Television viewing: Moderator or mediator of an adolescent physical activity intervention?

Abstract

Purpose

To determine whether amount of TV watched by participants enrolled in a physical activity intervention mediates or moderates program effectiveness

Design

Nine-month controlled school-based physical activity intervention

Setting

Public high school

Participants

One hundred twenty two sedentary adolescent females (mean age = 15.04 ± 0.79 years)

Intervention

Supervised in-class exercise, health education, and internet-based self-monitoring

Measures

Physical Activity - 3 Day Physical Activity Recall; Television Viewing – self-report; Cardiovascular Fitness – Cycle Ergometer

Analysis

T-tests were conducted to examine between-group differences. Linear regression equations tested the mediating and/or moderating role of television watching relative to the intervention.

Results

TV viewing moderated the intervention’s effect on vigorous activity; the intervention significantly predicted physical activity among high (β = −.45; p <.001), but not low (p >.05), TV watchers. TV viewing did not mediate the intervention effect.

Conclusions

Consistent with displacement theory, adolescents who watched more television prior to the intervention showed post-intervention increases in vigorous physical activity and concomitant decreases in television viewing, whereas those who watched less TV showed no change in physical activity or television viewing.

Purpose

Television (TV) has been implicated in the growing epidemic of childhood obesity in the United States.¹ Results of a nationally representative survey of 8- to 18-year-olds in the United States revealed that, on average, members of this demographic group use some form of media between six and seven hours daily, with TV accounting for approximately half of this total.² Since the mainstream introduction of TV, “displacement” theories have been proposed suggesting that time spent watching TV displaces time spent engaging in other pursuits.^3.4 Physical activity (PA) is one pursuit that may be displaced by TV viewing. Displacement of PA by TV viewing could harm public health by altering energy balance and promoting weight gain. The present study seeks to determine whether the amount of TV watched by participants in a PA intervention will mediate or moderate program effectiveness.

TV viewing could behave as a mediator if the intervention leads to decreased television viewing and thence to increased PA. Decreasing TV viewing was not an intervention goal; however, the displacement hypothesis implies that this mediation is a possibility worthy of investigation. Another possibility that will be investigated is that TV viewing might act as an intervention moderator. If TV viewing functions as a moderator, high TV viewers would show a different PA outcome following the intervention than low TV viewers. Based on the notions that a) TV viewing might displace some PA time and b) adolescents who spend a lot of time watching TV have a relatively larger reservoir of potentially active time, we hypothesize that participants who are more prolific TV consumers at baseline (i.e. those who have more available minutes to potentially reallocate from TV viewing to PA) will show a greater increase in PA following the intervention than those who watch less TV.

Methods

Design

A 9-month controlled school-based PA intervention was conducted at two Southern California high schools. Intervention details are presented elsewhere.⁵ The schools were similar in size, demographics, and academic achievement (i.e., the ethnic distributions, Academic Performance Indicator scores, and percent of students qualified for meal subsidies were very similar at the two schools). Both schools offered Physical Education five days per week for one class period. All tests were performed at a University-based General Clinical Research Center.

Sample

Participants were sedentary adolescent females (N = 122, mean age = 15.04 ±0.79 years) recruited via flyers at two public high schools. Sedentary was defined, as not meeting CDC physical activity criteria, (i.e. 3 vigorous or 5 moderate exercise sessions per week⁶). Sedentary status was confirmed by excluding volunteers who scored in the top 25% of the baseline cardiovascular fitness test. Parents/guardians provided informed consent and adolescents assented to the procedures. The study was approved by the Institutional Review Board at the University of California, Irvine. Adolescent females are of special concern because: a) they engage in less PA compared to boys and are therefore at greater risk for the negative health consequences of inactivity;⁷ b) 72% of high school girls do not meet current CDC PA recommendations;⁸ and c) during high school, females move from active to sedentary at a disproportionately high rate.⁹ Participants were assigned to intervention (N = 63) or comparison (N = 59) group based on the school attended. The participants’ average VO_{2 max} was 23.6 ± 4.4 ml/min/kg and average BMI was 23.0 ± 4.2. Table 1 includes participants’ ethnic distribution.

Table 1.

TV Viewing, Physical Activity, and Participant Ethnicity^*

	All Participants			Intervention Participants Only			Comparison Participants Only
	Low TV	High TV	Low and High	Low TV	High TV	Low and High	Low TV	High TV	Low and High
Sample size	62	57	119	25	35	60	37	22	59
TV viewing, h/d (SD)†
Baseline	0.98a (0.49)	3.20a,k (1.56)	2.04 (1.59)	1.11b (0.40)	3.10b,l (1.69)	2.27 (1.65)	0.89c (0.53)	3.35c (1.36)	1.81 (1.51)
9 months	1.19d (0.86)	2.37d,k (1.98)	1.76 (1.61)	1.31o (0.98)	2.26o,l (1.49)	1.86 (1.37)	1.10f (0.77)	2.56f (2.62)	1.65 (1.83)
Vigorous PA, METS/d (SD)‡
Baseline	0.59 (0.73)	0.57m (0.68)	0.58 (0.70)	0.63 (0.73)	0.71n (0.70)	0.68 (0.71)	0.57 (0.73)	0.35 (0.61 >	0.49 (0.69)
9 months	0.70 (0.60)	0.84m (0.55)	0.77 (0.58)	0.79 (0.57)	1.03g,n (0.39)	0.93h (0.48)	0.64 (0.62)	0.55g (0.63)	0.61h (0.62)
Ethnicity, No. (%)
Caucasian	35 (57)	32 (56)	67 (56)	18 (72)i	21 (60)	39 (65)	17 (46)i	11 (50)	28 (48)
Asian	10 (16)	11 (19)	21 (18)	3 (12)	7 (20)	10 (17)	7 (19)	4 (18)	11 (19)
Latina	13 (21)	10 (18)	23 (19)	3 (12)	4 (11)	7 (12)j	10 (27)	6 (27)	16 (27)j
Other	4 (6)	4 (7)	8 (7)	1 (4)	3 (9)	4 (7)	3 (8)	1 (5)	4 (7)

^*TV viewing and physical activity reported at baseline and at 9 months.

^† SD indicates standard deviation; PA, physical activity; METS, metabolic equivalents.

^‡ METS were log-transformed because of skewness in the distrbution.

^a–np < 0.05 for pairwise differences between corresponding lettered values.

Measures

Cardiovascular fitness was measured via cycle ergometer. A ramp-type progressive exercise test on the electronically braked ergometer was conducted to determine each participant’s peak oxygen consumption. Participants warmed up by cycling for 3 minutes with no resistance (i.e. 0W) after which they pedaled at a rate of 70 revolutions per minute while the power output increased progressively by 15W/minute. The test ended after 8–12 minutes, when the participant became too fatigued to pedal at 70rpm. The SensorMedics Vmax 229 metabolic cart (Yorba Linda, CA) measured peak oxygen consumption (VO_2peak in L/min) using a method designed for children and adolescents.¹⁰

BMI (weight/height²) was assessed using standard, calibrated scales and stadiometers.

Activity level was assessed via 3 Day PA Recall (3DPAR). Activities were converted into metabolic equivalents (METs) using the compendium published by Ainsworth and colleagues¹¹ and aggregated to calculate average daily minutes spent engaging in moderate (3–6 METs) and vigorous (>6 METs) activity.

Television viewing, assessed via self report, included hours spent watching both TV programming and videos. A median split divided the participants into high- and low-TV groups.

Intervention

Participants attended supervised exercise sessions 4 days per week throughout the school year (approximately 40 minutes of PA per session). Health education was provided during class on the fifth day. The supervised physical activities were determined, in part, by participant preferences and included both aerobic (three times per week, including aerobic dance, kickboxing, and brisk walking) and strength-building (one time per week, including weightlifting and yoga) activities. The intervention effectively increased average levels of PA and cardiovascular fitness among the participants.⁵ Participants at the comparison school were given no special instructions regarding physical activity; many participated in standard PE.

Analysis

Three participants were removed from data analyses due to outlying (>3 S.D. from the mean) data values on time spent watching TV or engaging in vigorous PA. Student’s independent-samples t-tests were used to test for participant differences (age, ethnicity, VO_{2 max}, BMI, TV viewing time, and amount of vigorous activity) between groups (high- and low-TV groups overall, intervention and comparison groups overall, and high- and low-TV groups within the intervention and comparison groups) at baseline. Regression analyses were conducted using SPSS (SPSS version 13.0, SPSS, Chicago, Illinois). In a hierarchical equation predicting post-intervention vigorous activity, baseline vigorous PA and ethnicity were entered on the first step. Group (intervention vs. comparison) and TV viewing (high vs. low) were added on the second step; the Group by TV viewing interaction was entered on the third step to determine whether the effect of the intervention was moderated by level of TV viewing. Baron and Kenny’s three step regression test for mediation¹² was also employed.

McClelland and Judd’s Statistical Difficulties of Detecting Interactions and Moderator Effects suggests adopting an alpha greater than .05 as one way researchers can “improve their chances of detecting interactions.”¹³ Accordingly, the possibility of moderation was examined for any interaction that trended towards conventional levels of significance at the p < .10 level.

Following the hierarchical regression test of moderation, separate regression models of post-intervention PA were fit for high- and low-TV groups. Baseline vigorous activity and ethnicity were again entered as control variables.

Post-hoc analyses were conducted to examine change over time in PA and TV viewing. Paired t-tests examined the change in vigorous activity and TV viewing over time within each of the four subgroups: high-TV intervention, low-TV intervention, high-TV comparison, and low-TV comparison.

Results

Group Comparisons

Table 1 presents the between group comparisons for ethnicity, TV viewing, and vigorous PA. Comparisons were also made for age, VO_{2 max}, and BMI, but this data is not included in Table 1 as there were no significant differences between any of the groups for these variables. The only significant difference between intervention and comparison groups at baseline was in ethnic distribution. The intervention group had a smaller proportion of Latinas (t(117) = −2.16, p <.05) than the comparison group.

There were no significant differences between high-TV viewers at the intervention vs. the comparison school. The low-TV group at the intervention school included a higher percentage of Caucasian students than the low-TV group at the comparison school (t(60) = 2.07, p < .05), likely due to the greater proportion of Caucasian students enrolled in the intervention school compared to the comparison school.

As expected, t-tests comparing the high- and low-TV groups within each school revealed significant differences in TV viewing time. At the intervention school, high-TV group members watched an average of 3.1 ± 1.7 hours of TV per day at baseline, and the low-TV group watched 1.1 ± 0.4 hours per day (t(58) = −5.8, p < .001). At the comparison school, the high-TV group watched 3.4 ± 1.4 hours of TV per day at baseline versus 0.9 ± 0.5 hours for the low-TV group (t(57) = −9.9, p < .001). There were no other differences between the high and low-TV groups within either school.

Test for Moderation

Based on the hierarchical multiple regression analysis, the interaction between intervention condition and television viewing approached conventional levels of significance (p <.08), suggesting that the intervention’s impact on vigorous PA depended upon baseline TV viewing (see Table 2). Therefore the association of the intervention with exercise behavior was examined separately in the high- and low-TV groups.

Table 2.

Results of Hierarchical Regression Analyses for Variables Predicting Postintervention Vigorous PA^† (N = 119)

Variable	Model 1			Model 2			Model 3
	B	SE B	β	B	SE B	β	B	SE B	β
Baseline vigorous PA	0.047	0.076	0.057	0.027	0.074	0.033	0.012	0.074	0.015
Ethnicity	0.196	0.107	0.168*	0.157	0.105	0.135	0.177	0.105	0.153
Intervention condition				−0.281	0.107	−0.243**	−0.103	0.146	−0.089
Television viewing				0.097	0.105	0.084	0.280	0.146	0.242*
TV × intervention							−0.375		− 0.252**
F(df,df)		2.078 (2,116)			3.435 (4,114)**			3.439 (5,113)***
D R2					0.073			0.025
R2 adjusted		0.018			0.076			0.094

^† PA indicates physical activity.

^*p < 0.08;

^**p < 0.05;

^***p < 0.01 (two-tailed tests).

The regression model for the high-TV group revealed that the intervention was a significant predictor of PA among high-TV watchers (β = −.45; p < .001). The regression model for the low-TV group did not reveal a significant effect of the intervention on PA for low-TV watchers (p > .05).

Test for Mediation

The first step of Baron and Kenny’s test of mediation¹² (regressing the independent variable on the dependent variable) revealed that intervention status was a significant predictor of post-intervention vigorous PA (β = −.28; p < .01). The second step of the mediation test revealed that the potential mediator (change in TV viewing) did not significantly predict PA (p>.05), suggesting that TV viewing was not a mediator of the relationship between intervention status and vigorous PA.

Post-hoc Analyses

Average hours spent watching TV decreased significantly (from 3.1 ± 1.7 hours per day to 2.3 ± 1.5 hours per day) for the intervention/high-TV group (t(34) = 2.1, p < .05). For low-TV participants in the intervention group, TV viewing time did not change over the course of the intervention (p > .05). Television viewing time did not change for participants in either the high- or low-TV group at the comparison school (p > .05).

Vigorous PA increased significantly in the intervention/high-TV group (from 12.0 ± 14.7 to 13.1 ± 7.3 METs (t(33) = −2.1, p < .05)). The intervention/low-TV group did not show any significant change in vigorous PA. Vigorous activity did not change significantly for participants in the high- or low-TV group at the comparison school (p > .05).

Discussion

Summary

The regression results and the pre/post comparisons of activity level and television viewing provide evidence that TV viewing may be displacing some PA among these adolescents. Analyses supported a moderating, but not a mediating, role of TV viewing in the relationship between the intervention and PA. Specifically, moderation was evidenced in that adolescents who were above the median in TV-viewing at baseline showed post-intervention increases in vigorous PA and simultaneous decreases in television viewing, whereas those who were below the median for TV viewing showed no change in vigorous PA or television viewing. These results suggest possible displacement of PA with TV viewing. The displacement is not one-to-one (i.e., PA did not increase as much as TV viewing decreased), yet the increase in PA is meaningful and, if sustained, would be likely to translate into health benefits.

These results and a recent review article¹⁴ addressing the influence of the media environment on PA, which reports that the evidence for displacement is mixed, suggest that further investigation of the displacement hypothesis is warranted. Indeed, while displacement has been demonstrated in individual studies of child and adult populations as well as in a recent meta-analysis examining children’s media use and PA, some of the most methodologically rigorous research in the area has turned up only limited support for the displacement hypothesis.¹⁴

Limitations

Participants in this study were not randomly assigned to watch high or low amounts of television. While the high- and low-TV groups were comparable along key characteristics, it is possible that another factor that differed between these groups could have accounted for the differential response to the intervention. Further, while key comparisons were made between the high- and low-TV groups within the intervention school, it is important to note that participants were not randomly assigned to control and intervention conditions, and the control school had a significantly larger proportion of ethnic minority students. The presence of a non-identical comparison sample may be a source of bias in the findings. Future studies would benefit from more nearly matching the intervention and comparison groups, ideally by randomly assigning participants to conditions. There may also have been floor effects for indicators of fitness and PA (owing to the study selection criteria) that resulted in the high- and low-TV groups showing no differences on these indicators. One might otherwise expect to see lower fitness and PA among high-TV viewers. Whether these results generalize to populations other than sedentary teenage girls has not yet been tested. The present analyses relied heavily upon self-reported data, which is subject to many sources of potential error. Consistent results from studies employing objective measures of PA (e.g. accelerometer) and television viewing (e.g. automatic viewing log) would strengthen our assertion of moderation.

Significance

These results suggest that sedentary adolescent girls who spend a lot of time watching television may derive greater benefit from a school-based PA intervention than peers who watch less TV. An implication is that specific types of PA interventions could be more effective among girls who watch more television, and other interventions would be more effective among girls who watch less TV. Matching individuals with appropriate interventions would improve both program effectiveness and cost effectiveness (i.e. money would not be spent unsuccessfully attempting to influence behavior of a group via means inappropriate for that population.)