The GUM intervention was developed and refined based on the GoM pilot study. GoM was conducted from 2015 to 2016 and included 2 middle schools, with a total of 24 participants aged 11 to 17 years. The purpose of the GoM intervention was to examine program and methodological feasibility and gain further insight concerning PA experiences, preferences, and self-concept among at-risk adolescent girls [32]. Findings from the GoM program indicated that the assessments used were appropriate, program delivery (ie, duration, geographic setting, and facilitation) was acceptable, and participants were satisfied with the activities and content. The findings also suggested that adolescent girls are aware of the importance of proper PA and nutrition in relation to health outcomes, believe that peer social support is important for PA engagement, and feel that enjoyment is crucial when dictating PA participation. Based on these outcomes, minor refinements were made to the GUM intervention protocol and research methodology, including the addition of alternative physical activities on and off school grounds, the use of personalized journals for participants to write in, the addition of measures to assess different contexts of social support and the impact of the leaders and facilitators, and the name change from Girls on the Move to Girls United on the Move.

The GUM intervention was a pragmatic PA and psychosocial program designed specifically for at-risk adolescent girls. Components of the self-determination theory, particularly autonomy, competence, and relatedness, were integrated throughout the intervention because these are known to affect intrinsic motivation and engagement in PA, especially among underserved youth [11,33]. The GUM intervention consisted of 9 weekly, group-based, 90-minute face-to-face sessions that were cofacilitated by a trained researcher and a registered social worker from a partnering community organization or key stakeholder. The partnering stakeholder was a nonprofit organization with the goal of bringing an end to violence and poverty and seeking justice for children and women. This nonprofit organization works extensively with adolescent populations, particularly those categorized as at risk. Each GUM group consisted of 7 to 14 at-risk adolescent girls from participating schools. Smaller groups were desired in order to create an intimate and supportive environment and provide a greater opportunity for participants to get acquainted with each other and the program facilitators.

Within each of the weekly 90-minute program sessions, 45 minutes of PA were delivered by a trained researcher and 45 minutes were allocated to social worker–led discussions on various psychosocial topics. Examples of physical/sport activities included on-campus activities such as yoga, team obstacle courses, and soccer, or off-campus activities such as rock climbing, self-defense classes, and fitness classes. Examples of psychosocial topics that were discussed within each group included healthy relationships, conflict resolution strategies, substance abuse issues, sexuality, and gender issues. Table 2 outlines the components of the GUM intervention.

In addition to the group discussions facilitated by the social worker, personal journals were distributed to each participant during week 1. These journals provided participants with an opportunity to disclose their thoughts and feelings about their current state, any personal struggles they may be undergoing, or anything to do with the GUM program itself. Participants returned the journals to the registered social worker at the end of each of the weekly GUM sessions, indicating if they wanted the social worker to read the entry and give a written response. This allowed participants to reach out and seek advice if they needed it, without requiring the participants to talk in person about their issues to either one of the program facilitators. Although not previously used in the GoM pilot study, disclosure journals have been used previously by the social worker and have been reported to be an important component of previous programs delivered to this population. The disclosure journals were used as a component of the intervention, but no data were extracted or evaluated from the journals. After program completion (ie, after week 9), a subsample of participants (30/101, 29.7%) were invited to participate in an individual semistructured interview so that the program facilitators could gain further information concerning program acceptability and satisfaction. Those who attended a minimum of 80% (7/9) of the program sessions were invited to participate in the voluntary semistructured interviews. Participants were informed that they were able to withdraw from the study at any time, for any reason, without consequence. All personal data were coded and handled with confidentiality.

Demographic information consisted of age, grade, number of siblings, housing location, mode of transportation to school, and whether the participant took part in any PA inside or outside of school.

Level of PA was assessed using the Physical Activity Questionnaire for Children, a 10-item questionnaire that assesses the frequency of PA participation over a 7-day recall period [36]. This assessment has been reported to be a valid and reliable measure of PA among children and adolescents as it is able to accurately measure general PA patterns regarding intensity, frequency, and duration [36].

The Self-Compassion Scale (SCS) was used to assess self-compassion through 6 subscales that include contrasting components: self-kindness vs self-judgement, mindfulness vs overidentification, and common humanity vs isolation [37,38]. The SCS is a 26-item scale that measures each of these components using a 5-point Likert scale (1=almost never, 5=almost always). The mean scores of each of the subscales can then be combined to yield a total score, which reflects a global self-compassion score [38].

Perceived social support was assessed using the Child and Adolescent Social Support Scale (CASSS) [39]. The CASSS consists of five 12-item subscales that measure social support received from a variety of sources, including parents, teachers, classmates, close friends, and the school environment [39]. However, a condensed version without the school environment subscale was used, as the actual sources of social support received in the school environment were not well defined and were deemed unnecessary for the purposes of the GUM study [40]. The CASSS measures each of the items on a 6-point Likert scale, ranking each item on how often the participant feels they receive social support from the indicated source, from never (1) to always (6). Additionally, each item has a perceived importance scale, which is measured using a 3-point Likert scale ranging from not important (1) to very important (3). The CASSS has been represented as a valid and reliable measure for its intended purposes [40].

The LCQ [35] was used to assess participants’ perceptions of support provided by the program facilitators. The LCQ is a 15-item measure with values ranging from strongly disagree (1) to strongly agree (7). It provides a good indication of the support that leaders provided for the 3 psychological needs (ie, autonomy, competence, and relatedness) [35]. Research shows that when these needs are supported, higher levels of psychological well-being are evident [35]. The LCQ has been shown to have extensive internal consistency and reliability, and it provides a valid indication of perceived supportiveness among youth outside of sporting contexts [41]. The LCQ was the only self-report questionnaire that was not administered at both T1 and T3; instead, it was distributed at T2.

Sport commitment and enjoyment was assessed using the Sport Commitment Model [42], a questionnaire that has been shown to be a valid and reliable measure of sport commitment and enjoyment among youth [42]. The model addresses the 5 components related to sport commitment, namely sport enjoyment, personal investments in the self-indicated activity, involvement opportunities, social constraints, and involvement alternatives. Each of the 58 items is assessed using a 5-point Likert scale, with scores ranging from strongly disagree (1) to strongly agree (5).

Face-to-face, individual, semistructured interviews were conducted with a subsample of participants (30/101, 29.7%) to gain greater knowledge concerning acceptability of and satisfaction with the GUM program. Participants were asked to share their thoughts, opinions, experiences, and recommendations concerning activities, education, and resources of the program. These interviews were conducted at postintervention by a trained researcher at a time that was convenient to the participants. Interview questions were informed by the study objectives and the research literature specific to this topic area and study population. Data from the interviews were audio recorded and transcribed verbatim with all possible identifiable information removed.

To enhance rigor, various verification strategies were used, such as having the data systematically checked by other trained researchers and closely monitoring gradually learned information as the research project progressed. Table 3 provides a summary of the measures used at the various data collection time points.