Suicide remains a major global public health challenge, accounting for approximately 727,000 deaths worldwide in 2021[1]. In South Korea, the burden is particularly pronounced, with one of the highest suicide rates among member countries of the Organization for Economic Cooperation and Development (OECD) [2,3]. Contemporary suicide research increasingly conceptualizes suicidal behavior as arising from multiple interacting individual and health-related conditions rather than from a single isolated determinant [4,5]. Accordingly, multidimensional frameworks have been emphasized in population-based epidemiological studies.

Sleep disorders have recently been considered health- related correlates of suicidal behaviors. Among these, obstructive sleep apnea (OSA) is of particular public health relevance because of its high prevalence and chronic course. OSA involves recurrent upper airway obstruction during sleep, producing intermittent hypoxia and sleep fragmentation that affect daytime functioning and car-diometabolic health [6]. Global estimates suggest that a substantial proportion of adults experience OSA, although prevalence varies depending on diagnostic criteria and population characteristics [7]. In Korea, reported prevalence ranges widely due to differences in screening approaches and study designs [8]. Despite this burden, OSA remains frequently undetected in community settings, limiting its consideration in population mental health research.

Accumulating epidemiologic evidence suggests an association between OSA and suicidal ideation and related behaviors. Using Danish nationwide registry data, Udholm et al.[9] reported elevated risks of suicide and deliberate self-harm among individuals with clinically diagnosed OSA. In a Korean population-based study using the 2019∼ 2020 KNHANES, Han et al. [10] found that adults with high OSA risk (STOP-Bang score ≥3) had a higher prevalence of suicidal ideation than those at low risk, with depression and short sleep duration identified as co-occurring factors.

However, prior studies have largely treated OSA as a single exposure and have not examined how biopsychosocial factors co-occur with suicidal ideation within different levels of OSA risk. This distinction is important because individuals at elevated OSA risk often present with greater burdens of comorbid physical, psychological, and behavioral conditions, suggesting that their correlate profiles may differ from those at lower risk. Furthermore, because OSA is frequently undetected in community settings, a screening-based stratified approach may better reflect the population-level distribution of suicide-related vulnerability. Such an approach may capture heterogeneity in correlate patterns that cannot be identified in a single pooled model.

Prior studies have described biological and psychosocial mechanisms associated with OSA and suicidal behaviors. Recurrent hypoxia and sleep disruption have been associated with inflammatory and stress-response pathways [11-13], and untreated OSA has been linked to psychiatric symptoms, emotional dysregulation, and social functioning difficulties [6,14,15]. Broader lifestyle and behavioral factors are also recognized contributors to suicide risk [4,16,17], suggesting that sleep-related health conditions may coexist with psychological and social vulnerabilities within broader health profiles. These findings support the consideration of OSA risk as a health context in which multiple vulnerability factors may cluster.

The biopsychosocial model provides a framework for examining such heterogeneity by conceptualizing suicide risk as the product of interacting biological, psychological, and social processes [18]. From this perspective, examining how associated factors are distributed across health contexts may provide additional understanding beyond single risk-factor approaches. Conceptualizing OSA risk as a sleep-related health context may therefore help characterize how biopsychosocial correlates of suicidal ideation are distributed across the population.

Therefore, the present study describes biopsychosocial factors associated with suicidal ideation among Korean adults across OSA risk groups using data from the 2022∼ 2023 Korea National Health and Nutrition Examination Survey (KNHANES). This study aims to characterize biopsychosocial correlates of suicidal ideation across OSA risk groups and to inform the contextual understanding of suicide risk at the population level.

This study described biopsychosocial correlate profiles of suicidal ideation within OSA risk groups using nationally representative Korean data. Depression and anxiety were consistently associated with suicidal ideation across both groups, while smoking and social isolation were observed within the high-risk group, and short sleep duration and mental health consultation within the low-risk group.

Suicidal ideation prevalence did not differ significantly between the high-risk (4.0%) and low-risk (3.1%) OSA groups, consistent with Han et al.[10], who reported a similar null difference across STOP-Bang risk strata in a Korean population-based sample. Instead, the correlate profiles observed within each group are described below without formal between-group comparison.

Depression and anxiety were consistently associated with suicidal ideation in both groups. The absence of depressive symptoms carried the largest protective effect in both strata, reinforcing the well-established centrality of depression in suicidal behavior [4,5,21]. Elevated anxiety scores were similarly associated across groups, consistent with prior studies reporting associations between anxiety symptoms and suicidal ideation [22]. Mental health consultation was associated with suicidal ideation in the low-risk group but not in the high-risk group, suggesting that individuals at lower OSA risk who have sought professional mental health care may represent a subgroup warranting closer attention in suicide surveillance. This consistency across OSA risk levels suggests that psychological distress screening retains its relevance regardless of sleep-related health status, a point with direct implications for population-level assessment.

Behavioral and social correlates were observed within the high-risk group, with patterns described at the stratum level. Current smoking and the absence of shared meals were each associated with suicidal ideation in this stratum but not in the low-risk group. The smoking finding aligns with prior meta-analytic evidence linking current smoking to elevated odds of suicidal ideation in the general population [23,24], with proposed mechanisms involving nicotine-related serotonergic dysregulation [25]. The meal-sharing finding is consistent with Timkova et al.[15], who reported in a clinical OSA sample that lower social support was independently associated with suicidal ideation after adjusting for depression and insomnia severity. That both factors clustered in the high-risk group, which also carried higher rates of obesity, chronic disease, and hazardous drinking, suggests they may reflect a clustering of co-occurring vulnerabilities within this group rather than independent risk indicators.

In the low-risk group, short sleep duration was associated with suicidal ideation, whereas this association was observed in the low-risk group; no formal comparison across groups was conducted. This finding is consistent with longitudinal evidence linking short sleep to suicidal ideation and behavior [26,27]. The lack of an association in the high-risk group may reflect differences in how sleep-related characteristics are expressed across OSA risk groups. Because OSA itself produces fragmented and non-restorative sleep, additional variation in sleep duration may carry less discriminative value within an already sleep-compromised population.

Biological and metabolic indicators, including inflammatory markers and glycemic status, were not significantly associated with suicidal ideation after adjustment. Prior studies implicating inflammatory pathways in suicidal behavior have largely been conducted in clinical psychiatric samples [11-13], where inflammatory dysregulation may be more pronounced than in a general community population. The non-significance of biological indicators is interpreted as negative evidence suggesting that psychosocial factors account for a greater share of variance in suicidal ideation at the population level. Therefore, psychosocial interventions may warrant prioritization in community-based suicide prevention over biological screening

Taken together, these findings support treating OSA risk as a contextual health variable rather than a causal exposure. Prior studies have asked whether OSA increases suicidal ideation risk [9,10]. The present study asks whether the correlate profile of suicidal ideation varies by sleep-related health context. The observation that smoking and social isolation were observed within the high-risk group, and that short sleep duration was observed within the low-risk group, illustrates that correlate compositions may vary across health contexts. This descriptive heterogeneity may have implications for how population-level surveillance tools are designed, though the present data cannot establish the mechanisms underlying these patterns.

From a public health standpoint, these findings suggest that incorporating sleep health indicators into population-level suicide risk assessment may improve the characterization of at-risk groups. Social connectedness and smoking status may represent relevant screening targets within individuals at elevated OSA risk, and sleep duration warrants attention within those at lower risk. These observations are exploratory and require confirmation through longitudinal studies with adequate statistical power before informing clinical practice.

This study has several limitations that bear on the interpretation of findings. The cross-sectional design precludes any temporal inference, as all associations reflect concurrent states rather than predictive relationships. OSA risk was classified by the STOP-Bang questionnaire rather than polysomnography, which is not feasible in large-scale population surveys. Screening-based classification introduces misclassification that would likely attenuate observed group differences, meaning actual correlate heterogeneity may be somewhat larger than reported. Suicidal ideation was measured with a single past-year item that does not capture severity, chronicity, or current ideation status, limiting construct validity and potentially underestimating true prevalence. In addition, the 12-month recall period may introduce recall bias. Because KNHANES variables span heterogeneous reference periods, findings should be read as concurrent state descriptions rather than risk predictions. Information on psychiatric and sleep treatment was unavailable, which may have confounded associations between psychological correlates and suicidal ideation.

Statistical precision is also a concern. With 65 and 70 suicidal ideation events in the high-risk and low-risk groups against 24 model parameters, the events-per-variable (EPV) ratio fell well below recommended thresholds. Individual adjusted odds ratios are therefore subject to sparse-data instability, particularly for variables with low prevalence, and should be read as descriptive indicators of within-stratum correlate patterns rather than precise effect estimates.

Depression and anxiety were consistently associated with suicidal ideation across both OSA risk groups, while smoking and social isolation were observed within the high-risk group, and short sleep duration and mental health consultation were observed within the low-risk group. These findings suggest that the biopsychosocial correlate profile of suicidal ideation may vary across sleep-related health contexts, and that sleep health status may be a meaningful variable when characterizing suicide risk in population-based research. Given the descriptive study design and limited suicidal ideation events within each group, these findings should be interpreted as hypothesis-generating rather than confirmatory, and longitudinal studies with adequate power are needed to determine whether correlate patterns differ across OSA risk groups and to clarify their temporal relationships.