Skip to main content
BMC Psychology logoLink to BMC Psychology
. 2026 Aug 6;14:1584. doi: 10.1186/s40359-026-05316-5

Symptom-level associations among sleep quality, depression, and social rhythm in people living with HIV: a network analysis

Jiali Nie 1,#, Xuedi Dai 1,#, Houyu Huang 1, Qianqian Yuan 1, Honghong Wang 2,✉, Jingjing Meng 1,✉
PMCID: PMC13729331  PMID: 42972607

Abstract

Background

Sleep disturbance, depression, and social rhythm irregularity commonly co-occur in people living with HIV (PLWH), yet their symptom-level associations remain insufficiently understood. This study aimed to examine the network structure linking sleep quality, depressive symptoms, and social rhythm in PLWH and to identify central and bridge symptoms within this system.

Methods

A cross-sectional sample of 737 people living with HIV (PLWH) in Changsha, China, was assessed using the Pittsburgh Sleep Quality Index, Self-Rating Depression Scale, and Brief Social Rhythm Scale. A 32-node Gaussian graphical model was estimated using EBICglasso, with correlation types selected according to variable measurement levels. Expected influence (EI) and bridge expected influence (BEI) identified central and bridge symptoms, respectively, and were z-standardized for presentation. Community-size-normalized BEI was calculated to account for unequal community sizes, and its agreement with raw BEI rankings was assessed using Spearman’s rank correlation. Node predictability was estimated using a separate Gaussian model, with the regularization parameter selected by 10-fold cross-validation. Edge-weight accuracy and centrality stability were evaluated using nonparametric and case-dropping bootstraps, respectively. Goldbricker screening and reduced-network sensitivity analysis assessed node redundancy and the robustness of centrality findings.

Results

The network consisted of 32 nodes and 174 nonzero edges, indicating substantial interconnections among sleep quality, depression, and social rhythm. Subjective sleep quality had the highest z-standardized expected influence (EI = 2.40) and bridge expected influence (BEI = 2.87). After community-size normalization, depressive sleep disturbance had the highest community-size-normalized bridge expected influence (nBEI = 0.045). Subjective sleep quality, daytime dysfunction, depressive sleep disturbance, and fatigue consistently comprised the four leading bridge nodes in both the bridge expected influence and community-size-normalized bridge expected influence analyses, although their ordering differed.

Conclusions

Sleep quality, depression, and social rhythm were closely interconnected at the symptom level among PLWH. Subjective sleep quality, daytime dysfunction, depressive sleep disturbance, and fatigue may warrant prioritization in clinical assessment. Longitudinal and interventional studies are needed to determine whether targeting these symptoms can improve other symptoms across the network.

Keywords: HIV/AIDS, Sleep quality, Depression, Social rhythm, Network analysis

Background

Human immunodeficiency virus (HIV) infection remains a major global public health challenge [1], with an estimated 40.8 million people living with HIV (PLWH) worldwide [2]. With advances in antiretroviral therapy (ART), HIV has increasingly become a chronic condition, shifting the focus of HIV care from extending survival to promoting long-term psychological health, daily functioning, and quality of life [3]. In this context, depression and sleep disturbances are among the most common and clinically significant psychological problems in PLWH. A recent meta-analysis reported that the prevalence of depression among PLWH is approximately 35% [4], while a systematic review found that the pooled prevalence of sleep disturbances in this population is 52.29% [5]. Both conditions are associated with poorer psychosocial functioning [6], reduced quality of life [7], and adverse health outcomes [8]. Clarifying the interrelationship between depression and sleep disturbances is therefore important for improving psychiatric assessment and informing targeted intervention strategies in HIV care.

Sleep disturbances and depression frequently co-occur in PLWH and may reinforce one another. Sleep disturbances can increase emotional vulnerability and worsen depressive symptoms, whereas depression often manifests with prominent sleep-related symptoms [9]. Sleep disturbances and depression constitute a critical symptom cluster that affects quality of life, treatment adherence, and even immune function in PLWH [10]. However, previous studies have predominantly employed total scale scores and regression-based approaches to examine the relationship between sleep disturbances and depression at a global level [10]. Such approaches conceptualize sleep disturbances and depression as relatively homogeneous constructs, overlooking symptom heterogeneity and the potential for specific symptoms to exert disproportionate influence within a broader comorbidity system.

Social Rhythm Theory provides a useful symptom-level framework for understanding the comorbidity between sleep disturbances and depression. This theory proposes that regular daily routines, including sleep-wake schedules, work/study, entertainment/socializing, and meals, function as social zeitgebers that help regulate endogenous circadian rhythms and emotional functioning [11]. Disruptions in these rhythms may increase vulnerability to sleep and mood disturbances [12]. This framework may be particularly relevant to PLWH, whose daily routines can be affected by chronic disease management demands [13], stigma-related stress [14], and treatment schedules [15]. Within this theoretical framework, social rhythm disruption may not merely coexist with sleep disturbances and depression, but may occupy a structurally important position within their symptom network.

Network analysis conceptualizes symptom clusters as systems of interconnected symptoms, thereby enabling the investigation of symptom-level conditional associations, central symptoms, and bridge pathways among sleep quality, depressive symptoms, and social rhythm [16]. To date, symptom-level evidence on the joint network structure of sleep quality, depression, and social rhythm in PLWH remains limited. Therefore, this study employs network analysis to examine the interrelationships among sleep quality, depression, and social rhythm in PLWH, identify the most central and bridging symptoms within the network, and evaluate the stability of the estimated network structure. By elucidating the structural role of social rhythm in relation to sleep quality and depression, this study aims to provide more precise, theoretically grounded evidence to inform integrated assessment and future intervention research for PLWH.

Methods

Setting and participants

This cross-sectional study was conducted from December 2020 to February 2021 at an HIV-designated hospital and an HIV outpatient clinic of the Center for Disease Control and Prevention in Changsha, China. Consecutive sampling was used to recruit all eligible participants presenting at the two study sites during the prespecified study period. Eligible participants met the following inclusion criteria: (1) were diagnosed with HIV infection by the Center for Disease Control and Prevention at least one month before enrollment; (2) were aged 18 years or older; and (3) were willing to participate and provide informed consent. The exclusion criteria were as follows: (1) impaired consciousness or communication difficulties that prevented questionnaire completion; (2) current diagnosis of a neurological or psychiatric disorder (e.g., schizophrenia, depressive disorder, bipolar disorder); and (3) severe acquired immunodeficiency syndrome (AIDS)-related opportunistic infections (e.g., active tuberculosis, Pneumocystis jirovecii pneumonia).

Data collection

Data were collected by five trained investigators, including one doctoral-level nursing researcher, two master’s-level nursing researchers, one HIV specialist, and one psychotherapist. All investigators received standardized training before data collection to ensure consistency in questionnaire administration and adherence to ethical procedures. Before participation, participants were informed about the study objectives, procedures, voluntary nature of participation, confidentiality protections, and their right to withdraw. For participants with limited literacy or difficulty completing the questionnaires independently, investigators provided standardized assistance to ensure comprehension while minimizing interviewer-induced bias. Questionnaires were completed in quiet, private, and well-lit consultation rooms at the participating sites. The questionnaire package required approximately 20–30 min to complete. Participants received nominal compensation of 20 Chinese yuan in cash or via WeChat after completion. Immediately after submission, each questionnaire was reviewed for completeness by a trained investigator before the respondent left the study site. If an unanswered item was identified, the investigator informed the participant of the omission and asked whether they were willing to provide a response. Responding to any omitted item remained voluntary, and participants were allowed to leave items unanswered. Questionnaires with excessive missing responses or major logical inconsistencies were excluded during data cleaning.

Sociodemographic and clinical characteristics

A structured questionnaire developed by the research team was used to collect sociodemographic and clinical characteristics, including sex, age, time since HIV diagnosis, and ART treatment status.

Sleep quality

Sleep quality was measured using the Pittsburgh Sleep Quality Index (PSQI), which assesses sleep quality and sleep-related difficulties over the past month [17]. The Chinese version has demonstrated good reliability and validity among PLWH [18]. The PSQI comprises 18 scored items aggregated into seven components: subjective sleep quality, sleep latency, sleep duration, sleep efficiency, sleep disturbances, sleep medication, and daytime dysfunction [19]. Each component is scored from 0 to 3, and the global score ranges from 0 to 21, with higher scores indicating poorer sleep quality [20]. A global score > 5 indicates poor sleep quality. In this study, the Cronbach’s α coefficient for the PSQI was 0.737, which is considered acceptable given that the PSQI assesses multiple distinct dimensions of sleep rather than a single homogeneous construct [21, 22]. For the network analysis, the seven established component scores were entered as sleep-quality nodes, consistent with the original scoring structure of the PSQI. Individual item responses were not modeled separately because they are designed to derive these seven clinically meaningful components rather than to function as independent indicators [17].

Depression

Depression was measured using the Self-Rating Depression Scale (SDS) [23], developed by Zung and adapted for Chinese populations by Wang et al. [24]. The SDS contains 20 items rated on a 4-point scale, with raw scores ranging from 20 to 80 [23]. Standardized scores are obtained by multiplying the raw total score by 1.25, with higher scores indicating more severe depressive symptoms. Standardized scores of < 50, 50–59, 60–69, and ≥ 70 are conventionally interpreted as indicating normal, mild, moderate, and severe depressive symptoms, respectively. Positively worded SDS items were reverse-scored according to the standard scoring procedure, such that higher scores on all items indicated greater depressive symptom severity. In this study, Cronbach’s α was 0.822. The SDS was used to assess the severity of depressive symptoms rather than to establish a clinical psychiatric diagnosis. Therefore, self-reported depressive symptoms identified by the SDS were not considered an exclusion criterion. For network analysis, all 20 SDS items were retained as individual depression nodes to preserve symptom-level heterogeneity and clinical interpretability, consistent with the symptom-level network approach [16, 23].

Social rhythm

Social rhythm was assessed using the Brief Social Rhythm Scale (BSRS), developed by Margraf et al. [25]. The scale includes 10 items covering rhythms related to sleep, wake-up, study/work, entertainment/socializing, and meals over a 7-day period. Each item is rated on a 6-point scale from 1 (very regular) to 6 (very irregular), and total scores range from 10 to 60, with higher scores indicating greater irregularity in daily routines. In this study, the Cronbach’s α coefficient for the BSRS was 0.923. For network analysis, the paired weekday and weekend items for each activity were averaged to generate five activity-specific rhythm nodes: sleep, wake-up, study/work, entertainment/socializing, and meals. This approach preserved the original content structure of the BSRS by representing each daily activity as a single node while integrating its regularity across weekday and weekend contexts. These activity-specific nodes were treated as content-based indicators rather than psychometrically validated subscales.

Ethical considerations

The study was approved by the Nursing and Behavioral Medicine Research Ethics Review Committee of Xiangya School of Nursing, Central South University (Approval No. E2020182) and by administrative departments of the participating institutions. All participants provided written informed consent before participation.

Statistical analysis

Descriptive statistics were conducted using IBM SPSS Statistics version 25.0. Continuous variables are presented as mean ± standard deviation (SD) or median and interquartile range (IQR), as appropriate. Categorical variables are presented as frequencies and percentages.

Network estimation

All network analyses were conducted in R version 4.4.2 using the qgraph, bootnet, mgm, networktools, and NetworkComparisonTest packages. The analytic network included 32 nodes: 7 PSQI components, 20 SDS symptom items, and 5 BSRS rhythm-domain nodes. Node definitions followed the established scoring or content structure of each instrument and the symptom-level objective of the study. The 32-node model was retained to preserve the prespecified representation of the constructs and clinically meaningful symptom heterogeneity. Given the differing measurement levels of the variables, correlations were selected automatically using the cor_auto procedure, which computes polychoric, polyserial, or Pearson correlations as appropriate. The resulting correlation matrix was used for network estimation. The network was estimated using the graphical least absolute shrinkage and selection operator (graphical LASSO) [26] with model selection based on the Extended Bayesian Information Criterion (EBIC) [27]. The EBIC tuning parameter gamma was set to 0.5 [28]. This regularization procedure shrinks small associations to zero, yielding a sparse and interpretable network. Edges represent EBICglasso-regularized partial correlations between two nodes after controlling for all other nodes [28]. The sign indicates the direction of the conditional association, and the absolute magnitude indicates its strength. Edge weights should not be interpreted as percentages or proportions of explained variance. The network was visualized using the Fruchterman–Reingold algorithm in qgraph, with nodes grouped into the three predefined communities: sleep quality, depression, and social rhythm communities. Positive associations are shown as green edges, negative associations as red edges, and thicker edges indicate stronger associations.

Centrality and predictability

Expected influence (EI) was calculated to identify influential nodes in the overall network [29]. EI was selected because it retains the signs of edge weights and is therefore appropriate for networks containing both positive and negative associations. Higher EI values indicate stronger overall net influence within the network. Bridge expected influence (BEI) was calculated using the networktools package to identify nodes connecting the predefined sleep quality, depression, and social rhythm communities [30]. Higher BEI values indicate stronger cross-community net influence. EI and BEI values were z-standardized for presentation. Because the three communities differed in size, community-size-normalized bridge expected influence (nBEI) was additionally calculated using normalize = TRUE in the bridge function. Agreement between the BEI and nBEI rankings was assessed using Spearman’s rank correlation.

Node predictability was estimated using a Gaussian graphical model implemented in the mgm package [31]. For this complementary analysis, all node scores were treated as continuous variables, and the regularization parameter was selected using 10-fold cross-validation based on the minimum cross-validated deviance. A fixed random seed was used before model estimation to ensure reproducibility. Predictability was quantified as explained variance (R²), representing the proportion of variance in each node explained by all remaining nodes in the fitted Gaussian model. Higher R² values indicate greater statistical predictability within this complementary model.

Network accuracy, stability, and sensitivity analyses

Network accuracy and stability were evaluated using bootstrap procedures implemented in the bootnet package [32]. Nonparametric bootstrapping with 1,000 resamples generated 95% confidence intervals for edge weights, with narrower confidence intervals indicating greater estimation precision. Case-dropping bootstrapping with 1,000 resamples assessed the stability of EI and BEI. Because nBEI was additionally examined to account for unequal community sizes, its stability was evaluated in a separate case-dropping bootstrap analysis using the same settings. Stability was quantified using the correlation stability coefficient (CS-C), with values greater than 0.25 indicating acceptable stability and values greater than 0.50 indicating good stability.

As a sensitivity analysis, potential node redundancy was assessed using the Goldbricker procedure in the networktools package [33], with p = 0.05, the Hittner method, a difference threshold of 0.25, and a minimum correlation of 0.50. For each potentially redundant pair, one node was retained using the best_goldbricker option in the net_reduce function, and the reduced network was re-estimated using the primary EBICglasso settings. Spearman’s rank correlations were used to compare the rankings of EI, BEI, and community-size-normalized BEI between the primary and reduced networks among nodes retained in both networks.

To explore whether the network differed by time since HIV diagnosis, the sample was divided into a ≤ 36-month group (n = 247) and a > 36-month group (n = 490). Time since HIV diagnosis was initially categorized into 1–3, 4–12, 13–36, and > 36 months. Because the first two categories included only 18 and 36 participants, respectively, estimating separate 32-node networks for all four categories could result in imprecise estimates [32] and insufficient statistical power for between-group comparisons [34]. The first three categories were therefore combined to form two analytically feasible groups. Because the 36-month cutoff was not clinically validated, the subgroup analysis was considered exploratory. Networks for the two groups were estimated using the primary EBICglasso procedure and compared in terms of network structure and global strength using the Network Comparison Test with 1,000 permutations for independent samples [34]. Given the cross-sectional design, all network edges and centrality indices were interpreted as conditional associations rather than causal effects. The study was reported in accordance with the STROBE (Strengthening the Reporting of Observational Studies in Epidemiology) statement for cross-sectional studies.

Results

Participant characteristics

Of 1,097 PLWH screened, 32 were excluded because they were younger than 18 years, and 287 declined participation. Among the 778 participants who completed the questionnaire package, 41 questionnaires with excessive missing responses or major logical inconsistencies were excluded, yielding a final analytic sample of 737. The final analytic dataset contained no missing values; therefore, no imputation was performed. The sample was predominantly male (n = 649, 88.1%). Most participants were aged 18–60 years (n = 680, 92.3%), and the mean age was 38.47 years (SD = 13.66). The mean time since HIV diagnosis was 55.59 months (SD = 37.24), and 490 participants (66.5%) had been diagnosed with HIV more than 36 months previously. Most participants were receiving ART (n = 715, 97.0%). Participant characteristics are presented in Table 1.

Table 1.

Participant characteristics (N = 737)

Variables Category n (%) Mean (SD)
Sex Male 649 (88.1)
Female 88 (11.9)
Age (years) 38.47 (13.66)
Age group (years) 18–60 680 (92.3)
> 60 57 (7.7)
Time since HIV diagnosis (months) 55.59 (37.24)
1–3 18 (2.4)
4–12 36 (4.9)
13–36 193 (26.2)
> 36 490 (66.5)
ART status Yes 715 (97.0)
No 22 (3.0)

Continuous variables are presented as mean (SD), and categorical variables are presented as n (%)

Network structure

The estimated tri-domain symptom network consisted of 32 nodes and 174 nonzero edges, indicating substantial interconnections among sleep quality, depression, and social rhythm. Of these, 30 edges (17.2%) had absolute weights < 0.01 and were displayed as dashed lines in Fig. 1, whereas the remaining 144 edges (82.8%) were displayed as solid lines. Among the 30 dashed edges, 29 were positive and one was negative. The only negative edge was observed between PSQI2 (sleep latency) and SDS2 (mood circadian rhythm) (weight = − 0.0089). Most edges were positive, indicating that greater symptom severity or irregularity in one node was conditionally associated with greater severity or irregularity in the connected node. Negative edges were sparse and generally weaker than positive edges. Hereafter, EBICglasso-regularized partial correlations are reported as edge weights. The strongest edge was observed between BSRS3 (study/work) and BSRS4 (entertainment/socializing) (weight = 0.67), followed by BSRS1 (sleep) and BSRS2 (wake-up) (weight = 0.54), and SDS11 (confusion) and SDS12 (psychomotor retardation) (weight = 0.46). Other relatively strong associations included SDS5 (decreased appetite) and SDS6 (decreased libido) (weight = 0.35), SDS17 (unworthiness) and SDS18 (emptiness) (weight = 0.32), and PSQI1 (subjective sleep quality) and SDS4 (depressive sleep disturbance) (weight = 0.31). The five smallest nonzero edges in absolute magnitude were observed between PSQI4 (sleep efficiency) and SDS4 (depressive sleep disturbance; weight = 0.000026), SDS15 (irritability) and SDS18 (emptiness; weight = 0.001022), PSQI7 (daytime dysfunction) and SDS8 (constipation; weight = 0.001238), PSQI1 (subjective sleep quality) and BSRS5 (meals; weight = 0.001422), and BSRS3 (study/work rhythm) and SDS18 (emptiness; weight = 0.001616). All five edges were positive. The network structure is shown in Fig. 1.

Fig. 1.

Fig. 1

Network structure of sleep quality, depression, and social rhythm. The network includes components or symptoms from the Pittsburgh Sleep Quality Index (PSQI), Self-Rating Depression Scale (SDS), and Brief Social Rhythm Scale (BSRS). Green edges indicate positive associations, red edges indicate negative associations, and thicker edges indicate stronger absolute regularized partial correlations. Rings surrounding the nodes represent node predictability (R²) obtained from the complementary Gaussian model, defined as the proportion of variance in each node that was statistically predictable from the remaining nodes in that model. Dashed lines represent edges with absolute regularized partial correlations < 0.01

Central symptoms

As shown in Fig. 2A, PSQI1 (subjective sleep quality) had the highest expected influence (EI = 2.40), indicating that it was the most central node in the network. Other highly central nodes included SDS18 (emptiness; EI = 1.20), BSRS1 (sleep; EI = 1.09), SDS1 (sad mood; EI = 1.08), PSQI7 (daytime dysfunction; EI = 1.00), SDS10 (fatigue; EI = 0.98), and BSRS3 (study/work; EI = 0.83).

Fig. 2.

Fig. 2

Centrality and bridge-centrality indices of network nodes. A Z-standardized expected influence (EI) in the sleep quality–depressive symptoms–social rhythm network. B Z-standardized bridge expected influence (BEI) across the three instrument-based communities

Bridge symptoms

Bridge expected influence results are shown in Fig. 2B. Based on the z-standardized BEI values, PSQI1 (subjective sleep quality; BEI = 2.87) had the highest BEI, followed by PSQI7 (daytime dysfunction; BEI = 2.67), SDS4 (depressive sleep disturbance; BEI = 2.34), SDS10 (fatigue; BEI = 1.14), PSQI2 (sleep latency; BEI = 1.03), BSRS1 (sleep; BEI = 0.67), and BSRS5 (meals; BEI = 0.40). The BEI and nBEI rankings were strongly correlated (Spearman’s rs = 0.959). In the nBEI analysis, SDS4 (depressive sleep disturbance; nBEI = 0.045) ranked first, followed by SDS10 (fatigue; nBEI = 0.028), PSQI1 (subjective sleep quality; nBEI = 0.025), PSQI7 (daytime dysfunction; nBEI = 0.024), and PSQI2 (sleep latency; nBEI = 0.013). Thus, subjective sleep quality, daytime dysfunction, depressive sleep disturbance, and fatigue consistently comprised the four highest-ranked bridge nodes across both analyses, although their ordering differed.

Predictability

The rings surrounding the nodes in Fig. 1 represent node predictability. The mean predictability across all nodes was 0.392, indicating that, on average across nodes, 39.2% of node variance was statistically predictable from the remaining nodes. BSRS3 (study/work rhythm) had the highest predictability (R² = 0.748), followed by BSRS4 (entertainment/socializing rhythm; R² = 0.716), BSRS1 (sleep rhythm; R² = 0.704), BSRS2 (wake-up rhythm; R² = 0.693), PSQI1 (subjective sleep quality; R² = 0.606), SDS18 (emptiness; R² = 0.529), SDS17 (unworthiness; R² = 0.509), SDS12 (psychomotor retardation; R² = 0.472), SDS11 (confusion; R² = 0.464), and SDS14 (hopelessness; R² = 0.450). In contrast, SDS2 (mood circadian rhythm) showed the lowest predictability (R² = 0.136), followed by SDS8 (constipation; R² = 0.148) and PSQI6 (sleep medication; R² = 0.155).

Network accuracy and stability

Bootstrap 95% confidence intervals for edge weights are presented in Fig. 3A. The bootstrap results indicated acceptable edge-weight accuracy. Case-dropping bootstrap results for EI and BEI are shown in Fig. 3B. The CS-C values for both EI and BEI were 0.75, corresponding to the highest case-dropping proportion tested and indicating good stability. In a separate case-dropping bootstrap analysis, nBEI also showed a CS-C of 0.75, indicating similarly stable estimates. In an exploratory subgroup analysis stratified by time since HIV diagnosis, the Network Comparison Test did not detect a statistically significant difference between the ≤ 36-month and > 36-month groups in overall network structure (M = 0.198, p = 0.460) or global strength (S = 2.732, p = 0.456).

Fig. 3.

Fig. 3

Network accuracy and stability. A Nonparametric bootstrap 95% confidence intervals for edge weights. B Case-dropping bootstrap results for expected influence and bridge expected influence

Node-redundancy sensitivity analysis

Goldbricker screening identified four potentially redundant node pairs: SDS11–SDS12 (r = 0.636; proportion of significantly different correlations = 0.000), SDS14–SDS17 (r = 0.565; proportion = 0.167), BSRS3–BSRS4 (r = 0.834; proportion = 0.200), and SDS5–SDS6 (r = 0.514; proportion = 0.233). After BSRS4, SDS6, SDS11, and SDS17 were removed, the reduced network contained 28 nodes. Among the 28 common nodes, the centrality rankings of the primary and reduced networks were strongly correlated for EI (rs = 0.917), BEI (rs = 0.995), and nBEI (rs = 0.983). PSQI1 remained the highest-ranked node for EI and BEI, and the leading bridge nodes remained largely unchanged. Table 2 presents the potentially redundant node pairs and node-retention decisions.

Table 2.

Goldbricker screening results and node-retention decisions

Potentially redundant pair Zero-order correlation, r Proportion of significantly different correlations Retained node Removed node
SDS11–SDS12 0.636 0.000 SDS12 SDS11
SDS14–SDS17 0.565 0.167 SDS14 SDS17
BSRS3–BSRS4 0.834 0.200 BSRS3 BSRS4
SDS5–SDS6 0.514 0.233 SDS5 SDS6

The proportion refers to the proportion of correlations with all other nodes that differed significantly between the paired nodes

Discussion

This study examined the symptom-level network structure of sleep quality, depression, and social rhythm in PLWH. The three domains demonstrated strong interconnections rather than operating independently. Within the network, subjective sleep quality emerged as both the most central symptom and the strongest bridge symptom, positioning perceived sleep quality at the intersection of the sleep, depression, and social rhythm domains. The high bridge expected influence of daytime dysfunction, depressive sleep disturbance, and fatigue further demonstrated the interconnected network structure, highlighting the cross-domain relevance of sleep-related functional impairment, mood-related sleep problems, and subjective exhaustion.

Subjective sleep quality was strongly connected with multiple symptoms in the network. This finding may reflect the broad clinical relevance of perceived sleep quality among PLWH, as sleep may be associated with multiple factors, including antiretroviral therapy [35], stigma [36], and symptoms of anxiety and depression [5]. Subjective sleep quality reflects a cognitive and experiential appraisal rather than a purely objective sleep parameter. Among PLWH, poorer subjective sleep quality has been associated with greater depressive symptom severity, poorer quality of life, and impaired working-memory performance [37]. This interpretation is consistent with the strong edge between PSQI1 (subjective sleep quality) and SDS4 (depressive sleep disturbance). Subjective sleep quality showed the highest bridge expected influence, whereas depressive sleep disturbance ranked first after community-size normalization, suggesting complementary cross-community roles within this network: subjective sleep quality may reflect broad connections across symptom domains, while depressive sleep disturbance may more specifically capture the overlap between sleep complaints and depressive symptoms. Both symptoms may therefore warrant prioritization in clinical assessment and evaluation as candidate targets in future longitudinal and interventional studies. Cognitive behavioral therapy for insomnia, which addresses maladaptive sleep-related cognitions and behaviors, can improve insomnia and may help prevent worsening depressive symptoms [38].

Daytime dysfunction and fatigue emerged as prominent bridge symptoms, suggesting that sleep-related functional impairment and subjective exhaustion may constitute important cross-domain manifestations linking sleep quality, depression, and social rhythm. Daytime dysfunction, characterized by low energy, reduced concentration, and decreased engagement in daily activities, may co-occur with difficulties in fulfilling social roles, maintaining regular routines, disrupted social rhythms, and depressive symptoms [39]. Fatigue represents a broader subjective experience of physical and mental exhaustion and has been associated with poorer cognitive and daily functioning, as well as sleep-related difficulties among PLWH [40]. Although closely related, these symptoms capture distinct aspects of functioning: daytime dysfunction primarily reflects impaired daytime performance, whereas fatigue captures the subjective burden of reduced energy and exhaustion. From the perspective of Social Rhythm Theory, the prominent bridge positions are consistent with close cross-sectional associations among disrupted sleep, reduced activity engagement, functional impairment, and mood disturbance. Collectively, these findings suggest that both daytime functioning and fatigue should be considered in the integrated assessments of sleep quality, depression, and social rhythm in PLWH.

The social rhythm community showed a high degree of interdependence among its components. In particular, the strong associations between study/work rhythm and entertainment/socializing rhythm, as well as between sleep rhythm and wake-up rhythm, suggest that different daily routines are closely interconnected. Accordingly, instability in one domain may co-occur with disruptions in other daily routines. These findings highlight the importance of incorporating the regularity of daily routines into the assessment of sleep quality and depression, as irregular sleep-wake patterns, work/study schedules, and social activity routines may be associated with greater sleep and mood-related difficulties.

Our findings empirically delineate symptom-level associations among sleep quality, depression, and social rhythm, providing evidence consistent with Social Rhythm Theory in PLWH. The present network adds symptom-level detail to previous domain-level evidence by specifying how these constructs are interconnected through specific symptoms related to sleep quality and depression. However, because the data were cross-sectional and did not include objective circadian measures, the findings should be interpreted as being consistent with, rather than constituting a direct test of, Social Rhythm Theory.

For PLWH, whose daily lives often involve long-term medication adherence [41], regular clinic visits [42], stigma-related stress [43], and adaptation to chronic illness [44], disruptions in daily routines may be particularly relevant to the co-occurrence of sleep quality and depression. Current World Health Organization guidance emphasizes the integration of mental health care and adherence support into routine HIV services, highlighting the ongoing clinical importance of addressing the psychological and behavioral health needs of PLWH [45]. In this context, the symptom-level associations identified in the present study remain relevant to contemporary HIV care and may inform integrated assessment and future intervention research addressing sleep, mood, and daily routines.

Limitations and future directions

Several limitations should be acknowledged. First, the cross-sectional design precludes inferences about temporal ordering or causal relationships. Longitudinal studies, ecological momentary assessment, and intervention designs are needed to determine whether changes in central or bridge symptoms are associated with improvements in other symptoms across the network. Second, participants were recruited consecutively from two clinical settings in Hunan Province, and the sample was predominantly male. Although the sex distribution is broadly consistent with the epidemiological profile of HIV infection in many clinical settings in China [46], recruitment from a single province using consecutive sampling may limit the geographic representativeness of the sample and the generalizability of the identified symptom network. Future studies should recruit more geographically diverse and gender-balanced samples from multiple regions to examine the reproducibility and generalizability of the findings. Third, data collection overlapped with the COVID-19 pandemic. Although routine outpatient services and most daily activities in Changsha had largely resumed during the study period, residual public health measures may have influenced the overall levels of sleep quality, depressive symptoms, and daily rhythms among PLWH [47, 48]. However, because all participants were assessed within the same period and the present study focused on symptom-level associations rather than symptom prevalence, such contextual factors were less likely to systematically bias the observed relationships among variables. Nevertheless, the findings should be interpreted within the 2020–2021 context and validated in more recent post-pandemic samples. Fourth, all variables were assessed using self-report questionnaires, which may be subject to recall bias and social desirability effects. Future studies would benefit from incorporating objective sleep measures, such as actigraphy, alongside clinical interviews to obtain a more comprehensive and multidimensional assessment.

Conclusions

This study provides symptom-level evidence that sleep quality, depression, and social rhythm are closely interconnected in PLWH. Subjective sleep quality emerged as the most central symptom and the leading bridge symptom in the BEI analysis, while depressive sleep disturbance, daytime dysfunction, and fatigue also showed prominent bridge positions across these symptom domains. These findings are consistent with Social Rhythm Theory and suggest that the integrated assessment of sleep, mood, and daily rhythms may help identify candidate symptoms for further assessment and intervention research. Longitudinal and interventional studies are needed to further elucidate the temporal dynamics and clinical implications of these associations.

Acknowledgements

The authors thank all the volunteers who participated in this study.

Abbreviations

AIDS

Acquired immunodeficiency syndrome

ART

Antiretroviral therapy

BEI

Bridge expected influence

BSRS

Brief Social Rhythm Scale

CS-C

Correlation stability coefficient

EBIC

Extended Bayesian information criterion

EI

Expected influence

HIV

Human immunodeficiency virus

IQR

Interquartile range

LASSO

Least absolute shrinkage and selection operator

PLWH

People living with HIV

PSQI

Pittsburgh Sleep Quality Index

R²

Explained variance

SD

Standard deviation

SDS

Self-Rating Depression Scale

STROBE

Strengthening the Reporting of Observational Studies in Epidemiology

Authors' contributions

JN and XD jointly completed the data analysis and manuscript writing. JM played a key role in the collection of the survey data. HW contributed to study design and project supervision. HH and QY provided important revision suggestions. All authors read and approved the final manuscript.

Funding

This work was supported by the Anhui Provincial Department of Education - Higher Education Research Project [grant numbers 2025AHGXZK40429], the Nursing Project of the Anhui Institute of Translational Medicine and the Nursing International Joint Research Center of Anhui [grant numbers 2024zhyx-hl-B13], and the Research Foundation of Anhui Medical University [grant numbers 2023xkj264]. The funding bodies had no role in the study design; data collection, analysis, or interpretation; or in the writing of the manuscript.

Data availability

The datasets used and analysed during the current study are available from the corresponding author on reasonable request.

Declarations

Ethics approval and consent to participate

This study strictly adhered to the ethical principles outlined in the Declaration of Helsinki and was approved by the Nursing and Behavioral Medicine Research Ethics Review Committee of the Xiangya School of Nursing, Central South University (Ethics Approval No. E2020182). Additionally, the study protocol was approved by the relevant administrative departments of the participating institutions—a designated AIDS treatment hospital and the Center for Disease Control and Prevention in Changsha. All participants provided voluntary informed consent before taking part in the study.

Consent for publication

Not applicable.

Competing interests

The authors declare no competing interests.

Footnotes

Publisher’s note

Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.

Jiali Nie and Xuedi Dai contributed equally to this work.

Contributor Information

Honghong Wang, Email: honghong_wang@hotmail.com.

Jingjing Meng, Email: mengjingjing@ahmu.edu.cn.

References

  • 1.World Health Organization. HIV and AIDS. https://www.who.int/news-room/fact-sheets/detail/hiv-aids. Accessed 17 March 2026.
  • 2.World Health Organization. HIV. https://www.who.int/data/gho/data/themes/hiv-aids. Accessed 5 April 2026.
  • 3.Lazarus JV, Safreed-Harmon K, Kamarulzaman A, Anderson J, Leite RB, Behrens G, et al. Consensus statement on the role of health systems in advancing the long-term well-being of people living with HIV. Nat Commun. 2021;12(1):4450. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 4.Yang Y, Chen B, Zhang H, Huang P, Qian J, Lin L, et al. Global prevalence of depressive symptoms among people living with HIV/AIDS: a systematic review and meta-analysis of the past five years. AIDS Care. 2024;36(2):153–64. [DOI] [PubMed] [Google Scholar]
  • 5.Lee S, Oh JW, Park KM, Ahn JY, Lee S, Lee E. The prevalence and moderating factors of sleep disturbances in people living with HIV: a systematic review and meta-analysis. Sci Rep. 2024;14(1):14817. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 6.Xu Y, Li M, Su C, Zhu Q, Liu Q, Zhang Y, et al. Associated factors of cognitive frailty in people living with HIV aged 50 and older: a cross-sectional study. Infect Dis Ther. 2025;14(10):2429–44. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 7.Safreed-Harmon K, Fuster-RuizdeApodaca MJ, Pastor de la Cal M, Lazarus JV. Problems undermining the health-related quality of life of people living with HIV in Spain: a qualitative study to inform the development of a novel clinic screening tool. Health Qual Life Outcomes. 2022;20(1):84. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 8.Daubert E, French AL, Burgess HJ, Sharma A, Gustafson D, Cribbs SK, et al. Association of poor sleep with depressive and anxiety symptoms by HIV disease status: Women’s Interagency HIV Study. J Acquir Immune Defic Syndr. 2022;89(2):222–30. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 9.Fang H, Tu S, Sheng J, Shao A. Depression in sleep disturbance: A review on a bidirectional relationship, mechanisms and treatment. J Cell Mol Med. 2019;23(4):2324–32. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 10.Chen Z, He K, Chen Y, Zhang X, Ye Z, Xie C, et al. Sleep quality mediates the effect of medical social support on depression symptoms in patients with HIV/AIDS. BMC Public Health. 2024;24(1):1429. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 11.Ehlers CL, Frank E, Kupfer DJ. Social zeitgebers and biological rhythms: a unified approach to understanding the etiology of depression. Arch Gen Psychiatry. 1988;45(10):948–52. [DOI] [PubMed] [Google Scholar]
  • 12.Sabet SM, Dautovich ND, Dzierzewski JM. The rhythm is gonna get you: social rhythms, sleep, depressive, and anxiety symptoms. J Affect Disord. 2021;286:197–203. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 13.Swendeman D, Ingram BL, Rotheram-Borus MJ. Common elements in self-management of HIV and other chronic illnesses: an integrative framework. AIDS Care. 2009;21(10):1321–34. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 14.Drabarek K, Rzeszutek M. Profiles of HIV/AIDS stigma and psychological well-being in relation to embodiment among people living with HIV. Qual Life Res. 2025;34(11):3257–68. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 15.Sheehan DM, De La Cruz Y, Ramírez-Ortiz D, Swendeman D, Muñoz-Laboy M, Duncan DT, et al. A qualitative study of antiretroviral therapy adherence interruptions among young Latino men who have sex with men with HIV: Project D.A.I.L.Y. AIDS Care. 2021;33(7):962–9. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 16.Borsboom D. A network theory of mental disorders. World Psychiatry. 2017;16(1):5–13. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 17.Buysse DJ, Reynolds CF, Monk TH, Berman SR, Kupfer DJ. The Pittsburgh Sleep Quality Index: a new instrument for psychiatric practice and research. Psychiatry Res. 1989;28(2):193–213. [DOI] [PubMed] [Google Scholar]
  • 18.Yan D, Huang Y, Chen X, Wang M, Li J, Luo D. Application of the Chinese version of the Pittsburgh Sleep Quality Index in people living with HIV: preliminary reliability and validity. Front Psychiatry. 2021;12:676022. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 19.Zitser J, Allen IE, Falgàs N, Le MM, Neylan TC, Kramer JH, et al. Pittsburgh Sleep Quality Index (PSQI) responses are modulated by total sleep time and wake after sleep onset in healthy older adults. PLoS ONE. 2022;17(6):e0270095. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 20.Carpi M. The Pittsburgh Sleep Quality Index: a brief review. Occup Med (Lond). 2025;75(1):14–5. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 21.Mollayeva T, Thurairajah P, Burton K, Mollayeva S, Shapiro CM, Colantonio A. The Pittsburgh sleep quality index as a screening tool for sleep dysfunction in clinical and non-clinical samples: A systematic review and meta-analysis. Sleep Med Rev. 2016;25:52–73. [DOI] [PubMed] [Google Scholar]
  • 22.Manzar MD, BaHammam AS, Hameed UA, Spence DW, Pandi-Perumal SR, Moscovitch A, et al. Dimensionality of the Pittsburgh Sleep Quality Index: a systematic review. Health Qual Life Outcomes. 2018;16(1):89. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 23.Zung WWK. A Self-Rating Depression Scale. Arch Gen Psychiatry. 1965;12(1):63–70. [DOI] [PubMed] [Google Scholar]
  • 24.Wang ZY, Chi YF. Self-rating depression scale (SDS) [in Chinese]. Shanghai Arch Psychiatry. 1984;(2):71–2.
  • 25.Margraf J, Lavallee K, Zhang X, Schneider S. Social rhythm and mental health: a cross-cultural comparison. PLoS ONE. 2016;11(3):e0150312. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 26.Friedman J, Hastie T, Tibshirani R. Sparse inverse covariance estimation with the graphical lasso. Biostatistics. 2008;9(3):432–41. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 27.Chen J, Chen Z. Extended Bayesian information criteria for model selection with large model spaces. Biometrika. 2008;95(3):759–71. [Google Scholar]
  • 28.Epskamp S, Fried EI. A tutorial on regularized partial correlation networks. Psychol Methods. 2018;23(4):617–34. [DOI] [PubMed] [Google Scholar]
  • 29.Robinaugh DJ, Millner AJ, McNally RJ. Identifying highly influential nodes in the complicated grief network. J Abnorm Psychol. 2016;125(6):747–57. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 30.Jones PJ, Ma R, McNally RJ. Bridge Centrality: A Network Approach to Understanding Comorbidity. Multivar Behav Res. 2021;56(2):353–67. [DOI] [PubMed] [Google Scholar]
  • 31.Haslbeck JMB, Waldorp LJ. mgm: Estimating Time-Varying Mixed Graphical Models in High-Dimensional Data. J Stat Softw. 2020;93(8):1–46. [Google Scholar]
  • 32.Epskamp S, Borsboom D, Fried EI. Estimating psychological networks and their accuracy: A tutorial paper. Behav Res Methods. 2018;50(1):195–212. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 33.Levinson CA, Brosof LC, Vanzhula I, Christian C, Jones P, Rodebaugh TL, et al. Social anxiety and eating disorder comorbidity and underlying vulnerabilities: Using network analysis to conceptualize comorbidity. Int J Eat Disord. 2018;51(7):693–709. [DOI] [PubMed] [Google Scholar]
  • 34.van Borkulo CD, van Bork R, Boschloo L, Kossakowski JJ, Tio P, Schoevers RA, et al. Comparing network structures on three aspects: A permutation test. Psychol Methods. 2023;28(6):1273–85. [DOI] [PubMed] [Google Scholar]
  • 35.Chen CC, Liu HY, Chen YC, Ko NY. Relationships among trajectories of sleep disturbance, depression, and antiretroviral therapy in persons newly diagnosed with HIV: a one-and-a-half-year observational longitudinal study. Nat Sci Sleep. 2022;14:2133–42. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 36.Dong L, Bogart LM, Klein DJ, Phaladze N, Kgotlaetsile K, Goggin KJ, et al. Internalized HIV stigma, sleep problems, and depressive symptoms in people with HIV in Botswana. Behav Med. 2025;51(3):175–9. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 37.Zahr NM, Sullivan EV, Pfefferbaum A. Poor subjective sleep reported by people living with HIV is associated with impaired working memory. NeuroImmune Pharm Ther. 2023;2(2):127–37. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 38.Leerssen J, Lakbila-Kamal O, Dekkers LMS, Ikelaar SLC, Albers ACW, Blanken TF, et al. Treating insomnia with high risk of depression using therapist-guided digital cognitive, behavioral, and circadian rhythm support interventions to prevent worsening of depressive symptoms: a randomized controlled trial. Psychother Psychosom. 2022;91(3):168–79. [DOI] [PubMed] [Google Scholar]
  • 39.Voss J, Barroso J, Wang T. A critical review of symptom management nursing science on HIV-related fatigue and sleep disturbance. Int J Environ Res Public Health. 2021;18(20):10685. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 40.Campbell LM, Sun-Suslow N, Heaton A, Heaton RK, Ellis RJ, Moore DJ, et al. Fatigue is associated with worse cognitive and everyday functioning in older persons with HIV. AIDS. 2022;36(6):763–72. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 41.Okesanya OJ, Ayeni RA, Amadin P, Ngwoke I, Amisu BO, Ukoaka BM, et al. Advances in HIV treatment and vaccine development: emerging therapies and breakthrough strategies for long-term control. AIDS Res Treat. 2025;2025:6829446. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 42.Pereira B, Mazzitelli M, Milinkovic A, Casley C, Rubio J, Channa R, et al. Evaluation of a clinic dedicated to people aging with HIV at Chelsea and Westminster Hospital: results of a 10-year experience. AIDS Res Hum Retroviruses. 2022;38(3):188–97. [DOI] [PubMed] [Google Scholar]
  • 43.Nutor JJ, Gyamerah AO, Duah HO, Asakitogum DA, Thompson RGA, Alhassan RK, et al. The association of HIV-related stigma and psychosocial factors and HIV treatment outcomes among people living with HIV in the Volta region of Ghana: a mixed-methods study. PLOS Glob Public Health. 2024;4(2):e0002994. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 44.Sematlane N, Knight L, Dube TL, Delport A, Campbell L, Masquillier C, et al. Adapting to, integrating and self-managing HIV as a chronic illness: a scoping review. Discover Public Health. 2024;21(1):163. [Google Scholar]
  • 45.World Health Organization. WHO guideline on HIV service delivery: updated guidance on the integration of diabetes, hypertension and mental health services, and interventions to support adherence to antiretroviral therapy. Geneva: World Health Organization; 2025. Report No.: 978-92-4-011387-9. [PubMed] [Google Scholar]
  • 46.Zhang Y, Cai C, Wang X, Li Y, Tang H, Ma J. Disproportionate Increase of New Diagnosis of HIV/AIDS Infection by Sex and Age - China, 2007–2018. China CDC Wkly. 2020;2(5):69–74. [PMC free article] [PubMed] [Google Scholar]
  • 47.Joint prevention and control mechanism of the state council. Guidance on regular prevention and control of COVID-19 [in Chinese]. https://www.nhc.gov.cn/jkj/c100063/202005/3081f720bf314da88900bdbf51c71937.shtml. Accessed 23 July 2026.
  • 48.Hunan Provincial People’s Government. Major hospitals in Changsha resume routine outpatient medical services [in Chinese]. https://www.hunan.gov.cn/hnyw/szdt/202003/t20200317_11813934.html. Accessed 23 July 2026.

Associated Data

This section collects any data citations, data availability statements, or supplementary materials included in this article.

Data Availability Statement

The datasets used and analysed during the current study are available from the corresponding author on reasonable request.


Articles from BMC Psychology are provided here courtesy of BMC

RESOURCES