Use and consequences of exercise tracking technology on exercise psychopathology and mental health outcomes in adolescents

ABSTRACT

Exercise tracking technology use is associated with exercise psychopathology in adolescents; however, research is yet to identify components of such technology that can predict maladaptive exercise at this age. This research assessed the relationship between exercise tracking technology use and exercise psychopathology in adolescents. Development of a new measure of exercise tracking behaviours/attitudes was also conducted. Adolescents (N = 327; aged 12–15, mean = 13.64 years (SD = .95); n = 168 girls) participated in this multi-phase study. Following factor analysis to develop and validate the new measure, relationships between exercise tracking behaviours/attitudes and compulsive exercise were explored. Key components of such technology (e.g. pressure to achieve exercise-related goals) were significantly associated with higher compulsive exercise in adolescents. However, using technology to simply monitor their own exercise behaviours was significantly associated with positive exercise and mental wellbeing outcomes. Prospective research should assess how exercise tracking can predict exercise psychopathology changes and mental wellbeing throughout adolescent development.

KEYWORDS:

• Exercise tracking
• fitness wearables
• exercise technology
• compulsive exercise
• young people
• adolescent

Introduction

Physical activity has many health benefits, such as supporting cardio-metabolic and bone health, and enhancing mental wellbeing and body satisfaction (e.g. Biddle & Asare, 2011; Fernández-Bustos et al., 2019; Janssen & LeBlanc, 2010). Adolescence, defined by the World Health Organization (WHO, 2023) as the period between 10 and 19 years of age, is a key developmental phase for establishing health behaviours (e.g. eating and exercise). However, more than 80% of adolescents, globally, do not reach the recommended levels of physical activity (81%; World Health Organization [WHO], 2023), with a particular decline in physical activity between 12 and 15 years (e.g. Marques et al., 2020), and concurrent increase in maladaptive health behaviours (e.g. disordered eating, Breton et al., 2022). There are a wide variety of efforts to increase exercise behaviours in young people, with exercise tracking technology (ETT) such as wearable devices (e.g. Fitbit) and mobile applications (e.g. Strava) often forming a key component of interventions (e.g. Böhm et al., 2019). Such tools have been considered beneficial in increasing adolescents’ exercise behaviours (e.g. Carlin et al., 2015), through using common self-regulatory behaviour change techniques such as allowing users to track and monitor their exercise behaviours, and to set targets and promote achievements (e.g. Dute et al., 2016; Gordon et al., 2019).

Self-monitoring and goal-setting behaviours are considered key components of physical activity behaviour change interventions that can increase activity (e.g. Michie et al., 2009; Ridgers et al., 2016). Higher ETT use has been associated with higher levels of exercise activity in adolescents and young adults (e.g. AlSayegh et al., 2023; McFadden & Li, 2019), with sustained physical activity behaviour changes occurring when ETT provides users with positive feedback on exercise activities (Edwards et al., 2013). However, the reciprocal nature of ETT use and exercise frequency remains unclear. While some research has identified that ETT use only enhances exercise levels in adolescents who are already physically active (e.g. Gaudet et al., 2017), other studies have suggested that ETT use only enhances exercise in physically inactive adolescents (e.g. Creaser et al., 2021). Approximately 30% of adolescents report using ETT (e.g. Creaser et al., 2023). However, research has highlighted mixed findings in relation to continued ETT use with adolescents. A recent systematic review found that while ETT increased motivation for adolescents to exercise via self-monitoring and goal-setting, sustained use was limited due to technical issues and novelty of the ETT (Creaser et al., 2021). Some adolescents report a lack of motivation to continue with ETT due to a perceived pressure to achieve the prescribed exercise-related goals (Kerner & Goodyear, 2017; Reynolds, Plateau, et al., 2022), while other adolescents report continued use of ETT due to feelings of guilt when not achieving the prescribed goals, or to make up for missed exercise sessions (Reynolds, Plateau, et al., 2022). These feelings are characteristic of unhealthy relationships with exercise (e.g. Taranis et al., 2011).

Compulsive exercise is defined as a severe or intense drive to exercise, despite illness or injury and is often co-morbid with eating disorders in adolescents (e.g. Stiles-Shields et al., 2012). Indeed, approximately 40% of adolescents diagnosed with an eating disorder report engaging in compulsive exercise (e.g. Levallius et al., 2017). The multidimensional model of compulsive exercise highlights key components of exercise psychopathology such as exercising for weight control and rigidity of exercise behaviours despite lack of enjoyment or illness/injury (Meyer et al., 2011). Whilst dedication to training and routine around exercise is important in athlete populations (e.g. Plateau et al., 2014), features of compulsive exercise (e.g. exercising for weight control and to manage negative mood) have still been found to be relevant in this population. Importantly, more frequent engagement with ETT has been found to significantly predict higher compulsive exercise outcomes in adolescents, with introjected regulation (e.g. internal sense of obligation to exercise) a key predictor of compulsive exercise in this instance (Bratland-Sanda et al., 2022). Similar findings are apparent in young adults, such that ETT users report higher levels of compulsive exercise and disordered eating behaviours and attitudes (e.g. Hahn et al., 2022; Plateau et al., 2018), and pressure from others to lose weight significantly predicts both eating and exercise tracking behaviours in young adults (O’Loughlin et al., 2023). As adult eating and exercise behaviours and attitudes are often established in adolescence (e.g. Patton et al., 2016), and adolescents are particularly susceptible to social pressures in their environment (e.g. Blakemore & Mills, 2014), it is crucial to explore the relationships between key components of ETT use and exercise psychopathology in adolescents.

To date, research aiming to explore ETT use has focused predominantly on frequency of exercise tracking behaviours. While qualitative research has begun to identify the attitudes and motives behind ETT use in adolescents, such as exercise monitoring and guilt around not attaining the prescribed exercise-related goals, there is no measure available to quantitatively assess these key components of ETT use. Given that a significant association has been identified between ETT use and compulsive exercise in both young adults and adolescents, having a measure that assesses ETT motives and attitudes will enable researchers to assess the potential wider impact of exercise interventions that employ ETT in adolescents, prior to any consolidation of such maladaptive exercise behaviours and attitudes in adulthood. A measure of this nature will also facilitate the identification of potentially problematic relationships with ETT use in adolescents and provide a clearer understanding of the motives and barriers associated with ETT on exercise and eating behaviours in adolescents over time.

Research aims

The aims of this study were threefold. The first aim (Part 1) was to construct and validate a self-report tool designed to measure exercise tracking behaviours and attitudes (e.g. goal setting and monitoring, and perceptions around ETT use) in adolescents: Exercise Tracking in Adolescents Questionnaire (ETAQ). The study aimed to identify the factor structure of the ETAQ using exploratory factor analysis (EFA) and to refine the item pool.

Part 2 sought to use the ETAQ to cross-sectionally explore relationships between exercise tracking and compulsive exercise in adolescents. It was hypothesized that, after accounting for related constructs (e.g. disordered eating, mental wellbeing), higher ETAQ scores would statistically predict higher compulsive exercise in adolescents.

Confirmatory factor analysis was conducted (Part 3) to explore the construct validity of the ETAQ and to confirm the factor structure identified through EFA.

Part 1: Development of the ETAQ and item refinement via exploratory factor analysis

Method

Initial development of the ETAQ

The ETAQ was designed to assess exercise tracking behaviours and attitudes in adolescents. Initial item generation was informed by focus group discussions with adolescents aged 12–16 years (see Reynolds, Haycraft, et al., 2022) and via a systematic review of the existing literature (e.g. Reynolds, Plateau, et al., 2022) and engagement with experts in the area. Constructs such as exercise monitoring, feelings of guilt when not reaching prescribed ETT goals, and social comparison of exercise activity were therefore incorporated in initial item development. Eighteen items were initially included, and respondents were asked to rate how true each statement was for them on a Likert scale, with anchors 1 (never true) to 5 (always true). Higher scores indicate greater (and potentially problematic) engagement in exercise tracking behaviours (e.g. higher monitoring of exercise activity) and attitudes (e.g. higher social comparisons, negative affect).

The 18-item version of the ETAQ was pilot tested with a separate sample of adolescents (n = 60) to gather feedback on comprehension and relevance of the questionnaire items. Feedback via semi-structured interviews, focus group discussions, or open-ended and closed feedback questions was collated after adolescents had completed the measure. Content analysis was employed to analyse feedback and percentages for multiple choice (closed) questions were calculated to analyse frequencies of responses. Feedback received was positive, for example, all adolescents reported understanding the questionnaire items and were totally or fairly sure what to do when following the questionnaire instructions. As feedback on the items was favourable, no further changes were made to the measure at this stage.

Participants and recruitment

The participants for this study were a part of a larger study (Reynolds et al., in press). Participants in the larger study were provided with the closed question: ‘do you use exercise tracking technology (e.g. Fitbit, Apple watch, Strava, exercise apps etc…)?’. Only those who reported using ETT by selecting ‘yes’ (36.1%, n = 327) were included in this study. Using purposive sampling, adolescents aged 12–15 years were recruited from five secondary schools in the East Midlands, UK, via an email sent to the headteacher. Schools were recruited from diverse areas, according to postcode. Using the Ministry of Housing (2019) deprivation indices (1=most deprived, 10=least deprived), recruited schools scored between 2 and 9. Headteachers notified adolescents from one or more of the following year groups about the opportunity to take part in the study: year 8 (12–13 years), year 9 (13–14 years) and/or year 10 (14–15 years).

Procedure

Following institutional ethical review board approval [2022–4960–8593], each school was provided with electronic or paper study information letters for parents. Parent consent was obtained on an opt-out basis. Schools could choose between a paper or online version of the questionnaire. Questionnaires were completed by participants during school time, and supervised by their classroom teacher. Participants provided informed consent and were instructed by their teacher to complete the measure independently, without discussing their responses with others. Paper questionnaires were returned to the teacher and collected by the researcher. In addition to the ETAQ, participants provided demographic information (gender, age, ethnicity), and also completed measures of related constructs (e.g. eating and exercise psychopathology, physical activity, psychological wellbeing). Participants were invited to report the type of ETT they used (free-text response). Questionnaires took approximately 30–45 min to complete.

Results

Descriptive statistics

A total of 327 adolescent participants (boys n = 150, girls n = 168, ‘prefer not to say’ n = 9) were recruited. The mean age of the participants was 13.68 years (SD: 0.99, range 12–15 years). Over half of the participants (56.6%, n = 185) identified their ethnicity as White British, 20.8% (n = 68) identified as Asian British, 6.4% (n = 21) Asian other, 5.2% (n = 17) ‘Other’, 4.9% (n = 16) White other, 2.8% (n = 9) Black other, and 0.6% (n = 2) Black British. Ethnicity was not reported by nine (2.7%) participants.

A total of 76 (23.2%) adolescents reported using wearable ETT. Of these participants, the majority wore a Fitbit (n = 34, 44.7%), or an Apple watch (n = 30, 39.5%). Of the 21 (6.4%) adolescents who reported using a mobile application, most did not report the app they used (n = 13; 61.9%), with a further six reporting using Strava (28.6%). The remaining adolescents (70.3%; n = 230) in the sample did not disclose the type of ETT they used, but just indicated that they used it.

Exploratory factor analysis

To explore the factor structure of the ETAQ, exploratory factor analysis (EFA) was conducted using IBM SPSS Statistics 27, through principal axis factoring. As it was assumed that correlations would exist between factors, as all items reflected exercise tracking behaviours and attitudes, a direct oblimin rotation method was conducted (Field, 2018).

As the determinant was above 0.00001 for correlations between measure items, multicollinearity in the dataset was not a cause for concern (Field, 2018). Bartlett’s Test of Sphericity was significant (p < .001) for the ETAQ, suggesting that correlations between variables were significantly different from zero and the variables were adequately related to find clusters within the dataset (r=.120–.714). The Kaiser–Meyer–Olkin (KMO) measure of sampling adequacy scored .900, suggesting a ‘marvellous’ sample size (Kaiser & Rice, 1974), and was above the recommended score of .50 for EFA (e.g. Field, 2018).

Item retention and elimination

Item retention and elimination followed several criteria. Factor loadings ≥ .40 are considered substantial for EFA (e.g. Field, 2018), and it is advised that all factors have a minimum of three, non-cross-loading items (e.g. Knekta et al., 2019). Items with loadings less than .40 were therefore removed, in addition to instances where there were fewer than three items on a factor, or where cross-loading was evident. These criteria resulted in four items being removed due to item loadings less than .40 (see Supplementary Material 1), leaving 14 items for factor analysis.

Analysis of the remaining items

Principal axis factoring was conducted with these 14 items on the ETAQ (direct oblimin rotation). According to Kaiser’s (1960) criteria, factors with an eigenvalue greater than 1 should be retained. Four factors with an eigenvalue greater than 1 were identified, which explained 68.50% of the total variance (see Supplementary Material 2). In contrast, the scree plot was ambiguous and showed inflections that would satisfy a three to five-factor structure (Cattell, 1966). Eigenvalues can tend to overestimate the number of factors (e.g. Izquierdo et al., 2014) and scree plot analysis inflections reflect an element of subjectivity, leading researchers to disagree on the number of factors (e.g. Norman & Streiner, 2014). Parallel analysis (Horn, 1965) was therefore also conducted. Parallel analysis statistically simulates a random dataset with the same number of participants and variables as the real dataset to create a set of eigenvalues. Parallel analysis suggested a three-factor structure. To reach a plausible factor solution, it is suggested that the final factor structure is developed from the exploration of a combination of methods (e.g. Izquierdo et al., 2014; Lloret et al., 2017). The eigenvalues, scree plot and parallel analysis were therefore analysed in turn and then combined. Based on these analyses, all four factors were retained for further analysis. All items were deemed to conceptually fit within their loading factor and all item loadings were greater than .40. Factor loadings, means and ranges of scores (1–5) for the four-factor structure were evaluated using expert guidance and are presented in Supplementary Material 2.

Reviewing and naming the subscales

Factor one (5 items) included questions related to monitoring exercise in accordance with the prescribed exercise-related goals on devices (e.g. ‘I regularly use my exercise tracking device to monitor my exercise’) and was labelled ‘Exercise Monitoring’. Factor two (3 items) included items related to feelings of guilt when not reaching prescribed exercise-related goals on the devices (e.g. ‘I feel guilty if I do not reach my exercise goals’) and was labelled ‘Exercise-Related Obligation’. Factor three (3 items) reflected guidance of exercise behaviours (e.g. ‘I use my exercise tracking device/apps to guide how much exercise I should do’) and was therefore labelled ‘Exercise Guidance’. The fourth factor (3 items) included items related to comparing exercise behaviours with others (e.g. ‘I use my exercise tracking device/apps to compare how much I exercise with others’) and was labelled ‘Exercise Comparison’.

Internal consistency

Internal consistency was calculated for the global score and each subscale of the ETAQ. All four subscales were found to have good (range = .80–.84) internal consistency (e.g. Nunnally & Bernstein, 1994; see Supplementary Material 2) and the global Cronbach's alpha score was good (α = .89). Moderate, significant positive correlations were also identified between each of the four factors of the ETAQ (r = .306–.609, p < .001; see Supplementary Material 3).

Summary: Part 1

The first part of this study aimed to develop and validate a new measure of exercise tracking behaviours and attitudes in adolescents (ETAQ). EFA identified four factors: (i) exercise monitoring; (ii) exercise-related obligation; (iii) exercise guidance; (iv) exercise comparison. All factors demonstrated good internal consistency (Nunnally & Bernstein, 1994; Ursachi et al., 2015). The next step, described in Part 2, was to establish the convergent validity of the ETAQ with constructs such as eating and exercise psychopathology, and to explore the predictive validity of the ETAQ in relation to compulsive exercise outcomes, eating practices and mental wellbeing in adolescents.

Part 2: Convergent validity of the ETAQ and exploring exercise tracking as a predictor of compulsive exercise

Method

Participants and procedure

The same participant sample as in Part 1 was used for Part 2 of this study. However, only participants who completed the ETAQ and the Compulsive Exercise Test (CET; Taranis et al., 2011) were included in this part, as these were the key variables of interest for analyses. Six participants were therefore removed for Part 2. The sample for this part therefore comprised 321 adolescents aged 12–15 years (mean = 13.55; SD = .96; male: n = 147; female: n = 165; ‘prefer not to say’: n = 9).

Measures

An overview of the measures completed in addition to the ETAQ, in the order in which they were completed, is provided below. A total score on the ETAQ was calculated as the mean of the four subscales (potential score range: 1–5; see Supplementary Material 1).

Compulsive exercise test (CET; Taranis et al., 2011)

The CET comprises five subscales which measure an individual’s compulsivity towards exercise: (1) avoidance and rule-driven behaviour; (2) weight control exercise; (3) mood improvement; (4) lack of exercise enjoyment; and (5) exercise rigidity. The CET contains 24 items, each with a 6-point Likert scale ranging from 0 (never true) to 5 (always true). Greater scores represent greater levels of compulsive exercise. The CET has shown good levels of internal consistency with adolescent populations (e.g. α = .88 for boys, α = .89 for girls; Goodwin et al., 2014). Internal consistency for this sample for the total score was .89, with individual subscales ranging from .72 to .90.

Dutch eating behavior questionnaire – restrained eating subscale (DEBQ; van Strien et al., 1986)

The DEBQ measures three eating styles: (1) emotional eating; (2) external eating; and (3) restrained eating. As restrained eating is the component of disordered eating most strongly associated with compulsive exercise (e.g. Dalle Grave et al., 2008), this subscale was used for the present study (10 items). Each item is measured using a 5-point Likert scale ranging from 1 (never) to 5 (very often). An overall restrained eating score is calculated as the mean of the 10 items. The DEBQ restrained eating subscale has shown good levels of reliability with adolescent samples (e.g. α = .88; Hunot-Alexander et al., 2019). Internal consistency for the current sample was .96.

Short Warwick-Edinburgh mental wellbeing scale (SWEMWBS; Stewart-Brown et al., 2009)

The WEMWBS (14 items) measures mental wellbeing with positively worded statements (e.g. ‘I’ve been thinking clearly’), and participants are required to rate their experiences over the last 2 weeks on a 5-point Likert scale ranging from 1 (none of the time) to 5 (all of the time). The 7-item version of the WEMWBS (SWEMWBS) was used for this study. Total scores range from 7 to 35 and raw total scores are converted into metric scores (Stewart-Brown et al., 2009). Higher scores indicate higher levels of mental wellbeing. As per the scoring instructions, metric scores were calculated and presented in all analyses for this study. The SWEMWBS has shown good internal consistency with adolescent populations (α=.78; McKay & Andretta, 2017). Internal consistency for the current sample was good (α=.87).

International physical activity questionnaire – short form (IPAQ-SF; Craig et al., 2003)

The IPAQ-SF was used to determine the level of physical activity of participants. The IPAQ-SF consists of seven questions that assess daily time spent sitting, walking and engaging in moderate and vigorous physical activity over the last 7 days. Only the questions related to walking, moderate and vigorous physical activity were used for this study. Total minutes reported by participants are converted into Metabolic Equivalent Task minutes per week (MET-min/week). The MET assignments for each level of physical activity intensity are walking (3.3 METs), moderate (4 METs), and vigorous (8 METs). To avoid younger samples inaccurately reporting physical activity levels on the IPAQ-SF (e.g. Rääsk et al., 2017), the ordering of the questions was reversed (Hagströmer et al., 2008), such that the ‘walking’ question was asked first, then ‘moderate’ physical activity, then ‘vigorous’ physical activity. The IPAQ-SF has shown good internal consistency with adolescent populations (α=.80; He & Qiu, 2022). Internal consistency for the current sample was .94.

Data analysis

All data were analysed using IBM SPSS Statistics 27. Data were initially screened for normality. Distribution of the study variables were assessed using Shapiro–Wilk tests; an appropriate method for this sample size (e.g. Mishra et al., 2019). Shapiro–Wilk tests identified a non-normal distribution of the study variables; therefore, non-parametric tests were used where possible. Where non-parametric tests were not possible, parametric tests were employed. A Kruskal–Wallis H-test explored preliminary gender differences. No significant differences were identified between groups for any study variables. Subsequent analyses were therefore conducted on the whole sample. A p value of p < .01 was employed for all analyses to reduce the risk of a Type 1 error.

Spearman’s two-tailed correlations were initially conducted to explore convergent validity of the ETAQ with related constructs (e.g. compulsive exercise, restrained eating, mental wellbeing, physical activity). Interrelationships between these related constructs were also explored. This study hypothesized that, after accounting for related constructs, higher scores on the ETAQ would be positively associated with higher compulsive exercise in adolescents. Therefore, where significant associations were identified, these covariates were controlled for in the first step of the subsequent regression analyses. Hierarchical stepwise regressions were conducted to identify the extent to which exercise tracking behaviours and attitudes uniquely predicted compulsive exercise, beyond known correlates (i.e. restrained eating, mental wellbeing and physical activity). In this second step of the hierarchical regressions, ETAQ subscales, significantly correlated with each CET subscale, were included.

Results

Descriptive statistics

Descriptive statistics for the sample are presented in Table 1. CET and DEBQ scores were generally in line with existing research conducted with non-clinical adolescent samples (e.g. Goodwin et al., 2014; Hunot-Alexander et al., 2019). SWEMWBS scores were slightly lower in comparison to existing research with UK adolescents (e.g. McKay & Andretta, 2017). Physical activity levels in the current sample were broadly in line with existing research with adolescents (e.g. Aktürk et al., 2019; Fernández-Bustos et al., 2019).

Table 1. Means, standard deviations, median and interquartile range for study variables by gender.

	Boys (n = 147)		Girls (n = 165)		Prefer not to say (n = 9)
	Mean (SD)	Median (IQR)	Mean (SD)	Median (IQR)	Mean (SD)	Median (IQR)
CET Avoidance & rule-driven behaviour	1.24 (1.02)	1.00 (0.00–2.00)	1.26 (1.17)	1.00 (0.00–2.00)	1.54 (1.17)	1.50 (0.00–3.00)
CET Weight control exercise	2.11 (1.04)	2.00 (1.00–3.00)	2.43 (1.34)	2.00 (1.00–4.00)	1.98 (1.38)	2.00 (0.50–3.00)
CET Mood improvement	2.12 (1.31)	2.00 (1.00–3.00)	2.13 (1.23)	2.00 (1.00–3.00)	1.91 (1.11)	1.00 (0.00–2.50)
CET Lack of exercise enjoyment	1.29 (0.97)	1.00 (0.00–2.00)	1.56 (1.17)	1.00 (0.00–2.00)	1.85 (1.38)	2.00 (0.00–2.00)
CET Exercise rigidity	1.85 (1.31)	2.00 (1.00–3.00)	1.84 (1.27)	2.00 (0.00–3.00)	1.70 (0.98)	2.00 (0.50–2.50)
DEBQ restrained eating^1	2.00 (0.95)	2.00 (1.00–2.50)	2.39 (1.21)	2.00 (1.00–3.00)	2.49 (1.30)	2.00 (1.00–3.00)
SWEMWBS^1	19.91 (5.06)	19.98 (17.16–23.66)	18.97 (5.11)	18.59 (15.84–21.54)	17.52 (5.68)	18.29 (14.82–21.01)
IPAQ-SF (MET-min/week)^1	3685.42 (3811.18)	2724.00 (738.00–5026.50)	2762.77 (2861.20)	2079.00 (408.75–4211.50)	724.00 (1254.00)	0.00 (0.00–1086.00)
ETAQ mean total	2.12 (0.74)	2.00 (1.00–3.00)	2.1 (0.83)	1.50 (1.00–3.00)	2.20 (0.81)	2.50 (1.00–3.00)
Exercise monitoring	2.73 (1.08)	3.00 (2.00–4.00)	2.80 (1.86)	3.00 (1.00–4.00)	2.29 (0.76)	2.00 (1.00–3.00)
Exercise related obligation	1.69 (0.82)	1.00 (2.00–4.00)	1.89 (1.08)	1.00 (1.00–2.50)	2.37 (1.47)	2.00 (1.00–3.00)
Exercise guidance	2.25 (1.02)	2.00 (1.00–3.00)	2.06 (1.09)	2.00 (1.00–3.00)	2.07 (0.83)	2.00 (1.00–3.00)
Exercise comparison	1.81 (0.85)	2.00 (1.00–2.00)	1.68 (0.87)	2.00 (1.00–2.00)	2.07 (0.89)	2.00 (1.00–3.00)

CET = Compulsive Exercise Test; DEBQ = Dutch Eating Behavior Questionnaire; ETAQ = Exercise Tracking in Adolescents Questionnaire; IPAQ-SF – International Physical Activity Questionnaire Short Form; SWEMWBS = Short Warwick-Edinburgh Mental Well-Being Scale.

¹Mean scores reflective of the participants who completed the measure: DEBQ restrained eating – boys n = 109 (74.1%), girls n = 144 (87.3%), prefer not to say n = 7 (77.8%); SWEMWBS – boys n = 109 (74.1%), girls n = 143 (86.7%), prefer not to say n = 6 (66.7%); IPAQ-SF - boys n = 81 (55.1%); girls n = 102 (61.2%); prefer not to say n = 5 (55.6%).

Convergent validity

Two-tailed Spearman’s correlations were conducted to explore convergent validity between the ETAQ and other related constructs (i.e. DEBQ, IPAQ-SF, CET, SWEMWBS; see Supplementary Material 4). Most associations identified between the ETAQ and CET were small to moderate (r ≥ .150, p < .01). Almost all ETAQ subscales positively, significantly correlated with CET subscales, aside from ETAQ exercise monitoring and exercise guidance where the correlations with CET lack of exercise enjoyment were non-significant. Weak-to-moderate significant positive correlations were also identified between the DEBQ and all four ETAQ subscales (r ≥ .199, p < .001). Correlations between the ETAQ and IPAQ-SF identified a small significant, positive correlation with ETAQ exercise monitoring only (r = .297, p < .001). ETAQ exercise monitoring was the only subscale of the ETAQ to significantly, positively correlate with mental wellbeing (r = .189, p < .01). These findings suggest that exercise monitoring using ETT is significantly associated with higher levels of physical activity and higher mental wellbeing in adolescents; however, these effect sizes were small.

Two-tailed Spearman’s correlations were further conducted to explore relationships between compulsive exercise outcomes and all other related constructs (e.g. disordered eating, mental wellbeing, physical activity) to inform regression analyses.

DEBQ restrained eating significantly, positively correlated with all CET outcomes (r ≥ .231, p < .01). Mental wellbeing and physical activity levels significantly, positively correlated with CET mood improvement and exercise rigidity (r ≥ .211, p < .01), and significantly negatively correlated with CET lack of exercise enjoyment (r ≤ -.301, p < .01).

Regression analysis

Hierarchical regression analyses were conducted to determine the extent to which exercise tracking behaviours predicted compulsive exercise outcomes in adolescents. As significant relationships were identified between ETAQ subscales and covariates such as restrained eating, mental wellbeing, and physical activity, these covariates were therefore controlled for in the first step of the regressions to enable exploration of the unique variance explained by the ETAQ in relation to compulsive exercise outcomes. ETAQ subscales which were significantly correlated with each compulsive exercise outcome were entered into the second step, using a stepwise regression (see Table 2).

Table 2. Hierarchical stepwise regression analysis predicting compulsive exercise scores from covariates and exercise tracking (n = 321).

Predictor	CET avoidance			CET weight control			CET mood improvement			CET lack of enjoyment			CET exercise rigidity
	β	Adj R^2	F change	β	Adj R^2	F change	β	Adj R^2	F change	β	Adj R^2	F change	β	Adj R^2	F change
Step 1		.23	79.00**		.43	196.30**		.20	13.64**		.19	12.76**		.23	15.84**
DEBQ restrained eating	.49**			.66**			.29**			.22*			.29**
SWEMWBS							.26**			−.22*			.22*
IPAQ-SF							.30**			−.26**			.37**
Step 2		.42	11.05*		.46	12.99**		.29	19.12**		.22	6.78*		.37	10.65*
DEBQ restrained eating	.23**			.60**			.20*
SWEMWBS							.21*			−.20**			.21*
IPAQ-SF							.28**			−.25**			.35**
ETAQ exercise related obligation	.42**									.22*			.26*
ETAQ exercise guidance	.17*						.32**						.25**
ETAQ exercise comparison				.18**

DEBQ = Dutch Eating Behaviour Questionnaire; IPAQ-SF = International Physical Activity Questionnaire Short Form.

ETAQ exercise tracking behaviours italicised. Only significant predictors at p < 0.01 reported.

**p < .001; *p < .01.

Regression results

The regression models to assess the predictive roles of covariates and exercise tracking behaviours on compulsive exercise outcomes were significant for all five CET subscales (Table 2). Significantly correlated covariates that were entered in step 1 differed for each CET subscale and explained 23% (Avoidance and Rule-Driven Behaviour), 43% (Weight Control Exercise), 20% (Mood Improvement), 19% (Lack of Exercise Enjoyment) and 23% (Exercise Rigidity) of the initial total variance respectively. ETAQ variables explained an additional 3–19% of the variance. For CET avoidance and rule-driven behaviour and CET exercise rigidity, ETAQ exercise-related obligation made the largest contribution, followed by ETAQ exercise guidance. For CET lack of exercise enjoyment, ETAQ exercise-related obligation made the largest contribution. For CET mood improvement, ETAQ exercise guidance made the largest contribution. ETAQ exercise comparison was the strongest predictor of CET weight control exercise. ETAQ exercise monitoring was not a significant predictor for any compulsive exercise outcomes.

Summary: Part 2

Part 2 of this study assessed convergent validity of the ETAQ and explored exercise tracking as a statistical predictor of compulsive exercise outcomes. Correlation analyses identified significant positive relationships between exercise tracking behaviours and attitudes and maladaptive health outcomes (i.e. compulsive exercise and disordered eating) in adolescents. Regression analyses further identified exercise-related obligation, guidance and comparisons as significant predictors of compulsive exercise. However, in contrast, exercise monitoring was significantly correlated with higher mental wellbeing and higher physical activity levels, suggesting positive associations with ETT use. The final part of this study sought to establish the construct validity of the ETAQ and to confirm the four-factor structure of the ETAQ through CFA.

Part 3: Construct validity of the ETAQ factor structure using confirmatory factor analysis (CFA)

Data analysis

CFA was employed to assess the fit of the four-factor model of the ETAQ in a separate adolescent sample. The CFA was conducted using IBM SPSS AMOS 27, using the maximum likelihood estimation procedure. To assess the factorial validity of the model, multiple goodness-of-fit indices were used such as chi-square and significance of chi-square, the Tucker-Lewis Index (TLI), the Comparative Fit Index (CFI), the Incremental Fit Index (IFI), the root mean square error of approximation (RMSEA), and the standardized root mean square residual (SRMR). A good fit model should have a non-significant chi-square (p > .05; Kline, 2005). An RMSEA value of < .05 is indicative of a good model fit and values ≥ .06 and below .10 can be considered acceptable (e.g. Hu & Bentler, 1999; MacCallum et al., 1996). For TLI, CFI and IFI, values greater than .95 indicate a good fit of the data (Hu & Bentler, 1999); however, a cut-off value of .90 is considered more conventional in the literature (Russell, 2002). A SRMR of .05 and below is considered a good model fit and a SRMR between .05 and .09 indicates an adequate fit (MacCallum et al., 1996). A cut-off of .40 and above was used to identify significant factor loadings (Stevens, 2002).

Procedure and descriptive statistics

Adolescents who participated in Part 1 of this study were contacted after 6 months and invited to complete a follow-up questionnaire (see Supplementary Material 5 for flowchart of sampling process). Of the 327 participants who initially completed Part 1, 132 participants (boys n = 61, girls n = 68, ‘prefer not to say’ n = 3) responded. The mean age of the participants was 14.30 years (SD: 0.81, range 12–16). Over 30% of the participants (30.3%, n = 40) identified their ethnicity as Asian British, 28.8% (n = 38) identified as White British, 17.4% (n = 23) Asian other, 9.8% (n = 13) ‘other’, 5.3% (n = 7) White other, 4.5% (n = 6) Black other, and 3% (n = 4) Black British. One participant (0.8%) did not report their ethnicity. All participants reported using a device and/or mobile application to track their exercise activities. Mean ETAQ scores were as follows: exercise monitoring (2.59 [SD = 1.06]); exercise-related obligation (1.89 [SD = .97]); exercise guidance (1.04 [SD = .98]); exercise comparison (1.81 [SD = .94]). Tests of difference were conducted on the first time point data (T1) between those who completed both time points and those who only completed T1. No significant differences were identified for age, exercise tracking behaviours, compulsive exercise, restrained eating, physical activity or mental wellbeing scores between the two participant groups.

Results

Factor loadings for the items are shown in Supplementary Material 6. The chi-square was significant (χ²(199) = 1196.55, p < .001) and did not meet the recommendation for a good fit model; however, the CFI (.925), TLI (.903) and IFI (.926) all reflected a good fitting model. The RMSEA (.088; 90% confidence interval: .067–.108) and SRMR (.057) also approximated an acceptable fit according to the aforementioned criteria. All factor loadings met the cut-off value of .40 and above and were therefore retained in the final factor model.

Discussion

This three-part study developed and validated a new measure of exercise tracking behaviours and attitudes in adolescents (ETAQ). As no standardized measure of ETT psychopathology currently exists, thus limiting the ability to reliably assess the potential psychological impacts of engaging with such technology, the development of the ETAQ has addressed this gap in the research and expanded on simply understanding frequency of ETT use and its links to exercise engagement. We incorporated a comprehensive range of items which had been theoretically and empirically linked with engagement with (and potentially problematic) exercise tracking behaviours and attitudes in adolescents. Following EFA, four factors were identified for the ETAQ: (i) exercise monitoring; (ii) exercise-related obligation; (iii) exercise guidance; and (iv) exercise comparisons. All factors demonstrated good levels of internal consistency (Nunnally & Bernstein, 1994; Ursachi et al., 2015) and the four-factor model fit was considered adequate through a subsequent CFA. Preliminary evidence from this study suggests that the ETAQ is a reliable tool in assessing exercise tracking behaviours and attitudes and is useful in exploring the relationship between ETT use and adolescent exercise psychopathology, and wider mental wellbeing.

While existing research has identified a significant relationship between frequency of ETT use and higher compulsive exercise in adolescents (e.g. Bratland-Sanda et al., 2022), this study is the first to identify the key components of ETT use that significantly predict compulsive exercise in adolescents. One of the strongest predictors of compulsive exercise outcomes in adolescents was exercise-related obligation (e.g. feeling guilty for not achieving exercise goals), which reflects adolescent perceptions of ETT use in existing qualitative research (Kerner & Goodyear, 2017; Reynolds, Plateau, et al., 2022). Adolescents reported that using ETT to determine how much or what type of exercise to do (‘Exercise guidance’) was another strong predictor of compulsive exercise. While research has also identified that ETT engagement improves when behavioural feedback from the devices is positive (Edwards et al., 2013), our research has extended this to identify that potentially negative feedback related to missed exercise sessions can predict compulsive exercise outcomes in adolescents, specifically feelings of guilt. Using ETT to model frequency and type of exercise activities, or to compare exercise tracking activity with others, can also predict compulsive exercise outcomes in adolescents. As setting and achieving goals is considered a key component of how ETT can facilitate positive behaviour change (e.g. Gordon et al., 2019), it is therefore crucial that interventions which incorporate ETT use to promote exercise in adolescents are aware of the potential negative psychological implications of prescribed exercise goals on such devices, and how conforming to, and comparing achievement of, prescribed goals may promote maladaptive exercise behaviours and attitudes in adolescents. In addition, qualitative research will be beneficial to understand whether comparison of goal achievement and exercise behaviours occurs in an upward or downward direction. The significant, positive correlations identified between the ETAQ and maladaptive health behaviours (i.e. disordered eating and compulsive exercise) in Part 2 of this study moves beyond previous research which has only identified frequency of ETT use to be significantly linked to higher levels of disordered eating and compulsive exercise in young people (e.g. Bratland-Sanda et al., 2022; Hahn et al., 2022), and suggests that the ETAQ is a suitable tool to use in future longitudinal studies to unpack temporal relations between key components of ETT use and the potential development of maladaptive health behaviours in some adolescents.

The only ETAQ subscale related to physical activity was ‘exercise monitoring’, with adolescents who reported higher levels of monitoring/tracking of exercise behaviours when using ETT engaging in higher levels of physical activity. Exercise monitoring was also the only ETAQ subscale which did not significantly predict compulsive exercise outcomes in adolescents. Higher exercise monitoring was related to better mental wellbeing, which suggests that using ETT to simply monitor and/or track personal exercise activities could indeed promote a healthy and positive attitude towards exercise engagement. Previous research suggests that the most effective behaviour change interventions for promoting physical activity are those which incorporate self-monitoring techniques (Michie et al., 2009). While this and the present findings are beneficial for public health interventions that aim to use ETT to increase exercise in adolescents, exercising to achieve prescribed exercise-related goals on the devices can deter continued use of ETT (e.g. Kerner & Goodyear, 2017; Reynolds, Plateau, et al., 2022), and promote compulsive exercise outcomes as outlined in this study. Future research should therefore acknowledge the circumstances in which self-monitoring of exercise behaviours alone, through autonomous monitoring and goal setting on the devices, could therefore promote healthier attitudes towards physical activity and increase general mental wellbeing. Understanding the significance of individual differences (e.g. frequency/type of exercise tracking behaviours), and how ETT behaviours change over time, is also particularly crucial to acknowledge when evaluating intervention strategies employing ETT to encourage physical activity among adolescents.

Existing research exploring the relationship between ETT use and exercise psychopathology in adolescents (e.g. Bratland-Sanda et al., 2022) has largely focused on the frequency, as opposed to the motives and attitudes around ETT use, and how this relates to exercise psychopathology. Research conducted in this area with adolescent populations has been predominantly qualitative (e.g. Kerner & Goodyear, 2017; Reynolds, Plateau, et al., 2022) and while this present study explores adolescent perceptions regarding the social nature of ETT use and the concerns with their promotion of unhealthy exercise behaviours, future longitudinal research is needed to quantitatively explore the key contributors of ETT use in compulsive exercise with a larger sample of adolescents over time. The ETAQ is therefore particularly beneficial for researchers to thoroughly assess the core aspects of ETT use most significantly related to exercise psychopathology in adolescents.

This study has successfully developed a new measure that reliably assesses key components of exercise tracking behaviours and attitudes in adolescents. EFA was implemented to develop the ETAQ, and each factor had a good level of internal consistency. CFA demonstrated a good model fit for the factor structure, and the ETAQ was significantly correlated with related constructs (e.g. disordered eating and mental wellbeing), supporting the convergent validity of the ETAQ. While the ETAQ can contribute to the key gaps in existing literature, the development of the ETAQ is also particularly beneficial for assessing the wider impacts of exercise-related interventions which incorporate ETT (e.g. exercise wearables, mobile apps) for adolescents. Using the ETAQ in future research to evaluate such interventions will support researchers in understanding the perceived benefits and barriers associated with ETT in promoting physical activity. A further strength of this study is the ethnic diversity of the sample, as exercise activity levels can vary among ethnic groups (e.g. Eyre & Duncan, 2013).

While there are many strengths, this study is not without its limitations. The sample size was adequate for EFA and met the criteria of a minimum of 10 participants per item (Tabachnick & Fidell, 2007); however, the sample size for the CFA fell slightly short of the recommended minimum sample size. The factor model, however, was still considered an adequate fit through CFA and supported the construct validity of the measure. It would be beneficial to explore whether a larger sample size could further strengthen the model fit. Future research to explore the stability of the factor structure both cross-culturally and prospectively, and to further explore measurement invariance between genders in a larger sample, is also warranted. This, in turn, will facilitate further research to understand how ETT may be linked to exercise psychopathology, eating practices and broader adolescent mental wellbeing over time. A further limitation is the self-report nature of the measures administered. However, while there are limitations to self-report measures (e.g. responder bias; Demetriou et al., 2015), they are considered to be more cost-effective and easier to administer (e.g. Corder et al., 2008). It is noteworthy that only a small number of adolescents completed the free-text question to report the type of ETT they used. A potential adaptation to this question could be to use a multiple-choice response so the question format is in keeping with the structure of the rest of the measure. While the development of the ETAQ focused more on ETT psychopathology, rather than the well-researched ‘frequency’ of ETT in relation to maladaptive health behaviours, future research using the ETAQ would benefit from also identifying the frequency with which adolescents use ETT and what type of ETT they use to further understand the extent to which ETT is used throughout adolescence. In addition, physical activity levels were controlled for in analyses and, generally, physical activity levels were in line with research conducted with adolescent populations (e.g. Aktürk et al., 2019). However, athlete data was not collected. As differences in compulsive exercise outcomes between athlete and non-athlete populations are apparent (e.g. Plateau et al., 2014), it would be of interest for future research to use the newly developed measure to explore differences in the relationship between ETT use and exercise psychopathology in athlete and non-athlete populations. This would also be beneficial for future research to explore when using the ETAQ to understand whether exercise tracking as a predictor of exercise psychopathology is dependent on frequency of ETT use.

Conclusion

This study has successfully developed and validated a measure of exercise tracking behaviours and attitudes in adolescents, which encompasses a multitude of motives that are theoretically and empirically associated with exercise tracking behaviours in adolescent populations. As previous research has assessed the relationship between exercise tracking technology use and frequency of exercise behaviours, this new measure can be used to assess the underlying drivers of exercise tracking technology use and their potential role in the promotion of maladaptive exercise behaviours and attitudes in adolescents. Future research with the new measure should focus on prospective research to assess how exercise tracking may be linked to eating and exercise psychopathology and wider mental wellbeing in adolescents over time, with a view to informing public health interventions to prevent the onset of unhealthy exercise attitudes or behaviours, and promote healthier attitudes towards exercise at this age.

Credit author statement

Kalli A Reynolds: Conceptualisation, Formal Analysis, Investigation, Writing – Original Draft Emma Haycraft: Conceptualisation, Writing – Review & Editing, Supervision Carolyn R Plateau: Conceptualisation, Writing – Review & Editing, Supervision.

Disclosure statement

No potential conflict of interest was reported by the author(s).

Supplementary material

Supplemental data for this article can be accessed online at https://doi.org/10.1080/02673843.2024.2371397.

Additional information

Funding

Kalli A. Reynolds is funded by a PhD studentship awarded by the School of Sport, Exercise and Health Sciences at Loughborough University, UK.