*2.1. Participants and Procedure*

The Hong Kong Family and Health Information Trends Survey (FHInTS) is a periodic territory-wide telephone survey on the general public's behaviors and views regarding information use, individual health, and family well-being, under the project named "FAMILY: A Jockey Club Initiative for a Harmonious Society." The target population was Cantonese-speaking Hong Kong residents aged 18 years old or above. We have conducted five waves of FHInTS since 2009 and reported details of the study design elsewhere [9,11,16]. The present survey was conducted from February to August 2017, as part of the fifth FHInTS. A two-stage probability-based sampling procedure was used. In the first stage, landline telephone numbers were randomly generated using known prefixes assigned to telecommunication services providers under the numbering plan provided by the Office of the Communication Authority. Invalid numbers were then removed according to the computer and manual dialing records. Numbers of successful cases from previous FHInTS were also filtered. In the second stage, once a household was successfully reached, an eligible family member who would have the nearest next birthday was invited to the survey. Interviews were conducted by trained interviewers from the Public Opinion Programme (POP) at the University of Hong Kong. Of 5773 respondents invited, 4054 were successfully interviewed, yielding a response rate of 70.2%.

#### *2.2. Measurements*

The ten-item Smartphone Addiction Scale-Short Version (SAS-SV) measured five addiction-like symptoms of PSU, including daily-life disturbance, withdrawal, cyberspace-oriented relationship, overuse, and tolerance [22]. Each item scores on a Likert scale of 1 (strongly disagree) to 6 (strongly agree), with a higher total score (range 10 to 60) indicating a higher PSU level [22]. The Chinese version of SAS-SV was valid and reliable (Cronbach's alpha 0.84) and indicated acceptable fit for the one-factor model (comparative fit index [CFI] 0.98 [> 0.90 acceptable, > 0.95 excellent]; root mean square error of approximation [RMSEA] 0.08 [< 0.08 acceptable, < 0.05 excellent]; non-normed fit index [NNFI] 0.96 [> 0.95 acceptable]) in our previous study using the same sample of the Hong Kong general population [11]. The convergent validity was adequate, with composite reliability of 0.85 (CR > 0.70 acceptable) and average variance extracted of 0.37 (AVE > 0.40 acceptable if CR > 0.60).

The four-item Patient Health Questionnaire (PHQ-4) includes the two-item General Anxiety Disorder screener (GAD-2) and the two-item Patient Health Questionnaire (PHQ-2) [23]. PHQ-2 has two DSM-IV diagnostic core criteria for major depression disorder, and GAD-2 has two core criteria for generalized anxiety disorder that can also screen for panic and social anxiety disorders [23]. Each item scores on a Likert scale of 0 (not at all) to 3 (nearly every day), with total scores of each subscale ranging 0 to 6 [23]. GAD-2 and PHQ-2 Scores of ≥ 3 are recommended to screen positive for anxiety and depression symptoms, respectively [24]. The Chinese version of PHQ-2 has been validated in our previous study in the Hong Kong general population [25]. In the present sample, confirmatory factor analysis (CFA) indicated an excellent fit for the two-factor model of PHQ-4 (Relative Chi-Square 0.24; *p* = 0.63; incremental fit index [IFI] 1.00 [> 0.95 excellent]; goodness of fit index [GFI] 1.00 [> 0.95 excellent]; adjusted goodness of fit index [AGFI] 1.00 [> 0.95 acceptable]; CFI 1.00; RMSEA 0.01; standardized root mean square residual [SRMR] 0.001 [< 0.08 acceptable, < 0.05 excellent]). GAD-2 had a Cronbach's alpha of 0.74, CR of 0.65, and AVE of 0.48. PHQ-2 had a Cronbach's alpha of 0.73, CR of 0.63, and AVE of 0.46.

The four-item Subjective Happiness Scale (SHS; mean score range 1 to 7) measured the cognitive and affective state characterized by pleasure or satisfaction from hedonic well-being aspect [26]. The Chinese version of SHS has been validated in our previous study in the Hong Kong general population [27]. In the present sample, CFA indicated an excellent fit for the one-factor model (relative Chi-Square 21.68; *p* < 0.001; IFI 0.993; GFI 0.999; AGFI 0.995; CFI 0.993; RMSEA 0.10; SRMR 0.02). SHS had a Cronbach's alpha of 0.75, CR of 0.77, and AVE of 0.47. The seven-item Short Warwick-Edinburgh Mental Well-Being Scale (SWEMWBS; total score range 7 to 35) was adapted by the WEMWBS that covers both hedonic and eudemonic aspects [28]. A randomly selected subset (*n* = 1331, 32.8%) were asked the frequency of feeling optimistic about the future, useful, relaxed, close to other people, dealing with problems well, thinking clearly, making up their own mind about things in past two weeks [28]. The Chinese version of SWEMWBS was valid and reliable (Cronbach's alpha 0.85) and indicated excellent fit for the one-factor model (CFI 0.995; RMSEA 0.03; SRMR 0.04; normed-fit index [NFI] 0.991 [> 0.95 acceptable]) in our previous study using the same sample of the Hong Kong general population [16]. The convergent validity was adequate, with CR of 0.85 and AVE of 0.44.

Sociodemographic and lifestyle-related variables included sex, age, marital status, employment status, educational attainment, monthly household income, smoking, and alcohol drinking. This list of potential confounders was selected based on prior associations with PSU and mental health outcomes [11,29,30].

#### *2.3. Statistical Analysis*

We checked the distributions of all variables independently, with a skewness value of ≤ |2.0| and a kurtosis value of ≤ |7.0| indicating the normality [31]. All data were weighted by age, sex, and educational attainment distribution of the Hong Kong general population using the random iterative method [32]. Missing data were handled by available case analyses as there were minimal missing values for all variables (< 2.6%). We examined the Spearman correlations of anxiety and

depression with mental well-being outcomes, as scores of GAD-2 and PHQ-2 were not normally distributed. Moderate correlation coefficients (*r*) were observed with scores of SHS (*r* range −0.35 to −0.38; both *p* < 0.001) and SWEMWBS (both *r* = −0.42; both *p* < 0.001). We examined the associations of SAS-SV score with the odds of severity of anxiety and depression symptoms using bivariate and multivariable logistic regression analyses adjusting for sociodemographic and lifestyle-related variables. Bivariate and multivariable linear regression analyses examined the associations of SAS-SV score with scores of SHS and SWEMWBS. We further stratified the associations of SAS-SV score with scores of SHS and SWEMWBS by symptom severity of anxiety (GAD-2 cutoff of 3) and depression (PHQ-2 cutoff of 3). The interaction effects of anxiety and depression symptoms on the associations of SAS-SV score with mental well-being outcomes were examined using adjusted Wald tests. Values of *p* < 0.05 were considered statistically significant. All analyses were conducted using Stata/MP 15.1 (StataCorp LP, College Station, TX, USA), except for CFA that were conducted using LISREL 9.30 with diagonally weighted least square estimation.
