Physical and mental health in adolescence: novel insights from a transdiagnostic examination of FitBit data in the ABCD study

The study addresses whether objective physical fitness metrics derived from wearable devices are associated with mental health symptoms in early adolescence, and whether these associations are transdiagnostic or specific to particular symptom dimensions (psychosis-like experiences, internalizing, externalizing). Adolescence is a key developmental period marked by metabolic, hormonal, neural, and social changes, with rising sedentary behavior and declining physical activity. Prior research suggests links between cardiovascular fitness, sedentary behavior, and physical activity with mental health outcomes, but much of it relies on recall/self-report and often examines single symptom domains in parallel rather than jointly. The purpose is to leverage the ABCD cohort with wearable-derived fitness data to evaluate the specificity of associations between fitness behaviors and multiple symptom dimensions within the same individuals, hypothesizing that poorer cardiovascular fitness (higher resting heart rate) would be associated with poorer mental health.

Prior work across psychosis spectrum, internalizing, and externalizing psychopathologies has independently linked lower cardiovascular fitness and greater sedentary behavior with increased symptoms, and moderate-to-vigorous physical activity with reduced symptom severity, including observational and interventional studies. Reviews indicate physical activity benefits cognitive performance, brain development, and stress reduction in adolescence, and ABCD findings link optimal BMI/body morphology to better health, cognition, and neurodevelopment. However, much literature relies on self-report and does not jointly model multiple symptom dimensions, leaving unclear whether relationships are broad (transdiagnostic) or dimension-specific. Evidence also suggests distinct mechanisms: cardiovascular health, sedentary behavior, and physical activity may influence symptoms via different metabolic pathways (e.g., inflammation), implying potential specificity in associations and intervention targets.

Design and cohort: The study draws from the ABCD Study, a large, geographically diverse U.S. cohort recruited via schools across 21 sites, with standardized sampling to reflect U.S. demographics and include underrepresented groups. The subsample analyzed included adolescents who completed clinical symptom assessments and opted into Fitbit monitoring during Year 2. Participants: The current subsample included 507 individuals (ages 10.3–13.5 years; M=11.50, SD=0.65), 48.41% female, with BMI mean 20.17 kg/m² (SD=4.32). Demographics (sex assigned at birth, age, household income categories, race/ethnicity) were collected from parents at baseline. Measures: - Psychosis-like experiences: Prodromal Questionnaire–Brief Child Version (PQ-CB), 21 items; distress-weighted severity score sums items with distress ≥2. - Internalizing and externalizing: Child Behavior Checklist (CBCL) self-report; total internalizing (depression, anxiety subscales) and externalizing symptom scores; only complete responders included. Wearable fitness metrics: Fitbit Charge HR2 provided one-second heart rate sampling (photoplethysmography) and activity classifications. Derived metrics included resting heart rate (RHR), minutes spent sedentary, and minutes in moderate-to-intense physical activity. Weekly summaries were aggregated across days, then averaged across weeks for individuals with sufficient data per ABCD quality thresholds (at least one week of continuous daily data for metrics and at least one week for three weeks). Follow-up checks examined reliability versus hours recorded and proportions of moderate/vigorous activity. Statistical analysis: Separate multilevel models (lme4 in R) assessed each fitness metric’s association with PLE severity, internalizing, and externalizing symptoms in the same model, including random effects for familial relatedness and fixed effects for sex, age, household income, and BMI. When significant composite associations emerged, follow-ups examined internalizing subdomains (depression, anxiety). Log-transformations were tested to address zero-inflation; results were consistent. Multiple comparisons correction applied across three fitness metrics (Bonferroni p<0.017). Model diagnostics indicated low multicollinearity (fixed effects intercorrelations rs<0.10; VIF<4). Analytic code was made available for reproducibility.

- Resting heart rate (cardiovascular fitness): Higher RHR was associated with greater internalizing symptoms (t=3.632, p=2.84e−04; Bonferroni-significant). Association with PLE severity was weaker (t=2.127, p=0.033) and did not survive Bonferroni correction. No significant association with externalizing symptoms (t=1.758, p=0.079). - Sedentary time: Greater sedentary time was associated with higher PLE severity (t=5.492, p=4.02e−08; Bonferroni-significant). No significant associations with internalizing (t=−0.182, p=0.885) or externalizing (t=−0.813, p=0.99). - Moderate-to-intense physical activity: More moderate activity was associated with lower PLE severity (t=−2.698, p=0.007; Bonferroni-significant) and lower internalizing symptoms (t=−6.286, p=3.57e−10; Bonferroni-significant). The association with externalizing was not significant after correction (t=2.159, p=0.031). - Internalizing subdomains: Higher RHR was specifically linked to higher depression symptoms (t=3.073, p=0.002; Bonferroni-significant), not anxiety (t=1.451, n.s.). Greater moderate activity was linked to lower depression symptoms (t=−4.043, p=5.37e−05; Bonferroni-significant); the anxiety association did not survive correction (t=−2.392, p≈0.068). Effect sizes were small (e.g., d≈−0.17 to −0.19) but comparable to other adolescent risk markers. - Covariates: Several models showed significant effects of age, sex, BMI, and income on fitness metrics, consistent with developmental and sociodemographic influences (e.g., sex and BMI strongly related to moderate activity; age inversely related to RHR).

Findings support dimension-specific links between physical fitness/behavior and mental health in early adolescence. Elevated resting heart rate, reflecting lower cardiovascular fitness, was uniquely related to internalizing symptoms—particularly depression—after accounting for other symptom dimensions. Sedentary behavior showed a specific positive association with PLE severity, suggesting it may serve as an early risk marker for psychosis-like experiences, potentially via inflammatory or metabolic pathways. Moderate physical activity related to lower PLE and internalizing severity but not externalizing, aligning with literature on exercise benefits for psychosis spectrum and mood symptoms. Together, the results argue against a purely transdiagnostic effect and emphasize precision: different fitness metrics relate to different symptom dimensions. This specificity underscores the utility of wearable-derived digital health metrics to inform targeted prevention and intervention strategies during a sensitive developmental window.

This study leverages wearable-derived fitness metrics in a large adolescent cohort to demonstrate that physical health behaviors relate to mental health in a dimension-specific manner: higher resting heart rate links to internalizing (especially depression), sedentary time to PLE severity, and moderate physical activity to reduced PLE and internalizing symptoms. These insights highlight the promise of digital biomarkers for risk stratification and the need to tailor physical health interventions to target specific symptom dimensions in early adolescence. Future research should employ longitudinal designs to clarify temporal dynamics and causality, examine mechanistic pathways (e.g., inflammation, neuroplasticity), evaluate dose–response relationships for externalizing symptoms, and test precision exercise interventions tailored to specific symptom profiles.

- Cross-sectional, single time point wearable data precludes causal inference and limits insight into developmental dynamics across adolescence. - Potential sampling bias and limited generalizability: the analytic sample differed from the full ABCD cohort in household income and racial composition; demographic representativeness may be limited. - Small effect sizes, though comparable to other adolescent risk markers, warrant careful interpretation. - Observational design and aggregated fitness metrics may mask intra-individual variability and dose–response effects; externalizing outcomes may depend on activity dosage not captured here. - Wearable data reflect adherence and device classification constraints; although reliability checks were conducted, measurement artifacts cannot be entirely ruled out. - Some measures relied on self-report (e.g., PQ-CB, CBCL), which can introduce reporting biases despite the use of objective fitness metrics.