Circadian rhythm reflects an individual’s preference for sleep timing. Influenced by factors such as age, gender, and geographical latitude, the most distinct types of chronotypes are morningness and eveningness (9). Chronotype is strongly associated with sleep patterns; morning types typically maintain regular sleep schedules, demonstrating greater resilience to environmental disruptions and a greater tolerance for sleep disruptors (10). Conversely, evening types exhibit heightened vulnerability to disturbances in sleep-wake patterns. This irregularity constitutes a significant risk factor for mental disorders, with conditions such as sleep disorders and schizophrenia showing strong associations with the evening chronotype (11). The biological circadian system is governed by intricate regulatory mechanisms involving genes, hormones, and environmental cues (5). Key regulatory substances, such as cortisol, exhibit robust diurnal variation, typically peaking in the morning and declining throughout the day (12). Research has shown that individuals with relatively low daytime cortisol levels and elevated nocturnal levels demonstrate an increased risk for externalizing problem behaviors (12). Substantial evidence has linked sleep disturbances—particularly impaired quality or shortened duration—to externalizing behaviors like aggression, as shown in studies of both healthy populations and clinical groups (13). Notably, individuals with evening chronotypes exhibit a higher incidence of aggressive behavior.

Circadian dysregulation may affect brain circuits involved in emotion regulation, such as prefrontal-limbic system pathways, potentially increasing aggression risk (14). The hippocampus is especially critical; patients displaying aggression and antisocial personality disorder often show unilateral hippocampal atrophy and functional impairment (15). Notably, evening chronotypes exhibit specific associations with localized atrophy in the right hippocampal subiculum, which may contribute to their heightened aggression propensity through structural alterations.

Aggression in schizophrenia patients poses a substantial public health challenge, resulting in physical harm, perpetuating stigma, inflicting psychological distress, and escalating healthcare costs (16). It arises from multifactorial determinants including genetics, neurotransmitter systems, and neuroanatomical structure. Circadian rhythms may further modulate this aggression through hormonal/neurotransmitter regulation, as well as neurostructural plasticity. Building on this framework, we posit the following hypothesis:

Circadian rhythms are intrinsically linked to the regulation of sleep. The suprachiasmatic nucleus (SCN), the central circadian pacemaker, governs sleep timing, duration, architecture, and initiation, while also influencing wakefulness, alertness, cognitive function, and brain activity patterns (17). Research has indicated that a greater eveningness chronotype is linked to a higher risk of sleep disorders.

Sleep disturbances have been linked to increased impulsivity. Relevant studies have demonstrated that this is related to dysfunction of the medial prefrontal cortex (PFC) and the loss of top-down inhibitory control over the amygdala. This subsequently leads to excessive emotional responses to negative stimuli and an increase in emotional instability (18). According to the General Aggression Model, sleep disturbances may promote aggression through emotional, cognitive, and inhibitory control pathways (19). Additionally, the theory of hostile attribution bias suggests that poor sleep quality exacerbates the tendency to interpret ambiguous social cues as hostile, thereby triggering aggressive responses (20). Substantial evidence supported this connection (7, 21). Studies conducted by Stephen and Van et al. found a significantly higher aggressive behavior propensity in individuals with sleep disturbances, with poor sleep quality increasing aggression risk by 3.61-fold (22, 23). Given that sleep patterns are a primary circadian manifestation and sleep disturbances elevate aggression risk, we propose that sleep disturbances mediate the relationship between circadian rhythm dysfunction and aggression in patients with schizophrenia. Therefore, the impact of circadian disruption on aggression may operate partially through induced sleep pathology. Based on this rationale, we hypothesize:

Although circadian rhythms significantly influence aggressive behavior in schizophrenia patients, their impact varies substantially across individuals. Notably, even among patients sharing the same chronotype (e.g., evening type), the severity and likelihood of aggression differ. Consequently, identifying key moderating variables in this relationship is essential.

Social cognitive theory posits that social relationships profoundly shape cognition and behavior (24). Research has indicated that perceived social support effectively mitigates negative cognitions and associated maladaptive behaviors. Empirical work by Patti et al. further demonstrated that social support influences the manifestation of aggression by regulating cognitive processes (25). Simultaneously, the buffering model of social support suggested that perceived social support exerts beneficial effects regardless of stress levels (3), facilitating the reappraisal of stressful events. When excessive stress triggers negative emotions or behavioral tendencies (e.g., aggression), available social support buffers these adverse reactions.

Specifically in schizophrenia, circadian rhythms can elevate the risk of aggression via induced sleep disturbances. However, this risk manifests heterogeneously across patients, potentially moderated by their level of perceived social support. While sleep disturbances are a contributing factor to aggression, their severity is modulated by individual characteristics such as psychological resilience, stress tolerance, and adaptability (26). Crucially, support from family, friends, and social networks can enhance these capacities. Thus, perceived social support likely moderates the relationship between sleep disturbances and aggression: higher support may attenuate the adverse effects of sleep disturbances on aggressive behavior.

Recent studies corroborated the moderating role of perceived social support in the relationship between various factors and aggression (8, 27). For patients with schizophrenia who are experiencing significant stress and negative emotions, it plays a vital regulatory role in cognition and behavior. Based on this theoretical and empirical foundation, we propose:

In conclusion, this study employs a moderated mediation model to investigate the influence of circadian rhythm disruption on aggressive behavior in schizophrenia and its underlying mechanisms.

This was a cross-sectional study based on self-report questionnaires administered at a single time point. From September 2020 to August 2021, patients with schizophrenia were recruited from 28 communities in Pengzhou, China (including Tianfu Middle Road Community, Tianfu Middle Road Small District, Guangming Community, Jinyang Community, Longtan Community, Linjiang Community, and Qingping Community) using a convenience sampling approach. Questionnaires were administered via the Questionnaire Star platform, a widely used online survey tool in China that allows researchers to design, distribute, and collect questionnaire data electronically. Recruitment consisted of psychiatrists and community nurses identifying potentially eligible patients from their registers based on inclusion and exclusion criteria. Subsequently, eligible patients and their families were contacted by community team members to explain the study objectives and procedures. For those who initially expressed interest, face-to-face meetings were arranged to obtain written informed consent. Community staff then assisted participants in accessing and completing the questionnaire through the Questionnaire Star platform on their personal smartphones or provided a tablet device for this purpose to ensure that participants with limited digital literacy could participate. The inclusion criteria for participants were as follows:

Exclusion criteria were as follows:

Regarding drug use, some participants are receiving stable antipsychotic medication treatment, using 1–3 types of psychotropic drugs as part of community therapy. The lack of systematic collection of drug details is considered a limitation of the potential confounding effects of psychotropic drugs on chronotype and aggression. This study was approved by the Ethics Committee of the Fourth People’s Hospital of Chengdu, with review number [2017], ethical examination number (16), and China Clinical Trial Registration Number ChiCTR1800015219, registration date March 15, 2018. Upon completion of the study, the researchers exported all survey data from the Questionnaire Star platform. A total of 818 participants were initially enrolled in the study. Prior to formal data collection, a pilot study was conducted to establish a minimum threshold for valid questionnaire completion. Under researcher supervision, 25 community-dwelling patients with stable schizophrenia completed the full set of questionnaires as instructed. The 10th percentile of completion time was 605 seconds. Given that the questionnaires covered multiple dimensions requiring deliberate responses (e.g., demographics, chronotype, aggression, sleep disturbance and social support), and based on the pilot data, responses completed in less than 600 seconds (10 minutes) were flagged as rapid and excluded. This conservative threshold, set just below the pilot P10 value, aimed to screen out inattentive responses while preserving valid data. After applying this criterion, 785 valid questionnaires were obtained, yielding an effective response rate of 95.97%.

A priori sample size calculation was performed using G*Power 3.1. For a linear multiple regression model with up to 6 predictors (including covariates), assuming a small to medium effect size (f² = 0.05), α = 0.05, and power = 0.80, the required sample size was 146. Our final sample of 785 exceeds this requirement, providing adequate statistical power.

Data were analyzed using SPSS Statistics version 25.0 and the PROCESS macro (version 3.3). Prior to the analyses, all continuous variables were mean-centered. First, descriptive statistics and Spearman’s rank correlation analyses were conducted using SPSS for all variables. Next, PROCESS Model 4 was employed to examine the mediating effect of sleep disturbance. Subsequently, moderated mediation was tested using PROCESS Model 59. Finally, simple slope analysis was performed to examine whether the mediation effect of sleep disturbance (and chronotype) differed across different levels of the moderator variable. The following demographic variables were included as covariates in all statistical models: gender, body mass index (BMI categorized as<18.5,18.5-23.9,≥24kg/m²), family residence (rural/urban), occupation (categorized as unemployed, part-time/full-time employed, student, or other), monthly family income (categorized as ≤1000, 1001-3000, ≥3001 yuan), and marital status (single, married, widowed, divorced).Percentile bootstrap confidence intervals (CIs) were calculated based on 5,000 bootstrap samples.

Data for this study were collected via self-report questionnaires, which may introduce common method bias (35). To mitigate this potential issue, procedural remedies (including participant anonymity and reverse-scoring of items) were implemented during data collection. Additionally, Harman’s single-factor test was conducted prior to analysis to assess standard method variance statistically. An exploratory factor analysis, which included all items measuring the four constructs (chronotype, sleep disturbance, perceived social support, and aggressive behavior), revealed six factors with eigenvalues greater than 1. The first factor accounted for 23.21% of the total variance, which is below the commonly accepted threshold of 40%. These results suggest that common method bias is unlikely to have substantially affected the findings of this study.

This study included 785 community-dwelling patients with schizophrenia. The sample demographics were as follows: 414 (52.74%) male, 669 (85.22%) residing in rural areas, 555 (70.70%) unemployed, and 374 (47.64%) married. Regarding income distribution, the most significant proportion (n=402, 51.21%) reported a monthly income of ￥1001-3000. Based on BMI categories(<18.5kg/m²:underweight;18.5-23.9kg/m²:normal weight;≥24kg/m²: overweight/obese), 317 patients (40.38%) were overweight or obese. Marital status categories included single, married, widowed, divorced and others (After divorce or widowhood, having a new spouse). Aggressive behavior showed statistically significant differences across gender subgroups (χ² = 11.720, p = 0.001), with a higher percentage of females (18.06%) reporting aggressive incidents compared to males (9.66%). See Table 1.

Partial correlation analyses among chronotype, sleep disturbances, aggressive behavior, and perceived social support, controlling for all demographic covariates (gender, BMI, residence, occupation, income, marital status), revealed significant associations. Sleep disturbances exhibited significant negative correlations with both chronotype and perceived social support, but a significant positive correlation with aggressive behavior. Chronotype demonstrated a significant positive correlation with perceived social support and a significant negative correlation with aggressive behavior. Additionally, perceived social support was significantly negatively correlated with aggressive behavior. Although the associations were statistically significant, their small effect sizes, all r < 0.2, suggest that the relationship is modest and may be influenced by other unmeasured variables. The mean MOAS score was low (0.32 ± 1.04). Only 13.63% of participants reported any aggressive incident (MOAS subscale score >0) in the past month, indicating a low base rate of overt aggression in this clinically stable community sample. Mainly because study entry criteria required patients to be clinically stable (PANSS total score <60), which may have selected for a less symptomatic population. Second, the community setting itself may represent a patient group with better disease control. And that patients may have social expectation bias or lack of insight when using self-rating instruments to measure aggressive behavior. Complete correlation coefficients and significance values are presented in Table 2.

Using Hayes’ PROCESS macro (Model 4) in SPSS (36), we examined sleep disturbance as a mediator between chronotype and aggression. After controlling for demographic covariates (e.g., gender), chronotype significantly predicted sleep disturbance (a=-0.218, SE = 0.0347, p < 0.001). In the regression model including both predictors, chronotype retained a significant direct effect on aggression (c’ = -0.187, SE = 0.0347, p < 0.001), while sleep disturbance also significantly predicted aggression (b=0.178, SE = 0.0344, p < 0.001). Bias-corrected bootstrapping (5,000 samples) confirmed a significant indirect effect (ab=-0.032, BootSE=0.0122, 95% CI [-0.0594, -0.0117]), indicating that 14.22% of the total effect was mediated (ab/(ab + c’)). This suggests that disturbed sleep, although a statistically significant mediator, is not the only mechanism responsible for the total effect of chronotype on aggressive behavior. See Table 3 for complete results.

In the second analytical phase, moderated mediation analysis was conducted using Model 59 of the SPSS PROCESS macro to test the moderating role of perceived social support. Three standardized regression equations were estimated simultaneously while controlling for demographic covariates (e.g., gender). Variance inflation factors (VIFs) remained ≤ 3.0 across models, indicating no concerns regarding multicollinearity.

As shown in Table 4, all conditions for moderated mediation were satisfied: (1) Chronotype significantly negatively predicted aggressive behavior (β = -0.19, p < 0.001); (2) Chronotype significantly negatively predicted sleep disturbance (β = -0.17, p < 0.001); and (3) Perceived social support demonstrated a significant main effect negatively predicting aggressive behavior (β = -0.13, p < 0.001), along with significant interaction effects were observed between social support and both chronotype (β = -0.09, p < 0.01) and sleep disturbance (β = 0.09, p < 0.01). These findings confirm that perceived social support moderates both the direct circadian rhythm-aggression pathway and the sleep disturbance-aggression pathway within the mediation framework. The complete moderated pathways are illustrated in Figure 1.

To elucidate the interaction effects of perceived social support on chronotype and sleep disturbance, we conducted a simple slope analysis by dichotomizing perceived social support into high (M+SD) and low (M-SD) groups. We created a visualization of the simple effects.

Simple slope analysis revealed distinct predictive patterns based on levels of perceived social support. For patients with schizophrenia in the low perceived social support group (M-SD), chronotype showed a significant negative prediction of aggressive behavior (Bsimple=-0.0265, t=-5.8482, p < 0.001). Conversely, this negative relationship was non-significant in the high perceived social support group (Bsimple=-0.756, t=-1.5947, p>0.05). These differential moderation effects are illustrated in Figure 2.

In the sleep-disturbed chronotype pathway, sleep disorder significantly predicted aggressive behavior in schizophrenia patients with low scores (Bsimple=0.2559, t=5.4295, p < 0.001), whereas this prediction was not significant in patients with high scores (Bsimple=0.0720, t=1.4863, p>0.05) (see Figure 3 for details).

Collectively, the findings indicated that the pathway linking chronotype disturbances to aggressive behavior via sleep disorder is mediated by perceived social support. Specifically, for schizophrenia patients with low perceived social support, the indirect effect of chronotype disturbances on aggressive behavior through sleep disorder was statistically significant (index=-0.0394, BootSE=0.0223, 95%CI [-0.0924, -0.0064]). In contrast, for patients with high perceived social support, this indirect effect was not statistically significant (index=-0.0128, BootSE=0.0102, 95% CI [-0.0359, 0.0041]).

The overall framework of this study was cross-sectional, limiting the ability to establish causal relationships between variables. Future studies should employ longitudinal designs or intervention experiments to further verify causal mechanisms. Second, our sample was only from schizophrenia patients in Sichuan Province, limiting the generalizability of the results. Future studies should expand the sampling range to patients nationwide to improve representation. Third, the assessment of aggressive behavior relies entirely on self-assessment and may be affected by recall bias and underreporting, especially since individuals may not fully perceive or acknowledge the aggressiveness of their behavior. Fourth, the exclusion of patients with significant physical comorbidities (e.g., metabolic syndrome) may limit the representativeness of our sample regarding the broader community-dwelling schizophrenia population, which often presents with multimorbidity. Fifth, all key variables were assessed using self-rating instruments, which are susceptible to common method variance and biases such as social expectations and inaccurate recall. Future studies using objective measures (e.g., actigraphy for circadian rhythms and sleep, collateral reporting of aggressive behavior) would be more persuasive. Sixth, we did not systematically collect information on antipsychotic medications, which may influence circadian rhythms and aggression, thus representing a potential confounding factor. Additionally, data collection during the COVID-19 pandemic (2020-2021) may have influenced sleep patterns, social support, and aggression due to pandemic-related stressors, potentially affecting result generalizability. Finally, studies were not pre-registered, increasing the risk of analytical flexibility. Future studies should incorporate objective measures, use nationally representative samples, and consider other potential mediating and moderating variables to construct more comprehensive causal models.