Unraveling links between lifestyle behaviors and depressive symptoms: a network analysis using NHANES 2007–2018 data

Abstract

Objective

Implementing effective interventions targeting specific depressive symptoms is essential for reducing the overall burden of depression. Our study employed network analysis to explore the complex interrelationships between multiple lifestyle behaviors and depressive symptoms, and to compare network differences across different age groups.

Methods

Using data from the National Health and Nutrition Examination Survey (NHANES), we analyzed 4,040 participants between the ages of 20 and 80 from the 2007–2018 survey cycles. Depressive symptoms were assessed using the 9-item Patient Health Questionnaire (PHQ-9), and data on five lifestyle behaviors—healthy diet, alcohol consumption, screen time, smoking, and physical activity—were collected. Network models were estimated using R version 4.4.3 to calculate strength, bridge strength, and age-related differences.

Results

The network analysis revealed that smoking (bridge expected influence; BEI = 2.59), alcohol use (BEI = 2.19), and screen time (BEI = 2.17) were the most prominent bridging nodes linking lifestyle behaviors and depressive symptoms. In addition, healthy diet (LB1) was negatively associated with appetite problems (PHQ5; edge weight = -0.02), and psychomotor agitation or retardation (PHQ8; edge weight = -0.02); screen time (LB3) was positively linked to trouble sleeping (PHQ3; edge weight = 0.04), and appetite problems (PHQ5; edge weight = 0.03); smoking (LB4) was associated with appetite problems (PHQ5; edge weight = 0.04) and depressed mood (PHQ2; edge weight = 0.03); physical activity (LB5) was negatively related to fatigue (PHQ4; edge weight = -0.05). Finally, global network strength differed significantly across age groups, with a clear age-related decline in overall connectivity: the Youth group exhibited the highest global strength (4.56), followed by the Middle-aged group (4.07) and the Older group (3.45).

Conclusions

These findings contribute to the development of more nuanced and effective public health strategies for the prevention and treatment of depression.

Supplementary Information

The online version contains supplementary material available at 10.1186/s12991-026-00656-3.

Introduction

Depression is a significant public health issue, affecting over 300 million individuals worldwide, approximately 4.4% of the global population [1, 2]. According to the World Health Organization, it is the leading cause of disability and a major contributor to the global disease burden. Depression is a complex disorder arising from the interaction of genetic and environmental factors [3]. Growing evidence indicated that depression significantly impairs quality of life and increases the risk of cardiovascular disease, reduced physical functioning, and all-cause mortality [4–6]. Although the substantial impairment associated with depression is increasingly recognized and its prevalence continues to rise, available treatment options remain limited, encompassing both pharmacological and psychological interventions [7]. It is estimated that 30%–60% of patients do not respond adequately to existing antidepressant therapies [7, 8]. Furthermore, due to limited availability and accessibility of psychological services, high treatment costs, and insufficient mental health resources, particularly in low- and middle-income countries, a substantial proportion of individuals with depression worldwide lack access to appropriate psychological treatment [9]. Therefore, identifying modifiable risk factors for depression is essential for developing effective public health interventions aimed improving mental health.

The etiology of depression is multifactorial, involving complex interactions between social, psychological, and biological factors [10–12]. Prior studies indicate that depression is often triggered by adverse life events, including elevated stress and interpersonal difficulties [13, 14]. In addition, engagement in healthy lifestyle behaviors may play a significant role in the prevention, management, and treatment of mental health disorders [15–17]. Lifestyle behaviors refers to modifiable health-related habits shaped by prolonged social interactions, such as dietary choices, physical activity, and alcohol consumption [11, 18]. A recent systematic review provides moderate-to-strong evidence supporting the crucial role of healthy lifestyle behaviors in preventing and treating depression [19]. However, most studies to date have focused on specific populations [20–22]. Therefore, comprehensive national analyses of lifestyle factors and their impact on depression are urgently needed to guide more effective prevention and treatment strategies [23].

The relationship between lifestyle and depression can be understood within the framework of lifestyle medicine (LM) [24]. LM primarily studies how modifications to lifestyle behaviors—such as diet, physical activity, sleep, stress management, and avoidance of harmful substances—can prevent, manage, and treat diseases by addressing their root causes [25]. As an emerging medical discipline, LM holds considerable potential for the prevention, management, and treatment of mental health disorders [19, 25, 26]. A growing body of research has assessed the effectiveness of LM in reducing depressive symptoms [23, 24, 27]. On the one hand, recent studies have indicated that engaging in regular physical activity and maintaining a healthy diet are strongly inversely associated with depressive symptoms [28, 29]. In contrast, alcohol consumption, cigarette smoking, and excessive screen time have been positively associated with depressive symptoms [30, 31]. However, most existing studies on lifestyle modifications for depression have focused on individual behaviors, without comprehensively examining lifestyle as an integrated construct [30, 32–34]. Accumulating evidence indicates that these behaviors often co-occur and may have synergistic effects [35–37]. Given the complex and multidimensional nature of lifestyle behaviors, analyzing them in isolation may inadequately capture their relationship with depression [38]. To address these limitations, we employed emerging network analysis techniques to explore the associations between lifestyle behaviors and depressive symptoms [39–41].

This study employs a network analysis approach to examine the relationship between lifestyle behaviors and depressive symptoms, while also analyzing their clustering structure within national-level data. Importantly, this method allows for a comprehensive exploration of how various health-related lifestyle behaviors are concurrently associated with depressive symptoms. Moreover, recent studies have examined the associations between lifestyle behaviors and depressive symptoms within specific age groups, but have largely neglected comparative analyses across age groups [42–45]. Therefore, this study also includes an age-based comparison of the network structure, aiming to reveal whether the associations between lifestyle behaviors and depressive symptoms differ across age groups. These findings may offer valuable insights for informing more effective depression prevention and treatment strategies through health promotion programs and policies, while also addressing existing research gaps in national survey data.

Methods

Study design and populations

The National Health and Nutrition Examination Survey (NHANES), conducted by the National Center for Health Statistics (NCHS) [46], is a nationally representative, cross-sectional survey designed to assess the health and nutritional status of non-institutionalized civilian population in the United States [47]. Data are collected on a continuous basis and released in biennial cycles. The survey sample comprises individuals from diverse racial, age, and socioeconomic backgrounds to ensure national representativeness [48]. In this study, data from 4,040 participants were extracted from the 2007–2018 NHANES cycles (Fig. 1). Health examinations were performed at mobile examination centers (MECs), and dietary intake was evaluated using a computerized 24-h dietary recall method [49]. The first interview was conducted in person, followed by a second one conducted via telephone. Additional data on demographics, dietary patterns, tobacco and alcohol use, physical activity, and screen time were obtained through self-reported questionnaires. All study protocols were approved by the NCHS and the Centers for Disease Control and Prevention (CDC).

Fig. 1.

Flow chart showing the steps involved in the study sample

Lifestyle behavior assessments

The structured questionnaire and 24-h dietary recalls in NHANES were used to collect all lifestyle factors [50]. Each of the healthy lifestyle behavior was scored according to specific criteria. Dietary quality in NHANES was assessed using 24-h dietary recalls and evaluated with Planetary Health Diet Index (PHDI) scores. The PHDI is based on the EAT-Lancet Commission’s recommendations and reflects adherence to a sustainable and healthy diet, with higher scores indicating healthier dietary patterns [51, 52]. Smoking status was assessed by querying usage frequency [53–55]. Smoking status was classified into four categories: non-smokers (fewer than 100 cigarettes smoked lifetime), former smokers, occasional smokers, and frequent smokers [49]. Alcohol consumption was measured by participants’ average daily intake over the past 12 months, with higher scores reflecting more frequent use [56, 57]. Physical activity was quantified using the metabolic equivalent of task (MET), with higher scores indicating greater activity levels [58–60]. Screen time was recorded as the total hours spent watching television/videos and using a computer over the past 30 days, with higher scores denoting longer screen exposure [61–63].

Depressive symptoms

Depressive symptoms were assessed using the Patient Health Questionnaire-9 (PHQ-9), a validated self-report instrument based on Diagnostic and Statistical Manual of Mental Disorders (DSM) criteria and widely used in clinical and research settings to screen for and quantify depression severity [46]. The questionnaire asked participants to report the frequency of nine symptoms experienced over the preceding two weeks, including lack of interest, depressed mood, trouble sleeping, fatigue, appetite problems, feelings of worthlessness, difficulty concentrating, psychomotor agitation or retardation, and suicidal thoughts [64]. Each item was rated on a 4-point Likert scale ranging from 0 (not at all) to 3 (nearly every day). The total score is calculated by summing all item responses, ranges from 0 to 27, with higher scores indicating greater depressive symptom severity [65, 66].

Statistical analysis

All statistical analyses complied with CDC guidelines and were conducted using R version 4.4.3. Independent variables included five lifestyle behaviors: healthy diet, physical activity, alcohol consumption, screen time, and cigarette smoking. Dependent variables comprised nine depressive symptoms from the PHQ-9. Continuous variables were reported as means with standard deviations or medians with interquartile ranges; categorical variables were reported as frequencies or percentages. Network analysis was subsequently performed using the same R version. This approach integrates multiple factors into a unified network, enabling identification of the most “central” or “bridge” nodes [67].

To quantitatively evaluate the predefined communities, we applied the Spinglass community detection algorithm via the igraph package, which is commonly used in psychological network research [68]. Detected communities were then compared with the hypothesized theoretical communities (lifestyle vs. depressive symptom nodes) to assess the validity of our predefined classification [17].

To construct and visualize the network, to calculate BEI to assess node predictability, and to evaluate robustness, we employed the qgraph [69], mgm [70], igraph [71], bootnet [72], networkcomparisontest [73], and networktools packages [74]. The analysis included all participants and was based on nodes (questionnaire items) and undirected edges (partial correlations). The network was generated using the Extended Bayesian Information Criterion (EBIC)glasso function in qgraph [69], which computed Spearman’s rank correlations and constructed a Graphical Gaussian Model (GGM).

Regularization was conducted via the graphical Least Absolute Shrinkage and Selection Operator (LASSO), yielding a sparse network by excluding weak correlations and minimizing false positives [75]. Correlation strength was assessed with EBIC using a tuning parameter (γ = 0.5) to ensure conservatism. The Fruchterman-Reingold algorithm determined the network layout based on node connectivity [69]. Blue edges represented positive relationships, and red edges represented negative relationships. Node predictability, defined as the proportion of a node’s variance explained by adjacent nodes, was assessed using mgm [70].

Two communities were predefined: healthful lifestyle behaviors and depressive symptoms. The bridge function from Networktools identified key bridge variables linking these communities [76]. Given the presence of both positive and negative edges, the BEI index was selected to quantify a node’s connectivity across communities. Higher BEI values suggest nodes may activate neighboring communities, underscoring their bridging role [76].

Network stability was evaluated using bootnet [72]. Edge weight accuracy was assessed by estimating 95% confidence intervals via nonparametric bootstrapping with 1,000 samples; narrower intervals indicate greater reliability. Stability of node bridge centrality indices and BEI was examined through correlation stability (CS) coefficients using a case-dropping bootstrap (1,000 samples). A CS coefficient above 0.5 indicates ideal stability, while values above 0.25 indicate acceptable stability [72]. Bootstrapping with 1,000 samples also tested differences in node BEIs and edge weights between node pairs (α = 0.05).

Finally, differences in network characteristics across age groups were examined using the Network Comparison Test (NCT), a permutation test for comparing two networks. As NCT supports only pairwise comparisons, three tests compared networks of young adults (20–39 years), middle-aged adults (40–59 years), and older adults (60–80 years), based on age group classifications used in previous studies [77]. The NCT was applied to age-stratified subsamples with 1,000 permutations, as recommended [78]. This procedure evaluated global network strength by comparing the absolute sum of all edge weights. Subsequently, edge weight distributions within each network were compared to characterize network structure. Differences in edge strength across networks were assessed, with Holm–Bonferroni correction applied for multiple comparisons.

Results

Descriptive statistics

Table 1 summarized the demographic characteristics, mean scores, and standard deviations of the variables analyzed in the network study.

Table 1.

Sociodemographic and study variables characteristics

Variables		Frequency	%
Gender	Female	1966	48.66
	Male	2074	51.34
Race	Mexican American	393	9.72
	Other Hispanic	357	8.84
	Non-Hispanic White	1838	45.50
	Non-Hispanic Black	808	20.00
	Other Race	644	15.94
	Abbreviations	Mean	SD
Age(years)		45.13	16.86
Independent Variables
Healthful lifestyle
Healthy diet	LB1	72.21	12.97
Alcohol use	LB2	2.43	2.14
Smoking	LB3	5.16	2.33
Screen time	LB4	1.68	0.99
Physical activity	LB5	236.00	255.93
Dependent Variables
Depressive symptoms	PHQ	2.47	3.45
Lack of interest	PHQ1	0.27	0.63
Depressed mood	PHQ2	0.24	0.57
Trouble sleeping	PHQ3	0.52	0.84
Fatigue	PHQ4	0.63	0.81
Appetite problems	PHQ5	0.28	0.63
Feelings of worthlessness	PHQ6	0.19	0.54
Difficulty concentrating	PHQ7	0.20	0.56
Psychomotor agitation or retardation	PHQ8	0.10	0.42
Suicidal thoughts	PHQ9	0.03	0.19

Network structures of lifestyle behaviors and depressive symptoms

The Spinglass algorithm identified two clusters corresponding to lifestyle and depressive symptoms, consistent with our predefined categories and supporting the validity of the hypothesized communities. Figure 2 presented the regularized partial correlation network illustrating the associations between lifestyle behaviors and depressive symptoms, with node labels and descriptions shown in the figure. The network comprised 14 nodes and 52 non-zero edges out of 91 possible between-community edges (57.14%). The mean predictability was 0.232, indicating that 23.2% of the variance in each node is explained by variations in other nodes. Notably, several strong between-community edges were observed, including healthy diet (LB1) associated with appetite problems (PHQ5; edge weight = -0.02) and psychomotor agitation or retardation (PHQ8; edge weight = -0.02); screen time (LB3) related to trouble sleeping (PHQ3; edge weight = 0.04) and appetite problems (PHQ5; edge weight = 0.03); smoking (LB4) linked to depressed mood (PHQ2; edge weight = 0.03) and appetite problems (PHQ5; edge weight = 0.04); and physical activity (LB5) connected to fatigue (PHQ4; edge weight = -0.05).

Fig. 2.

Network structure for lifestyle and depressive symptoms. Note: Edges between nodes signify associations. Blue edges indicate positive associations, while red edges denote negative associations. Predictability was depicted as a filled part of a circle surrounding each node

Bridge expected influence analysis

Figure 3 identified three nodes with the highest bridge strength: smoking (2.59), alcohol use (2.19), and screen time (2.17). These findings suggested that inter-community connections within the network were predominantly driven by lifestyle behavior-related factors (Fig. S1). Notably, these three nodes demonstrated significantly greater bridge strength than approximately 77% of the other nodes, underscoring their pivotal roles in linking distinct communities. The remaining nodes showed moderate to low bridge strength, indicating a less central role in inter-community connectivity.

Fig. 3.

Bridge expected influence values for healthful lifestyle and depressive symptom. Note: LB1, Healthy diet; LB2, Alcohol; LB3, Screen time; LB4, Smoking; LB5, Physical activity; PHQ1, Lack of interest; PHQ2, Depressed mood; PHQ3, Trouble sleeping; PHQ4, Fatigue; PHQ5, Appetite problems; PHQ6, Feelings of worthlessness; PHQ7, Psychomotor agitation or retardation; PHQ8, Psychomotor disturbances; PHQ9, Suicidal thoughts

Network stability and accuracy

The results of the edge weight bootstrap procedure are shown in Fig. S2, and the 95% confidence intervals for all edge weights indicate good accuracy. The results of the bootstrapped difference tests are presented in Fig. S3, confirming the network structure through the edge-weight difference test. Bridge expected influence (BEI) demonstrated moderate stability (CS = 0.67), indicating high network stability (Figs. S4 and S5). These results support the reliability of the network structure and centrality measures for subsequent analyses.

Age groups-based network comparison

We conducted pairwise comparisons of network models across three age groups (Fig. 4). As shown in Fig. S6, no significant differences were found in the overall network structure between any group pairs (Youth vs. Middle-aged: M = 0.149, p = 0.166; Youth vs. Older: M = 0.139, p = 0.459; Middle-aged vs. Older: M = 0.138, p = 0.653). However, significant differences emerged in global strength. Specifically, overall network connectivity progressively declined from the Youth group (global strength = 4.56) to the Middle-aged group (4.07) and further to the Older group (3.45). A significant difference was observed between the young and older adult groups (S = 1.109, p = 0.005), and a marginally significant difference was noted between the young and middle-aged groups (S = 0.490, p = 0.056). However, no significant difference was found between the middle-aged and older adult groups (S = 0.619, p = 0.135). This pattern indicated a gradual decline in overall network connectivity with advancing age. These findings suggested that although the overall network structure remains stable across age groups, the strength of associations among variables diminishes with age. In contrast, no significant differences in bridge centrality were observed across the three groups, indicating that key bridging symptoms connecting distinct communities remain relatively consistent regardless of age (Fig. 4d). To further investigate group differences (Fig. 4a–c), we conducted edge-wise comparisons and identified several edges exhibiting significant variations in both p-values and test statistics. From these, we selected a subset of representative edges to illustrate group-level distinctions. Between the Youth and Middle-aged groups, the Alcohol–Lack of interest edges (t = 0.017, p = 0.006) demonstrated significantly stronger connectivity in the Youth group. In comparisons between the Youth and Older groups, the Smoking–Lack of interest edges (t = 0.030, p = 0.007) showed significantly stronger associations in the Older group.

Fig. 4.

Bridge expected influence models for participants of different age groups, (a) Young adult group network; (b) Middle-aged group network; (c) Older adult group network; (d) Age-group comparisons of bridge expected influence centrality.Note: LB1, Healthy diet; LB2, Alcohol; LB3, Screen time; LB4, Smoking; LB5, Physical activity; PHQ1, Lack of interest; PHQ2, Depressed mood; PHQ3, Trouble sleeping; PHQ4, Fatigue; PHQ5, Appetite problems; PHQ6, Feelings of worthlessness; PHQ7, Psychomotor agitation or retardation; PHQ8, Psychomotor disturbances; PHQ9, Suicidal thoughts.

Discussion

The present study constitutes an initial effort to investigate the complex relationships between lifestyle behaviors and depressive symptoms using a symptom-level network analysis. Through a novel network psychometric approach, we uncovered intricate associations between various lifestyle behaviors and specific depressive symptoms. Notably, the results identified smoking, alcohol consumption, and screen time as key bridging factors connecting lifestyle behaviors and depressive symptoms within the network. Furthermore, our findings indicated age-related differences in the associations between lifestyle behaviors and depressive symptoms. These findings revealed key links between lifestyle and depression, suggesting potential targets for early prevention and effective treatment strategies.

Our findings revealed several specific associations between lifestyle behaviors and depressive symptoms. First, a healthy diet was negatively associated with both appetite disturbances and psychomotor agitation or retardation. Based on previous studies, diets particularly rich in fruits and vegetables may have anti-inflammatory effects, which could partly explain these associations [79]. On the other hand, depressive symptoms, such as appetite disturbances and psychomotor agitation or retardation, may contribute to poorer dietary behaviors. Future studies should employing more rigorous designs, particularly longitudinal studies, are warranted to clarify the causal direction of these associations. Furthermore, our findings indicate an inverse association between physical activity and fatigue. This relationship is complex and potentially bidirectional: insufficient physical activity may contribute to fatigue, whereas individuals experiencing fatigue may engage in reduced daily activity [80]. Such a self-reinforcing cycle could exacerbate fatigue. One possible explanation is that physical activity enhances dopamine secretion, potentially reducing perceived fatigue or fatigability [81]. However, in this study, physical activity was measured using total metabolic equivalent of task (MET) values without distinguishing between activity domains (e.g., occupational versus recreational), which may introduce residual confounding and limit understanding of the specific factors underlying this association. As fatigue is a non-specific depressive symptom, the link between physical activity and fatigue may not truly represent the core psychopathology underlying the depression-activity relationship.

In contrast, screen time showed positive associations with both sleep disturbances and appetite disturbances. This pattern aligns with prior research suggesting that excessive screen exposure may increase the risk of depressive symptoms [31, 82]. However, total screen time was assessed without differentiating between occupational and leisure-related use, which may have distinct associations with sleep and appetite outcomes. Likewise, smoking was also positively associated with both depressed mood and appetite dysregulation. One potential mechanism is that smoking may disrupt sex hormone balance, thereby increasing susceptibility to depressive symptoms [83]. Conversely, depressive symptoms, such as anhedonia, may contribute to smoking behavior, indicating potential bidirectional associations that could form a self-reinforcing cycle [84]. In contrast, no significant association was observed between alcohol consumption and specific depressive symptoms. Previous studies have suggested that light to moderate alcohol intake may reduce the risk of depression [85]. Therefore, our findings may reflect the complex and dynamic relationship between alcohol use and depression.

Notably, the bridge expected influence identified smoking as the most central factor linking lifestyle behaviors to depressive symptoms, followed by excessive alcohol consumption and screen time. Consistent with these findings, previous studies have shown that smoking and screen time are consistently associated with an increased risk of depression [86, 87]. Therefore, interventions targeting the aforementioned factors may be more effective for preventing and alleviating depression than those focused on other lifestyle factors. Interestingly, although no strong symptom-level associations were observed between alcohol consumption and specific depressive symptoms, alcohol consumption nevertheless demonstrated relatively high bridge expected influence within the network. This elevated bridge centrality may reflect its associations with multiple other lifestyle behaviors, including a positive association with smoking and inverse associations with healthy dietary behaviors and physical activity. Rather than implying direct symptom-level associations or causal relationships, this pattern suggests that alcohol consumption plays a structural bridging role between the lifestyle behavior community and the depressive symptom community within the network. Therefore, these findings enhance our understanding of the network-level relationships between lifestyle behaviors and depressive symptoms and underscore the potential value of network-informed approaches for hypothesis generation and exploratory intervention design.

It should be noted that, although the overall network structure and node centrality did not differ significantly across age groups, our findings revealed a clear age-related decline in global network strength, with younger adults exhibiting stronger overall connectivity. These findings underscore the importance of tailoring interventions to age-specific characteristics: younger adults may benefit from comprehensive, multidimensional lifestyle strategies, whereas older adults may require more targeted and individualized support. Furthermore, edge-level comparisons revealed age-related variability in the associations between specific lifestyle behaviors and depressive symptoms. First, the Alcohol–Lack of Interest edge demonstrated significantly stronger connectivity in the younger group compared to the older group, potentially reflecting the greater influence of interpersonal dynamics and social pressures during youth, which may lead to increased alcohol consumption [88]. In contrast, the Smoking–Lack of Interest edge was significantly stronger in the older group. This finding is consistent with previous research suggesting that smoking is more strongly associated with anhedonia among older adults [89]. One possible explanation is that anhedonia may emerge and worsen as a consequence of long-term smoking [90]. Taken together, these findings underscore the importance of adopting age-sensitive approaches in both research and interventions.

Limitations and future research

Several limitations of the current study should be acknowledged. The first limitation of this study is its cross-sectional design, which precludes causal inferences. Longitudinal studies are necessary to clarify potential causal relationships between lifestyle behaviors and depression. Second, the sample was limited to adults from the United States, all aged 20 years or older. Future research should replicate this study with more diverse and broader populations to validate the findings. The single-country context restricts the generalizability of the results. Therefore, future studies should include participants from multiple countries and examine the interrelationships across different national contexts. Third, our study did not account for all relevant lifestyle behaviors. Previous research has shown that other factors, such as social activity and sleep, also affect depression [91–93]. Future research should incorporate these variables to improve the validity of the findings. Fourth, while smoking showed the strong association with depression, our categorization of smoking status may not fully capture cumulative exposure, particularly within the heterogeneous ‘former smoker’ group. Future studies should incorporate more granular assessments, such as nicotine delivery methods (e.g., e-cigarettes), duration, and intensity of past use, to better understand the impact of cumulative smoking history on mental health. Finally, although the PHQ-9 is a widely used DSM-based tool, certain items (e.g., fatigue, appetite changes, psychomotor symptoms) lack etiological specificity and reflect general somatic or psychological [94]. Thus, in this non-clinical sample, the observed symptom-level associations may represent broader health or psychological burdens rather than mechanisms unique to major depressive disorder. Future studies should replicate these findings in clinical populations using supplementary measures to better isolate depression-specific symptoms.

Conclusions

The present study, using data from a nationally representative cohort, examined associations between lifestyle behaviors and depressive symptoms through network analysis. Smoking, alcohol use, and screen time emerged as central nodes linking lifestyle and depressive symptom variables. Notably, the strength of associations between lifestyle behaviors and depressive symptoms appeared to decrease with age. Overall, by revealing the complex pattern of statistical associations between lifestyle factors and depressive symptomatology, this study provides insights that may inform future research and the exploratory development of strategies for addressing depression, particularly among adults.

Abbreviations

PHQ-9

9-Item Patient Health Questionnaire

LM

Lifestyle medicine

NHANES

National Health and Nutrition Examination Survey

NCHS

National Center for Health Statistics

MECs

Mobile examination centers

CDC

Centers for Disease Control and Prevention

PHDI

Planetary Health Diet Index

MET

Metabolic equivalent of task

DSM

Diagnostic and Statistical Manual of Mental Disorders

BEI

Bridge expected influence

GGM

Graphical Gaussian Model

LASSO

Least Absolute Shrinkage and Selection Operator

EBIC

Extended Bayesian Information Criterion

CS

Correlation stability

NCT

The Network Comparison Test

Author contributions

Conceptualisation/design: L.J. and Y.F. Data analysis/interpretation: L.L., Z.Y., and C.D. Supervision: Y.R. and L.J. Drafting article: Y.F. and L.L. Critical revision of the article: Y.R. and L.J. All authors have read and agreed to the published version of the manuscript.

Funding

This study was supported by the Joint Key Project of Yunnan Federation of Social Sciences and Yunnan University of Traditional Chinese Medicine (LHZX202403), the Key Project of the National Traditional Chinese Medicine Examination Research Fund (TA2024001), the Yunnan Provincial Department of Education Research Fund (2024J0412), the 2025 Open Research Project of the Yunnan Provincial Key Laboratory of Dai and Yi Medicine (2025ZD2501), Key Research and Development Program of Yunnan Provincial Science and Technology Department, Social Development Special Plan(202403AC100017), and the Inheritance, Innovation, and Key Technology Research of Yi Medicine (202402AA310035)

Data availability

The datasets generated and analyzed for the current study are available in the NHANES repository. These data can be accessed using the following link: https://wwwn.cdc.gov/nchs/nhanes/Default.aspx.

Declarations

Ethics approval and consent to participate

The study analyzed data downloaded from the National Health and Nutrition Examination Survey public database. The National Center for Health Statistics Ethics Review Committee granted ethics approval. The methods involved in this study were conducted in accordance with relevant guidelines and regulations (Declaration of Helsinki). All individuals provided written informed consent before participating in the study. Details are available at https://www.cdc.gov/nchs/nhanes/irba98.htm. The current study was deemed exempt from further review because the data used are deidentified and publicly accessible.

Consent for publication

Not applicable.

Competing interests

The authors declare no competing interests.

Clinical trial number

Not applicable.