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

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


BMC Public Health. 2015;15(839) 

In This Article



At baseline, there were 60 % women and 40 % men (Table 1). Sex differences were observed in all descriptive variables but age. More than half of the males (55 %) reported work as their main occupation, while among the women, an equal proportion worked (43 %) and studied (44 %). Nine and 13 %, respectively, were categorized as "other occupation", i.e. were unemployed, on sick leave, parental leave, or for other reasons not in work or studies. Mean BMI was 23.7 kg/m 2 in men, and 22.7 kg/m 2 in women, and the prevalence of overweight and obesity was 28 % in men and 21 % in women, which corresponds relatively well to the public health statistics from self-reported weights and heights of 16–29 year old Swedes in 2007.[27]

Computer gaming was more common among males; more than half of the men played computer games on a daily basis, while a clear majority (77 %) of the females did not. Twenty-five percent of the males spent 1 h or more on computer gaming compared to 6 % of the women. Using the computer for leisure time communication (emailing/chatting) was more common and similar in both sexes, with 30 % of both males and females spending an hour or more on this activity and only about 10 % reported no daily emailing/chatting. About half of the males and 40 % of the females reported that they exercised on a regular basis, while 17 % of the males and 13 % of the females reported being mainly physically inactive at leisure time. Mean reported sleep duration over the week was 7.7 h per night in males and 7.8 in females.

In Table 1, the study demographics are also shown stratified for the highest category (i.e. >2 h) of the leisure time computer variables, and statistical differences between the high groups and the rest can be seen in the table (p-values). The "high gamers" were less often employed or students, compared to the total group. One of four women gamers was unemployed, compared to 7 % in the total group of women, and they more than twice as often reported being on sick leave. Also, the high gaming men as well as the high emailing/chatting men, reported being unemployed about twice as often as the total group of men. The "high gamers" of both sexes less often reported regular physical activity and more often a predominantly physically inactive leisure. The "high gamers" were twice as likely to be "high email/chatters" compared to the total study group, and vice versa. Both "high" groups also reported a total daily computer use of >4 h per day more than twice as often. The high gaming women, and to some degree also the high emailing/chatting women, reported lower social support in private life than the total female study group. Also, among the high emailing/chatting men a slightly lower social support was reported (Table 1).

Furthermore, data on the exposure and outcome variables at the three time points are presented for those who remained in the study from baseline to the 5-year follow-up (n = 2593) (Table 2). While the proportion of high gaming males seemed to be steady over the 5-year follow-up period, there was an increase in computer gaming among the women. A decrease of non-gamers can be seen in both men and women. However, the questionnaire item was changed at 5-year follow-up to include also the use of smartphones and tablets, which makes the comparison formally inaccurate. The same applies to the emailing/chatting item. Keeping this in mind, the reported emailing/chatting increased in both groups over the follow-up period; at the 5-year follow-up there were only a few (4 and 1 %, respectively) who reported no daily emailing/chatting. Mean BMI increased about 1 unit over the five years of study in both sexes, and an increase in the prevalence of overweight and obesity can also be seen (Table 2).

Those who remained in the study at the 5-year follow-up were more likely to be female, students, to have higher educational level, and to report higher computer use and higher physical activity at baseline, compared to those who did not participate after 5 years. The males were also slightly older and had lower BMI at baseline, and the females were slightly more often gamers, compared to those who did not participate after 5 years. No statistical significant different drop-out rates were seen in emailing/chatting, social support or sleep duration at baseline.

Cross-sectional Associations Between Leisure Time Computer use and Overweight

There were cross-sectional associations between leisure time computer gaming and overweight (BMI ≥ 25) at baseline in both men and women (Table 3). For men, only the highest category (>2 h gaming per day) had a clear association with overweight in the crude analysis (OR 1.7), but after adjusting for demographic and lifestyle factors in Models I and II, also the next-to-highest category (1–2 h gaming per day) was associated with increased overweight (OR 1.4, Model II). For women, all the gaming categories were associated with overweight (ORs 1.6–2.2, Model II, Table 3).

Leisure time emailing and chatting >2 h per day was associated with overweight in the women (OR 1.4, Model II, Table 3), but not in the men. For the men, medium (1–2 h daily) emailing or chatting was actually associated with a lower prevalence of overweight in the crude analysis (OR 0.8). However, when adjusting for demographic and lifestyle factors the negative association was no longer statistically significant.

A supplementary complete case analysis was performed to check possible influence of partially missing data in the crude and Model 1 analyses. No major effects on results were seen except for the loss of statistical significance of medium emailing/chatting in the men in the crude analysis. All in all, the ORs changed in the range of −0.2 to +0.1, and there were slight changes in the confidence limits, which in most cases meant a widening of the CIs.

An additional, exploratory third model, which included total time spent on computer, was tested, and seemed to strengthen the existing associations over all (data not shown).

Cross-sectional Associations Between Leisure Time Computer use and Obesity

The patterns when cross-sectionally analyzing obesity (BMI ≥ 30) were similar to the above analyses of overweight. There were clear associations between computer gaming and obesity in both sexes (data not shown in table). For the men, the associations were amplified, compared to overweight being the outcome; the two highest categories of gaming (>2 h and 1–2 h) generated ORs 1.8 (CI 1.02–3.04) and 2.0 (CI 1.17–3.38) after adjusting for demographic and lifestyle factors. For the women, the three gaming categories (>2 h, 1–2 h, <1 h) generated ORs 2.1 (CI 1.02–4.47), 2.9 (CI 1.64–4.96), and 2.2 (CI 1.54–3.15) in Model II. No clear associations were found between leisure time emailing/chatting and obesity in either sex. For the men, high and medium emailing/chatting generated ORs of 0.7 (CI 0.34–1.36) and 0.8 (CI 0.49–1.41) and for the women 1.5 (CI 0.95–2.47) and 1.1 (CI 0.75–1.62) in Model II.

Prospective Associations Between Leisure Time Computer use and Overweight

In the prospective analyses, i.e. after excluding those with BMI ≥ 25 at baseline, high computer gaming (>2 h per day) at baseline meant higher odds for overweight at 1-year follow-up for the women (OR 3.2, Model II) (Table 4). Furthermore, the two highest categories of gaming (>2 h and 1–2 h) at baseline, were steadily associated with new cases of overweight at 5-year follow-up (ORs 3.0 and 2.7, Model II). The additional, third model which adjusted for total daily computer use, slightly amplified the associations. For the men, no statistically significant prospective associations were seen between computer gaming and overweight at either follow-up. Moreover, no statistically significant prospective associations between leisure time emailing and chatting and overweight were seen in either sex (Table 4). Complete case crude and Model I analyses did not change results notably. Due to a low number of cases, no prospective analyses were done with obesity (BMI ≥ 30) as a separate outcome.

Change in BMI From Baseline to 5-year Follow-up

Linear regressions with change in BMI from baseline to 5-year follow-up as the outcome showed a higher BMI-increase for all computer gaming categories in all the models for the women (Table 5). A dose–response relationship emerged, however, with overlapping CIs in most analyses. Belonging to the highest category of female gamers implied an additional BMI-increase of an estimated 1.33 BMI units (Model II), while even gaming less than 1 h per day implied an estimated increase of 0.51 BMI units (Model II). The estimated total increase in BMI from baseline to 5-year follow-up in the female group, i.e. when also taking the "natural" change in BMI into account, was 0.79 BMI-units in no-gamers, 1.30 in <1 h gamers, 1.66 in 1–2 h gamers, and 2.12 in >2 h gamers. No clear associations were seen between change in BMI and computer gaming for men or for emailing/chatting in either sex (Table 5). Complete case crude and Model I analyses did not change results notably.

Because adjusting for baseline BMI (Model 1b) showed an increase in the estimated change in BMI, we also split the dataset and performed analyses in the subsets of those with BMI ≥ 25 and those with BMI < 25. The increased change in BMI was only statistically significant in the females with BMI < 25 (data not shown).