Adolescent health behavior profiles and associations with mental health in a longitudinal study

Abstract

To characterize adolescent health behavior profiles and associations with mental health, mixture models using data from two assessment points (N = 201; Female = 53.7%, Time 1 m-age = 15.4 years; Time 2 m-age = 16.3 years) identified three distinct health behavior patterns. Profile 1 (27.9% of sample) had overall healthy behaviors (e.g. diet, physical activity, sleep), except nearly half tried e-cigarettes. Profile 2 (51.9%) had above average physical activity, minimal substance use, but diets high in sugar and below average sleep. The smallest, most concerning group (20.2%, Profile 3), had high caffeine and sugar consumption, low fruit/vegetable intake, below-average sleep duration, were physically inactive, and had the most substance use. Profile 3 adolescents had greater parent psychopathology and co-occurring and future mental health symptoms (p < 0.001). The findings provide important insights into the interrelated nature of adolescent health behaviors and their connection to mental health. Identifying high-risk groups allows for tailored, efficient intervention efforts.

Keywords

diet health behaviors mental health sleep vaping

Adolescence is a critical time for establishing healthy behaviors and a time for experimentation with risky behaviors as adolescents gain increasing autonomy. Substance use, diet, physical activity, sedentary behavior, and sleep are among the most important health/risk behaviors for adolescents (Spear and Kulbok, 2001) and often co-occur (Gallant et al., 2020). Different patterns of health behaviors may be associated with distinct mental or physical health outcomes (Akasaki et al., 2019; Hartz et al., 2018; Loewen et al., 2019). In the short term, adolescents engaging in multiple unhealthy behaviors (e.g. substance use, physically inactive) may be at an increased risk for mental health issues such as depression or externalizing behaviors (Walsh et al., 2020). In the long term, poor adolescent health behaviors may persist into adulthood (Mikkilä et al., 2005) and increase the risk of lifestyle-related chronic diseases such as obesity, cardiovascular disease, and certain cancers (Hartz et al., 2018; Marttila-Tornio et al., 2020; Suvarna et al., 2020). To identify common behavioral patterns and better understand risk profiles in adolescence, adolescent health behavior research needs high-quality measurement and sophisticated analyses, allowing for the development of efficient prevention and intervention strategies.

Jessor’s (1982) Problem Behavior Theory outlines that adolescence is a time of risk behavior initiation and posits that engaging in one risky behavior increases the likelihood of engaging in other risky behaviors (Jessor, 1982). Jessor (1984) theorized that adolescent risk behaviors are interrelated in ways that form syndromes or “organized constellation of behaviors” that tend to co-occur (Jessor, 1984). Researchers have employed cluster analysis to identify patterns of adolescent health behaviors (Borner et al., 2018; Parker et al., 2019). Fitting with the types of behaviors initially described within Problem Behavior Theory, substance use has been the most frequently assessed behavior in research thus far (Akasaki et al., 2019; Bartlett et al., 2006; van Nieuwenhuijzen et al., 2009), followed by delinquent behaviors (e.g. early initiation of sex, truancy, bullying behaviors; Akasaki et al., 2019; Bartlett et al., 2006; Busch et al., 2013; Walsh et al., 2020). Quickly after its inception, Problem Behavior Theory expanded to include health behaviors, such as physical activity, diet, and sleep (Colorodo et al., 1991), with physical activity and diet now fairly well represented in the literature (Borner et al., 2018; Hartz et al., 2018; Pearson et al., 2009). However, sleep has not been included in cluster analysis research as often (Bourdeaudhuij and Oost, 2007; de Moraes et al., 2016; Dumuid et al., 2017), despite sleep being a foundational health behavior that has implications for nearly every facet of well-being.

Adolescent alcohol, cigarette, and cannabis use have been the center of research within the substance use realm (Babbin et al., 2015; Chabrol et al., 2012). Based on a review of 70 studies, researchers have consistently identified three patterns of adolescent substance use: (1) low use of alcohol, cigarettes, and cannabis, (2) single or dual use only, and (3) moderate-high multi-use of alcohol, cigarettes, and cannabis (Halladay et al., 2020). Although there is a robust adolescent substance use literature, less is known about electronic cigarette (e-cigarette) use or vaping. E-cigarettes have quickly gained traction among adolescents (Chadi et al., 2019), with the U.S. Surgeon General declaring e-cigarette use to be a public health epidemic (U.S. Department of Health and Human Services Office of Surgeon General, 2018). Because e-cigarettes are relatively new, researchers have just begun to characterize how e-cigarette use clusters with other problematic health behaviors (Jackson et al., 2020). E-cigarette use patterns may differ from traditional cigarettes, as adolescents may perceive e-cigarettes to be less risky than traditional cigarettes (Russell et al., 2020). Adolescent perceptions and health beliefs may underlie different patterns of e-cigarette use compared with traditional cigarettes.

Although Problem Behavior Theory does not extend into mental health risks, problem behaviors may signal underlying mental health concerns or may predict worsening mental health over time. Existing developmental science research suggests that adolescent engagement in risky and poor health behaviors may increase the risk for poor mental health (Benton et al., 2021; Sampasa-Kanyinga et al., 2020). The theoretical foundation for examining the relationships between health behavior patterns and mental health outcomes rests upon the biopsychosocial model, in which health behaviors and mental health have an intricate relationship. In a literature review, Sampasa-Kanyinga et al. (2020) identified associations between children and adolescents meeting recommendations for physical activity, sedentary time, and sleep duration with better mental health indicators. Studies using cluster analysis to characterize adolescent and young adult health behaviors have identified correlations between risky health behaviors and worse mental health (Brooks et al., 2002; Busch et al., 2013; Jao et al., 2019). However, most researchers, but not all, have tended to include mental health problems in the cluster rather than as a separate outcome (Logan-Greene et al., 2019; Noel et al., 2013; Peltzer, 2009). A longitudinal analysis examining how health behavior profiles predict mental health would show developmental changes and provide a better understanding of temporal order than purely cross-sectional designs.

Problem Behavior Theory highlights the importance of the “environment system” which includes parent expectations and norms. Within this socioecological context, parental mental health may be associated with adolescents’ well-being (Giannakopoulos et al., 2009) and health behavior engagement (Dimitratos et al., 2022; Voisin et al., 2020). Previous literature has demonstrated associations between parental mental health and adolescent risk behavior (e.g. substance use; Ali et al., 2016; McGovern et al., 2023). With effective intervention, parental mental health may be a modifiable correlate, with implications for adolescent risk behavior prevention efforts (Barrett et al., 2024).

The extant literature on health behavior profiles is limited in several ways. First, many existing cluster analysis studies tend to lack breadth, either focusing on substance use (Tomczyk et al., 2016) or diet and physical activity (Borner et al., 2018; de Moraes et al., 2016; Dumuid et al., 2017; Hartz et al., 2018). A study combining substance use, diet, physical activity, and sedentary behaviors, alongside understudied behaviors such as vaping and sleep would provide more comprehensive understanding of various patterns of behavior. Second, rigorous measurement and analysis would increase confidence in the findings. Many existing studies have prioritized large sample sizes with retrospective self-reports (Akasaki et al., 2019; Bartlett et al., 2006; Busch et al., 2013; Marttila-Tornio et al., 2020; Walsh et al., 2020) rather than objective recording of behaviors. Objective measures of diet are rare in this literature (Borner et al., 2018; Cuenca-Garcia et al., 2013; Pearson et al., 2009). Further, the field has most often relied on k-means cluster analysis or the creation of a cumulative risk score (collapsing all behaviors as Yes/No and summing affirmative responses; Bartlett et al., 2006; Pearson et al., 2009; Peltzer, 2009; Petridou et al., 1997; Smpokos et al., 2014) rather than latent class analysis (LCA) or mixture models if including both continuous and categorical variables (Akasaki et al., 2019; Borner et al., 2018; Noel et al., 2013; Parker et al., 2019). These latter models offer additional advantages (Tomczyk et al., 2016), for example, in comparison to other techniques, LCA and mixture models are more robust to skewness (Bakk and Vermunt, 2016), which is important in situations in which only a smaller proportion of adolescents have engaged in a behavior (Tomczyk et al., 2016). LCA is well-suited for identifying naturally occurring subgroups of adolescents engaging in patterns of health behaviors. While regression analyses are powerful for hypothesis testing about linear predictor-outcome relationships, LCA allow exploration of heterogeneity without assuming how health behaviors will relate to each other. The field needs research that combines breadth of behaviors (i.e. substance use and health behaviors), rigorous measurement approaches (e.g. actigraphy and 24-hour dietary recalls), and modern analytic approaches.

Research using objective measurement of health behaviors is limited, and more research is needed to better understand adolescent health behavior profiles and their nuanced relationships with mental health concurrently and longitudinally. To address this gap, the current study employed robust, objective measurement of health behaviors and sophisticated analyses to identify profiles of adolescent health behavior engagement (substance use, diet, physical activity, sedentary behavior, and sleep). This study also examined the associations between adolescent health behaviors and mental health to contribute novel insights and address a research gap that has received limited attention in cluster/latent class analysis research. First, regarding the specific latent profiles, it was hypothesized that risky and unhealthy behaviors would cluster together, resulting in a profile of adolescents with relatively unhealthy behaviors and a profile of adolescents with relatively healthy behaviors across a range of indicators. Second, it was hypothesized there would also be a profile of adolescents who had unhealthy behaviors in some domains (e.g. diet) but not necessarily others (e.g. substance use), but there were not specific a priori hypotheses regarding the patterns that would emerge. Third, it was hypothesized that adolescents with more risky and unhealthy behavioral profiles would (a) report higher levels of concurrent internalizing and externalizing mental health symptoms, (b) have parents with higher levels of mental health problems, and (c) report higher levels of mental health problems approximately 1 year later. A better understanding of how adolescent health behaviors are interrelated and their connection to mental health is needed to allow for more efficient and effective health interventions.

Methods

Participants and design

Adolescents who had enrolled in a lagged cohort sequential longitudinal study beginning in preschool were invited to participate in a follow-up study on health behaviors. Participants who had self-reports of health behaviors and actigraphy data from at least one adolescent time point were included in this study (n = 201). Participants were initially recruited in preschool through community flyers in a small Midwestern city from 2006 to 2012. Eligibility criteria for the preschool phase of the study included English as the primary language spoken in the home and absence of a diagnosed developmental, behavioral, or language disorder at the time of initial recruitment (Wiebe et al., 2008). Diagnosis after initial recruitment was not cause for exclusion. The adolescent phase of the study began in 2017 with adolescents invited to attend annual data collection visits roughly coinciding with their birthdays. The mean age of participation at Time 1 was 15.4 years (SD = 1.1, range = 14–18). About half the participants were female (53.7%). The majority of participants reported their race as White (72.1%, inclusive of Hispanic ethnicity), 23.9% Multiracial/biracial, 3.5% Black, and 0.5% Asian American. In terms of ethnicity 12.9% of the sample was Hispanic. For socioeconomic status, 41.3% of the sample received public health insurance or fell below the federal poverty guidelines. All health behavior assessments contributing to the profiles were from Time 1. To examine health profiles as longitudinal predictors of mental health, data from a second time point was pulled which included mental health assessments. At Time 2, 163 participants had complete mental health data (81% of the T1 sample). At Time 2, participants were 16.3 years (SD = 1.0, range = 15–18). Longitudinally, 92% of participants completed the two study visits 1 year apart, with 7.4% completing the visits with a lag of 2 years. Participants who had Time 2 data did not differ in terms of sex, race, ethnicity, maternal education, income to needs ratio, nor health behavior indicators from adolescents at Time 1, compared to those who did not have Time 2 data (all p values >0.05).

At the initial laboratory visit, parents provided written consent, and adolescents provided written assent for participation. Adolescents then completed measures in a quiet room at the university laboratory, while parents completed pen/paper forms in the waiting room. Trained graduate students and research assistants gave the adolescents actigraphs (wearable accelerometer sensors worn on the wrist that record sleep and physical activity metrics) and scheduled dietary recalls to be completed from home. Adolescents were compensated up to $95 in gift cards for completing lab and phone procedures, and up to $92 in cash for out-of-lab procedures such as completing online dietary recalls, wearing the actigraph, and completing brief daily diaries. Parents were compensated up to $125 in gift cards for completing lab-based procedures. The University of Nebraska-Lincoln’s Institutional Review Board approved all procedures.

Measures

Health behaviors

The following categories of health behaviors were examined: (1) substance use, (2) diet, (3) sleep, and (4) physical activity and sedentary behavior. All health behaviors were from the same time point (Time 1) for analyses.

Substance use

Adolescents completed a standard substance use interview over the phone assessing history of ever using the following substances: cigarettes, electronic cigarettes, alcohol, and marijuana. An example of the question format was as follows: “Have you ever vaped or used electronic cigarettes, also called e-cigs? 1 = Yes, 0 = No.” These questions were modeled after the Monitoring the Future survey questions (Miech et al., 2018). Phone interviews (as opposed to in-person interviews) may reduce social desirability bias.

Diet

Participants completed 24-hour dietary recalls on three non-consecutive days (prompts sent for two weekdays, one weekend day) using the Automated Self-Administered 24-hour Dietary Assessment Tool (ASA24) developed by the National Cancer Institute (Subar et al., 2012). Dietary recalls in general are a reliable and valid method for acquiring dietary data from adolescents (Drapeau et al., 2024; Rankin et al., 2010), including the ASA24 (Hughes et al., 2017; Storey, 2015). The first 24-hour recall was completed during the first laboratory visit. Adolescents were not told ahead of time that they would be reporting on their dietary intake during the session. A random generator was used to schedule the next two recall days during the week following the laboratory session. If any scheduled assessments were not completed during the first week following the session, make-up prompts were sent during the next week. If participants still had not completed three ASA24 prompts at the end of 2 weeks, they were given an opportunity to complete another ASA24 when they returned to the lab. The ASA24 team developed SAS scoring syntax to calculate a Total Fruits and Vegetables Healthy Eating Index score, Total Sugar (g), and Total Caffeine (mg) averaged across the 3 days (NCI, 2017). For the Total Fruits and Vegetables Healthy Eating Index, the maximum possible score is a 10, which is the equivalent of eating one cup of fruit and one cup of vegetables per 1000 calories. Adolescents with at least 2 days of recalls were included in analyses (178 adolescents completed 3 of 3 dietary recalls, 23 adolescents completed 2 of the 3 dietary recalls (included in sample) and 5 participants completed 1 dietary recall (excluded from sample)). There were no outliers to exclude.

Sleep

Actigraphy is a valid and reliable method to assess sleep duration in adolescents (Ancoli-Israel et al., 2015; Meltzer et al., 2012). Participants were instructed to wear a wGT3X-BT accelerometer (ActiGraph, Pensacola, FL) for 14 days on their non-dominant wrist. Participants also completed a daily survey, received via SMS or email, indicating bedtimes and wake times to resolve any conflicts when processing the data. The Sadeh algorithm was applied to the data using ActiLife version 6.13. The grand mean of total sleep time per day across the 14 days, excluding naps, was calculated for participants with at least three nights of data. Adolescent actigraphy adherence was good, with adolescents wearing the actigraph for an average of 12 nights out of 14 (SD = 2.95).

Physical activity

Actigraphy is a valid and reliable method to assess physical activity (Kim et al., 2012; Romanzini et al., 2012, 2014). The Evenson algorithm was applied to the actigraphy data to calculate daily moderate to vigorous physical activity (minutes/day). The grand mean of moderate to vigorous activity across the 14 days was calculated, only including days that met a 70% wear time threshold. Two outliers with improbable values were removed and treated as missing.

Sedentary behavior

Participants completed the National Cancer Institute’s Family Life, Activity, Sun, Health, and Eating (FLASHE) sedentary time survey (Nebeling et al., 2017). Total sedentary time was calculated by averaging five items assessing the amount of time per day watching television, playing video games, using a computer, and using a phone, as well as overall sedentary habits assessing amount of time spent sitting on a 1–5 scale (example item: “How much time did you spend using your cell phone? This includes time spent talking or texting?”). Internal consistency for the sedentary behaviors scale was low (Cronbach’s α = 0.39).

Mental health

Adolescent mental health

At each study visit, participants completed the Youth Self-Report (YSR), which is a validated and commonly used measure of adolescent mental health and problem behaviors (Achenbach and Rescorla, 2014). The YSR provides two broadband scales, Internalizing Problems and Externalizing Problems. The Internalizing Problems scale consists of 31 items assessing symptoms of anxiety (e.g. “I worry a lot”), depression (e.g. “I am unhappy, sad, or upset”), and somatization (e.g. headaches, stomachaches without medical cause). The Externalizing Problems scale consists of 32 items (29 non-substance use items). These items tap into two main categories of behavior including Rule-Breaking Behavior (e.g. lying, stealing, cheating, swearing, truancy) and Aggressive Behavior (e.g. fighting, arguing, mean to others). Internal consistency was excellent for both scales (internalizing Cronbach’s α = 0.93; externalizing α = 0.90). The T-scores from two time points were included in concurrent and longitudinal analyses as correlates of the health behavior profiles. Including adolescent mental health as a correlate and predictor provides a characterization of how different patterns of health behaviors are related to mental health symptoms and if certain patterns of health behaviors would increase risk for worsening mental health over time.

Parental mental health

Parents completed the 53-item Brief Symptom Inventory (BSI), the short version of the Symptom Checklist-90-Revised (SCL-90-R). Parents indicated how much symptoms (e.g. “trouble concentrating,” “easily annoyed or irritated”) were distressing on a 5-point Likert scale to compute the Global Severity Index. The BSI has excellent reliability and validity, as well as good internal consistency (Cronbach’s α = 0.75–0.89; Boulet and Boss, 1991). The parents’ Global Severity Index was included in analyses as a correlate and longitudinal outcome of the adolescents’ health behavior profiles.

Analysis plan

Overview of plan

We used mixture modeling (McLachlan et al., 2019) to identify distinct patterns of health behaviors in this sample of adolescents using health data from Time 1. A sample size of 200 is sufficient for mixture modeling, as simulation research has demonstrated that sample size is not related to power for latent profile analysis (power is more strongly driven by the distance between factors, with sample size unrelated to statistical power for latent profile model indicators; Tein et al., 2013). We correlated the Time 1 health profiles with Time 1 mental health and examined Time 1 health profiles as predictors of Time 2 mental health.

Details of plan

The following health behavior and substance use variables were included in the profiles: sugar, caffeine, fruits/vegetables, physical activity, sedentary activity, sleep duration, traditional cigarettes, e-cigarettes, alcohol, and cannabis. To identify the health profiles, we tested three models within each multi-class solution (e.g. 2-class): (1) means differ but variance and covariance parameters were constrained to be equal across profiles, (2) means and variances differ across profiles but covariances were constrained, and (3) means, variance, and covariances differ across profiles. (Note only variances/covariances of continuous indicators vary across profiles.) Class enumeration (i.e. optimal number of profiles) was evaluated using the Akaike Information Criteria (AIC; Akaike, 1987), and Bayesian Information Criteria (BIC; Schwarz, 1978), and we selected the model with the smallest values that also produced substantively interpretable profiles of adequate size (i.e. profiles ideally containing at least 3% of total sample (Spurk et al., 2020)). We also evaluated Entropy (E; Celeux and Soromenho, 1996) to determine whether there was adequate class/profile separation (>0.80; Clark and Muthén, 2009). As a final check, we conducted bootstrapped likelihood ratio tests (BLRT; McLachlan and Peel, 2000) to compare class/profile solutions (e.g. one-profile vs two-profiles). A significant BLRT suggests that the more complex solution, with more classes, is a better fit. To characterize profiles once selected, we converted continuous indicators to z-scores which facilitated comparison across different indicators and relative to sample means.

After determining the optimal number of profiles, we used Vermunt’s (2010) three-step approach to identify significant predictors (i.e. parental psychopathology) of probability of profile membership (p < 0.05) and examined the magnitude of effects using odds ratios (OR; Chen et al., 2010). To examine whether profiles significantly differed in levels of adolescent psychopathology (concurrent and future), we used Bolck et al.’s (2004) three-step approach. Externalizing Problems includes three items assessing substance use. We conducted two analyses: first examining the Externalizing Problems scale as developed with the three substance use items included and second, removing the three substance use items to avoid overlapping content in the health behavior profiles and Externalizing Problems scale. In auxiliary analyses, internalizing and externalizing change scores were computed to control for baseline mental health differences across profiles.

Results

Please see Table 1 for participant demographic characteristics. Fit statistics across the models are presented in Table 2. The best fitting model was the 3-class/profile solution with heterogenous means and variances as indicated by the smallest AIC (9070.81) and BIC (9235.98) values. This model demonstrated adequate class separation with Entropy = 0.88. A 3-class solution was superior to a 2-class solution as evidenced by significant BLRT (210.97, df = 17, p < 0.001). Please refer to Figures 1 and 2 for characteristics of each of the three health profiles presented separately for physical activity, sleep, and diet (Figure 1 and Supplemental Figure 1) and substance use (Figure 2).

Table 1.

Descriptive statistics of participant characteristics.

Participant characteristics	m ± SD
Participant characteristics	n (%)
Age Time 1	15.37 ± 1.08
Age Time 2	16.29 ± 0.98
Sex
Female	108 (53.7%)
Male	93 (46.3%)
Race/ethnicity
Asian American	1 (0.5%)
Biracial/multiracial	48 (23.9%)
Black or African American	7 (3.5%)
White^a	145 (72.1%)
Ethnicity
Hispanic	26 (12.9%)
Non-Hispanic	175 (87.1%)
Maternal education
Some high school	6 (3%)
High school or GED	16 (8%)
Some college	83 (41.5%)
College graduate	67 (33.5%)
Graduate or professional school	28 (14%)
Economically at-risk	83 (41.3%)
Income-to-Needs Ratio	2.43 ± 1.70

When race and ethnicity are examined together, n = 119 (59.2%) of participants were non-Hispanic White.

Table 2.

Mixture model fit statistics and comparisons.

Class solution	FP	LL	Entropy	AIC	BIC
1	16	−4747.117	0.00	9526.23	9579.09
2 (means only)	27	−4657.092	0.90	9368.18	9457.37
2 (means and variances)	33	−4590.893	0.82	9247.79	9356.80
3 (means only)	38	−4603.155	0.93	9282.31	9407.84
3 (means and variances)*	50	−4485.407	0.88	9070.81	9235.98
4 (means only)	No convergence

Fully heterogenous models with means, variances, and covariances free to vary across classes would not converge. The best fitting model (indicated by*) was the 3-class solution with heterogenous means and variances as indicated by the smallest AIC and BIC values. This model also demonstrated adequate class separation with Entropy >0.80 (Clark and Muthén, 2009). The 3-class solution with both means and variances free to vary across classes was a superior fit to the 3-class solution with only means varying across classes.

FP: number of free parameters; LL: log likelihood; AIC: Akaike information criteria; BIC: Bayesian information criteria.

Profile 1 (27.9% based on estimated posterior probability) included adolescents who had relatively healthier behavior compared to their peers, except nearly half had tried e-cigarettes (45%) and nearly a quarter had tried cannabis (23%). In this profile, no one smoked traditional cigarettes, and alcohol use was low (11%). They did not consume substantial amounts of caffeine (m profile 1 = 0.51 mg vs m total sample = 24.71 mg). Adolescents in Profile 1 had the healthiest diet of adolescents in this sample, including relatively less sugar consumption than adolescents in the other profiles (m profile 1 = 64.39 g vs m profile 2 = 113.04 g vs m profile 3 = 101.26 g). Adolescents in Profile 1 were engaged in relatively more physical activity than adolescents in Profile 3 (m profile 1 = 109.91 minutes vs m profile 3 = 95.43 minutes) and less sedentary behavior than Profile 3 (on a 1–5 scale m profile 1 = 2.61 vs m profile 3 = 2.95). They had slightly longer sleep durations than adolescents in Profile 3 (m profile 1 = 6.52 hours vs m profile 3 = 6.38 hours), though their sleep still fell below recommended durations.

Profile 2 (51.9% based on estimated posterior probability) included adolescents with less healthy diets but minimal substance use. They consumed the most sugar of all the groups (m profile 2 = 113.04 g), although they consumed moderate amounts of fruits and vegetables. They consumed an average amount of caffeine compared to the overall sample (m profile 1 = 23 mg; m total sample = 24.71 mg). These adolescents did not smoke cigarettes and had low alcohol (6%), cannabis (2%), and e-cigarette use (12%). They had relatively more physical activity than Profile 3 (m profile 2 = 108.77 minutes vs m profile 3 = 95.43 minutes) but also average sedentary behavior (1–5 scale m profile 2 = 2.83; m total sample = 2.80). Similar to adolescents in Profile 1, they had slightly longer sleep durations than adolescents in Profile 3 (m profile 2 = 6.57 hours vs m profile 3 = 6.38 hours).

Profile 3 (20.2% based on estimated posterior probability) included adolescents with the most substance use (50% smoking, 60% alcohol, 82% vaping, 74% cannabis). They had the most caffeine and high amounts of sugar (however, less sugar than Profile 2). They consumed the least fruits and vegetables (HEI combined fruits/vegetables scale 1–10: m profile 3 = 4.36 vs m profile 1 = 4.98 vs m profile 2 = 4.59). They had below average physical activity and above average sedentary behavior compared to the overall sample. They had below average sleep duration (m = 6.38 hours) compared to the sample.

Predictors of profile membership: Parental psychopathology

Higher parental BSI global severity assessed in early adolescence was associated with a lower probability of being in Profile 1 (healthy overall but had tried e-cigarettes; odds ratio = 0.53, SE = 0.21, p = 0.024) and Profile 2 (unhealthy diets but minimal substance use; odds ratio = 0.55, SE = 0.17, p = 0.007) relative to Profile 3 (unhealthy diets, high substance use, sedentary, poor sleep).

Profile differences in concurrent and future adolescent psychopathology

Cross-sectional analysis

As depicted in Figure 3, Profile 3 (unhealthy diets, high substance use, sedentary, poor sleep) had significantly higher levels of concurrent internalizing symptoms measured with the YSR relative to Profile 1 (healthy but tried e-cigarettes; χ²(2) = 7.61, p = 0.006) and Profile 2 (unhealthy diets, but no substance use; χ²(2) = 8.58, p = 0.004; m profile 3 = 60.96, SE = 1.74; m profile 1 = 53.76, SE = 1.91; m profile 2 = 54.64, SE = 1.22). The same pattern emerged for externalizing problems measured with the YSR such that Profile 3 had higher symptom levels relative to Profile 1 (χ²(2) = 13.74, p < 0.001) and Profile 2 (χ²(2) = 19.81, p < 0.001; m profile 3 = 58.47, SE = 1.51; m profile 1 = 50.26, SE = 1.59; m profile 2 = 50.35, SE = 0.96). Importantly, these findings replicated when removing the three items of the YSR externalizing scale that reflect substance use (3 vs 1, χ²(2) = 7.75, p = 0.005; 3 vs 2, χ²(2) = 11.90, p = 0.001), suggesting that higher levels of externalizing problems observed in Profile 3 were not due to potential confounds from overlapping content across Profile indicators and YSR items.

Longitudinal analysis

As depicted in Figure 3, adolescents in Profile 3 (unhealthy diets, high substance use, sedentary, poor sleep) had significantly higher levels of future internalizing symptoms measured with the YSR relative to Profile 1 (healthy but have tried e-cigarettes; χ²(2) = 5.34, p = 0.021) and Profile 2 (unhealthy diets, no substance use; χ²(2) = 11.33, p = 0.001; m profile 3 = 64.16, SE = 2.10; m profile 1 = 57.46, SE = 1.97; m profile 2 = 55.78, SE = 1.26). A similar pattern emerged for externalizing problems measured with the YSR such that Profile 3 had higher symptom levels relative to Profile 2 (χ²(2) = 10.23, p < 0.001); however, this difference did not reach significance when comparing Profile 1 with Profile 3 with regard to future externalizing problems (χ²(2) = 2.47, p = 0.116; m profile 3 = 57.60, SE = 2.19; m profile 1 = 53.14, SE = 1.77; m profile 2 = 49.81, SE = 0.97). Importantly, the finding that Profile 3 had higher externalizing problems than Profile 2 replicated when removing the three items of the YSR externalizing scale that reflect substance use (χ²(2) = 7.21, p = 0.007). Auxiliary mental health change score results converged with the primary longitudinal analyses. Profile 1 had a mean increase of 1.48 from T1 to T2 in externalizing problems which was significantly greater than the −0.64 change in Profile 3. No other significant differences were observed.

Discussion

The current study characterized adolescent health behavior profiles using longitudinal and objective measurement and mixture modeling, along with examining mental health correlates, which makes a valuable contribution to research on adolescent health risks. We identified three health behavior profiles characterizing adolescents. Profile 1 (27.9% of the sample) included adolescents with the healthiest diet, physical activity, and sleep compared to their peers, but this profile had a high rate of ever using e-cigarettes. Compared to recommended guidelines for sugar, sleep duration, and physical activity, Profile 1 adolescents were not meeting recommendations as a group; they were healthier than others in the sample, but still not at optimal levels. The largest group, Profile 2 (51.9% of the sample), was characterized by minimal substance use but diets high in added sugar. These adolescents were physically active but also spent above average time engaging in sedentary behavior. Finally, Profile 3 was marked by the least healthy behaviors: high substance use, high sugar, high caffeine, lowest fruit and vegetable intake, low physical activity, and high sedentary behavior (20.2% of the sample). Profile 3 adolescents were the most at risk for both concurrent and longitudinal mental health problems. These adolescents also were more likely to have parents self-report mental health problems.

Patterns of substance use and health behavior engagement

The substance use findings are consistent with the pattern of results in the extant literature (Halladay et al., 2020), such that the largest profile was characterized by generally abstaining from substance use, followed by single or co-use of only one or two substances (in this case, e-cigarettes and cannabis), with a smaller profile characterized by an increased probability of having tried multiple substances. The current results extend beyond the extant substance use literature to paint a fuller picture of how diet, physical activity, sedentary behavior, and sleep occur alongside substance use patterns. These results highlight clear areas of concern (Profile 3—poor diet, poor sleep, low physical activity, multiple types of substance use), but also more subtle patterns of problematic health behaviors (Profile 2—greater probability of abstaining from substance use, but also of consuming high amounts of sugar and caffeine and engagement in high physical activity as well as high screen time/sedentary behavior).

New data on how e-cigarette use profiles with other health behaviors enrich the extant literature. E-cigarette use did not cluster in a distinct way with other substances or other unhealthy behaviors. Rather, e-cigarette use was prevalent (45% having ever tried e-cigarettes), even in adolescents who otherwise engaged in healthy behaviors such as abstaining from other substances, being physically active, and eating healthy foods (Profile 1). This finding complements results from an existing study using Monitoring the Future data (Jackson et al., 2020). Jackson et al. (2020) found that although adolescents who had tried traditional cigarettes had greater odds of having low fruits and vegetables, low physical activity, and short sleep, adolescents who used e-cigarettes did not have increased odds of other unhealthy behaviors. There are several plausible reasons why adolescents who have not tried alcohol, cannabis or other substances have tried e-cigarettes, including high accessibility (Chaffee et al., 2022), low perceptions of harm (Russell et al., 2020), and robust social media marketing campaign (Collins et al., 2019). A robust e-cigarette marketing campaign, including advertisements on social media (Collins et al., 2019), has been associated with adolescents’ having lower perceptions of harm and greater intention to try and actual use of e-cigarettes (Collins et al., 2019).

Associations between health behavior profiles and mental health

Additionally, the findings of this study provide important insights into the interrelated nature of adolescent health behaviors and their connections to adolescent mental health. The finding that the health behavior profiles correlated with both parent and adolescent mental health inventories are consistent with, and expand upon, extant research. For instance, in a large prospective study of early adolescents, adolescents who met fewer of the standard recommendations for diet, sleep, screen time, and physical activity had more mental health physician visits 3 years later (Loewen et al., 2019). In the current study, notably, the poorest dietary, sleep, and activity behaviors clustered with substance use and were associated with psychopathology. General psychopathology risk or underlying transdiagnostic factors, such as emotion dysregulation, impulsivity, or sensation seeking tendencies may explain the associations between the health behaviors and mental health.

The findings reveal nuanced differences in the constellations of adolescent health behaviors and mental health. There was not a group of adolescents who were high in psychopathology but otherwise healthy in terms of substance use, sleep, diet, and physical activity. The largest probability of group membership (Profile 2) engaged in some unhealthy behaviors like high sugar use and screen time but were not high in psychopathology nor substance use.

Interestingly, when followed-up after 1 year, adolescents with overall good health behaviors, with the exception that many had ever tried e-cigarettes, (Profile 1) also had significant risk for later externalizing problems at similar levels as adolescents who had unhealthy behaviors and had tried multiple substances (e-cigarettes, alcohol, tobacco, and cannabis; Profile 3). In other words, Profile 1 (characterized by overall healthy behaviors but nearly half trying e-cigarettes) may be at risk for escalating externalizing problems from Time 1 to Time 2, as they no longer differ in externalizing problems from Profile 3 by Time 2. These results support regular screening for e-cigarette use and interventions to promote and maintain positive health behaviors and optimal mental health.

Clinical implications

Health psychologists are poised to intervene with adolescents who are engaging in less healthy behaviors, as these risky and unhealthy behaviors can have lasting implications for their physical and mental health. The identification of high-risk groups, characterized by a constellation of risky and unhealthy behaviors, is valuable for targeting intervention efforts to the adolescents most in need. For example, for the adolescents in the high-risk group (Profile 3), intervention strategies must be broad, addressing underlying issues with risk-taking and regulation. Health psychologists may intervene with adolescents and their parents to address mental health comorbidities and unhealthy behaviors (e.g. poor diet, sedentary behavior/low physical activity, etc.), which may have their etiology in or are reinforced by family behavior patterns. In contrast, intervention efforts for adolescents in Profile 1 (generally healthier with some having tried e-cigarettes) could be tailored to educate and motivate abstaining from e-cigarettes. Finally, across all groups, adolescents had higher than recommended sugar intake and insufficient sleep. These results highlight the importance of universal intervention across all adolescents to reduce sugar intake and increase sleep duration.

Limitations and future directions

The study has several limitations that must be acknowledged. First, the substance use assessment was limited to questions regarding having ever tried the specific substances. The current study is unable to answer questions regarding quantity of substance use across individuals. Second, although representative of the surrounding area and representing diverse socioeconomic status, the sample is less representative of the overall U.S. racial and ethnic composition with more than half of the current sample identifying as non-Hispanic White. Future research is warranted to replicate these findings in a more inclusive sample. Third, the analytical strategy is a major strength of the study, but mixture models are limited in the number of covariates that can be included while still allowing the model to converge. Despite limitations, the study is strengthened by the use of actigraphy for sleep and physical activity, as well as multiple 24-hour recalls for diet.

Conclusions

This study enriches the extant literature on adolescent health behaviors. It is novel in its breadth of behaviors, rigorous measurement approach, and sophisticated/preferred analytic strategy. The findings of this study provide important insights into the interrelated nature of adolescent health behaviors and their connection to adolescent mental health. Broad intervention efforts across health behaviors and mental health are needed for a subset of adolescents, targeted intervention on e-cigarette use is needed for adolescents who present with otherwise healthy behaviors, and universal intervention is needed for all adolescents to reduce sugar intake and increase sleep duration.

Supplemental Material

sj-docx-1-hpq-10.1177_13591053251314328 – Supplemental material for Adolescent health behavior profiles and associations with mental health in a longitudinal study

Supplemental material, sj-docx-1-hpq-10.1177_13591053251314328 for Adolescent health behavior profiles and associations with mental health in a longitudinal study by Katherine M Kidwell, Rebecca L Brock, Cara Tomaso, Eric Phillips, Tiffany D James, Amy Lazarus Yaroch, Jennie L Hill, Jennifer Mize Nelson, Terry T-K Huang, W Alex Mason, Kimberly Andrews Espy and Timothy D Nelson in Journal of Health Psychology

Footnotes

Acknowledgements

We are grateful to the participating families, as well as the research technicians, undergraduate and graduate students, and lab coordinators who made this research possible.

Author contributions

Katherine M Kidwell—visualization (equal), writing (original draft, lead), writing (review & editing, lead). Rebecca L Brock—formal analysis (lead), writing (original draft, supporting), writing (review & editing, supporting). Cara Tomaso—conceptualization (supporting), formal analysis (support), writing (original draft, supporting). Eric Phillips—formal analysis (support). Tiffany D James—data curation (lead), methodology (supporting). Amy Lazarus Yaroch—writing—review & editing (equal). Jennie L Hill—writing—review & editing (equal). Jennifer Mize Nelson—data curation (lead), methodology (supporting), project administration (lead), supervision (lead). Terry T-K Huang—writing—review & editing (equal). W Alex Mason—methodology (supporting), funding acquisition (equal), writing—review & editing (equal). Kimberly Andrews Espy—funding acquisition (supporting), methodology (equal). Timothy D Nelson—conceptualization (lead), funding acquisition (equal), writing—review & editing (lead), investigation (equal), supervision (lead), writing-original draft (supporting).

Data sharing statement

The data are not publicly available because participants did not consent to having their data shared at the time of participation. However, summary data are available in (i.e. correlation matrix). Analytic code will be provided upon reasonable request.

Declaration of conflicting interests

The authors declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Funding

The authors disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: This work was supported by the National Institute of Diabetes and Digestive and Kidney Diseases [R01DK116693, R01DK125651], the National Institute on Drug Abuse [R01DA041738], and the National Institute of General Medical Sciences [P20GM130461] of the National Institutes of Health. The content is solely the responsibility of the authors and does not necessarily represent the official views of the National Institutes of Health.

Ethics approval

The University of Nebraska’s Institutional Review Board approved all procedures (Protocol number: 18411; IRB approval #: 20180718411EP).

Informed consent

All adolescents completed written assent and parents provided written consent.

ORCID iDs

Katherine M Kidwell

Cara Tomaso

References

Achenbach

Rescorla

(2014) The Achenbach system of empirically based assessment (ASEBA) for ages 1.5 to 18 years. In The Use of Psychological Testing for Treatment Planning and Outcomes Assessment. Routledge, pp. 179–213.

Akaike

(1987) Factor analysis and AIC. Psychometrika 52: 317–332.

Akasaki

Ploubidis

Dodgeon

, et al. (2019) The clustering of risk behaviours in adolescence and health consequences in middle age. Journal of Adolescence 77: 188–197.

Ali

Dean

Jr Hedden

(2016) The relationship between parental mental illness and/or substance use disorder on adolescent substance use disorder: Results from a nationally representative survey. Addictive Behaviors 59: 35–41.

Ancoli-Israel

Martin

Blackwell

, et al. (2015) The SBSM guide to actigraphy monitoring: Clinical and research applications. Behavioral Sleep Medicine 13(sup1): S4–S38.

Babbin

Velicer

Paiva

, et al. (2015) Replicating cluster subtypes for the prevention of adolescent smoking and alcohol use. Addictive Behaviors 40: 57–65.

Bakk

Vermunt

(2016) Robustness of stepwise latent class modeling with continuous distal outcomes. Structural Equation Modeling: A Multidisciplinary Journal 23(1): 20–31.

Barrett

Muir

Burns

, et al. (2024) Interventions to reduce parental substance use, domestic violence and mental health problems, and their impacts upon children’s well-being: A systematic review of reviews and evidence mapping. Trauma, Violence, & Abuse 25(1): 393–412.

Bartlett

Holditch-Davis

Belyea

, et al. (2006) Risk and protection in the development of problem behaviors in adolescents. Research in Nursing & Health 29(6): 607–621.

10.

Benton

Boyd

Njoroge

WFM

(2021) Addressing the global crisis of child and adolescent mental health. JAMA Pediatrics 175(11): 1108–1110.

11.

Bolck

Croon

Hagenaars

(2004) Estimating latent structure models with categorical variables: One-step versus three-step estimators. Political Analysis 12(1): 3–27.

12.

Borner

Mitchell

Carlson

, et al. (2018) Latent profile analysis of young adolescents’ physical activity across locations on schooldays. Journal of Transport & Health 10: 304–314.

13.

Boulet

Boss

(1991) Reliability and validity of the Brief Symptom Inventory. Psychological Assessment: Journal of Consulting and Clinical Psychology 3(3): 433.

14.

Bourdeaudhuij

Oost

(2007) Family characteristics and health behaviours of adolescents and families. Psychology & Health 13(5): 785–803.

15.

Brooks

Sion

Thrall

, et al. (2002) Association of adolescent risk behaviors with mental health symptoms in high school students. Journal of Adolescent Health 31(3): 240–246.

16.

Busch

Van Stel

Schrijvers

AJP

, et al. (2013) Clustering of health-related behaviors, health outcomes and demographics in Dutch adolescents: A cross-sectional study. BMC Public Health 13(1118): 1–11.

17.

Celeux

Soromenho

(1996) An entropy criterion for assessing the number of clusters in a mixture model. Journal of Classification 13: 195–212.

18.

Chabrol

Chauchard

Goutaudier

, et al. (2012) Exploratory study of the psychopathological profiles of adolescent cannabis users. Addictive Behaviors 37(10): 1109–1113.

19.

Chadi

Hadland

Harris

(2019) Understanding the implications of the “vaping epidemic” among adolescents and young adults: A call for action. Substance Abuse 40(1): 7–10.

20.

Chaffee

Halpern-Felsher

Croker

, et al. (2022) Preferences, use, and perceived access to flavored e-cigarettes among United States adolescents and young adults. Drug and Alcohol Dependence Reports 3: 100068.

21.

Chen

Cohen

Chen

(2010) How big is a big odds ratio? Interpreting the magnitudes of odds ratios in epidemiological studies. Communications in Statistics—Simulation and Computation^® 39(4): 860–864.

22.

Clark

Muthén

(2009) Relating latent class analysis results to variables not included in the analysis. Los Angeles, CA.

23.

Collins

Glasser

Abudayyeh

, et al. (2019) E-cigarette marketing and communication: How e-cigarette companies market e-cigarettes and the public engages with e-cigarette information. Nicotine and Tobacco Research 21(1): 14–24.

24.

Colorodo

Jessor

Costa

(1991) Adolescent health behavior and conventionality-unconventionality: An extension of problem-behavior therapy. Health Psychology 10(1): 52–61.

25.

Cuenca-Garcia

Huybrechts

Ruiz

, et al. (2013) Clustering of multiple lifestyle behaviors and health-related fitness in European adolescents. Journal of Nutrition Education and Behavior 45(6): 549–557.

26.

de Moraes

Fernandez-Alvira

Rendo-Urteaga

, et al. (2016) Effects of clustering of multiple lifestyle-related behaviors on blood pressure in adolescents from two observational studies. Preventive Medicine 82: 111–117.

27.

Dimitratos

Swartz

Laugero

(2022) Pathways of parental influence on adolescent diet and obesity: A psychological stress–focused perspective. Nutrition Reviews 80(7): 1800–1810.

28.

Drapeau

Laramée

Lafreniere

, et al. (2024) Assessing the relative validity of a web-based self-administered 24-hour dietary recall in a Canadian adolescent’s population. Nutrition Journal 23(1): 66.

29.

Dumuid

Olds

Lewis

, et al. (2017) Health-related quality of life and lifestyle behavior clusters in school-aged children from 12 countries. The Journal of Pediatrics 183: 178–183.e172.

30.

Gallant

Thibault

Hebert

, et al. (2020) One size does not fit all: Identifying clusters of physical activity, screen time, and sleep behaviour co-development from childhood to adolescence. International Journal of Behavioral Nutrition and Physical Activity 17(1): 58.

31.

Giannakopoulos

Dimitrakaki

Pedeli

, et al. (2009) Adolescents’ wellbeing and functioning: Relationships with parents’ subjective general physical and mental health. Health and Quality of Life Outcomes 7: 1–9.

32.

Halladay

Woock

El-Khechen

, et al. (2020) Patterns of substance use among adolescents: A systematic review. Drug and Alcohol Dependence 216: 108222.

33.

Hartz

Yingling

Ayers

, et al. (2018) Clustering of health behaviors and cardiorespiratory fitness among U.S. adolescents. Journal of Adolescent Health 62(5): 583–590.

34.

Hughes

Summer

Ollberding

, et al. (2017) Comparison of an interviewer-administered with an automated self-administered 24 h (ASA24) dietary recall in adolescents. Public Health Nutrition 20(17): 3060–3067.

35.

Jackson

Boccio

Leal

(2020) Do youth who vape exhibit risky health lifestyles? Monitoring the future, 2017. Preventive Medicine 136: 106101.

36.

Jao

Robinson

Kelly

, et al. (2019) Unhealthy behavior clustering and mental health status in United States college students. Journal of American College Health 67(8): 790–800.

37.

Jessor

(1982) Problem behavior and developmental transition in adolescence. Journal of School Health 52(5): 295–300.

38.

Jessor

(1984) Adolescent development and behavioral health. In: Matarazzo

Weiss

Herd

al.

(eds) Behavioral Health: A Handbook of Health Enhancement and Disease Prevention. New York, NY: John Wiley & Sons, pp.69–90.

39.

Kim

Beets

Welk

(2012) Everything you wanted to know about selecting the “right” Actigraph accelerometer cut-points for youth, but…: A systematic review. Journal of Science and Medicine in Sport 15(4): 311–321.

40.

Loewen

Maximova

Ekwaru

, et al. (2019) Lifestyle behavior and mental health in early adolescence. Pediatrics 143(5): e20183307.

41.

Logan-Greene

Linn

Hartinger-Saunders

, et al. (2019) Understanding the ecological context of mental, emotional, and behavioral health problems: A person-centered approach. Journal of Community Psychology 47(4): 833–855.

42.

McGovern

Bogowicz

Meader

, et al. (2023) The association between maternal and paternal substance use and child substance use, internalizing and externalizing problems: A systematic review and meta-analysis. Addiction 118(5): 804–818.

43.

McLachlan

Lee

Rathnayake

(2019) Finite mixture models. Annual Review of Statistics and Its Application 6: 355–378.

44.

McLachlan

Peel

(2000) Mixtures of factor analyzers. In: Proceedings of the seventeenth international conference on machine learning (ed P

Langley

), San Francisco, CA, 29 June–2 July, pp.599–606. San Francisco, CA: Morgan Kaufmann.

45.

Marttila-Tornio

Ruotsalainen

Miettunen

, et al. (2020) Clusters of health behaviours and their relation to body mass index among adolescents in Northern Finland. Scandinavian Journal of Caring Sciences 34(3): 666–674.

46.

Meltzer

Montgomery-Downs

Insana

, et al. (2012) Use of actigraphy for assessment in pediatric sleep research. Sleep Medicine Reviews 16(5): 463–475.

47.

Miech

Johnston

O’Malley

, et al. (2018) Monitoring the Future National Survey Results on Drug Use, 1975-2017: Volume I, Secondary School Students. Ann Arbor, MI: Institute for Social Research, The University of Michigan.

48.

Mikkilä

Räsänen

Raitakari

, et al. (2005) Consistent dietary patterns identified from childhood to adulthood: The cardiovascular risk in Young Finns Study. British Journal of Nutrition 93(6): 923–931.

49.

NCI (2017) The Healthy Eating Index - Scores for describing dietary intake. Available at: https://epi.grants.cancer.gov/hei/hei-scores-for-describing-dietary-intake.html (accessed 4 May 2024).

50.

Nebeling

Hennessy

, et al. (2017) The FLASHE study: Survey development, dyadic perspectives, and participant characteristics. American Journal of Preventive Medicine 52(6): 839–848.

51.

Noel

Denny

Farrant

, et al. (2013) Clustering of adolescent health concerns: A latent class analysis of school students in New Zealand. Journal of Paediatrics and Child Health 49(11): 935–941.

52.

Parker

Salmon

Brown

, et al. (2019) Typologies of adolescent activity related health behaviours. Journal of Science and Medicine in Sport 22(3): 319–323.

53.

Pearson

Atkin

Biddle

, et al. (2009) Patterns of adolescent physical activity and dietary behaviours. International Journal of Behavioral Nutrition and Physical Activity 6: 45.

54.

Peltzer

(2009) Health behavior and protective factors among school children in four African countries. International Journal of Behavioral Medicine 16(2): 172–180.

55.

Petridou

Zavitsanos

Dessypris

, et al. (1997) Adolescents in high-risk trajectory: Clustering of risky behavior and the origins of socioeconomic health differentials. Preventive Medicine 26: 215–219.

56.

Rankin

Hanekom

Wright

, et al. (2010) Dietary assessment methodology for adolescents: A review of reproducibility and validation studies. South African Journal of Clinical Nutrition 23(2): 65–74.

57.

Romanzini

Petroski

Ohara

, et al. (2014) Calibration of ActiGraph GT3X, Actical and RT3 accelerometers in adolescents. European Journal of Sport Science 14(1): 91–99.

58.

Romanzini

Petroski

Reichert

(2012) Accelerometers thresholds to estimate physical activity intensity in children and adolescents: A systematic review. Revista Brasileira de Cineantropometria & Desempenho Humano 14: 101–113.

59.

Russell

Katsampouris

Mckeganey

(2020) Harm and addiction perceptions of the JUUL e-cigarette among adolescents. Nicotine and Tobacco Research 22(5): 713–721.

60.

Sampasa-Kanyinga

Colman

Goldfield

, et al. (2020) Combinations of physical activity, sedentary time, and sleep duration and their associations with depressive symptoms and other mental health problems in children and adolescents: A systematic review. International Journal of Behavioral Nutrition and Physical Activity 17(1): 1–16.

61.

Schwarz

(1978) Estimating the dimension of a model. The Annals of Statistics 6: 461–464.

62.

Smpokos

Linardakis

Papadaki

, et al. (2014) Clustering of chronic disease behavioral risk factors among adolescents in Crete (Greece): Associations with biological factors and cardiorespiratory fitness levels. Journal of Public Health 22(5): 433–442.

63.

Spear

Kulbok

(2001) Adolescent health behaviors and related factors: A review. Public Health Nursing 18(2): 82–93.

64.

Spurk

Hirschi

Wang

, et al. (2020) Latent profile analysis: A review and “how to” guide of its application within vocational behavior research. Journal of Vocational Behavior 120: 103445.

65.

Storey

(2015) A changing landscape: Web-based methods for dietary assessment in adolescents. Current Opinion in Clinical Nutrition & Metabolic Care 18(5): 437–445.

66.

Subar

Kirkpatrick

Mittl

, et al. (2012) The automated self-administered 24-hour dietary recall (ASA24): A resource for researchers, clinicians and educators from the National Cancer Institute. Journal of the Academy of Nutrition and Dietetics 112(8): 1134.

67.

Suvarna

Phillips

, et al. (2020) Health risk behaviours and allostatic load: A systematic review. Neuroscience & Biobehavioral Reviews 108: 694–711.

68.

Tein

J-Y

Coxe

Cham

(2013) Statistical power to detect the correct number of classes in latent profile analysis. Structural Equation Modeling: A Multidisciplinary Journal 20(4): 640–657.

69.

Tomczyk

Isensee

Hanewinkel

(2016) Latent classes of polysubstance use among adolescents—A systematic review. Drug and Alcohol Dependence Reports 160: 12–29.

70.

U.S. Department of Health and Human Services, Office of Surgeon General (2018) Surgeon General’s advisory on e-cigarette use among youth. Available at: https://e-cigarettes.surgeongeneral.gov/documents/surgeon-generals-advisory-on-e-cigarette-use-among-youth-2018.pdf (accessed 21 January 2024).

71.

van Nieuwenhuijzen

Junger

Velderman

, et al. (2009) Clustering of health-compromising behavior and delinquency in adolescents and adults in the Dutch population. Preventive Medicine 48(6): 572–578.

72.

Vermunt

(2010) Latent class modeling with covariates: Two improved three-step approaches. Political Analysis 18(4): 450–469.

73.

Voisin

Kim

Bassett

, et al. (2020) Pathways linking family stress to youth delinquency and substance use: Exploring the mediating roles of self-efficacy and future orientation. Journal of Health Psychology 25(2): 139–151.

74.

Walsh

Sela

De Looze

, et al. (2020) Clusters of contemporary risk and their relationship to mental well-being among 15-year-old adolescents across 37 countries. Journal of Adolescent Health 66(6S): S40–S49.

75.

Wiebe

Espy

Charak

(2008) Using confirmatory factor analysis to understand executive control in preschool children: I. Latent structure. Developmental Psychology 44(2): 575.

Supplementary Material

Please find the following supplemental material available below.

For Open Access articles published under a Creative Commons License, all supplemental material carries the same license as the article it is associated with.

For non-Open Access articles published, all supplemental material carries a non-exclusive license, and permission requests for re-use of supplemental material or any part of supplemental material shall be sent directly to the copyright owner as specified in the copyright notice associated with the article.

0.00 MB

0.02 MB