INTRODUCTION
Depressive disorders are a major public mental health issue worldwide [1-3]. Many subtypes of depressive disorders have been defined because of heterogeneous symptoms [4], and age might be an influential factor in the clinical presentation and classification of these disorders. However, inconsistencies in agerelated features have been reported in several previous studies [5]. Notably, the severity of depression tends to be lower in older individuals than in younger individuals. In terms of specific symptoms, younger patients are characterized by severe suicidal ideation, guilt, and irritability, whereas older patients predominantly exhibit somatic symptoms such as illness anxiety and insomnia, particularly early morning awakening [6-9]. The pervasive influence of ageism necessitates a critical evaluation of its impact on the perception and treatment of depression and aging. Ageism, which affects both older and younger populations, denotes stereotypes, prejudices, and discrimination based on age. In terms of depressive disorders, older adults are often perceived as frail or dependent, which contributes to their marginalization in mental health care and research. Conversely, younger individuals with depressive disorders are frequently regarded as impulsive or irresponsible. Such biases can exacerbate care inequities and reinforce internalized societal stereotypes, manifesting as self-perceived ageism that further influences the mental health outcomes of patients [10-12].
A recent neuroimaging study on age-related changes in brain networks supports the complexity of depression manifestations across the lifespan. In older adults, structural connectivity exhibits compensatory reorganization in response to cognitive function decline, taking advantage of the redundant mechanisms in large-scale brain networks. It has been suggested that these networks can provide alternative pathways for information processing. In addition, they could be protected by factors that mitigate age-related changes, particularly in areas associated with cognitive control and default mode activity [13]. These compensatory processes align with the age-associated clinical features of depressive disorders. In addition, anxiety symptoms are closely associated with depression symptoms and might influence the network structure of depression symptoms in patients with depressive disorders [14]. Network analysis can reveal symptom patterns unrecognized by conventional reductionistic approaches because it is a bottom-up approach. Thus, this study uses data from the Research on Asian Psychotropic Prescription Patterns for Antidepressants, Phase 3 (REAP-AD3) study to compare the network structures of depression and anxiety symptoms among older (>65 years), middle-aged (35-64 years), and young adult (18-34 years) Asian patients with depressive disorders [15,16]. Specifically, this study characterizes the network structure of older Asian patients with depressive disorders in the context of ageism.
METHODS
Study participants
The REAP-AD3 study, conducted from August 2023 to February 2024 across 11 Asian countries (China, India, Indonesia, Iran, Japan, Korea, Malaysia, Pakistan, Singapore, Taiwan, and Thailand), was a large-scale research project that enrolled 4,587 psychiatric patients receiving antidepressant treatment. In the REAP-AD3 study, participants were selected based on their antidepressant use. Ethical reviews of the study protocol were conducted, and informed consent forms were collected and approved by the institutional review boards of all study centers, including Hanyang University Guri Hospital in Guri, Republic of Korea (receipt number: 2023-05-021-006). All participants provided written informed consent for their data to be used for research purposes. Coordinated by the REAP consortium, data collection was securely managed via a web-based platform hosted by Taipei City Hospital.
The inclusion criteria for this study were as follows: 1) a diagnosis of depressive disorder, such as depressive episode (F32), recurrent depressive disorder (F33), or persistent depressive disorder (F34), according to the International Classification of Diseases, tenth edition [17], as determined by clinical psychiatrists; 2) the use of at least one antidepressant classified under the Anatomical Therapeutic Chemical classification system [18]; and 3) age of at least 18 years. The exclusion criteria were as follows: 1) severe medical or neurological conditions, 2) inability to provide informed consent, and 3) age less than 10 years. Using those criteria, a sample of 2,785 Asian patients with depressive disorders was gathered from the REAP-AD3 study population: 264 older, 1,277 middle-aged, and 1,244 young adults.
Variables and data collection
REAP-AD3 data were collected through personal interviews and self-administered questionnaires by clinical coordinators at the research centers. The variables considered included age, sex, country, treatment setting (outpatient or inpatient), psychiatric diagnosis, and the use of psychotropic medications (antidepressants, antipsychotics, mood stabilizers, anxiolytics, or hypnotics). Depression and anxiety symptoms were evaluated using the Patient Health Questionnaire-9 (PHQ-9) [19] and Generalized Anxiety Disorder-7 (GAD-7) [20], respectively. Higher scores on the PHQ-9 and GAD-7 indicate greater severity of depression and anxiety symptoms. Both the PHQ-9 and GAD-7 have been formally translated into various Asian languages, and their reliability and validity are considered to be favorable [21-25].
The abbreviations for the PHQ-9 items are as follows: APE, poor appetite or overeating; CNC, trouble concentrating on things, such as reading the newspaper or watching television; DEP, feeling down, depressed, or hopeless; ENG, feeling tired or having little energy; HUR, thoughts that you would be better off dead or of hurting yourself in some way; INS, trouble falling or staying asleep, or sleeping too much; PLE, little interest or pleasure in doing things; SEL, feeling bad about yourself, or feeling that you are a failure or have let yourself or your family down; SLW, moving or speaking so slowly that others could have noticed, or the opposite—being so fidgety or restless that you have been moving around much more than usual. The abbreviations for the GAD-7 items are as follows: ANX, feeling nervous, anxious, or on edge; AWF, feeling afraid, as if something awful might happen; DFF, worrying too much about different things; IRR, becoming easily annoyed or irritable; RLX, trouble relaxing; RTL, being so restless that it is hard to sit still; and WOR, not being able to stop or control worrying.
Statistical analysis
To examine the network structure of depression and anxiety symptoms in older Asian patients with depressive disorders, the qgraph R package was used [26]. Network nodes (depression and anxiety symptoms) and edges (associations between symptoms) were constructed based on the PHQ-9 and GAD-7 item responses. The least absolute shrinkage and selection operator minimized false-positive edges, with very small edges set to zero [27]. The edges represent partial correlation coefficients, which reflect some symptom relationships while controlling for others. The extended Bayesian information criterion optimized the shrinkage parameter [28], and the Fruchterman-Reingold algorithm visualized the network structures by clustering related nodes [29]. The PHQ-9 and GAD-7 items were treated as ordered categorical variables, ranging from 0 to 4, with network estimation based on polychoric correlations. Community structures were assessed using the spin-glass algorithm from the igraph R package, which detects clusters based on edge strength and density [30]. Centrality metrics were calculated focusing on node strength (the sum of all edge weights), closeness (symptom integration), and betweenness (nodes acting as bridges). Stability was evaluated using bootstrapping, with correlation stability (CS) coefficients>0.250 (preferably >0.500), indicating robust results [31].
To compare the network structures of depression and anxiety symptoms across different age groups, this study used a network comparison test with 100 iterations via the Network-ComparisonTest package [32]. Structural and global strength invariance were tested, and Spearman’s correlation coefficients were used to assess differences in node centrality and edge weights, with p<0.05 considered significant (two-tailed). All statistical analyses were conducted using R 4.3.3 (R Foundation for Statistical Computing).
RESULTS
Sample description
As shown in Table 1, 9.5% (n=264) of the 2,785 study participants were classified as the older group. The average age of the older group was 68.4 years (standard deviation=16.4). The older group was characterized by a high proportion of outpatients (79.8%) and women (70.4%). Most of the older group was East Asian (43.5%) or Southeast Asian (31.0%). In addition, the group was prescribed with adjunctive use of antipsychotic (36.3%) and anxiolytic (47.0%) medications. As shown in Tables 2 and 3, the response frequency distributions indicate that the older group presented relatively lower scores on the individual PHQ-9 and GAD-7 items than the middle-aged and young adult groups.
Network structure of depression and anxiety symptoms in older patients with depressive disorders
As shown in Figure 1A, 69 (57.5%) of the 120 possible edges were identified within the estimated network structure of depression and anxiety symptoms in the older group. All identified edges were estimated to be greater than 0. In terms of edge statistics, PLE-DEP (weight=0.487) had the most significant interconnection, followed by WOR-RLX (0.442), RLXRTL (0.356), CNC-SLW (0.276), APE-SEL (0.266), RTL-IRR (0.261), and INS-ENG (0.251). Those edges were assessed as favorable, with a CS coefficient of 0.441. In terms of node statistics, as illustrated in Supplementary Figure 1A, DEP was the most central node within the network, followed by RLX, WOR, ANX, RTL, and SEL, with INS being the least central. Node strength centrality demonstrated a meaningful level, with a CS coefficient of 0.281. The community detection method indicated that depression and anxiety symptoms were organized into two meaningful clusters. Cluster A consisted of PLE-DEP-INS-ENG-APE-SEL-CNE-SLW (i.e., depression symptom interconnection), and cluster B comprised HURANX-WOR-DFF-RLX-RTL-IRR-AWF (i.e., anxiety symptom interconnection). HUR was the only PHQ-9 item in cluster B.
Network structures of depression and anxiety symptoms in middle-aged and young adult patients with depressive disorders
As shown in Figure 1B, 81 (67.5%) of the 120 potential edges were observed within the estimated network structure of depression and anxiety symptoms in the middle-aged group. Most of those edges were estimated to be greater than 0, with the exceptions of SEL-WOR (-0.010), INS-AWF (-0.048), ENGHUR (-0.055), and HUR-ANX (-0.061). In terms of edge statistics, PLE-DEP (weight=0.367) represented the most significant interconnection, followed by ANX-WOR (0.272), SLWHUR (0.257), and DEP-HUR (0.250). The edge was estimated to be at the highest level, with a CS coefficient of 0.750. Regarding node statistics, as illustrated in Supplementary Figure 1B, DEP was the node most centrally located within the network, followed by ANX, WOR, RLX, HUR, and RTL, and APE was the least central. Node strength centrality exhibited a favorable level, with a CS coefficient of 0.517. The community detection method indicated that the depression and anxiety symptoms were organized into two meaningful clusters. Cluster A contained PLE-DEP-INS-ENG-APE-SEL-CNESLW-HUR (representing the interconnection of depression symptoms), and cluster B encompassed ANX-WOR-DFFRLX-RTL-IRR-AWF (reflecting the interconnection of anxiety symptoms).
As shown in Figure 1C, 79 (65.8%) of the 120 possible edges were identified within the estimated network structure of depression and anxiety symptoms in the young adult group. Most of the identified edges were estimated to be greater than 0, with the exceptions of PEL-RTL (-0.010), ENG-SLW (-0.015), and DEP-RTL (-0.062). Regarding edge statistics, PLE-DEP (weight=0.423) represented the most significant interconnection, followed by SEL-HUR (0.329), WOR-DFF (0.322), ANX-WOR (0.283), and INS-APE (0.258). The edge with the highest estimate (PLE-DEP) was characterized by a CS coefficient of 0.750. In terms of node statistics, as illustrated in Supplementary Figure 1C, DEP was the node most centrally located within the network, followed by WOR, RTL, DFF, RLX, and ANX, with AWF being the least centrally located. Node strength centrality exhibited a favorable level, with a CS coefficient of 0.516. The community detection method indicated that the depression and anxiety symptoms were organized into two meaningful clusters. Cluster A contained PLEDEP-INS-ENG-APE-SEL-CNE-SLW-HUR (representing interconnections among depression symptoms), and cluster B consisted of ANX-WOR-DFF-RLX-RTL-IRR-AWF (representing interconnections among anxiety symptoms).
Comparing the network structures of older, middle-aged, and young adult patients with depressive disorders
As shown in Table 4, the comparison between older (Network 1) and middle-aged (Network 2) patients with depressive disorders revealed no significant differences in network structure invariance, network structure invariance with equal sample sizes, global strength invariance, or global strength invariance with equal sample sizes. Spearman correlations of node strength centrality and edges indicated a strong correlation (0.83) and a very strong correlation (0.94), respectively. Similarly, the comparison between older (Network 1) and young adult (Network 3) patients with depressive disorders demonstrated no significant differences in terms of network structure invariance, network structure invariance with equal sample sizes, and global strength invariance with equal sample sizes. However, a significant difference was found in the global strength invariance between the two network structures (pvalue=0.01). Spearman correlations of node strength centrality and edges indicated a strong correlation (0.74) and a very strong correlation (0.92), respectively.
Lastly, as shown in Figure 2A, the network structure of anxiety and depression symptoms was estimated using data from all age groups to visualize the similarities and differences in the edge connections across groups. This network analysis of GAD-7 and PHQ-9 items revealed that 83 (69.1%) of the 120 possible edges had an estimated weight greater than zero. Several strong interconnections were identified in the network, particularly the following item pairs: PLE-DEP (0.429), WORDFF (0.282), ANX-WOR (0.273), SEL-HUR (0.250), INS-ENG (0.229), RLX-RTL (0.227), and WOR-RLX (0.222). Edge statistics indicated a high level of interpretability, with a CS coefficient of 0.750. As shown in Figure 2B, the node strength centrality analysis showed that DEP was the most central symptom, followed by WOR, ANX, RLX, HUR, and RTL. In contrast, RTL was the least interconnected symptom in the network, followed by APE, AWF, IRR, and CNC. The node statistics indicated a high level of interpretability, with a CS coefficient of 0.750.
DISCUSSION
This study found no significant differences in the network structure of anxiety and depression symptoms across adult age groups, although the symptom severities varied with age in patients with depressive disorders. Although the symptom network structures are consistent, subtle differences are present, particularly in the interconnection and expression of symp-toms. Lamers et al. [33] used a factor analysis and demonstrated that the structure of depression in adolescents is similar to the structure in adults. Those findings, combined with ours, support the use of age-neutral diagnostic frameworks that accommodate both shared and age-specific symptom interconnections while minimizing the impact of ageist biases. An ageneutral diagnostic framework would focus on identifying and evaluating symptoms based on their intrinsic characteristics, rather than relying on assumptions or stereotypes about how symptoms typically manifest in different age groups. Advanc-ing equitable and inclusive integrative care for depressive disorder is needed due to consistent symptomatic patterns [34-36]. In older adults, ageism can lead to the misinterpretation of depressive disorders as a natural part of aging because depressive symptoms can manifest as physical complaints [37]. In addition, the findings of this study can be discussed as follows. First, the findings reported here challenge ageist assumptions that age is a primary influence on clinical presentation, which reinforce biases in treatment priorities. The subtle differences observed in global strength and symptom connectivity indicate that depression and anxiety manifest in consistently interconnected ways throughout the lifespan. In our findings, the strong centrality of depressive mood and its connections to worry and restlessness across age groups highlight that treatment approaches should target shared symptom nodes regardless of the patient’s age. Second, against the ageist tendency to over- or under-pathologize certain age groups, our findings advocate for a more inclusive perspective when providing mental health care for depressive disorders. Ageism often manifests as exaggerated distinctions between younger and older adults, and it frequently leads to the marginalization of older populations in clinical trials and the allocation of resources [10-12,38]. By demonstrating the structural similarity of depression and anxiety symptom networks across age groups, this study should prompt a paradigm shift toward age-neutral diagnostic and therapeutic strategies to protect older adults with depressive disorders against ageist biases. Third, although symptom networks are similar across age groups, the causes of anxiety and depression might differ by age. In older adults, the strong interconnection between insomnia and fatigue could be causatively associated with physical illness and social isolation [39]. Contrastingly, in middle-aged adults, the strong interconnection between depressive mood and excessive worry could be caused by workplace stress and family caregiving burden [40], and in young adults, the strong interconnection between anxiety and self-criticism could be caused by economic insecurity, social comparison, and interpersonal stress [41]. Therefore, individualized approaches might be needed to prevent age-related biases when treating depressive disorders.
The network structure of depression and anxiety symptoms in older patients presents relatively strong interconnections among somatic symptoms (i.e., insomnia-energy), core depressive symptoms (e.g., loss of interest or pleasure-depressive mood), and anxiety-related symptoms (e.g., worry-restlessness). These findings can contribute to a reconsideration of ageist stereotypes that often marginalize older adults as frail and cognitively declining. The identification of depressive mood as the most central symptom highlights its pivotal role in the symptom network of older patients, emphasizing its potential as a primary therapeutic target. In addition, this finding can be discussed as follows. Unlike depression in young adulthood, late-life depression can be categorized based on differences in structural neuroimaging, distinct clinical features, genetic patterns, and susceptibility to future neurodegenerative diseases [42]. In addition, repeated depressive episodes can be associated with an acceleration of biological aging [43]. This perspective suggests that the core symptomatic phenomenon of depression resides in feelings of hopelessness, whereas age-related variations (i.e., pronounced cognitive decline, prominent physical symptoms) in older adults might represent unique characteristics of depressive disorders in that age group. On the other hand, the young adult group presented subtly different findings, including a relatively strong interconnection between selfcriticism and self-hurting. That finding might be explained by the unstable job security more often found in younger individuals than older adults [44]. The economic insecurity and social isolation associated with unstable job security could also contribute to the heightened risk of suicide in the young adult group [45]. Moreover, it is speculated that the biological interaction of aging and accelerating aging associated with depressive disorders present differently by age. In young adults, energy depletion can contribute to functional decline, which is associated with social vulnerability and can exacerbate job security and worsen economic conditions. Also, depressed young adults tend to feel more deprived and isolated than their nondepressed peers. Thus, these biological and social factors might explain the subtle variations found in symptom networks across age groups [43].
Our study has several limitations. First, the cross-sectional design limits our ability to infer causality or temporal changes in symptom networks. Longitudinal studies are required to examine how symptom interconnections evolve. Second, reliance on self-report measures could introduce response bias, particularly among older adults who might underreport emotional symptoms because of cultural or generational attitudes toward mental health. Third, variations in the sample size across age groups and countries might influence the generalizability of our findings. Further research is needed to investigate whether interventions targeting shared central symptoms, including depressive mood and excessive worry, can mitigate the effects of ageist biases in mental health care [46]. Fourth, PHQ-9 and GAD-7 might lack the sensitivity needed to capture age-specific psychiatric manifestations and the potential influence of cultural and generational factors on symptom reporting. Fifth, depression and anxiety symptom networks might vary across cultural and racial contexts, influencing symptom clustering and connectivity patterns. To enhance the generalizability of these findings, further study to examine whether similar network structures emerge in Western older adults might be needed. Comparing cross-cultural symptom networks could clarify whether these interconnections reflect universal aging-related mechanisms or culture-specific expressions of depression and thus inform more targeted diagnostic and therapeutic approaches. Sixth, anxiety/depression symptoms and other psychiatric conditions have potential overlap, and that might have influenced our findings. Lastly, the influence of antidepressants on symptom expression should be considered because both anxiety and depression symptoms can be alleviated by antidepressants such as selective serotonin reuptake inhibitors [47].
Despite those limitations, our network analysis has challenged ageist stereotypes about depressive disorders. Our study highlights the relatively consistent interconnection of depression and anxiety symptoms across different age groups, although the symptom severities vary with age. By identifying potentially central symptoms and advocating for inclusive and equitable mental health care, this study contributes to the growing movement to advance age-neutral approaches to the diagnosis and treatment of depressive disorders.
PDF Links
PubReader
ePub Link
XML Download
Full text via DOI
Download Citation
Supplement
Print
