Methods
Design and ethics
The present study was an observational prospective cohort study with three measurements (questionnaires) over a follow-up period of 12 months. This study was reported in accordance with the statement from STROBE (Strengthening the Reporting of Observational Studies in Epidemiology) and the Declaration of Helsinki.17 18
Participants and setting
Participants fulfilled the following inclusion criteria: (1) being an active professional player in rugby (Union, League, Sevens); (2) being 18 years old or older; (3) being male; (4) being able to read and comprehend texts fluently in either English, French or Spanish. Professional rugby players were defined as players committing significant time to rugby training and competing at the highest and second highest professional rugby level. The International Rugby Players’ Association asked nine national players’ associations in Australia, England, France, Ireland, Italy, New Zealand, Pacific Islands (including Fiji, Samoa, Tonga), South Africa and Wales to assist in recruiting participants. In addition, the rugby unions from Argentina, Canada and USA were asked to assist in recruiting participants.
Sample size calculation used for descriptive purposes (prevalence and incidence) indicated that at least 196 participants were needed after follow-up (CI of 95%; precision of 5%) to reach an anticipated and assumed population prevalence of 15% of mental health symptoms.19–21 This sample size is sufficient for testing the relationship between the independent and dependent variables under study as sample size calculation indicated that at least 66 participants were needed (n>50+8m where m is the number of independent variables).4 Considering a response rate of 33% (analogous with the response rate achieved in other prospective studies among professional athletes) and hypothesising a loss to follow-up rate of no more than 20%, we needed to invite at least 765 potential participants.20 21
Dependent variables: mental health symptoms
MHS were operationalised by assessing symptoms of distress, anxiety/depression, sleep disturbance, eating disorders and adverse alcohol use. Distress in the previous 4 weeks (baseline) and in the previous 6 months (follow-up) was measured using the Distress Screener (three items scored on a 3-point scale) which is based on the four-dimensional symptom questionnaire (4DSQ) (eg, “Did you recently suffer from worry?”).22 23 The 4DSQ, that is, Distress Screener in English, French and Spanish, has been validated for a recall period of up to several weeks (internal consistency: 0.6–0.7; test–retest coefficients: ≥0.9; criterion-related validity: area under ROC curve ≥0.79).22 23 A total score ranging from 0 to 6 was obtained by summing up the answers on the three items, a score of 4 or more indicating the presence of symptoms of distress.22 23
The 12-item General Health Questionnaire (GHQ-12) was used to assess psychological symptoms related to anxiety/depression in the previous 4 weeks (baseline) and in the previous 6 months (follow-up) (eg, “Have you recently felt under strain?”).5 The GHQ-12 in English, French and Spanish has been validated for a recall period of up to several weeks (internal consistency: 0.7–0.9; criterion-related validity: sensitivity ≥0.70, specificity ≥0.75, area under ROC curve ≥0.83).24 25 Based on the traditional scoring system, a total score ranging from 0 to 12 was calculated by summing the answers on the 12 items, with a score of 3 or more indicating the presence of symptoms of anxiety/depression (area under curve=0.88).24 25
Based on the (short form) Patient Reported Outcomes Measurement Information System (PROMIS), sleep disturbance in the previous 4 weeks (baseline) and in the previous 6 months (follow-up) was assessed through four single questions (eg, “Have you recently had problems sleeping?”) scored on a 5-point scale (from ‘not at all’ to ‘very much’).26 27 The PROMIS in English, French and Spanish has been validated for a recall period of up to several weeks (internal consistency: >0.9; construct validity: product–moment correlations ≥0.96) (for detailed information, see www.nihpromis.org). A total score ranging from 1 to 20 is obtained by summing the answers to the four questions, a score of 13 or more indicating the presence of symptoms of sleep disturbance.26 27
The Eating disorder Screen for Primary care (five items scored as ‘yes’ or ‘no’; ‘0’ for favourable answers, ‘1’ for unfavourable answers) was used as a screening instrument to detect eating disorders in the previous 4 weeks (baseline) and in the previous 6 months (follow-up) (eg, “In the past 4 weeks, were you satisfied with your eating patterns?”).28 The Eating disorder Screen for Primary care has been validated in several languages including English, French and Spanish (criterion-related validity: sensitivity 100%, specificity 0.71). A total score ranging from 0 to 5 is obtained by summing the answers on the five items, a score of 2 or more indicating the presence of eating disorders.28
Level of alcohol consumption at the present time (baseline) and in the previous 6 months (follow-up) was detected using the 3-item AUDIT-C (eg, “How many standard drinks containing alcohol do you have on a typical day?”).29 The AUDIT-C in English, French and Spanish has been validated for a recall period of up to several weeks (test–retest coefficients: 0.6–0.9; criterion-related validity: area under ROC curve 0.70–0.97).30 31 A total score ranging from 0 to 12 was obtained by summing the answers on the three items, a score of 5 or more indicating the presence of adverse alcohol use.29
Independent variables: musculoskeletal injuries and concussion
Participants were asked to report through a single question whether they had sustained musculoskeletal injuries in the previous 12 months (baseline) and in the previous 6 months (follow-up; eg, “How many severe injuries (other than concussion) have you had in the past 6 months?”). In our study, severe injury (not a concussion) was defined as any physical complaint that is sustained by a player during a rugby match or rugby training, irrespective of the need for medical attention or time‐loss from rugby activities, and that leads to either training or competition absence for more than 28 days.32 It was explicitly explained to the rugby players that an injury (defined as stated above) was not a concussion (as explained below) and consequently concussion was not an injury.7 The number of confirmed concussions that occurred in the previous 12 months (baseline) and in the previous 6 months (follow-up) was explored through a single question (eg, for follow-up, “How many concussions (other than muscle or joint injuries) diagnosed by a medical professional did you have in the past 6 months (training and competition)?”).33 The definition of injury and concussion was included in the questionnaire. For these questions, participants were recommended to consult their team doctor.
Procedures
A baseline and two follow-up electronic questionnaires were set up in English, French and Spanish (FluidSurveys), including all dependent and independent variables from the study. In addition, the following descriptive variables were added in the baseline questionnaire: age, height, body mass, duration of professional rugby career, field position (forward or back), level of play, level of education, other employment and family history of mental disorders. Each questionnaire took about 15 to 20 min to complete. Information about the study was sent via email to potential participants by the participating national associations. Players interested in participating in the study gave their informed consent and were given access to the baseline online questionnaire, which they were asked to complete within 2 weeks. At the end of the baseline questionnaire, participants could leave their email address and give their informed consent specifically for the follow-up online questionnaires. Follow-up questionnaires were sent per email 6 and 12 months later, being asked to complete them within 2 weeks. Reminders at baseline and follow-up were sent after 2 and 4 weeks. The responses to baseline and follow-up questionnaires were anonymised for reasons of privacy and confidentiality. The baseline and follow-up questionnaires were linked within participants through a unique code. Once completed, the electronic questionnaires were saved automatically on a secured electronic server that only the principal researcher could access. Players participated voluntarily in the study and did not receive any reward for their participation. Baseline questionnaires were distributed between March and August 2016.
Statistical analyses
All data analyses were conducted with the statistical software IBM SPSS Statistics V.23.0 for Windows. Descriptive data analyses (mean, SD, frequency) were performed with all variables included in our study. To explore whether loss to follow-up was selective, we compared baseline characteristics (all descriptive variables) of non-responders and responders at follow-up by means of independent t-tests.3 Only data without missing values were used for analyses. Two statistical models were used: (1) univariate logistic regression analysis, (2) multivariate logistic analysis (adjusted separately and combined for respectively age and family history of mental disorder). In these models, new injuries/concussion or MHS reported over the course of the follow-up period were considered. OR and 95% CI were calculated. For this analysis, only participants without a given mental health symptom at baseline were included.