Health-Related Social Control and Risk Perceptions in Emerging Adults

Method

Measures

Table 1 presents descriptive statistics for study measures.

OPEN IN VIEWER

Table 1.

Descriptive Statistics of Study Measures.

Measure	Mean (SD)	Actual Range	α
Positive social control	33.94 (9.9)	15–68	.91
Negative social control	8.64 (3.5)	6–23	.91
Positive affective responses	25.63 (8.8)	6–42	.94
Negative affective responses	10.61 (5.5)	6–29	.92
Behavior changea	3.42 (1.4)	1–7	—
Behavioral resistance	5.49 (2.9)	3–17	.83
Perceived likelihood	2.70 (.97)	1–5.4	.83
Perceived severity	5.95 (1.1)	1–7	.91
Worry	3.48 (1.5)	1–7	.91

^aBehavior change was measured using a single item.

Social control

Participants identified how often social network members used 15 positive and 6 negative social control strategies to influence their health behaviors (i.e., dietary, exercise, and sleep; Thorpe et al., 2008) over the past month using a 5-point scale (1 = never, 5 = all of the time/every day). To account for variability in the size of participants’ social networks, scores for each strategy were summed across social relationships and averaged based on the total number of social network members from whom a participant indicated receiving health-related social control. On average, participants reported interacting with 2.42 social network members (SD = 1.2; range = 1–7).

Affective responses to social control

Participants indicated the extent to which they experienced six positive and six negative emotions in response to social control over the past month on a 7-point scale (1 = not at all, 7 = very much; Derogatis, 1975; Fekete, Geaghan, & Druley, 2009). Scores for each subscale were summed across social relationships and averaged based on the total number of social network members from whom a participant indicated receiving social control.

Behavioral responses to social control

Participants indicated the extent to which they complied with requests to change their behaviors or resisted changing their behaviors in response to social control on a 7-point scale (1 = not at all, 7 = very much; Lewis & Rook, 1999). Items included a single item on behavioral change and three items on behavioral resistance. Scores for each behavioral response were summed across social relationships and averaged based on the total number of social network members from whom a participant indicated receiving health-related social control.

Health risk perceptions

Participants reported their health risk perceptions (perceived likelihood and severity) for chronic illnesses linked to dietary, exercise, and sleep behaviors, including heart disease, hypertension, stroke, breast/prostate cancer, colorectal cancer, any other form of cancer, and type 2 diabetes mellitus on a 7-point scale (1 = certain NOT to happen, 7 = certain TO happen). These measures were derived from scales utilized by Shiloh, Wade, Roberts, Alford, and Biesecker (2013). Responses were summed and averaged across health conditions to create an overall scale of each type of risk perception.

Analysis Plan

Any sociodemographic variables correlated with a mediator or outcome were retained as covariates. Thus, gender and relationship status were controlled for in all analyses. Men reported higher levels of behavioral resistance, and women reported higher levels of perceived likelihood of developing an illness. Individuals who were single reported higher levels of behavioral resistance, negative affective responses, and more behavior change than individuals in a serious relationship. Analyses were conducted using PROCESS in SPSS version 23 (Hayes, 2013), and separate models were conducted for each outcome. A moderated mediation analysis examined conditional indirect effects with the expected interaction occurring between social control and health risk perceptions in explaining affective responses to social control. To examine how the conditional indirect effects changed as health risk perceptions increased, the relationship between the predictor, mediators, and outcome variables were examined at different percentile ranks of the moderator (i.e., 10th, 50th, and 90th) for models with significant interactions. Bias-corrected bootstrapped confidence intervals (CIs) created with 10,000 samples were used to examine the significance of each percentile rank, and the index of moderated mediation was used as an indicator of the overall moderation effect. CIs that did not contain zero were considered to be significant (Gardner & Altman, 1986).

Results

Table 2 presents the sources of social control reported by emerging adults, and Table 3 presents the frequency that different health behaviors were targeted. As illustrated in Figures 1 and 2, positive control was associated with greater positive affective responses, whereas negative control was associated with fewer positive and more negative affective responses. Positive affective responses were associated with behavior change, whereas negative affect responses were associated with behavioral resistance. Figure 1 further reveals that significant interactions emerged between positive control and perceived severity in explaining the links between positive control and greater positive affective responses and between negative control and perceived severity in explaining the links between negative control and fewer positive affective responses. As perceived severity increased, the conditional indirect effect of positive control on behavior change through higher positive affective responses became weaker. In contrast, as perceived severity increased, the indirect effect of negative control on less behavior change through lower positive affective responses became stronger. No appreciable differences emerged in the strength or direction of these results when they were conducted without including covariates.

OPEN IN VIEWER

Table 2.

Sources of Social Control Reported by Emerging Adults.

Source of Social Control	Percentage of Participants Reporting Social Control
Parents	84.4
Friends	44.7
Significant others	41.1
Roommates	33.1
Siblings	18.9
Extended family	13.1
Other (coaches, coworkers, professors)	14.2

OPEN IN VIEWER

Table 3.

Most Frequently Targeted Health Behaviors by Social Network Members.

Health Behavior	Percentage of Participants Who Indicated That a Social Network Member Attempted to Influence Behavior
Get more sleep	82.9
Drink more water	62.5
Avoid eating fast food	57.8
Avoid eating high-fat foods	53.1
Eat more fruits and vegetables	52.0
Engage in more vigorous physical activity	45.1
Limit intake of sugary drinks	43.3
Eat more lean proteins	40.0
Engage in more moderate physical activity	35.3
Eat more foods made with whole grains	29.1
Engage in more light physical activity	26.9
Drink less alcohol	20.7
Consume more low-fat/nonfat dairy	16.7
Avoid eating foods made with refined grains	15.6

OPEN IN VIEWER

Figure 1.

Perceived severity as a moderator of the links between social control, affective responses to control, and behavior change/behavioral resistance. Model controls for gender and relationship status. Values represent unstandardized regression coefficients (b), standard errors (SE), and p values (p). Additionally, dotted lines are used to represent paths in the model that were not statistically significant. Bolded values represent the path coefficients between positive and negative affective responses and behavioral resistance. Conditional indirect effect of positive social control on behavior change through positive affective responses after controlling for gender and relationship status: IE _conditional = .048, SE = .01, confidence interval (CI)_bootstrap = [.032, .066] for 10th percentile; IE _conditional = .040, SE = .01, CI_bootstrap = [.029, .059] for 50th percentile; and IE _conditional = .034, SE = .01, CI_bootstrap = [.022, .050] for 90th percentile; index of moderated mediation = −.006, SE = .003, CI_bootstrap = [−.01, −.001]. Conditional indirect effect of negative social control on behavior change through positive affective responses after controlling for gender and relationship status: IE _conditional = −.052, SE = .01, CI_bootstrap = [−.083, −.029] for 10th percentile; IE _conditional = −.066, SE = .01, CI_bootstrap = [−.096, −.042] for 50th percentile; and IE _conditional = −.076, SE = .02, CI_bootstrap = [−.111, −.047] for 90th percentile; index of moderated mediation = −.01, SE = .006, CI_bootstrap = [−.024, −.002].

OPEN IN VIEWER

Figure 2.

Perceived likelihood as a moderator of the links between social control, affective responses to control, and behavior change/behavioral resistance. Model controls for gender and relationship status. Values represent unstandardized regression coefficients (b), standard errors (SE), and p values (p). Additionally, dotted lines are used to represent paths in the model that were not statistically significant. Bolded values represent the path coefficients between positive and negative affective responses and behavioral resistance. Conditional indirect effect of positive social control on behavior change through positive emotional responses after controlling for gender and relationship status: IE _conditional = .034, SE = .01, confidence interval (CI)_bootstrap = [.022, .050] for 10th percentile; IE _conditional = .04, SE = .01, CI_bootstrap = [.027, .054] for 50th percentile; and IE _conditional = .046, SE = .01, CI_bootstrap = [.031, .063] for 90th percentile; index of moderated mediation = .004, SE = .002, CI_bootstrap = [.0002, .001].

No significant interactions emerged in the model examining perceived likelihood as a moderator of the social control process (Figure 2). However, when the analyses were conducted without covariates, the marginally significant interaction between positive control and perceived likelihood in explaining positive affective responses became significant. In both the model with and without covariates, the conditional indirect effect of positive control on behavior change through positive affective responses to control became stronger as perceived likelihood increased.

Discussion

Consistent with prior research (e.g., Logic et al., 2009; Thorpe et al., 2008), when emerging adults received positive control, they experienced more positive affect and engaged in behavior change. In contrast, negative control was associated with more negative affect and behavioral resistance. Inconsistent with hypotheses, as emerging adults’ perceived severity of developing a future chronic illness increased, the relationship between positive control and more behavior change through positive affect decreased, rather than increased. Moreover, the relationship between negative control and less behavior change through lower positive affect became stronger, rather than weaker.

It is possible that receiving any type of social control sent a message that emerging adults were not managing their health well. This in combination with having higher perceptions about the severity of developing a chronic illness may have elicited anxiety in participants, which may have decreased their positive responses to social control. Some research suggests that in individuals with high levels of worry or anxiety about developing an illness such as cancer, risk perceptions are related to poorer rather than healthier behaviors (Ferrer, Portnoy, & Klein, 2013). Additionally, negative control promoted behavior change in men living with HIV, but only if they were not at risk of being depressed (Fekete et al., 2009). As depressive symptoms and anxiety are highly correlated (Moffitt et al., 2007), it is possible that anxiety may also influence interpretations of social control. Since emerging adults are still developing strategies to regulate their emotions (John & Gross, 2004), the links between social control, anxiety, and less positive affect may be pronounced in this age-group. Future research should examine whether anxiety plays a role in the social control process, particularly in the context of health risk perceptions.

Without the inclusion of gender and relationship status as covariates, the beneficial effects of positive control were enhanced for individuals with higher perceived likelihood of developing an illness. Other research confirms that increased perceived susceptibility to illnesses encourages behavior change and may have a stronger relationship to behavior change than perceived severity (Brewer, Chapman, Gibbons, Gerrard, & McCaul, 2007; Janz & Becker, 1984). Notably, this finding was reduced to marginal significance after including covariates, possibly because individuals who were single reported more behavior change. Future research should examine how risk perceptions influence the social control process in emerging adults who are in romantic relationships versus being single.

Given that our interactions were smaller in magnitude, it is likely that other motivating factors also influence the health behaviors of emerging adults. For example, emerging adults may engage in healthier behaviors to change their physical appearance, which can be related to factors like social acceptance and romantic appeal (Leary, Tchividjian, & Kraxberger, 1994). Among college students, men valued being in shape or toned, while women expressed fears about gaining weight (LaCaille, Dauner, Krambeer, & Pederson, 2011). Thus, risk perception related to more immediate factors as opposed to longer term health risks may be more influential in changing the health behaviors among emerging adults.

Interestingly, the indirect effect of negative control on behavioral resistance through increased negative affect was not moderated by health risk perceptions. Negative control was associated with feelings of anger, irritation, and resentment, which in turn was associated with reactive behaviors including resisting change, pretending to change, and changing in the opposite direction of what was requested. In general, emerging adults tend to experience higher levels of psychological reactance when their freedom of choice is threatened (Miller & Quick, 2010). It is possible that the negative affect and behavioral resistance aroused by negative social control was strong enough that participants ignored that they were engaging in behaviors that may have severe health consequences and instead focused on establishing their sense of independence.

The present study has several limitations. Our study did not assess participants’ health behavior knowledge, and it is possible that participants did not associate poor dietary, exercise, and sleep behaviors with risk of developing future chronic health conditions. Future research should assess how health behavior knowledge plays a role in both the social control process and the development of risk perceptions. It would also be interesting to examine if a different pattern of results emerged if social control for health risk behaviors (e.g., substance abuse and risky sex behaviors) was examined.

The study was self-report and cross-sectional in design. Causal relationships cannot be inferred, and results may be inflated due to common method variance. Additionally, we averaged across sources of social control rather than examining sources individually. It is possible that emerging adults may have responded differently to social control from different social network members. Finally, the majority of the participants were White female students, so the generalizability of our findings is unclear. Among college students, men tend to respond more favorably to social control than women (Brunson et al., 2014). Although this was not true for our data, it is possible different results may have emerged if the gender of our sample was more equally dispersed.

The present study adds to the growing research on moderators of the social control process. Additionally, this study examined health-related social control in emerging adulthood, a developmental period where declines in health behaviors are common (Nelson et al., 2006, 2008). Understanding how emerging adults’ social networks are involved in health promotion may be useful for developing effective interventions for health behavior change. Further, addressing health behavior change in emerging adults is important to prevent the development of chronic illness later in life, especially because many middle-aged and older adults fail to adopt healthy behavior change once they have been diagnosed with a chronic illness (Newsom et al., 2012).