure Time Computer use

Two aspects of leisure time computer use were examined: computer gaming and emailing/chatting. Self-reported data was collected from the cohort study questionnaire, through the items: a) On average, how much time per day have you spent on computer gaming (e.g., PC games or online games)?, and b) On average, how much time per day have you used on a computer for emailing and chatting? The items concerned leisure time use the past 30 days. There were four response categories: 1 = None at all, 2 = <1 h per day, 3 = 12 h per day, and 4 = >2 h per day. The questions were identical at baseline and at the 1-year follow-up. In the 5-year follow-up questionnaire, the questions were rephrased to also include the use of mobile phones and tablets. In regression analysis of computer gaming, response category 1 was used as the reference. In analysis of emailing/chatting, response categories 1 and 2 were merged into one and used as the reference, due to the low number of respondents in the lowest category.

Demographic Variables

Demographic information was collected from the baseline questionnaire, including age, highest completed educational level (elementary school, upper secondary school, or college or university studies), and occupation: working, studying, or other (other included being on long-term sick leave, or on parental or other leave, or being unemployed). Age and occupation were treated as potential confounders, while educational level was considered to be too closely associated with age in this age group to be included in the models.

Lifestyle Factors

Self-reported leisure time physical activity, sleep duration, social support, and total daily computer use were collected from the baseline questionnaire. Level of leisure time physical activity was measured with a slightly modernized version[23] of the Grimby-Saltin Physical Activity Scale:[24,25] How much do you move and exert yourself physically during leisure time? If your activity varies greatly between, for example summer and winter, try to estimate an average. The question concerns the past 12 months. There were four response categories: 1 = physically inactive: 2 = light physical activity, 3 = regular physical activity, and 4 = vigorous physical training or competitive sporting, each containing definitions and examples of activities.[23] In analysis, response categories 3 and 4 were merged into one (regular or vigorous physical activity), and thus three categories were used in analyses; physically inactive, light physical activity and regular or vigorous physical activity. The instrument has been validated in relation to serum cholesterol, blood pressure and BMI.[23,25]

Sleep duration was measured by a question, constructed for the cohort study: How many hours do you usually sleep per night on… a) weekdays (work or study days)?, and b) weekends (days off)? Responses were given in whole hours and concerned the past 30 days. Average sleep duration was calculated ((weekday hours × 5 + weekend hours × 2)/7). Reported sleep values equal or less than 1 h and equal or more than 24 h were considered unreasonable and coded as missing.

The variable social support was based on the item: When I have problems in my private life I have access to support and help. The item had been constructed for the cohort study as a one-item adaptation of the social support dimension in the Karasek-Theorell model of demands-control-social support,[26] here relating to private life (rather than working life). Response categories were: 1 = applies very poorly; 2 = applies rather poorly; 3 = applies rather well; 4 = applies very well. The responses were categorized as low (response categories 1–2), medium (response category 3), and high (response category 4). The high category was used as the reference category.

One questionnaire item was constructed to concern total daily computer use: On average, how much time per day, have you used a computer? The question concerns total time, i.e., for work, study, and leisure, the past 30 days. The three response categories were 1 = <2 h per day, 2 = 24 h per day, and 3 = >4 h per day.

A test-retest reliability study and a validating interview study were done in the process of developing the questionnaire. The variables showed moderate to high correlations at test-retest, and the validity was considered acceptable.[22]

BMI Outcomes

BMI [kg/m2] was calculated based on self-reported height and weight in the cohort questionnaire. Height was included only in the baseline questionnaire, while weight was reported at all three time points. Four implausible heights at baseline were excluded, as well as three unreasonable values for weight (0, 1, 10 kg) at the 5-year follow-up. BMI mean as a continuous variable and BMI categories underweight <18.5, normal range 18.5 – < 25, overweight ≥ 25 – < 30, and obesity ≥ 30, are presented in the descriptive statistics. The binomial outcome variable Overweight was defined as BMI ≥ 25, i.e. including the BMI categories overweight and obesity, and with BMI < 25 as reference. The binomial outcome variable Obesity was defined as BMI ≥ 30, with BMI < 30 as reference. The variable Change in BMI was calculated as the difference between BMI at the 5-year follow-up and baseline.

Analysis

All analyses were performed using the SAS statistical package, version 9.3 (SAS Institute, Cary, NC, USA). Descriptive statistics such as means and standard deviations (SD) for continuous variables, and percentages for categorical variables, were calculated with the procedures PROC UNIVARIATE and PROC FREQ, respectively. Descriptive statistics are also shown for the highest category (>2 h per day) of computer gaming and emailing/chatting. Further, leisure time computer use and BMI are presented for the three time points among those who remained at the 5-year follow-up (Table 2).

For all descriptive variables at baseline, sex differences, and differences between the "high gamers" and the rest, as well as between the "high emailers/chatters" and the rest, were analyzed using chi square tests for categorical variables and t-tests for continuous variables. The same tests were used to compare the baseline variables of those who remained in the study at 5-year follow-up and those who did not.

Logistic regression model (PROC LOGISTIC) was used for cross-sectional and prospective analysis. Overweight (BMI ≥ 25, i.e. including obesity), and obesity (BMI ≥ 30), were outcomes in the cross-sectional analysis of the baseline data. In the prospective analysis, those with BMI ≥ 25 (overweight or obese) at baseline were excluded and the outcomes were new cases of overweight (BMI ≥ 25) at the 1- and 5-year follow-ups. Due to a low number of cases, no prospective analyses were done with obesity as a separate outcome. First, crude analyses were done with the two primary explanatory variables gaming and emailing/chatting. Potentially confounding demographic (age, occupation) and lifestyle variables (leisure time physical activity, social support, sleep duration), were adjusted for if p-values ≤0.20 in at least one sex in the univariate analysis. Model I included the demographic variables and in Model II lifestyle factors were added. An additional exploratory model was tested by adding total computer time. Only baseline values of the explanatory variables were used.

Spearman correlation analysis showed no collinearity (defined as r >0.7) among the explanatory variables, i.e., gaming, emailing/chatting, total computer use, occupation, age, physical activity, sleep, and social support (the highest correlation was found between emailing/chatting and total computer use; r=0.44).

Supplementary analysis was performed to check the possible influence of partially missing data in the crude and Model I analyses, using the complete cases of Model II.

Furthermore, change in BMI, i.e. the difference between BMI at baseline and at 5-year follow-up, was analyzed as the outcome in regression analyses (PROC GENMOD). The same model building as for the logistic regressions was used, with the exception that BMI at baseline was adjusted for in Model Ib.

All analyses were done for men and women separately.

processing....