Introduction

Childhood and adolescence are marked by large declines in physical activity participation. By the age of 10 years, physical activity levels typically start to decrease at an average rate of 7% annually1. Although participation proportions change at different rates across various types of physical activity2,3, evidence suggests that adolescents are more likely to sustain participation in individual physical activities than group-based activities2. Beyond activity type, it is also important to describe longitudinal participation in physical activities from different contexts. For example, organized (e.g., structured activities requiring a coach/instructor and payment) and unorganized (e.g., unstructured activities without a coach/instructor and limited rules) forms of physical activity may have different probabilities of being sustained over time4,5,6 given unique barriers such as opportunity, availability, and cost7. Yet, sustainability and change in physical activity participation is not well-understood and no recent studies have been published2,8,9, existing evidence is limited by relatively short follow-up periods (e.g., 2 or 3 time points covering 2 to 4 years)3,10,11, and only a few physical activity types and/or contexts have been examined9,10,11. Furthermore, none of these studies identify predictors of sustainability or change in physical activity participation. It is critical to understand the types of physical activities that are the most likely to be maintained and when participation in specific physical activities is most likely to be discontinued to help inform and guide intervention efforts aimed at enhancing physical activity participation throughout adolescence.

One potential predictor of changes in physical activity participation during adolescence is the onset of puberty12,13. Puberty is the transition from childhood to adulthood, characterized by biological maturation. Inter-individual and between-gender differences in puberty timing are associated with multiple biological (e.g., increased height, weight, strength, fat mass) and psychosocial (e.g., self-esteem, body image concerns) changes and challenges for adolescents13,14,15. In relation to physical activity involvement, when comparing same-age youth, earlier-maturing boys might benefit from greater height, weight, and strength advantages16, which could facilitate continued participation14,17. However, earlier-maturing girls are typically less active due to a host of anthropometric and psychological variables including changes in body composition, feelings of self-consciousness13,16 and lower self-concept16,18. While the steep drop off in girls participation in sport around the time of puberty is often anecdotally described as challenges with developmental changes, associations between puberty timing and physical activity discontinuation from unorganized and individual physical activity contexts are currently unclear19.

Age difference between individuals within the same age group might be another factor which could explain variation in physical activity participation during adolescence. To provide developmentally appropriate physical activity experiences20,21, youth are often grouped by chronological age. However, those born earlier in the year (e.g., relatively older) have likely been exposed to a greater number of physical and psychological experiences than their relatively younger (i.e., born later in the year) counterparts22,23. This might represent a performance-related advantage for relatively older individuals within a same age group. An over-representation of participants from a given birth quartile (e.g., January-March), relative to participants from other birth quartiles (e.g., September-December) among same-age youth has been termed relative age effect20. While relative age effects are likely to depend on individual characteristics (e.g., birth date), task (e.g., specific physical activity), and environment (e.g., cultural importance of activity) constraints23, relative age effects have been documented in individual and team-based sports21, and organized and unorganized contexts22. However, this information is derived from mainly sport-specific cross-sectional studies20,21.

Examining physical activity-specific discontinuation and physical activity-general discontinuation concurrently will improve our understanding of physical activity involvement during adolescence. Therefore, we aimed to describe gender-specific longitudinal involvement, discontinuation, re-engagement, and uptake rates in various types and contexts of physical activity from the ages of 11 to 17 years, and to predict gender-specific discontinuation from organized, unorganized, individual, and group-based physical activities using puberty timing and relative age.

Methods

Data from the Monitoring Activities of Teenagers to Comprehend Their Habits (MATCH) study24 was used to test the research questions. Briefly, MATCH is an ongoing longitudinal study aimed at understanding physical activity behavior from childhood to early adulthood. At study inception (2011), participants in grades 5 or 6 (ages 10 to 13 years) were recruited from 17 schools in New Brunswick, Canada. Schools were purposely sampled to represent a mix of urban/rural locations and a variation in socioeconomic status. For the first 24 survey cycles, self-reported data were collected three times per year (September, January–February, and May–June) during the school year until the end of high school (2018). Since this analysis is aimed to better understand age-related discontinuation, re-engagement, and uptake rates of physical activities, we excluded grade 5 data so that the start point of analyses was grade 6 for all participants. To ensure accurate descriptions of physical activity discontinuation timing, we also excluded participants with gaps of ≥ 1 year (3 cycles) between consecutive data collection cycles. Ethics approval was obtained from the Université de Sherbrooke ethics committee and the study was conducted in accordance with recognized ethical standards and national/international laws. All participants provided written informed assent, and parents provided written informed consent.

Measures

Physical activity

To capture youths’ leisure-time involvement in physical activity, participants reported all physical activities they took part in outside of gym class at each cycle (every four months) using a checklist of 36 activities. Participants indicated the frequency (i.e., never, once a month or less, 2–3 times per month, once a week, 2–3 times per week, 4–5 times per week, or almost every day), and with whom (i.e., alone, with friends, with parents and/or siblings, or with an organized group or team) they most often participated in each activity. We only considered regular physical activities (e.g., at least once a week) to avoid counting spurious activities25. In addition, indoor and outdoor chores were excluded since they can be viewed as nonleisure/nonvolitional26. Walking was also excluded because it was reported by nearly all participants at each cycle. A detailed list of physical activities and their categorizations is presented in the Appendix.

Each physical activity was also categorized as organized or unorganized, using a validated method for the MATCH study27. Briefly, seven physical activities are classified as unorganized (home exercises, trampoline, games, skipping rope, weight training, indoor and outdoor chores). The remaining 29 activities are categorized as unorganized if participants reported taking part in the activity alone, with friends, or with parents and/or siblings. Alternatively, if participants reported involvement with an organized group or team, the activity was categorized as organized.

Using previously published definitions from the MATCH study28, each physical activity was also categorized as individual or group-based. Whereas 24 activities are always categorized as individual, 12 activities (ice hockey, street hockey, ringette, soccer, Canadian football, basketball, baseball, volleyball, handball, dance, ball games, or games) could be categorized as group-based if participants reported involvement with an organized group or team, with friends, or with parents and/or siblings. Otherwise, if participants reported participation alone, the activity was classified as individual.

Predictors of physical activity discontinuation

Puberty timing was categorized as early-maturation, on-time-maturation, or late-maturation, based on participant self-report using the pubertal development scale (PDS)29. All participants self-reported body hair growth using a 4-item Likert scale (not yet started, barely started, definitely started, seems completed). Using the same Likert scale, girls answered questions about breast size, whereas boys responded to questions about voice deepening and facial hair growth. Girls were also questioned whether menstruation had started (yes/no). Scores for each question were summated by cycle and PDS means and standard deviation (SD) were computed for girls and boys separately. Then, as was done previously30, we classified participants as early-maturation if their individual PDS score was one SD higher than the age- and gender-specific PDS mean, late-maturation if their individual PDS score was one SD lower than the age- and gender-specific PDS mean, or on-time-maturation for participants whose PDS score were within one SD of the age- and gender-specific PDS mean. The PDS demonstrates criterion validity with intraclass correlations (ICC) between physician assessment and self-rating [ICC (95% CI) 0.75 (0.72–0.85)] for girls and [0.72 (0.55–0.81)] for boys and has Cronbach’s α internal consistency coefficients for self-rating of 0.93 for girls and 0.91 for boys31.

To align with the majority of sport birthdate cut-offs, relative age was computed by classifying participants into Birth quartiles (i.e., Q1: January–March; Q2: April–June; Q3: July–September; and Q4: October–December) based on their birthdate20,21. Participants might take part in more than one activity with different birthdate cut-offs for registration. For example, in the province of New Brunswick, birthdate cut-offs for ice hockey (December 31st) are different than those for golf (August 1st) but participating in one activity does not preclude involvement in the other. Therefore, to understand the influence of activity type, sensitivity analyses (results not shown) were conducted where activities with birthdate cut-offs different than December 31 were excluded (e.g., ice skating, golf, and swimming). Since similar results were obtained in both sets of analyses, the current analysis includes all activities regardless of registration cut-off date.

Data analysis

We calculated the proportion of participants reporting involvement in each physical activity and organized, unorganized, individual, and group-based activities at each grade. We used the PROC LOGISTIC procedure in SAS to assess linear and quadratic trends in these proportions using logistic regressions. Specifically, we used participation (yes/no) in each activity as the outcome and tested age as a linear term. In a separate set of models, age was included both as a linear and quadratic term.

To describe discontinuation in each physical activity, and in organized, unorganized, individual, and group-based activities from grade 7 to grade 12, we determined the number of months that participants in a given physical activity in grade 6 (those who reported the activity at least once per week) continued to report participation in that activity. For each physical activity, we considered discontinuation to have occurred once participation in the given physical activity was not reported for at least 1 year (up to 3 consecutive data collection cycles) after grade 6. For organized, unorganized, individual, and group-based activities, we considered discontinuation to have occurred once participants no longer reported any activity classified into these groupings for at least 1 year after grade 6. Discontinuation rates were calculated with Poisson regressions and are expressed as incidence per 1000 person-month of follow-up with 95% confidence intervals using the PROC GENMOD procedure in SAS. To visualize discontinuation rates for each physical activity, we also produced a timeline indicating the months since baseline when most (i.e., ≥ 50%) participants discontinued participation.

Among participants who discontinued a given physical activity, re-engagement was calculated as the proportion of participants who reported involvement in that same activity at least once per week after an interruption of over one year. Finally, among those who were not involved in a physical activity in grade 6, uptake was calculated as the percentage of participants who started reporting participation in that physical activity in subsequent grades.

To assess puberty timing and relative age as predictors of discontinuation from organized, unorganized, individual, and group-based physical activities, we estimated gender-stratified bivariate Cox proportional hazard models using the “survival” package in R32. Discontinuation was considered to have occurred when an individual had a one year or more interruption in meeting the definition of participating in organized, unorganized, individual, or group-based activities. Time until discontinuation was computed as the number of months from the data collection dates between first involvement in a given physical activity type in grade 6 until the last time it was reported before discontinuation. Models were constructed separately for puberty timing and relative age. Puberty was treated as a time-dependent variable since pubertal status relative to peers might change across school grades whereas birth quartile was considered as time-invariant. Investigation of the proportional hazard assumption were conducted by visual inspection of Schoenfeld residuals and hypothesis testing of whether the effect of exposure differed over time. Analyses were conducted in SAS 9.4 (Cary, NC, USA) and in R version 4.2.1.

Results

Grade 6 physical activity participation data were available for 781 participants (84% of total sample; 57% girls) and were retained for analyses. These participants took part in a median (interquartile range) of 14 (8–20) data collection cycles. Participants were, on average, 11.5 (0.4 SD) years old at study onset and 17.4 (0.3) at study end. In grade 6, most participants (56.5%) had on-time puberty, whereas 14.2% were early-maturing and 29.3% were late-maturing. Twenty-five percent of participants were born between January and March (Q1), 27% between April and June (Q2), 22% between July and September (Q3), and 25% between October and December (Q4). There were no differences in participants’ gender between those who took part in < 5, < 10, or ≥ 15 cycles.

Prevalence of physical activity participation by school grade

Nearly all participants took part in at least one unorganized activity or individual activity in grade 6 (Table 1). In grade 6, the three most frequently reported activities among girls were bicycling (62%), games (chase, tag, hide and seek) (56%), and jogging/running (55%). For boys, the top activities were bicycling (78%), jogging/running (56%), home exercises (pushups, sit-ups) and soccer (54%, each). In grade 12, the top activities for girls were home exercises (46%), jogging/running (31%), and weight training (26%), whereas the top activities for boys were weight training (46%), home exercises (42%), and jogging/running (34%). Linear trends suggest that participation in most specific physical activities declined for both genders (Table 1). However, the proportion of boys and girls reporting aerobics and girls reporting home exercises remained stable from grades 6 to 12. The rates of decline in participation changed (significant quadratic trend) for unorganized activities, badminton, trampoline, and volleyball in both girls and boys. For most physical activities, the point at which the probability of discontinuing participation reached 50% occurred in the first years of follow up (Fig. 1).

Table 1 Percentage of participation in specific types of physical activities from grades 6 to 12 in the MATCH study (2011–2018).
Full size table
Figure 1
figure 1

Number of months at which the probability of discontinuing partcicipation in specific physical activities reaches 50%.

Full size image

Discontinuation of physical activity

By grade 12, 59% of girls were still involved in an individual activity, but only 46% had maintained their involvement in team-based activities (Fig. 2A). For boys, 68% maintained participation in individual activities throughout school years, while 51% continued with team-based activities (Fig. 2B). Slightly more than half (54%) of girls remained involved in unorganized and organized activities by grade 12 (Fig. 2C). More boys maintained their involvement in unorganized activities than organized activities (72% vs. 53% respectively in grade 12; Fig. 2D)).

Figure 2
figure 2

Probability of sustaining participation in organized, unorganized, team, and individual activities for girls and boys in the MATCH study.

Full size image

Among activities with at least 50 participants in grade 6, the physical activity most likely to be sustained for girls was ringette (incidence rate per 1000 person-month follow-up (IR), 95% CI 15, 10–22; Table 2), whereas for boys it was ice hockey (IR, CI 6, 5–9).

Table 2 Incidence rates (IR) and 95% confidence intervals (CI) per 1000 person-month follow-up of discontinuation among those who reported taking part in the activity in grade 6.
Full size table

Re-engagement and uptake of physical activities

For girls and boys, the three activity types most likely to be re-engaged after discontinuation were home exercises, weight training, and jogging/running (55%, 36%, 38% [girls]; 43%, 40%, 58% [boys]), respectively; Table 3). Otherwise, less than 1 in 4 participants re-engaged in activities after discontinuation.

Table 3 Percentage of participants who re-engaged in each activity after discontinuing for ≥ 1 year.
Full size table

The list of activities most likely to be initiated after grade 6 is similar to those with the highest level of re-engagement. For both girls and boys, bicycling, home exercises, jogging/running and weight training were activities most frequently associated with an uptake after grade 6 (51%, 60%, 62%, 31% [girls]; 62%, 47%, 52%, 49% [boys], respectively; Table 4).

Table 4 Percentage of uptake among participants who did not report the activity in grade 6.
Full size table

Association between puberty timing and relative age and discontinuation from various physical activities

Girls reporting early-maturation were more likely to discontinue organized sports than girls with on-time-maturation status (Hazard Ratio (HR), 95% confidence intervals (CI) 1.68 (1.05–2.69); Table 5). Puberty timing was not associated with dropout in boys.

Table 5 Hazard ratios and 95 confidence intervals for dropout of organized, unorganized, team, and individual physical activities for girls and boys according to puberty timing‡.
Full size table

Birth quartile was not associated with physical activity discontinuation in girls (Table 6). Compared to boys born in Q4, boys born in Q2 were less likely to discontinue from organized (HR [95% CI] 0.41 [0.23–0.74]), unorganized (HR [95% CI] 0.33 [0.16–0.70]), individual (HR [95% CI] 0.58 [0.34–0.98]), and group-based (HR [95% CI] 0.46 [0.23–0.91]) physical activities. Being born in Q3 was also associated with a decreased likelihood of discontinuation from unorganized physical activity (HR [95% CI] 0.41 [0.19 to 0.88]).

Table 6 Hazard ratios and 95 confidence intervals for dropout of organized, unorganized, team, and individual physical activities for girls and boys according to birth quartile.
Full size table

Discussion

In this study, we documented that participation in most physical activities decreased from ages 11 to 17 years. Our results also highlight that although most group-based and organized activities were not re-engaged after discontinuation, re-engagement was considerably more likely for individual-based and unorganized activities. This said, despite high rates of dropout from most specific physical activities during adolescence, many participants still maintained involvement in some individual and unorganized physical activity by the end of high school. Some activities had relatively higher likelihoods of being sustained throughout adolescence than others. We also found that earlier-maturing girls had a higher risk of discontinuing organized activities than other girls, and that boys born between April and June had a lower risk of discontinuing organized, unorganized, individual, and group-based activities.

Participation in most physical activities included in this study declined during adolescence. The similarity between the results of the current study and those conducted a decade ago suggest that teenage physical activity participation patterns have remained largely similar2,3. In line with the theory of biological regulation33, it is possible that declines in physical activity during adolescence are simply a natural part of aging. Despite this, our results clearly indicate that individual activities are generally sustained longer than group-based activities and that unorganized activities are more likely to be sustained than organized activities. Whereas previous studies also suggested that individual activities have a higher likelihood of being sustained longer than group-based activities2, this is the first study to describe sustainability in organized and unorganized activities during adolescence. These results highlight the importance of exposing and facilitating access to individual and unorganized physical activity for adolescents. Most specifically, our results point to fitness-focused activities (e.g., home exercises, weight training, running/jogging) as those with some of the highest potential for uptake, re-engagement, and maintenance during adolescence. Fitness-focused activities are also among the most frequently reported by active young adults34, which raises the hypothesis that they could contribute to lifelong physical activity participation if started during adolescence. These activities may display the most promise for long-term participation because they are associated with relatively few organizational barriers and are low-cost7, which also makes them appealing as targets for wide-reaching interventions.

Our observation that early maturation in girls predicts discontinuation from organized physical activities is in line with previous studies reporting an association between sport participation and puberty timing35,36. It is possible that girls’ earlier changes in physical factors, including increased fat mass, breast development and widening of the hips37 is accompanied by changes in psychosocial factors such as self-consciousness13,16, lower self-concept18 or teasing38 leading to increased likelihood of dropout. Such changes modulate girls’ idea of ideal body shapes and sizes and, therefore, often generate dissatisfaction with their bodies39. In turn, sub-optimal self-perceived body image is linked to lower physical activity levels, potentially owing to the avoidance of situations in which girls may be judged for their appearance39. In this respect, it may be advisable to promote physical activities that involve fewer opportunities for physical comparisons with peers (i.e., individual and unorganized physical activities) in advance of puberty so that girls can be active without worrying. Consequently, it is important to encourage girls to stay active and overcome self-imposed barriers associated with body image or misconceptions about how they should look during physical activities, as early evidence shows that interventions can improve satisfaction with body-image and increase intentions to engage in physical activity40.

In contrast, we found no associations between puberty timing and risk of dropping out of physical activities among boys. This conflicts with results from studies suggesting that boys with late-maturation are more likely to dropout from sports41,42,43,44. However, previous studies did not account for the impact of uptake of new physical activities following drop out. It is, therefore, possible that puberty-associated physical activity dropout among boys is nullified by engagement in a different activity.

Birth quartile was a predictor of physical activity drop out among boys, but not girls in this study. While relative age effects are typically identified as an increased proportion of participants born in Q1 vs. Q4, some activities display Q2 vs. Q4 differences20, similar to results found in this study. However, finding that birth quartile was protective against unorganized physical activity discontinuation was unexpected since unorganized physical activity is usually practiced by individual volition27 and therefore would not include a selection process. This suggests that talent selection is not the only driver of relative age effects23 and warrants future investigation into correlates of relative age effects and how they might present differently in unorganized physical activity than in organized physical activity contexts. The lack of relative age effects among girls in this study is different than previous findings21, but might reflect girls’ lower likelihood to participate in competitive sports involving a selection process.

The application of frequent assessments over a 7-year period provided the unique opportunity to present a comprehensive overview of changes in participation in various physical activities among adolescents. In addition, investigating sustainability in organized, unorganized, individual, and group-based physical activity provided information on domain-specific dropout. These data also allowed describing the contribution of puberty and relative age as predictors of change in physical activity participation. It nevertheless needs to be recognized that all measures were based on self-report and could therefore be associated to some over/under-estimation. Also, some activities classified as individual might have been taken part with others (e.g., swimming). Given our aim of documenting change in participation in different physical activity types during adolescence, the classification scheme of individual and group-based activities precluded the study of social context of participation. Future investigations of physical activity discontinuation that consider the social context of participation is warranted, given the importance of the social environment for the maintenance of physical activity45. In addition, since we defined discontinuation as not reporting an activity for at least 1 year, we were unable to capture activities that were discontinued during the final year of the study. We were unable to control for confounding factors such as BMI in associations between puberty timing and physical activity discontinuation, due to data unavailability. Further, although participants retained for the MATCH study were purposely sampled to represent a mix of urban/rural locations and a variation in socioeconomic status within a province, these results may not generalize to other samples from other countries given national and regional cultural differences in physical activity and sport participation. Researchers from other countries and/or areas could use the current methodology to replicate this study in other regions.

In conclusion, this study highlights that although there is a marked decline in participation in most physical activities during adolescence, general participation in physical activity may persist through sustainment of some activities, uptake of new ones and re-engagement in others. Physical activities most likely to be sustained, re-engaged after discontinuation, or initiated during adolescence were mostly unorganized and individual activities. This suggests that these activities have a potential to be carried over into adulthood. Specifically, interventions aimed at fostering individual and unorganized activity participation among adolescents are likely worthwhile since these types of activity are maintained longer and are more likely to be reengaged in than organized and group-based activities.