ETD

• actigraphy
• adolescents
• psychiatric acuity
• sleep

INTRODUCTION
Adolescence represents a critical period for sleep physiology, which is a fundamental biological process. In adolescents it is characterized by a biologically driven phase delay. As a result, over 60% of adolescents become chronically sleep-restricted. These normal developmental shifts are compounded by modern lifestyle factors, including evening smartphone use, social media, and gaming. Sleep disturbances have a prevalence of 20–30% in children and adolescents aged 0–17, and are severely underdiagnosed. Within psychiatric populations, sleep disturbances are a transdiagnostic vulnerability factor contributing to affective dysregulation, impulsivity, and suicidal risk.
The present study aimed to characterize objective and subjective sleep patterns in this population and to examine whether adolescents with suicidal ideation or behavior exhibit distinctive circadian features. A secondary objective was to assess the clinical utility of integrating systematic sleep assessment into acute inpatient care, given that sleep disturbances are less stigmatized than other psychiatric risk factors and represent a potentially modifiable therapeutic target.
METHODS
This cross-sectional observational study was conducted at the Unità Operativa Semplice di Emergenza Psichiatrica (UOSD) of the IRCCS Fondazione Stella Maris, between April 2025 and April 2026. The sample comprised two main groups. The inpatient cohort consisted of 59 adolescents aged 12–17 years admitted for acute psychiatric emergencies. Within this cohort, patients were subdivided based on suicidal risk: those with suicidal ideation or behaviors (C-SSRS score ≥ 3 or a suicide attempt within the past 24 months; n = 50) and those without suicidal risk (n = 11). The control cohort consisted of 28 healthy adolescents.
Sleep was assessed objectively through actigraphy using the Axivity AX6 device and raw actigraphic data were processed by the Dormi algorithm providing standard AASM sleep metrics: TST, SOL, SE, NAW, Duration of Awakenings (DAW), Movement Index (MI), Activity Index (AI), WASO, SRI, and estimated Dim-Light Melatonin Onset (DLMO).
Subjective sleep measures included three validated questionnaires: the reduced Morningness–Eveningness Questionnaire for Children and Adolescents (rMEQ-CA) for chronotype assessment; the School Sleep Habits Survey (SSHQ) for sleep habits and daytime sleepiness; and the Sleep Disturbance Scale for Children (SDSC), a parent-reported tool.
Psychiatric and suicidal risk assessment employed standardized scales for psychiatric assessment in children and adolescents.
RESULTS
The descriptive actigraphic analysis of the control cohort revealed that, while most healthy adolescents maintained normal SE (100%), normal DAW (82.1%), and normal DLMO (92.9%), the most critical finding was a pathological NAW in 89.3% of participants, indicating widespread sleep fragmentation. TST was borderline in 60.7% of controls, and WASO was borderline in 32.1%.
In the patient cohort, macro-sleep structure was relatively preserved: 72.9% of patients had normal TST and 91.5% maintained normal SE. However, motor activity during sleep was markedly elevated: 71.2% of patients showed borderline AI and 61.0% showed borderline MI. Group comparison between patients and healthy controls revealed statistically significant differences in TST (p < 0.001), with patients sleeping approximately 1 hour 40 minutes longer on average (mean TST: 552.5 vs. 452.0 minutes). Activity Index (AI; p < 0.001) and Movement Index (MI; p < 0.001) were both significantly higher in patients. Pathological WASO was found in 22.0% of patients versus 10.7% of controls. Conversely, pathological NAW was similarly elevated in both groups (86.4% of patients vs. 89.3% of controls), suggesting that sleep fragmentation via frequent short awakenings is a shared feature across both clinical and healthy adolescent populations in this context. Mean DLMO was earlier in patients (20:46) compared to controls (21:36).
Correlations between actigraphic parameters and K-SADS-PL diagnoses revealed several clinically meaningful associations. Alcohol use disorder showed a positive correlation with delayed DLMO (r = 0.59), bipolar disorder was associated with increased TST (r = 0.27) and reduced SRI (r = –0.34). Anxiety disorders (panic, generalized anxiety, separation anxiety) showed negative correlations with SRI (r = –0.25 to –0.28) and DLMO (r = –0.29). ADHD was characterized by greater nocturnal fragmentation (NAW: r = 0.43) and shorter TST (r = –0.35). Anorexia Nervosa was associated with reduced SRI (r = –0.42), earlier DLMO (r = –0.37), and longer DAW (r = 0.50).
In the preliminary interim analysis conducted on the first 26 inpatients, the correlation matrix between C-SSRS subscales and actigraphic parameters revealed a consistent pattern of negative associations between SRI, DLMO and suicidal indicators. Specifically, SRI showed moderate negative correlations with intensity of suicidal ideation (r = –0.47, p < 0.05), control over ideation (r = –0.40, p < 0.05), total number of suicidal behaviors (r = –0.38, p < 0.05), NSSI (r = –0.38, p < 0.05), and suicidal behaviors in the previous three months (r = –0.34, p < 0.05). DLMO showed negative correlations with total suicidal behaviors (r = –0.38, p < 0.05), suicide attempts (r = –0.45, p < 0.01), and suicidal behaviors in the previous three months (r = –0.43, p < 0.01).
Multivariate linear regression models confirmed the independent negative effect of SRI on suicidal ideation intensity, controlling for clinical diagnoses and pharmacological covariates.
When the full sample (n = 59) was analyzed, the SRI–C-SSRS associations were no longer present across all seasons. Seasonal stratification revealed that the negative correlation between SRI and ideation intensity was significant only in summer (r = –0.16, p = 0.044). Grouping data into two six-month clusters, the Summer–Autumn cluster showed a moderate negative trend (r = –0.332, p = 0.079), while the Winter–Spring cluster showed a weak positive trend (r = 0.282, p = 0.132). These findings suggest that seasonality moderates the SRI–suicidality relationship. Meteorological analysis showed that temperature anomalies correlated positively with DLMO (r = 0.29, p < 0.10) and AI (r = 0.28), while solar exposure correlated negatively with MI (r = –0.22) and AI (r = –0.24), and average temperature correlated negatively with TST (r = –0.26).
The analysis of subjective sleep measures (SSHQ, SDSC, rMEQ-A) in relation to C-SSRS scores revealed a broad and statistically coherent pattern of positive associations, whereby worse self- and parent-reported sleep quality was systematically linked to greater suicidal severity. Daytime sleepiness (SLS, SSHQ) showed the strongest associations: r = 0.55 with suicidal behaviors in the previous three months, r = 0.46 with NSSI, r = 0.42 with suicide attempts, and r = 0.41 with ideation intensity.
DISCUSSION
The most notable difference in objective sleep between patients and controls was the marked increase in TST among inpatients, likely attributable to a combination of factors. Earlier DLMO timing in patients, relative to controls, suggests a tendency toward an advanced rather than delayed circadian phase.
The diagnosis-specific actigraphic profiles are broadly consistent with the existing literature.
The convergence of subjective and objective findings regarding sleep and suicidality is of particular clinical relevance. From questionnaire data, daytime sleepiness emerged as the strongest correlate of suicidality across all C-SSRS dimensions, indicating that perceived sleep fragmentation and inefficiency represent sensitive markers of suicidal and self-injurious vulnerability.
From the actigraphic perspective, SRI emerged as a biologically coherent marker: lower sleep regularity was consistently associated with greater suicidal ideation intensity and a higher probability of preparatory acts, independently of diagnostic and pharmacological covariates. Similarly, delayed DLMO was associated with the full spectrum of suicidal behaviors, and maintained its correlation throughout the entire study period regardless of season. The seasonal moderation of the SRI–ideation association — significant only in summer, with a moderate negative trend in the Summer–Autumn cluster — suggests that environmental factors (heat, solar exposure) interact with circadian regulation to modulate suicidal vulnerability, possibly via effects of temperature anomalies on DLMO and nocturnal motor activity.
CONCLUSIONS
This study provides preliminary evidence that circadian dysregulation — particularly low sleep regularity (SRI) and delayed melatonin onset (DLMO) — constitutes a biologically coherent and clinically meaningful marker of suicidal vulnerability in hospitalized adolescents. Lower SRI independently predicted greater intensity of suicidal ideation and a higher likelihood of preparatory acts across multiple regression models.
The finding that sleep fragmentation is nearly universal in both clinical and healthy adolescent cohorts (pathological NAW in 86–89% of both groups) underscores the broader public health relevance of sleep hygiene in this developmental period.
The seasonal moderation of the SRI–suicidality relationship suggests that environmental factors may interact with circadian regulation to amplify suicidal vulnerability, warranting further investigation in longitudinal studies with meteorological covariates.
From a clinical standpoint, these results support the integration of systematic sleep assessment — including actigraphy and validated questionnaires — into the routine evaluation of adolescents admitted for psychiatric emergencies. Sleep disturbances are less stigmatized than other psychiatric risk factors, are measurable with non-invasive and well-tolerated devices, and represent a potentially modifiable therapeutic target: evidence-based interventions for insomnia and circadian dysregulation (sleep hygiene psychoeducation, melatonin, CBT-I) have demonstrated preliminary efficacy within a limited number of sessions, with positive effects extending to non-sleep outcomes including depression and suicidal ideation. Future research with larger samples, longitudinal designs, and polysomnographic validation should aim to define clinical risk thresholds and to establish the directionality and mediating mechanisms of the sleep–suicidality relationship in this vulnerable population.