While the previous chapter explored the prevalence of individual items in the addiction and engagement scale, and their validity in the form of associations with health and functioning, video game addiction is greater than the sum of its parts. In other words, video game addiction is a disease model which suggests an emergencearising from the combination and interaction of several of these features, rather than the presence of only individual features. Generally, it is expected that the greater the number and strength of these ‘symptoms’, the greater the decline in health and functioning. In statistical terms, that there will be a roughly linear relationship between addiction symptoms and declined health and functioning. Similarly, if addiction is to be successfully differentiated from close constructs, the same relationships with declined health and functioning will not be present with a
measure of engagement. Therefore, this chapter investigates relationships across the scale total scores for addiction and engagement with a diverse range of health and functioning indicators.
Few studies have explored differential predictive validity of video game addiction and engagement measures across a span of diverse measures of health and functioning, particularly in a sample of adults who fall at the more intense end of video gaming. So far, studies have found that addiction, but not engagement, was associated with poorer
academic performance, negative personality traits, attentional profile, and mental health in a small sample of participants (Charlton & Danforth, 2010; Metcalf & Pammer, 2011; Skoric et al., 2009). However, most studies of video game addiction have not attempted to test the differentiation between addiction and close but non-pathological constructs.
In terms of mental health, video game addiction has shown correlations with increased psychological distress, a range of psychopathologies, suicidal ideation, life trauma and health-related quality of life (Gentile et al., 2011; King & Delfabbro, 2009; King, Delfabbro, & Zajac, 2011; Mathers et al., 2009; Mentzoni et al., 2011; Primack et al., 2009; Rehbein et al., 2010; Starcevic et al., 2011; Van Rooij et al., 2011; Yang, 2005). Regarding social health, results suggest video game addicts have lower self-esteem, are less agreeable and
conscientious, more neurotic and socially anxious, have poorer relational maintenance skills, with less offline friends and are more comfortable meeting people online than offline
(American Medical Association, 2007; Chiu et al., 2004; Colwell et al., 1995; Ferguson et al., 2011; E. J. Kim et al., 2008; King & Delfabbro, 2009; King, Haagsma, et al., 2013; Kowert &
Oldmeadow, 2013; Lemmens et al., 2009, 2011; Lo et al., 2005; Peters & Malesky Jr, 2008; Starcevic et al., 2011). Some studies have also connected poorer physical health with video game addiction, including poorer health status, exercise frequency, physical functioning, vitality, general health, and sleep health (Griffiths et al., 2004; King & Delfabbro, 2009; Mathers et al., 2009; Mentzoni et al., 2011; Rehbein et al., 2010; Smyth, 2007; Tejeiro et al., 2012; Tejeiro Salguero & Morán, 2002). A meta-analysis investigating the average effect size of a range of problems connected with video game addiction concluded that
connections with social well-being showed the strongest effect size of r+=.25 (pooled
correlation coefficient), or roughly 6% of shared variance, compared with mental (r+=.19) and physical health (r+=.12) (Ferguson et al., 2011).
Yet in all instances, there is contrary evidence. Studies which included measures of the same or very similar variables have found an absence of relationships with video game play or addiction, or differences in the strength of the associations. These include mental health (King & Delfabbro, 2009; R. F. McClure & Mears, 1984; R. Z. McClure & Mears, 1986; Primack et al., 2009), social health (Colwell et al., 1995; King, Delfabbro, & Griffiths, 2011; Kowert & Oldmeadow, 2013; Loton, 2007; Sakamoto, 2005) and physical health (Mentzoni et al., 2011). Many deleterious correlates are what can be described as small in magnitude. Other studies included possible confounding variables, such as basic needs satisfaction, which in some instances explained almost all of the connections measured between video game addiction and health (Przybylski et al., 2009; Seay & Kraut, 2007). Further studies are needed comparing relationships to addiction versus engagement, and investigating the importance of video game addiction across a spread of well-established measures of health and indicators of functioning.
Focus of this Chapter
This chapter will investigate relationships between total scores for engagement, addiction and a range of health measures in a sample of adult, highly engaged and self-identified problematic gamers. The chapter will explore whether video game addiction is more
predictive of mental, social and physical health indicators, than engagement. Which aspects of health are most important will be investigated by including them as predictors in
regression models. Where possible, the size of relationships will be compared to other factors related to health, in order to contextualise the severity of associated problems with video game addiction.
Aims and Hypotheses
Aim:
2. Investigate the differentiation of video game addiction from engagement by analysing:
2.1. Relationships between video game addiction versus engagement with levels of social, physical and psychological health;
Hypotheses:
1. That video game addiction will be predicted by lower levels of mental, social and physical health;
2. That video game engagement will show no association with the same measures of health.
Method
Correlation matrix.
As a first step, a correlation matrix was produced reporting correlations between total scores for addiction, engagement and all health and functioning measures.
Predicting addiction versus engagement with health and functioning
indicators.
Two enter method multiple regression models were run. One model predicted total score for the addiction subscale and the other the engagement subscale, using the following
predictors: total scores for anxiety, stress and depression subscales, social activity index, total score for the social support scale, BMI, the physical exercise frequency item and total score for the sleep health scale.
Theoretical and statistical directionality of these models was thoroughly considered. Statistical directionality and model specification should usually follow lines of plausibility,
based on theory (Pallant, 2010). An exception is when the goal is to maximise the variance explained using statistical techniques, which may produce models which are inconsistent with theory when applied to variables. However, in the instance of video game addiction and health aspects, it is possible to construe these relationships in either direction, or in both directions. Video gaming can be a relaxing, enjoyable and distracting activity, which is easily accessible in the home environment. It is possible that in response to declined health, particularly mental health, which can cause people to feel unable to deal with even simple daily tasks, gamers may intensify their hobby, which may include increasing some aspects of addiction. Studies have shown that people experiencing poorer mental health, particularly depressive illnesses, often increase time spent at home (Dickerson, Gorlin, & Stankovic, 2011). This explanation characterises video game addiction as predominantly a response rather than cause of decreased health, at least in the initial stages. Alternatively, it is also plausible that a person experiencing stable health may develop video game addiction, which incurs negative consequences directly resulting from displacement of other activities and components of dependence, which then leads to poorer mental health. This explanation suggests the decrease in health would otherwise not have occurred, if not for the addiction. Likewise, it is theoretically plausible that any of these relationships are bi-directional and mutually reinforcing; i.e. a downward spiral (Gentile et al., 2011). This uncertainty regarding directionality is important to take into consideration when interpreting the results of cross- sectional analyses. While it was conducted subsequently, the longitudinal analysis in chapter 10 actually supports the statistical directionality of the models in this chapter, with
detrimental health predicting addiction.
Similarly, the connection between social support and video game addiction can be conceptualised in many ways. It could be considered a moderator, buffering negative outcomes of video game addiction. Alternatively, deficient social support could be a cause, with video games offering a substitute in the absence of fulfilling social relationships, as concluded by the literature review on the topic by the American Medical Association’s Council on Science and Public Health (American Medical Association, 2007). Yet another interpretation is that reduced social support could be an outcome of video game addiction, with play displacing relational maintenance strategies. A cross-sectional relationship
supports the latter interpretation in a study measuring media dependence (Chory & Banfield, 2009). As there is no clear rationale for controlling for social support, low support was
considered a possible cause of video game addiction and included, along with all other health measures, as a predictor in the two enter-method multiple regressions. As in the past models, scatterplots were produced displaying the standardised predicted versus actual values, which showed clear linearity, normality and quite highly homoscedastic errors. Once
again, in each model, a few cases were identified as univariate outliers on some measures based on standardised Z scores, and twelve cases were identified as multivariate outliers utilising Mahalanobis’ distance, with a Chi-Squared probability distribution applied. Given the sample size, all were retained and are not believed to contribute a large amount of bias to the final models.
Results
Correlation matrix.
Results of the correlation matrix are presented in Table 41. Correlations indicated significant relationships between the scale total scores and psychological, social and physical health indicators.
Table 41
Correlation Matrix (Pearson) Reporting Relationships Between Engagement, Addiction and Health at Wave 1
Eng Dep Anx Str Soc
supp Soc act Phy exer BMI Sleep Health Addiction .48** .42** .40** .44** -.27** -.24** -.15** .13** -.37** Engagement - .17** .17** .15** -.04 -.12** -.06 .19** -.18** Depression - .64** .68** -.49** -.25** -.16** .07 -.51** Anxiety - .71** -.34** -.16** -.13** -.00 -.52** Stress - -.31** -.19** -.17** .08 -.55** Social Support - .33** .13** -.05 .33** Social Activity - .17** -.08 .14** Physical Exercise - -.05 .12** BMI - -.15**
Predicting addiction versus engagement with health and functioning
indicators.
The model predicting addiction was significantly different from a null model (F(8)=23.30, p<.001), with adjusted R2 indicating 26.5% of addiction variance was explained, as detailed in Table 42. The same predictors explained substantially less variance of engagement scores, with adjusted R2 indicating 7.8% of variance was shared, but the model was still significant (F(8)=6.24, p<.001). This is reported in Table 43.
Multicollinearity diagnostics were run alongside the model. While there is no consensus on the multicollinearity indicators required to denote independence of predictor variables, variance inflation factors (VIF) under 10 is one rule of thumb (O’brien, 2007). As no predictor variable exceeded a VIF of 10, this suggests variables may be considered independent, and multicollinearity would not explain the lack of significance for depression. As receiving a mental illness diagnosis prior to the study (n=86) was associated with significantly higher scores on the DASS subscales (see Chapter 3, Method: Measures: Mental Health), a further hierarchical regression was run with prior mental illness as a dummy variable in the first step. Controlling for past mental illness made no substantial difference in the results, only changing the standardised beta coefficients at the third decimal point in two of the predictors. As such, this analysis is not reported here.
Table 42
Enter Method Multiple Regression Predicting Addiction with Measures of Social, Physical and Psychological Health.
Predictor b SE b β p VIF Social support -.06 (-.21, .06) .07 -.04 .26 1.44 Social activity -.09 (-.17, -.04) .02 -.13 .002** 1.15 Exercise frequency -.23 (-.63, .13) .20 -.05 .20 1.05 BMI .09 (-.02, .21) .05 .07 .08 1.04 Sleep health -.14 (-.24, -.02) .02 -.12 .02* 1.64 Stress .41 (.14, .70) .13 .18 .002** 2.67 Anxiety .31 (.01, .60) .14 .11 .02* 2.31 Depression .11 (-.09, .32) .11 .06 .24 2.50
Table 43
Enter Method Multiple Regression Predicting Engagement with Measures of Social, Physical and Psychological Health.
Predictor b SE
b
β p
Social support .12 (.01, .25) .07 .09 .05*
Social activity -.08 (-.13, -.001) .02 -.11 .01**
Physical exercise frequency -.17 (-.54, .22) .20 -.04 .38
BMI .22 (.08, .34) .05 .16 .001***
Sleep health -.12 (-.22, -.005) .06 -.10 .03*
Stress -.11 (-.37, .17) .14 -04 .44
Anxiety .28 (-.004, .57) .14 .13 .04*
Depression .11 (-.09, .32) .11 .08 .25
Note. Adjusted R2=.078. 95% confidence intervals for b reported in parentheses. Refer to previous table for VIF.
Discussion
Few studies have thus far investigated the differential connections between video game addiction versus engagement on a range of health indicators. As cut-off points are problematic and can artificially or inappropriately divide samples and mask relationships, relationship-based statistics are still preferable when investigating validity in this instance. As expected, many significant correlations were present between the total scores of addiction and engagement, and measures of health. As such, multiple regression analyses were conducted. These provide an indication of overall variance explained in addiction versus engagement by the combined measures of social, physical and psychological health, including which indicators were most predictive while taking into account the contribution of the other indicators. These models differ from the previous chapter as they consider scale total score, rather than item-level analyses, and they also assess the importance of all of the health measures simultaneously; unlike the previous chapter which separated models according to the specific health domain.
Addiction versus engagement.
The first and second hypotheses were partially confirmed. As noted, a key distinguishing feature of addiction is that it has demonstrable negative consequences. Both models were significant, indicating that a linear relationship is present between both addiction and engagement and measures of health. However, the difference in the variance explained provides some validity for the separation between addiction and engagement as designated by Charlton and Danforth (2007). A comparison of the overall adjusted R2 figures indicates that much less variance is explained by health indicators in the model predicting
engagement (7.8%) than the model predicting addiction (26.4%). Addiction predicted over three times the overall variance. However, significant relationships also emerged between engagement and detrimental health. In order to gain a greater understanding of the specific health indicators related to addiction and engagement, standardised beta coefficients are considered.
Video game addiction and mental health.
The multiple regression models give an indication of the relative importance of video game addiction on this spread of health indicators, taking into account the influence of the other health indicators. The results confirm the second hypothesis. Different indicators of health were clearly more important in predicting addiction, and engagement, when considered in the presence of all other health indicators. A review of the standardised beta coefficients reveal some interesting features, and facilitate a further exploration of the differences between engagement and addiction.
When accounting for all health indicators present, addiction was not predicted by social support, physical exercise frequency and, surprisingly, depression. Instead, stress arose as the most substantial predictor, followed by anxiety, which was equal to social activity and sleep disturbance, and finally BMI. The direction of all predicted variables is as expected. Relationships indicate that as video game addiction scores increase, it is generally
accompanied by increases in stress, anxiety and BMI; and decreases in social activity and sleep health.
The absence of relationship with depression contrasts past studies, which have consistently found significant relationships between depression and video game addiction, even in the presence of other mental health features (King, Delfabbro, & Zajac, 2011; Mentzoni et al.,
2011; Metcalf & Pammer, 2011; Rehbein et al., 2010; Starcevic et al., 2011). Interpretation of this result must take into account the nature of the sampling strategy, which targeted self- identified problematic adult gamers who were already above norms for these symptoms. This finding may indicate that in the presence of already elevated levels of mental illness symptomatology, increases in video game addiction is accompanied mostly by stress, and less-so anxiety – but not necessarily depression. This interpretation has some theoretical merit, as stress, in this case, refers to a general psychological distress. As addiction is thought to include increasing displacement of other activities, whereby important
responsibilities fail to be met, this may cause mounting pressure and subsequent distress and anxiety. However, it is equally plausible to conclude the opposite, that with increased psychological distress and anxiety brought about by other factors, some people become increasingly dependent on video gaming as an escape-based coping strategy, which posits video gaming as a response rather than the central cause, of declined. This result is difficult to interpret without further studies, especially longitudinal studies investigating temporal precedence. It must be noted that the DASS measures symptoms of these mental health conditions, rather than providing an actual diagnosis. Ideally, if resources permitted, an individual clinical assessment would help to confirm the relationship and provide accurate diagnosis.
Video game addiction and social health.
Further differences were apparent with regards to social health. Video game addiction was predicted by social activity, which is a structural measure, being a count of friendship
network size and social outings; but showed no significant relationship with social support, a functional measure. Social support is central to health as it taps subjective ratings of the quality of critical social support available, including informational, emotional and tangible support. This suggests that with increasing video game addiction, social relationships may be displaced, but not necessarily with those people who provide key forms of support, suggesting close and important relationships may be less affected. Alternatively, it may also be reflective of the provision of social support through video gaming. Many participants reported having made friends through video gaming, and a smaller proportion reported having current friendships that, if not for video gaming, they would have little in common (n=177, 35.6%). This finding may also point to a tendency for people who endorse video game addiction criteria to require comparatively less friends and social outings to satisfy their social needs. In this instance, introversion, which has been associated with internet and video game addiction in past studies (Charlton & Danforth, 2010; Gibb, Bailey, Lambirth, &
Wilson, 1983; R. F. McClure & Mears, 1984), may present a confounding variable that could be tested in future modelling exercises. This may explain why higher addiction scores were predicted by lower structural social support, but not by functional social support. Longitudinal studies may help to determine whether these more critical functional social connections erode with prolonged video game addiction.
Video game engagement and health.
Engagement also demonstrated some significant connections with measures of health, including one positive indicator – social support. Participants’ ratings of the quality of their own social support actually tended to increase with higher engagement. This may be in line with other research which described video gaming capacitating online communication and socialisation, and ultimately the provision of some forms of social support (Lieberman et al., 2011; Steinkuehler & Williams, 2006). This result also confirms the lack of a significant inverse prediction of social support in the model predicting addiction. In contrast to addiction, this analysis suggests social support is actually higher in those more engaged with video games, which may hint at the greater support provided by online social networks. It is also possible that some of this social support is derived by playing games with friends in-person. However, at baseline, only 12.4% of gaming time was spent with friends in-person,
compared with 39.6% online, suggesting that most of this additional social support garnered through engagement with games is derived online. Further, a large proportion of the sample reported having made friends through video gaming (n=335, 67.4%), and a smaller
proportion reported having current friendships that, if not for video gaming, they would have little in common (n=177, 35.6%), suggesting on-going friendships are made through video gaming. The differences between offline and online socialisation, including in terms of