*2.4. Ethics*

The Institutional Review Board of the University of Hong Kong/Hospital Authority Hong Kong West Cluster approved this study. All respondents provided verbal informed consent. Telephone interviews were tape-recorded for quality checking with respondents' consent. Records were then erased six months after completing the survey.

### **3. Results**

The weighted sample comprised 55.0% females, and the mean age ± standard deviation (SD) was 48.3 ± 18.3 years (Table 1). Smartphone owners (80.0%) reported a mean SAS-SV score ± SD of 28.9 ± 10.1. Using cutoff points ≥ 3 of GAD-2 and PHQ-2, 10.9% (*n* = 442) and 8.1% (*n* = 327) respondents screened positive for anxiety and depression symptoms, respectively. Mean scores of SHS and SWEMWBS were 5.2 ± 1.0 and 23.0 ± 4.2, respectively.


**Table 1.** Sociodemographic and lifestyle-related characteristics of the sample (*N* = 4054).


SAS-SV, Smartphone Addiction Scale-Short Version; GAD-2, Generalized Anxiety Disorder Questionnaire-2 (range 0–6); PHQ-2, Patient Health Questionnaire-2 (range 0–6); SHS, Subjective Happiness Scale; SWEMWBS, Short Warwick-Edinburgh Mental Well-Being Scale. <sup>a</sup> Weighted by age, sex, and educational attainment distributions of the Hong Kong general population. <sup>b</sup> US \$1 = HK \$7.8. <sup>c</sup> Subset sample (*n* = 1331).

Higher SAS-SV scores were observed for respondents who screened positive for symptoms of anxiety (31.9 ± 9.4 vs. 28.5 ± 10.2; *p* < 0.001) and depression (33.2 ± 10.2 vs. 28.5 ± 10.0; *p* < 0.001) than those with negative screening results (Table 2). Multivariable analyses showed that a 1-unit increase in SAS-SV score (range 10–60) was associated with a 3% increase in odds of severity of anxiety symptoms (adjusted odds ratio [AOR] = 1.03; 95% CI: 1.01, 1.04) and a 4% increase in odds of severity of depression symptoms (AOR = 1.04; 95% CI: 1.03, 1.06) after adjusting for sociodemographic and health-related variables.

**Table 2.** Odds of anxiety and depression associated with SAS-SV score (*N* = 4054).


SAS-SV, Smartphone Addiction Scale-Short Version (range 10-60); CI, Confidence Interval; GAD-2, Generalized Anxiety Disorder Questionnaire-2 (range 0–6); PHQ-2, Patient Health Questionnaire-2 (range 0–6). <sup>a</sup> Weighted by age, sex, and educational attainment distributions of the Hong Kong general population. <sup>b</sup> Adjusted for age, sex, marital status, employment status, educational attainment, monthly household income, smoking, and alcohol drinking.

Multivariable analyses showed that a 1-standard-deviation increase in SAS-SV score was associated with a 0.07-standard-deviation decrease in SHS score (adjusted B = −0.07; SE: 0.002; *p* < 0.001) and a 0.10-standard-deviation decrease in SWEMWBS score (adjusted B = -0.10; SE: 0.01; *p* = 0.002) (Table 3). Reduction in effect size of these associations were observed after stratifications by screening results of anxiety and depression symptoms, except for the stronger association of SAS-SV score with lower SHS score in respondents who screened positive for anxiety symptoms (adjusted B = −0.16; SE: 0.01; *p* = 0.013). The associations of SAS-SV score with lower SHS score remained in respondents who screened negative for anxiety symptoms (adjusted B = −0.04; SE: 0.002; *p* = 0.040) and depression symptoms (adjusted B = −0.05; SE: 0.002; *p* = 0.014). The associations of SAS-SV score with lower SWEMWBS score remained in respondents who screened negative for anxiety symptoms (adjusted B = −0.08; SE: 0.01; *p* = 0.022) and were marginally significant in those who screened negative for depression symptoms (adjusted B = −0.06; SE: 0.01; *p* = 0.054). We observed no interaction effects of anxiety or depression symptoms on the associations of SAS-SV score with lower scores of SHS and SWEMWBS.


**Table 3.** Standardized beta (B) of SHS and SWEMWBS scores associated with SAS-SV score (*N* = 4054).

SAS−SV, Smartphone Addiction Scale−Short Version (range 10–60); SE, Standardized Error; GAD−2, Generalized Anxiety Disorder Questionnaire−2 (range 0–6); PHQ−2, Patient Health Questionnaire−2 (range 0–6); SHS, Subjective Happiness Scale; SWEMWBS, Short Warwick−Edinburgh Mental Well−Being Scale. <sup>a</sup> Adjusted for age, sex, marital status, employment status, educational attainment, monthly household income, smoking, and alcohol drinking. <sup>b</sup> Subset sample (*n* = 1331).

#### **4. Discussion**

With a representative sample of Chinese adults in Hong Kong, we confirmed the associations of PSU with anxiety and depression in the general population. Few studies of potential mental health effects of PSU have incorporated both mental illness and mental well−being outcomes. We provided the first evidence of the associations of PSU with impaired hedonic and eudemonic well−being, which remained in respondents who screened negative for anxiety or depression symptoms.

Our study built on young people studies to indicate that the associations of PSU with anxiety and depression could have expanded to adults of all ages. The associations can be explained by the time displacement hypothesis that posits a possible tradeoff between smartphone activities and offline healthier activities such as social interactions [33]. Our previous study supported this explanation

by showing that PSU was associated with lower levels of perceived family communication and family well−being [9]. This lack of social support can induce the onset of affective disorders such as anxiety and depression [34]. Other studies showed that PSU symptoms such as overuse and tolerance could risk people to prolong the night−time smartphone usage, which might lead to sleep problems that could mediate the pathway to anxiety and depression [35]. Increasing evidence has suggested that the most problematic application correlated with PSU could be SNS, which could expose people to negative social comparisons with others in perceived more favorable lives and induce affective disorders [36–38]. However, people with symptoms of mental illness might be at higher risk for PSU given the smartphone could be the first and most obvious process to deflect negative cognition and affectivity [39]. Mechanisms in this potential reverse causality can include cognitive− and affective−related maladaptive coping strategies such as repetitive negative thinking and emotion dysregulation [39]. The bidirectional association was hence possible and evident by the reciprocal relations found in prospective cohort studies in young people [7,40].

We observed the association of PSU with lower scores of subjective happiness (i.e., hedonic well−being), which is characterized by affectivity of pleasure and cognition of satisfaction [16]. This finding was consistent with studies of Internet addiction with lower levels of happiness and life satisfaction in young people [41,42]. An intervention restricting night−time smartphone usage also reported the reduced PSU risk and increased levels of subjective happiness at one−week follow−up [43]. PSU was associated with lower scores of SWEMWBS that covers both hedonic and eudemonic aspects of mental well−being in the present study. In contrast, a study showed the improved mental well−being in self−concealers who intentionally withhold personal information in face−to−face settings but engaged more in online communication even driven by PSU [44]. These conflicting findings highlighted the important role of personality traits when evaluating the potential effects of PSU and suggested to balance our findings with potential benefits of ICTs usage such as fostering social inclusion among those who may feel excluded [45].

The magnitude of the association of PSU with impaired subjective happiness increased in respondents who screened positive for anxiety symptoms, which suggested the co−occurrence of lower levels of hedonic well−being with anxiety disorder. This finding can be supported by the cognitive−and affective−related coping processes in the pathway from anxiety symptoms to PSU [39]. Another explanation can be the moderate correlation between scores of SHS and GAD−2 found in the present sample. Previous studies also showed that people with symptoms of affective disorders had lower levels of mental well−being than those without [46]. Despite the correlated relations, the independence of mental well−being from mental illness was supported by the remained associations of PSU with mental well−being in respondents who screened negative for anxiety or depression. This finding provided insights that the absence of psychopathological symptoms might have non−buffering effects on the impaired mental well−being outcomes associated with PSU.

Our findings need to be interpreted with caution. Consistent with a systematic review that reported the small effect size associations of PSU with symptom severity of anxiety and depression (adjusted B range 0.12 to 0.18) [10], a 1−unit increase in SAS−SV score was associated with 3%–4% increase in the odds of positive screening results of anxiety and depression in the present study. The small effect size was also observed for the association of PSU with mental well−being outcomes (adjusted B range 0.07 to 0.10). A study across three large−scale datasets (total N = 355358) showed a much smaller association (median adjusted B = −0.07) of adolescents' digital technologies use with combined mental illness and well−being outcomes [47]. However, unaccounted factors might affect both PSU and mental health in such cross−sectional associations. Longitudinal and experimental studies are warranted to distill causal and predictive models.

One of the study's limitations is that the cross−sectional data restricts the causal inference of the findings. Residual confounding by unmeasured or unknown confounders might exist even after adjusting for many sociodemographic and lifestyle−related variables. We used the landline telephone survey. Sampling bias might exist due to the lack of data on mobile phone−only households that may have different smartphone use patterns. To increase the sample's representativeness, we weighted data according to the age, sex, and educational attainment distributions of the Hong Kong general population. We used self−reported data, which are subject to recall bias and social desirability bias. Future studies of PSU could include behavioral methods for collecting data on smartphone use, such as objectively examining participants' screen time and usage of individual apps. We used screening instruments rather than diagnostic instruments to measure PSU and mental health outcomes. However, more accurate diagnoses by face−to−face assessments in clinical settings would limit the generalizability of findings compared with the population−based study.
