Using a three-wave prospective cross-lagged panel design, the study examined the six-month stability of burnout, self-compassion, and social support among sports coaches in terms of measurement invariance, mean-level change, rank-order stability, and structural stability. The participating coaches (N = 422; M_age = 44.48, SD = 11.03) completed an online questionnaire measuring self-compassion, social support, coach burnout, and demographics at baseline and two follow-ups at three months and six months. The various forms of stability were assessed using structural equation modeling. There was no significant mean-level change in burnout, self-compassion, or social support, and all three constructs exhibited measurement invariance. Rank-order stability remained relatively high, ranging from 0.78 to 0.94 across the three time points. For all three constructs, covariances between latent factors were invariant over time, indicating high structural stability. While self-compassion and social support were positively related, both were negatively related to coach burnout. These results confirm the importance of preventing and addressing symptoms of burnout, low self-compassion, and poor social support in sports settings.

A sample of 422 sports coaches (20.7% female) was recruited in Switzerland through the national professional association of coaches, the national coach education, and the national department for youth and adult sport. 

Participants completed online questionnaires at baseline (T1) and at three and six months (T2 and T3, respectively), answering the same set of questions at each time point. Demographic items (age, gender, sport, level) were recorded only at T1.

We used longitudinal structural equation modeling, beginning with a longitudinal measurement model of three correlated latent factors (self-compassion, burnout, and social support) at the three time points (T1 to T3). For each of these latent variables, we used the item-to-construct balance technique to create parcels (i.e., aggregate-level indicators based on the average of several items) as manifest indicators rather than using single items. 

Variables are explained in the README file.