Coming back for more: individual participation patterns in the physical activity initiative parkrun in Australia.

Most physical activity (PA) maintenance research has concerned adherence to small-scale interventions or infrequent observation in cohort studies. We analysed individual attendance trajectories and their drivers in a large-scale 'real-world' community-based weekly PA event (parkrun) cohort in Australia. Data were weekly attendance (walking/running) of 223 224 unique parkrun participants over their first 3 years of participation. An unweighted moving average of participation in the preceding 12 weeks from the 12th week since the first participation to the 156th week was calculated and submitted to a cluster analysis of attendance patterns. Association of individual- (demographic, personal parkrun performance) and site-level (aggregated site-level participant characteristics and area-level measures) covariates with cluster membership was estimated with multinomial logistic regression models. We identified four groups: Few-Timers (76.4%), Decliners (12.4%), Low Maintainers (6.9%) and High Maintainers (4.3%). In the first 12 weeks, attendances averaged 2, 6, 5 and 7.5 times for each cluster, respectively, and by 52 weeks, they were 0.17, 1.9, 3.4 and 7.6 times, respectively. Continuing participation (vs Few-Timers) was strongly associated with faster personal finish times, but slower performance at the site level. Higher running club/group membership at a participant's parkrun predicted higher odds of being a High Maintainer. Our identification of a Low Maintainer group shows a community-based initiative may sustain interest, despite not requiring continuous or near-continuous attendance. Where someone is placed 'in the pack' locally and degree of identification with others in the group may be bidirectionally associated with attendance, underscoring the importance of considering social environment of PA maintenance.