Late childhood and adolescence represent critical developmental windows characterized by significant physiological, psychological, and social transformations. During this period, youth respond to these numerous biopsychosocial challenges with increasing personal autonomy, manifesting profound shifts across multiple health behavior patterns, including sleep [1-3], physical activity [4,5], dietary habits [6-8], and substance use [9]. Across these health behavior domains, research has made major advances in how health behaviors shape somatic health [10-14], but their role in the development of mental disorders during late childhood and adolescence remains poorly understood [15]. Nonetheless, the importance of modifiable health behaviors for mental health is gaining public awareness, and stakeholders are advocating their integration into public mental health strategies, including primary prevention [16], positive psychology [17,18], and nonpharmacological interventions for youth [19].

Extant research focused on the prospective and causal relationship of individual health behavior domains, including sleep, physical activity, dietary habits, and substance use, and regarding the onset and severity of diverse mental disorders, including depression, anxiety, bipolar disorder, and attention deficit hyperactivity disorder [20]. However, these isolated lines of research consistently show that the same behavior often relates to multiple disorders, and conversely, the same disorder associates with multiple behaviors. This lack of specificity limits our ability to infer which particular behavior underlies which symptom constellation [21]. In other words, 2 adolescents with the same health behavior profile at a single point of measurement may follow divergent clinical trajectories in terms of type and severity of symptoms over time. Such multifinality of risk factors underscores the need to move beyond domain-specific studies and toward integrative, intraindividual investigations that capture the temporal interplay of multiple behaviors and affective states within each person [21,22].

In this integrative approach, affective functioning, including momentary emotional reactions, mood fluctuations, and emotion-regulation processes [23], is particularly important, as it constitutes the critical interface linking health behaviors to the emergence of psychiatric symptoms. Taken together, these health behaviors and affective states form a dynamic regulatory system characterized by numerous pairwise interdependencies. For example, exercise improves sleep [24], which, in turn, affects activity levels [25] and eating preferences [26]. Emotions also influence physical activity [27,28] and sedentary behaviors [29], which subsequently impact eating behaviors, such as emotional eating [30,31]. Though these investigations remain limited to bivariate group-level associations. To capture the complexity of these interacting regulatory systems, research should adopt an intraindividual, multivariate paradigm that models complementary characteristics of their temporal dynamics [32-35]. Longitudinal studies confirm that adolescents exhibit distinct, person‐specific trajectories and intraindividual dynamics in sleep [36], physical activity profiles [37,38], eating habits [39,40], and affective functioning [41]. In sum, we posit that the interdependencies among these health behavior domains reflect dynamic, regulatory adjustment processes to life circumstances that are both specific to the individual and continuously evolving over time.

Providing evidence for the interrelated nature of these health behavior domains and their intraindividual dynamics is crucial for a holistic approach to youth health and requires advanced technical solutions for field research. Mobile digital technologies used for ambulatory assessments [42], such as actigraphy and frequently repeated self-reports collected via ecological momentary assessment (EMA; or experience sampling method), enable real-time monitoring of multiple health domains in naturalistic settings [43-45]. EMA has been widely used to track smoking, alcohol, and drug use [43], and emotional states [44] in youth, whereas actigraphy has been used primarily to assess patterns of sleep and, eventually, activity patterns [45]. Combining EMA and actigraphy allows investigating the temporal dynamic interactions between health domains over different timescales, for example, by investigating the temporal fluctuation of affective functioning and sleep disturbances from the daily [46] to the monthly timescale [47]. However, the inclusion of preadolescent children in EMA data collection depends on their access and autonomy in using a smartphone or similar connected mobile device. Moreover, the adequacy of this method remains subject to a normative ethical consideration regarding the inherent burden of intensive longitudinal designs as well as the role of parental involvement [48].

Taken together, to detect early signs of maladaptive health behaviors, etiological research should embrace a multivariate, intraindividual, and temporally sensitive approach. While large-scale epidemiological evidence requires significant resources, pilot studies can pave the way by examining the feasibility of intensive longitudinal protocols and unraveling the dynamic interplay between health domains and youth mental health.

This study aims to pioneer future intensive longitudinal designs for epidemiological research by addressing both feasibility and across different variations of a modular intensive longitudinal paradigm, as well as preliminary investigations of the dynamic links between health behaviors and mental health among youths. Our methodological ambition was to flexibly adapt a multiwave multitimescale paradigm to different target populations by adapting the intensity of data collection as well as the composition and modality of assessments. First, adolescents recruited in schools will monitor their sleep, physical activity, dietary behavior, substance use, and affective functioning across three 3-week assessment waves over a timeframe of approximately 1 school year. This protocol will be investigated for both feasibility targets as well as preliminary findings in the dynamic interplay among the target domains and mental health.

Second, the developmental and clinical imperative to understand how health behaviors influence youth mental health highlights the need for intensive longitudinal studies in preadolescent age groups, especially as they undergo critical clinical transitions. However, the substantial burden of such designs raises questions about their feasibility in these settings. It even raises ethical challenges to potentially overburden vulnerable youth without prior ground for assuming the feasibility of the essential design characteristics. Therefore, we extended this pilot study with a reduced intensive longitudinal protocol in 2 preadolescent settings: school‐recruited participants and psychiatric patients transitioning from intensive outpatient treatment to follow‐up care. For both preadolescent groups, we reduced the protocol to 2 waves spanning 10 days each. Further group-specific adaptations are described in the respective protocol descriptions.

For both the adolescent and preadolescent protocol, feasibility outcomes focus on participation metrics of response compliance (ie, the percentage of all answered among all planned EMA surveys) and participant retention. We will test the hypothesis that both remain at an acceptable level (ie, both at 70% or above) and stable across the different groups and waves. Achieving the required sample sizes for the evaluation of these outcomes represents a prerequisite for the evaluation of these outcomes (see “Sample size rationale”). Consequently, the corresponding number of invited youth and schools and the acceptance rate provide information on the feasibility and efficiency of the recruitment. Not achieving the targeted sample sizes or rates of retention and compliance will therefore lead to thorough reconsideration of central design parameters. Irrespective of the obtained outcome, we will collect feedback on the subjective experience and opinion of participants and young partners to identify weaknesses and barriers of the protocols that can orient future modification of the design.

Beyond assessing feasibility, we will leverage our comprehensive adolescent protocol and use advanced statistical approaches specifically designed for intensive longitudinal data to examine further dimensions of the bidirectional dynamics linking health behaviors and mental health. By complementing feasibility-related findings with these analytical techniques, we aim to strengthen methodological frameworks and optimize both the design and implementation of future research. Insights gained from this pilot study will therefore guide and refine the scope of subsequent epidemiological cohort studies, ultimately enhancing our understanding of how health behaviors influence youth mental health trajectories.

The design of this study closely aligns with the methods promoted by the mobile Motor Activity Consortium for Health (mMARCH), a collaborative international network coordinated by the National Institute of Mental Health, Intramural Research Program (ZIA MH002954), that was created to facilitate the coordination of procedures, analyses [49], and data sharing among research groups collecting mobile technology data. As an essential resource for intensive longitudinal research, the consortium provides a modular assessment framework that we followed to implement 2 age-specific assessment protocols, one for adolescents and one for preadolescents. The study is currently in data collection and continues to recruit participants. This study protocol for a pilot intensive longitudinal investigation follows the reporting guidelines for cohort studies as required by the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) Statement [50] as well as the adapted STROBE Checklist for Reporting Ecological Momentary Assessment Studies [51] (CREMAS). The following sections will introduce the methods for the adolescent and the preadolescent protocol separately. For the entire project, the initial goal was to enroll a total of 100 participants from the French-speaking region of Switzerland. Enrollment started in November 2022, and data collection will end in September 2025. Each assessment protocol is adapted and implemented to best meet the respective educational or clinical context, while respecting the capacity of each age group to engage in an intensive longitudinal study.

Participants from all groups are invited to feedback sessions at the end of the study, giving them the opportunity to express their honest opinion about the study. In addition, we will invite motivated adolescent participants to become young partners and contribute to a dedicated in-depth discussion session on the research project.

Our presentation of the preadolescent protocol primarily concentrates on differences relative to the adolescent protocol and eventual differences between the school-recruited and clinical subgroups among preadolescent participants. Design characteristics that are not explicitly specified for preadolescents align with the description of the adolescent protocol.

The targeted feasibility outcomes of acceptable and stable participation metrics will be examined by appropriate tests of equivalence regarding compliance and dropout against a priori benchmarks for both protocols [86]. We will further examine the stability of participation metrics by means of separate equivalence tests. We will also explore potential determinants of missing data and problematic response behaviors such as careless responding that might be clinically informative and that span different timescales as we test predictions from the retrospective baseline level (eg, ACEs) to the momentary level (eg, sedentary behavior).

To comprehensively address the inherent temporal and mechanistic complexity of dynamic interactions among health behaviors and affective functioning, multiple complementary statistical frameworks will be applied to data from the adolescent protocol. Each analytical model targets distinct characteristics of this interplay. We coregistered a first set of hypotheses claiming substantial heterogeneity in the daily within-person couplings of sleep patterns and daytime mood among adolescents [75,76]. Building upon prior research on internalizing etiology [87,88], we further hypothesize that the intensity of mood-sleep couplings is associated with levels of internalizing symptoms within the subclinical spectrum. These analyses apply Bayesian multilevel modeling [89,90].

To further contrast the relation between interindividual and intraindividual variability, time-series data will be analyzed to identify average and individual variability in motor activity, sleep patterns, eating habits, and associations with emotional states over different time periods (1 vs 2 vs 3 weeks; time of the day; day of the week; month and season). In order to provide preliminary findings on the different dynamic features of the interplay between health behaviors and mental health in youth, stability of sleep, physical activity, eating, and emotional states and their interrelationships will be assessed via different statistical approaches: (1) fragmentation analysis will be used to quantify subject-specific stability [91]; (2) functional principal component analyses and other machine learning approaches, such as the Joint and Individual Variance Explained (JIVE) [92] will be applied on the rich actigraphy dataset to explore the key components of sleep and physical activity, including average intensity, variability, and timing [93]; (3) Generalized Estimating Equations and network modeling [94] will be used to test the conditional estimate of the stationary dynamic associations between sleep, physical activity, eating, and emotional states, incorporating the lagged associations between these domains. Collectively, these approaches provide complementary rather than competing perspectives to deepen understanding of the dynamic interrelationships between health behaviors and mental health in youth.

Psychometric properties of the in-house translated scales (ie, PEARLS-CHILD, PEARLS-TEEN, PAQ-C, and PAQ-A) will be assessed based on internal consistency and construct validity. We will examine criterion validity by correlating ACEs with general psychopathology and self-reported physical activity with objectively recorded actigraphy. In addition, we will assess the test-retest reliability of the PAQ-C and PAQ-A scales across the 3 assessment waves based on changes observed in actimetry recordings.

Regarding the mixed populations and protocols within the present pilot study, we plan to balance data availability and statistical robustness for our feasibility targets and within-person effects. With respect to the limited annual intake of patients in the outpatient treatment who are eligible for group 1 (approximately 15 per year), our intention was to test feasibility per protocol with the combined preadolescent group reaching a sample size of 40 (ie, 20 each). Targeting the feasibility of the present multiwave design, we calculated statistical power for frequentist equivalence tests supporting the stability of response compliance rates across assessment waves. Based on equivalence bounds covering change scores within 70%±10%, a nominal alpha level of .05, and modest variability in change of response compliance (SD 0.1), statistical power of results in favor of stable compliance rates was considered adequate for the preadolescent protocol (power=0.87, n=40). This was also the case for the adolescent protocol (power=0.97, n=60), even under conservative Bonferroni correction accounting for 3 wave-specific equivalence tests (α=.015, power=0.91; see Figure S1 in Multimedia Appendix 1).

The rationale for determining sample size in the adolescent protocol, with regard to registered Bayesian multilevel models, was based on a reliability metric [95] for within-person effects between sleep patterns and daytime mood. A detailed explanation of this precision-oriented sample size calculation is available in the openly registered analysis plan [96]. Of note, precision-oriented determination of sample size aligns with principles of Bayesian analysis, emphasizing the accuracy of parameter estimates when comparing evidence for competing hypotheses [97].

Given the intensive longitudinal design and the likelihood of both intermittent and monotone missingness, our complementary analytic approaches will use listwise deletion or full-information maximum likelihood estimation to accommodate unbalanced data under a missing-at-random assumption. Where missingness may be informative, such as dropout being related to symptom severity, school holidays, or duration of participation, we aim to extend these models to jointly estimate the missingness mechanism and outcome trajectories using selection models within a structural equation modeling framework [98] or via joint multilevel survival models [99]. For variables exhibiting sporadic missingness under missing-at-random, we will perform multiple imputation by chained equations, generating at least 20 datasets. Finally, sensitivity analyses will compare these results to complete-case analyses to evaluate the robustness of our findings.

The study was approved by the Ethics Committee of the Vaud Canton, Switzerland (Project-ID 2021‐01821) and is in accordance with the Declaration of Helsinki. All participants received oral and written information about the study and received enough time to ask any questions. Adolescent participants provided written consent. Parents of adolescent participants received written information and were invited to reach out to the research team in case of any questions. Parents of preadolescent patients are met in person to explain the study and provide information letters and informed consent forms, whereas participants recruited in schools are asked to transmit information letters to their parents and return signed forms to the school staff or to the research team. Parents of both groups were encouraged to ask any questions regarding the study.

Adolescent participants will receive gift vouchers at a technology and culture retail chain of CHF 30 (US $37.72) at the end of the first and second, and CHF 40 (US $50.29) at the end of the third wave, respectively, when completing at least two-thirds of the self-report measures. Preadolescent participants receive gift vouchers of CHF 20 (US $25.15) and CHF 30 (US $37.72) at a technology and culture retail chain at the end of each respective wave, when completing at least two-thirds of the self-report measures.

During the ethics review, the local ethics committee recommended that such exclusion be avoided, noting that it could be perceived as unfair by students who had shown interest in participating. Participating schools opted to present the study to entire age cohorts, making oversampling a foreseeable outcome. This approach was formally documented in an amendment and received prior ethical approval.

This research highlights and attempts to capture the developmental dynamics of mental and physical health during emerging adolescence, using assessments across multiple timescales for in-depth, longitudinal analysis. Our feasibility-related findings will be crucially informative for future field research, especially in epidemiological paradigms. By examining the complex interplay between health behaviors across varying stages of late childhood and adolescence, we will gain preliminary insights into the transition from healthy states to those with mental health problems. This approach is pivotal for promoting health literacy, for developing age-appropriate early detection and prevention strategies, including transdiagnostic interventions in community and in clinical settings [100]. Using mobile technologies enables the creation of school-based programs for promoting healthy behaviors and discreetly identifying at-risk youth, particularly in older children and adolescents before the onset of many psychiatric disorders [101]. Because the study of these multiple health domains in youth using methods such as EMA and actigraphy is still struggling to converge on coherent frameworks of developmental psychopathology [102], the present methodological groundwork is essential across different clinical contexts and age groups.

Despite its commendable attributes, notably the rich and multifaceted longitudinal data, this research project has inherent limitations that need to be addressed. First, clinical participants were drawn exclusively from a single intensive outpatient treatment unit for preadolescents, and school‐recruited youth attended private institutions, resulting in unbalanced and unrepresentative sampling, which impedes group comparisons. Populations recruited from private schools likely differ systematically from public schools and community clinics in socioeconomic profile [103], access to beneficial environments (eg, parental support and calm and safe sleep environment) [104], and development of health behaviors [105]. For example, private‐school adolescents, who typically come from higher socioeconomic status (SES), have been shown to select healthier snacks and consume more fruits and vegetables than their public‐school peers [106], and high‐SES adolescents generally engage in greater levels of physical activity compared to lower‐SES youth [107]. Conversely, clinical patients represent a subgroup characterized by greater symptom severity and comorbidity, with adolescents accessing specialty mental health services tending to have more severe and persistent disorders than those in community settings [108]. Consequently, the temporal dynamics of sleep, activity, diet, and mood we characterize may not extend to adolescents in lower‐resource environments, public school settings, or clinical engagement. Our unbalanced design impeded corresponding group comparisons. Future large‐scale studies should purposively sample across diverse sociodemographic and educational contexts, including public schools, community mental health centers, and underresourced regions, to evaluate whether the longitudinal dynamics we observe hold across the broader population of youth. Second, while we aimed for a comprehensive evaluation of relevant health behavior domains, some areas may have been missed, such as interpersonal relations and peer dynamics. Third, the lack of detailed data on stressors and significant life events such as relationship break-ups or familial loss during the study is a limitation. Anticipated events, such as final school examinations, were also not examined. Fourth, the translation of selected questionnaires without formal validation due to the multilingual study cohort may be a limitation. Fifth, the reliance on self-reported measures, including both retrospective trait questionnaires and in situ EMA surveys, introduces potential biases such as recall error [109], social desirability [110], and measurement reactivity, which may affect the accuracy and generalizability of our findings. Although EMA minimizes recall intervals [111], participants’ awareness of ongoing monitoring can alter their responses. Future work should consider the most direct and unbiased assessment modalities available or under development (eg, passive dietary sensors and ecological smartphone usage logs) [102] or multimethod approaches to mitigate these self-report biases.

Despite these limitations, the comprehensive exploration of key health behavior domains and mental health in developing youth holds promise as a foundation for future studies. These findings can inform interventions and training programs to enhance prevention and intervention strategies for individuals.

The present multimodal ambulatory monitoring paradigm can be seamlessly embedded within routine therapeutic care by choosing the specific domain most relevant for a given clinical situation. At treatment initiation, EMA and actigraphy can be used collaboratively by patient and clinician to coconstruct a personalized case formulation, providing an “additional eye” on day-to-day symptom dynamics and functional patterns [112,113]. By repeating these assessments at midtherapy and during key care transitions (eg, stepping down from intensive support), clinicians and patients can discuss their perceptions in front of the changes captured between sessions to refine formulations, adjust treatment targets, and evaluate response in real time [6, 114,115].

Going forward, research should thoroughly test the validity of temporal and digital markers that emerge from fundamental psychopathology studies and translate into practical, in-session tools [113]. Such work will determine which patterns of sleep, activity, stress reactivity, or digital engagement most reliably map onto clinical processes (eg, remission and relapse) and other consequential outcomes, and how to present these markers to clinicians [116]. Ultimately, this line of research will bridge ambulatory assessment science and everyday practice, enhancing transparency, engagement, and, hopefully, the effectiveness of youth mental health interventions.

This research aims to strengthen the investigation of the intricate dynamics between mental and physical health during adolescence by investigating the interplay between health behaviors and affective functioning across multiple timescales. The findings will not only contribute to methodological advancements but also inform future large-scale epidemiological studies and interventions aimed at early detection and prevention of mental health issues, highlighting the importance of idiographic approaches in youth health research.