Risk factors for intentional doping
Doping moral disengagement
We used the Doping Moral Disengagement Scale – Short (DMDS-S) [18] to assess the extent to which one may rationalise the use of banned substances in sport. The DMDS-S contains six statements describing thoughts that athletes might have about doping (e.g., “Doping is okay if it helps an athlete advice other on how to do it right”) rated on a 7-point Likert scale from 1 (strongly disagree), 4 (neutral), to 7 (strongly agree). Standard translation-back translation strategies were followed to create the Chinese version DMDS-S (see Procedures). We generated mean scores for further analysis, with higher scores indicating greater (risk of) doping moral disengagement in sport.
Doping willingness
We employed the Doping Willingness in Sport Scale (DWISS) [19] to evaluate the level of one’s intention to use prohibited substances. The DWISS assesses athletes’ willingness to dope under eight hypothetical circumstances (e.g., “You thought everyone you were competing against was using a banned substance and getting away with it”) rated on a 5-point Likert scale ranging from 1 (not at all willing) to 5 (extremely willing). Standard translation-back translation strategies identical to the DMDS-S were followed to create the Chinese version DWISS (see Procedures). We generated mean scores for further analysis, with higher scores indicating greater doping willingness.
Self-concepts
Narcissism
We adapted the 16-item Narcissistic Personality Inventory (NPI-16) [26] and the Hypersensitive Narcissism Scale (HSNS) [27] to assess grandiose and vulnerable aspects of narcissism, respectively. More specifically, we adopted a 6-point Likert scale version of the 40-item NPI [28] to replace the forced-choice response of the original NPI. This is because the Likert-scale items yielded better psychometric properties than the force-response items of NPI in its Chinese version [29], whilst recent research has supported the advantages of Likert-scale over forced-choice response in assessing grandiose narcissism using NPI-based measures [30]. Participants rated the extent they agreed on sixteen narcissistic self-statements (e.g., “I like to be the centre of attention”) from 1 (strongly disagree) to 6 (strongly agree).
For the HSNS, we instructed participants to determine the extent to which the ten HSNS statements describe their feelings or behaviours (e.g., “My feelings are easily hurt by ridicule or by the slighting remarks of others”) from 1 (very uncharacteristic or untrue) to 5 (very characteristic or true). We implemented the original HSNS for the UK and US participants and used the Chinese version HSNS [29] which contains identical items of the original HSNS for the Chinese participants. Mean scores were generated for both the NPI-16 and the HSNS, with higher scores indicating greater levels of grandiose and vulnerable narcissism, respectively.
Self-compassion
We measured dispositional self-compassion and fears of compassion in sport using the 12-item Self-Compassion Scale Short (SCS-S) [31] and the 4-item Fear of Self-Compassion in Sport Scale (FSC) [12]. The SCS-based measure is the most commonly used inventory for assessing self-compassion in athletes [32]. Participants rated the frequency they behave in the stated manners described in the SCS-S (e.g., “When I fail at something important to me, I become consumed by feelings of inadequacy”) from 1 (almost never) to 5 (almost always). The Chinese version SCS-S validated in a recent study involving 21 countries [33] was used in this project. We adjusted the reversed items and generated mean scores with higher scores indicating greater self-compassion.
To supplement the use of SCS-S, the FSC offered an alternative, rather opposing or simply contrasting perspective on self-compassion, focusing on an individual’s fearful feeling towards rather than the disposition of being self-compassionate. It contains four items (e.g., “I fear that if I start to develop compassion for myself, I will become dependent on it”) rated from 1 (do not agree at all) to 5 (completely agree). We used the Chinese translation of the FSC items from Matos et al.’s Matos et al.’ (2021) multi-country study of fears of compassion [34] and instructed participants to rate the items considering themselves participating or competing in sport using the contextualised instructions provided by Zhang and McEwan [12]. Mean scores were generated for FSC with higher scores reflecting greater fears.
Psycho-behavioural factors
Resilient coping
We used the Brief Resilient Coping Scale (BRCS) [24] and its Chinese version [35] to assess resilient coping. The BRCS consists of four statements describing one’s behaviour and action under adversities (e.g., “I believe I can grow in positive ways by dealing with difficult situations”) rated from 1 (does not describe me at all) to 5 (describes me very well). Higher mean scores indicated greater resilient coping.
Fear of failure
We adopted the Performance Failure Appraisal Inventory – Short (PFAI-S) [23] and the Chinese translation-back translation version for assessing fear of failure. The PFAI-S consists of five statements describing one’s perceptions and beliefs when performance failure occurs (e.g., “When I am failing, I am afraid that I might not have enough talent”, “When I am failing, important others are disappointed”). Participants rated the PFAI-S items from 1 (do not believe at all) to 5 (believe 100% all the time). We generated mean scores based on all PFAI-S items, with higher scores representing greater fears.
Procedure
We obtained independent ethics approval from the World Anti-Doping Agency (WADA) and the lead author’s institution prior to starting data collection. To allow implementation of the study measures in all study countries (i.e., UK, China, US), we translated questionnaire measures that did not have a validated Chinese version (i.e., DMDS-S, DWISS, PFAI-S) using a translation-back translation strategy. Specifically, two PhD researchers in the field of sport psychology from a Chinese background first independently translated the relevant measures into Chinese. The lead Co-I from China then generated a test Chinese version based on the independently translated versions from the two PhD candidates. A professional Chinese-English translator was then employed to translate the test Chinese version of the targeted measures into English, followed by independent assessments of the back translated versions by the PI and the lead Co-I in the UK. Any potential issues or gaps between the back translated version and the original English version were identified and discussed between the reviewers, with feedback and comments sent back to the Chinese translators (i.e., the two PhD and the lead Co-I in China) for revision until satisfaction was reached between back translated and the original English versions for measures that required a Chinese translation.
With satisfaction translations for required study measures (see Supplementary Material – Scales), we built an online survey using Qualtrics with support from a project research assistant under the supervision of the PI and support from the Co-Is. The online survey distribution in all study countries was centrally managed at the lead institution by a project research assistant with support from the PI. The other Co-I’s facilitated the recruitment of participants in their corresponding country, which included but not limited to contacting gatekeepers for assistance, organising local research student helpers to facilitate participants recruitment and survey distributions. We created country-specific flyers that contains essential study information and a QR code for accessing the online survey. Prospective participants could scan the QR code using their mobile devices or obtain the survey URL link to access the online survey, which started with presenting full details of the study information and request for completing informed consent. The survey system would not direct a participant to the study questionnaires unless a formal consent was given. The whole survey took approximately 15 min to complete. Participants were encouraged to contact the project research assistant should they have any questions about the survey, which would be either answered by the research assistant straightway or forwarded to the PI or the Co-I in the relevant study country for a follow up. Following the completion of each survey, the research assistant would check the survey data for participants’ eligibility in receiving a cash incentive (i.e., e-voucher or prize; £6 or $8 equivalence) and provide a standard debriefing via email.
Data analysis
We used IBM SPSS Version 28 and Mplus Version 8 for data analysis. We first checked missing data and descriptive statistics (i.e., mean, SD, skewness, kurtosis) and examined the composite reliability of each study measure to assess internal consistency. We also examined zero-order correlations of study variables at the country level.
For testing the moderation hypothesis, we examined the three-way interaction between vulnerable narcissism (VN), grandiose narcissism (GN), and self-compassionate mind (i.e., self-compassion/SC, fear of self-compassion/FSC) on risk factors for doping. To explore potential cross-country differences, we conducted two multi-variate (i.e., doping moral disengagement/DM, doping willingness/DW) multi-group (i.e., UK, US, China) moderation models testing the VN × GN × SC and the VN × GN × FSC interactions whilst implementing model constraints on regressive paths to compare whether fixed (i.e., invariant across countries) vs. random (i.e., varying across countries) coefficient(s) achieved better fits (see Table 2). We used the robust maximum likelihood estimator (i.e., MLR) for model testing and compared robust Chi-Square (Rχ2) and degree of freedom (df) with a significant Rχ2 reduction suggesting better model fit [36]. We report comparative fit index (CFI), standardised root mean square residual (SRMR), and root mean square error of approximation (RMSEA) to assess model fit [37] and interpret the model with optimal fit only. For the final retained moderation models, we interpret main effects only when no 2-/3-way interactions were significant; we report 2-way interaction (i.e., VN × GN) only when no 3-way interaction (i.e., the moderation of SC/FSC on VN × GN) was significant. Simple slopes were assessed at mean ± 1SD [38].
For testing the mediation hypothesis, we conducted a multi-variate (i.e., DM, DW) multi-group (i.e., UK, US, China) mediation analysis, testing GN/VN/SC/FSC as predictors when examining their direct and indirect effects on DM/DW via resilient coping/RC and fear of failure/FOF. Identical approaches for cross-country comparison were taken (i.e., comparing models with fixed vs. random coefficients). All these moderation and mediation analyses were conducted with group mean centring of predictor variables and the operation of TYPE = COMPLEX command in Mplus to adjust for the nested nature of data (i.e., athletes nested in coaches) thus accounting for coach-/team-level confounds when assessing outcomes at the athlete level (Level 1) [6, 17, 39, 40]. These analytical strategies were pre-determined when registering for WADA’s Social Science Research Grant (Ref.19271).
