Clustering of cardiometabolic risk factors, physical activity, digital exclusion, and depressive symptoms trajectories in adults aged 50 years and older: findings from two prospective cohort studies Authors: Xiaosheng Dong, Jiaqiang Xiao, Xiangren Yi, Meng Ding, Mingyang Bai, Xinmeng Guo, Xiao Hou, Chengchao Zhou Published: 2025 ABSTRACT BackgroundTo explore the associations between clustering of cardiometabolic risk factors, physical activity (PA), and digital exclusion with depressive symptoms trajectories, and to assess whether PA and digital exclusion mediate these associations.MethodsThis study included 8999 participants from the Health and Retirement Study (HRS, n = 4380) and the English Longitudinal Study of Ageing (ELSA, n = 4519). Cardiometabolic risk factors included waist circumference, high-density lipoprotein (HDL) cholesterol, systolic and diastolic blood pressure, type 2 diabetes mellitus, glycated hemoglobin (HbA1c), and C-reactive protein (CRP). Depressive symptoms, PA, and digital exclusion were assessed via self-report. Statistical analyses used multinomial logistic regression and the Karlson-Holm-Breen (KHB) method.ResultsWe identified three clustering of cardiometabolic risk factors and four depressive symptoms trajectories. Compared to participants classified as healthy obesity, those in the obesity-hypertension group had an OR of 1.26 (95% CI: 1.09–1.46) for the moderately severe depressive symptom trajectory, while those in the complex cardiometabolic group had an OR of 1.31 (95% CI: 1.08–1.58), 1.79 (95% CI: 1.45–2.23), or 2.19 (95% CI: 1.66–2.90) for the moderate, moderately severe, or severe depressive symptom trajectory, respectively. Compared to those with PA insufficiency and digital exclusion, among PA sufficient participants, the OR of being in the moderate, moderately severe, or severe depressive symptom trajectory was 0.74 (95% CI: 0.65–0.84), 0.57 (95% CI: 0.50–0.66), or 0.35 (95% CI: 0.20–0.61), respectively; among digitally engaged participants, the OR of being in the moderately severe or severe trajectory was 0.63 (95% CI: 0.53–0.75) or 0.42 (95% CI: 0.33–0.54), respectively. Compared with the healthy obesity group, those with sufficient PA or digital inclusion in obesity-hypertension and complex cardiometabolic groups had lower odds of depressive symptoms trajectories. Mediation analysis showed that PA and digital inclusion accounted for 12.5% and 6.67% of the association between clustering of cardiometabolic risk factors and the severe depressive symptom trajectory, respectively.ConclusionsThe obesity-hypertension was only associated with increased odds of the moderately severe depressive symptom trajectory, while the complex cardiometabolic pattern was associated with increased odds of depressive symptoms trajectories; in contrast, sufficient PA and digital inclusion were associated with reduced odds.Supplementary InformationThe online version contains supplementary material available at 10.1186/s12916-025-04604-y. Background Depression is a major global public health challenge and has been identified by the World Health Organization as the leading cause of disability [1]. More than 300 million people worldwide are affected by depression [2], and the annual healthcare costs associated with depression in the United States have reached $382.4 billion [3]. Furthermore, depression is closely linked to adverse health outcomes such as cardiovascular disease and premature mortality [4–6], and it is projected to become the leading contributor to the global disease burden by 2030 [7]. Despite available psychological and pharmacological treatments, more than one-third of individuals with depression do not achieve an adequate therapeutic response [8–10]. Moreover, depressive symptoms fluctuate over time and exhibit substantial inter-individual variation [11, 12]; it is essential to investigate the longitudinal trajectories of depressive symptoms and their associated risk factors. Research has shown that metabolic risk factors such as diabetes are associated with an increased risk of depressive symptoms [13]. Elevated inflammatory markers associated with abdominal obesity and dyslipidemia contribute to neural and mood dysregulation, further increasing depression risk [14, 15]. Additionally, higher plasma levels of tumor necrosis factor-alpha (TNF-α) in individuals with hypertension are linked to depression via their effects on neurotransmission and synaptic regulation [16]. With the global population aging, the prevalence of coexisting cardiometabolic conditions continues to rise [17, 18], and the interaction of these conditions further compounds health risks [19, 20]. While prior studies have primarily focused on individual cardiometabolic risk factors, little is known about the effects of heterogeneous clustering of these factors on depressive symptoms trajectories [13, 21]. By identifying these clusters of cardiometabolic risk factors through data-driven latent class analysis (LCA), unobserved heterogeneous subgroups within the population based on combinations of multiple cardiometabolic risk factors can be revealed. This approach advances beyond prior research that merely counted the number of conditions, shifting instead toward a qualitative understanding of disease combinations that may be more significant etiologically and clinically [22]. Insufficient physical activity (PA) and digital exclusion are also important influencing factors for depressive symptoms [23]. PA is associated with the release of neurotrophic factors such as vascular endothelial growth factor, decreases pro-inflammatory cytokines like C-reactive protein (CRP), and supports the structure and function of brain regions such as the hippocampus, thereby alleviating depressive symptoms [24]. Similarly, Internet use can enhance social connectedness, strengthen interpersonal relationships, and offer cognitive stimulation through learning opportunities, all of which have been shown to reduce depressive symptoms [25–27]. However, the relationship between PA, digital exclusion, and the dynamic course of depression remains unclear [28, 29]. Studies have shown that behavioral and social factors may influence the association between cardiometabolic risk and mental health [30], providing a theoretical framework for understanding how cardiometabolic risk factors indirectly contribute to longitudinal depressive symptoms by reducing PA [31] and limiting digital inclusion [32] (Additional file 1: Fig. S1). Fatigue in diabetic patients can diminish PA participation [33]. Individuals with cardiometabolic risk factors are often exposed to high levels of stress, which may not only lead to impaired physical function [34] but also increase anxiety, frailty, and fatigue, thereby further reducing their PA levels [35]. Furthermore, cognitive or language impairments in stroke survivors may hinder Internet access, thereby reducing digital inclusion [36]. In addition, cardiometabolic risk factors may interact separately with PA and digital inclusion, each contributing to the joint influence on depressive symptoms. Cardiometabolic risk factors may elevate depressive symptoms by upregulating inflammatory levels [14, 15, 37], whereas physical activity can mitigate inflammation by reducing pro-inflammatory markers and increasing anti-inflammatory markers [24]. Digital inclusion is associated with reduced social isolation commonly experienced by individuals with cardiometabolic conditions [38, 39], enhancing well-being and social connectedness, which in turn reduces the risk of depression [38–42]. Nonetheless, the mediating and moderating roles of PA and digital exclusion in the relationship between clustering of cardiometabolic risk factors and depressive symptoms trajectories remain poorly understood. Currently, there is a lack of studies using large-scale, cross-national cohort data to systematically examine the relationships among specific clusters of cardiometabolic risk factors, PA, digital inclusion, and the longitudinal trajectories of depressive symptoms. Globally, several long-term studies on aging have been conducted, such as the US Health and Retirement Study (HRS) and the UK English Longitudinal Study of Ageing (ELSA). These studies have large sample sizes, adopt similar research designs, cover overlapping time periods, and maintain high data quality, providing a solid foundation for cross-national comparative research. Based on these data, we examined the associations between baseline clustering of cardiometabolic risk factors, PA, digital exclusion, and depressive symptoms trajectories during the follow-up period among middle-aged and older adults, and further explored the mediating and moderating roles of PA and digital exclusion in these relationships. Methods Design and sample Data for this study were obtained from the ELSA in the UK and the HRS in the USA. These are nationally representative samples of adults aged 50 years or older in the UK and the USA, respectively, with follow-up conducted biennially. Further methodological details are available elsewhere [43, 44]. The HRS was approved by the Institutional Review Board of the University of Michigan and the National Institute on Aging (HUM00061128). The ELSA was approved by the London Multicenter Research Ethics Committee (MREC/01/2/91). Written informed consent was obtained from all participants. Wave 2 of ELSA (2004/2005) and wave 9 of HRS (2008/2009) were selected as baseline data for this study. Follow-up data were obtained from waves 3 to 10 of ELSA (2006/2007 to 2021/2023) and waves 10 to 16 of HRS (2010/2011 to 2022/2023). The initial samples included 9432 ELSA and 17,217 HRS participants. Participants were excluded if they (1) were age under 50 years (261 ELSA and 360 HRS), (2) had missing main data at baseline (4176 ELSA and 11,986 HRS), and (3) lost follow-up (476 ELSA and 491 HRS); the final analytical samples consisted of 4519 ELSA and 4380 HRS participants. Flowcharts of sample selection are shown in Fig. 1.Fig. 1Flow chart of participant selection Flow chart of participant selection Cardiometabolic risk factors Based on existing studies [45], cardiometabolic risk factors included abdominal obesity, lowered high-density lipoprotein (HDL) cholesterol, elevated systolic blood pressure, elevated diastolic blood pressure, type 2 diabetes mellitus, elevated glycosylated hemoglobin (HbA1c), and elevated CRP. Specifically, abdominal obesity was defined as a waist circumference ≥ 94 cm in men and ≥ 80 cm in women; elevated systolic blood pressure as ≥ 130 mm Hg; elevated diastolic blood pressure as ≥ 85 mm Hg; and lowered HDL cholesterol as < 40 mg/dL in men and < 50 mg/dL in women. Type 2 diabetes mellitus was defined as self-reported, physician-diagnosed diabetes. Hyperglycemia was defined as an HbA1c level ≥ 48 mmol/mol (6.5%) according to the American Diabetes Association criteria [46], and elevated CRP was defined as a level > 3 mg/L, in line with guidelines from the American Heart Association and the Centers for Disease Control and Prevention [47]. All seven cardiometabolic risk factors were treated as dichotomous variables. Digital exclusion According to existing research [48], digital exclusion in the HRS was assessed using the following question: “Do you regularly use the Internet (or the World Wide Web) for sending and receiving e-mail or for any other purpose, such as making purchases, searching for information, or making travel reservations?” A response of “no” was classified as digital exclusion. In ELSA, numerical deprivation was assessed by asking participants how often they used the Internet, with response options ranging from 1 = “Every day or almost every day” to 6 = “Never”. Participants who reported using the Internet less frequently than once per week were classified as experiencing numerical deprivation. Physical activity Participants were asked about the frequency of their engagement in light-intensity, moderate-intensity, and vigorous-intensity PA, with response options including “every day,” “more than weekly,” “weekly,” “1–3 times monthly,” or “rarely/never.” Based on previous studies [49, 50], the reported frequency of each intensity level was converted into the equivalent number of activity days per month. The monthly frequency for each intensity category was then multiplied by the corresponding average metabolic equivalent (MET) value, and the resulting scores across all intensity categories were summed to obtain the total PA score. Subsequently, the total PA scores were categorized into two levels (insufficient PA and sufficient PA) based on a cut-off value of the 66th percentile (3rd tertile) of the range of the distribution in this cohort. Additional file 1: Table S1–S2 provides comprehensive details. Depressive symptoms Depressive symptoms in this study were assessed using the 8-item Center for Epidemiologic Studies Depression Scale (CESD-8) [51]. Both HRS and ELSA employ validated, English-language versions of the CESD-8 that have demonstrated good reliability and validity in participants [52]. The scale comprises eight symptoms, and participants reported whether they had experienced each one in the past week. 2 of the 8 items (i.e., “feel happy” and “have a happy life”) were reverse-coded prior to data analysis. Covariates Covariates included age, sex (male/female), marital status (married or partnered vs. unmarried or other), alcohol consumption (never vs. former or current), smoking status (never vs. former or current), educational attainment (higher education, high school or college, less than high school or no formal qualification), and presence of cardiovascular disease (yes/no), defined as a diagnosis of any of the following: stroke, myocardial infarction, coronary heart disease, angina pectoris, congestive heart failure, or other heart conditions. Statistical analysis Continuous variables in the descriptive statistics of this study were presented as means (standard deviations), and categorical variables as frequencies (percentages). First, LCA was used to identify patterns in the clustering of cardiometabolic risk factors [53–55]. For details and results of LCA, please refer to the Additional file 1: eMethods on LCA. Second, the trajectory of depressive symptoms was identified using group-based trajectory modeling (GBTM), and the best-fitting model was selected based on average posterior probabilities, sample size, and the Bayesian Information Criterion (BIC). Third, multinomial logistic regression was employed to examine the association between latent classes of cardiometabolic risk factors, PA, digital exclusion, and the trajectory of depressive symptoms, and the “Mild” trajectory was set as the reference group. Model 1 was unadjusted; model 2 adjusted for age and sex; and model 3 further adjusted for marital status, alcohol consumption, cigarette smoking, education level, and cardiovascular disease. Fourth, participants were stratified based on the combined classification of cardiometabolic risk factors and PA/digital exclusion into five groups. The healthy obesity group served as the reference for statistical comparisons, and analyses were further stratified by latent classes of cardiometabolic risk factors. Fifth, we assessed the mediating roles of PA (baseline) and digital exclusion (baseline) in the relationship between latent classes of cardiometabolic risk factors (baseline) and depressive symptoms trajectories (follow-up) using the Karlsson-Holm-Breen (KHB) method in Stata [56]. The KHB method allows the total effect to be decomposed into an indirect effect (the portion explained by the mediator) and a direct effect (the remaining association after controlling for the mediator). We conducted the analysis using multinomial logistic regression combined with the KHB method. To quantify the joint effects of cardiometabolic risk factor classes with PA or digital exclusion, participants were categorized into 5 groups based on all possible combinations of these variables, using the healthy obesity group as the reference. We did all primary analyses for each survey cohort and then pooled the results using a random effects model. In addition, we conducted five sensitivity analyses: (1) using the waist obesity criteria specific to the U.S. population in the HRS; (2) applying inverse probability weighting in the analysis; (3) categorizing PA levels into three subgroups, namely inactive PA, insufficient PA, and sufficient PA; (4) using imputed data; (5) defining the ELSA baseline as Wave 4 (2008/2009). All statistical analyses were performed using R 4.4.2 and Stata MP 17. Results Participant characteristics A total of 4380 participants from HRS and 4519 participants from ELSA were included in this study. The mean age of participants in HRS and ELSA was 68.2 (9.2) years and 65.2 (8.9) years, respectively. There were 40.2% male participants in HRS and 45.1% in ELSA. In HRS, 28.0% of participants had cardiovascular disease, compared to 21.95% in ELSA. Table 1 presents the baseline characteristics of participants in both cohorts. Table 1Characteristics of the participants from 2 independent cohortsVariablesHRS (n = 4380)ELSA (n = 4519)Age68.2 (9.2)65.2 (8.9)Gender Male1762 (40.2)2036 (45.1) Female2618 (59.8)2483 (55.0)Marital status Married or in a relationship2964 (67.7)3169 (70.1) Unmarried or other1416 (32.3)1350 (29.9)Education level Less than high school or no qualification945 (21.5)2095 (46.4) High school or college1490 (34.0)1167 (25.8) Higher education1945 (44.4)1257 (27.8)Drinking status Never2064 (47.1)402 (8.9) Ever or current2316 (52.9)4117 (91.1)Smoking status Never3796 (86.7)3783 (83.7) Ever or current584 (13.3)736 (16.3)Cardiovascular diseases Yes1226 (28.0)992 (22.0) No3154 (72.0)3527 (78.1)Values are n (%), or mean (SD)HRS The Health and Retirement Study, ELSA The English Longitudinal Study of Ageing Table 1Characteristics of the participants from 2 independent cohortsVariablesHRS (n = 4380)ELSA (n = 4519)Age68.2 (9.2)65.2 (8.9)Gender Male1762 (40.2)2036 (45.1) Female2618 (59.8)2483 (55.0)Marital status Married or in a relationship2964 (67.7)3169 (70.1) Unmarried or other1416 (32.3)1350 (29.9)Education level Less than high school or no qualification945 (21.5)2095 (46.4) High school or college1490 (34.0)1167 (25.8) Higher education1945 (44.4)1257 (27.8)Drinking status Never2064 (47.1)402 (8.9) Ever or current2316 (52.9)4117 (91.1)Smoking status Never3796 (86.7)3783 (83.7) Ever or current584 (13.3)736 (16.3)Cardiovascular diseases Yes1226 (28.0)992 (22.0) No3154 (72.0)3527 (78.1)Values are n (%), or mean (SD)HRS The Health and Retirement Study, ELSA The English Longitudinal Study of Ageing Characteristics of the participants from 2 independent cohorts Values are n (%), or mean (SD) HRS The Health and Retirement Study, ELSA The English Longitudinal Study of Ageing Latent class analysis Latent class analysis identified three latent classes of cardiometabolic risk factors as the best-fitting model (see Additional file 1: Table S3 for model fit statistics): Class 1 (C1), healthy obesity; Class 2 (C2), obesity-hypertension; and Class 3 (C3), complex cardiometabolic. C1 included 60.1% of participants in HRS and 61.5% in ELSA and was characterized by a low probability of all cardiometabolic conditions except abdominal obesity. C2 accounted for 26.0% of the HRS and 31.7% of the ELSA samples and exhibited a high probability of both obesity and hypertension. C3 comprised 13.9% of HRS and 6.9% of ELSA participants and was marked by a high probability of multiple cardiometabolic risks, including abdominal obesity, elevated systolic blood pressure, diabetes mellitus, and hyperglycemia (all with probabilities > 0.5). Figure 2 illustrates the distribution of the three cardiometabolic risk factor classes across the study cohorts.Fig. 2Predicted probabilities of cardiometabolic risk factors by latent classes from LCA in HRS and ELSA. HRS, The Health and Retirement Study; ELSA, The English Longitudinal Study of Ageing; CRP, C-reactive protein; DBP, diastolic blood pressure; HbA1c, hemoglobin A1c; HDL, high-density lipoprotein; SBP, systolic blood pressure Predicted probabilities of cardiometabolic risk factors by latent classes from LCA in HRS and ELSA. HRS, The Health and Retirement Study; ELSA, The English Longitudinal Study of Ageing; CRP, C-reactive protein; DBP, diastolic blood pressure; HbA1c, hemoglobin A1c; HDL, high-density lipoprotein; SBP, systolic blood pressure Group-based trajectory modeling We identified four best-fitting trajectories of depressive symptoms in both HRS and ELSA cohorts (Fig. 3), with model fit statistics detailed in Additional file 1: Table S4. Based on the pattern of symptom progression, the trajectories were labeled as mild, moderate, moderately severe, and severe. Participants in the mild group, who consistently exhibited low CESD-8 scores, accounted for 29.3% and 30.8% of the total participants in HRS and ELSA, respectively. The moderate group, with slightly elevated CESD-8 scores, represented 40.9% and 42.7% of participants in HRS and ELSA. The severe group, characterized by persistently high CESD-8 scores, included 7.7% and 6.4% of participants. The moderately severe group had CESD-8 scores lower than those of the Severe group but higher than those in the mild and moderate groups.Fig. 3Trajectories of depressive symptoms. CESD-8, the 8-item Center for Epidemiologic Studies Depression Scale; HRS, the Health and Retirement Study; ELSA, the English Longitudinal Study of Ageing Trajectories of depressive symptoms. CESD-8, the 8-item Center for Epidemiologic Studies Depression Scale; HRS, the Health and Retirement Study; ELSA, the English Longitudinal Study of Ageing Association between clustering of cardiometabolic risk factors, physical activity, and digital exclusion with trajectories of depressive symptoms Figure 4 presents the associations between clustering of cardiometabolic risk factors, PA, digital exclusion, and depressive symptoms trajectories, as assessed using multinomial logistic regression. After adjusting for all covariates and pooling the results, among participants classified as obesity-hypertension, the significant OR of being in the moderately severe depressive symptom trajectory (compared with the mild depressive symptom trajectory) was 1.26 (95% CI: 1.09–1.46, I2 = 0.0%, p = 0.828) compared to participants classified as healthy obesity. Among participants classified as complex cardiometabolic, the significant OR of being in the moderate, moderately severe, or severe depressive symptom trajectory (compared with the mild depressive symptom trajectory) was 1.31 (95% CI: 1.08–1.58, I2 = 0.0%, p = 0.613), 1.79 (95% CI: 1.45–2.23, I2 = 0.0%, p = 0.615), or 2.19 (95% CI: 1.66–2.90, I2 = 0.0%, p = 0.615) compared to participants classified as healthy obesity, respectively. Furthermore, among PA sufficient participants, the OR of being in the moderate, moderately severe, or severe depressive symptom trajectory (compared with the mild depressive symptom trajectory) was 0.74 (95% CI: 0.65–0.84, I2 = 34.6%, p = 21.6), 0.57 (95% CI: 0.50–0.66, I2 = 0.0%, p = 0.456), or 0.35 (95% CI: 0.20–0.61, I2 = 80.4%, p = 0.024) compared to PA insufficient participants, respectively. Compared to those who were digitally deprived, among digitally engaged participants, the significant OR of being in the moderately severe or severe depressive symptom trajectory (compared with the mild depressive symptom trajectory) was 0.63 (95% CI: 0.53–0.75, I2 = 25.7%, p = 0.246) or 0.42 (95% CI: 0.33–0.54, I2 = 0.0%, p = 0.504), respectively. Association between clustering of cardiometabolic risk factors, physical activity, and digital exclusion with trajectories of depressive symptoms in model 1 and model 2 are shown in Table S5 and S6. Predicted probabilities of belonging to the depressive symptoms trajectories across cardiometabolic latent classes are shown in Additional file 1: Table S7.Fig. 4Association between cardiometabolic risk factors, physical activity, and digital exclusion with trajectories of depressive symptoms. OR, odds ratio; CI, confidential interval; HRS, the Health and Retirement Study; ELSA, the English Longitudinal Study of Ageing Association between cardiometabolic risk factors, physical activity, and digital exclusion with trajectories of depressive symptoms. OR, odds ratio; CI, confidential interval; HRS, the Health and Retirement Study; ELSA, the English Longitudinal Study of Ageing Moderating effects of physical activity and digital exclusion on the association between cardiometabolic risk factors and depressive symptoms trajectories When stratified by cardiometabolic risk group, among participants with healthy obesity, the significant OR of being in the moderate, moderately severe, or severe depressive symptom trajectory was 0.74 (95% CI: 0.65–0.85, I2 = 0.0%, p = 0.494), 0.55 (95% CI: 0.46–0.66, I2 = 0.0%, p = 0.556), or 0.37 (95% CI: 0.26–0.54, I2 = 25.5%, p = 0.247) among PA sufficient participants, respectively; the significant OR of being in the moderately severe or severe depressive symptom trajectory was 0.68 (95% CI: 0.57–0.82, I2 = 0.0%, p = 0.813) or 0.42 (95% CI: 0.31–0.57, I2 = 0.0%, p = 0.818) among digitally engaged participants, respectively. Among participants in the obesity-hypertension group, the significant OR of being in the moderate, moderately severe, or severe depressive symptom trajectory was 0.70 (95% CI: 0.55–0.90, I2 = 35.7%, p = 0.212), 0.66 (95% CI: 0.51–0.86, I2 = 8.2%, p = 0.297), or 0.31 (95% CI: 0.11–0.84, I2 = 73.4%, p = 0.053) among PA sufficient participants, respectively; the significant OR of being in the moderately severe or severe depressive symptom trajectory was 0.64 (95% CI: 0.48–0.84, I2 = 21.2%, p = 0.260) or 0.33 (95% CI: 0.19–0.58, I2 = 36.6%, p = 0.209) among digitally engaged participants, respectively. Among those in the complex cardiometabolic group, the significant OR of being in the moderately severe depressive symptom trajectory was 0.57 (95% CI: 0.34–0.96, I2 = 0.0%, p = 1.000) among PA sufficient participants; the significant OR of being in the moderate or moderately severe depressive symptom trajectory was 0.53 (95% CI: 0.37–0.77, I2 = 0.0%, p = 0.455) or 0.40 (95% CI: 0.26–0.62, I2 = 0.0%, p = 0.608) among digitally engaged participants, respectively. (Additional file 1: Table S8; Fig. 5).Fig. 5Association between cardiometabolic risk factors, physical activity, and digital exclusion with trajectories of depressive symptoms by different cardiometabolic latent classes Association between cardiometabolic risk factors, physical activity, and digital exclusion with trajectories of depressive symptoms by different cardiometabolic latent classes Table 2 displays the joint associations of clustering of cardiometabolic risk factors, PA, and digital exclusion with the four depressive symptom trajectory groups. After pooling results, compared to participants with healthy obesity, among PA-insufficient participants in the obesity-hypertension group, the significant OR of being in the moderate, moderately severe, or severe depressive symptom trajectory (compared with the mild depressive symptom trajectory) was 1.25 (95% CI: 1.09–1.42, I2 = 0.0%, p = 0.725), 1.44 (95% CI: 1.22–1.69, I2 = 0.0%, p = 0.528), or 1.46 (95% CI: 1.07–1.99, I2 = 43.0%, p = 0.185), respectively; among PA-insufficient participants in the complex cardiometabolic group, the significant OR was 1.32 (95% CI: 1.07–1.64, I2 = 0.0%, p = 0.812), 1.98 (95% CI: 1.56–2.51, I2 = 0.0%, p = 0.676), or 2.36 (95% CI: 1.75–3.20, I2 = 0.0%, p = 0.609), respectively. Among PA-sufficient participants in the obesity-hypertension group, the significant OR of being in the severe depressive symptom trajectory (compared with the mild depressive symptom trajectory) was 0.46 (95% CI: 0.24–0.88, I2 = 44.6%, p = 0.179). Among PA-sufficient participants in the obesity-hypertension group, the estimated OR of being in the moderate, moderately severe depressive symptom trajectory was not statistically significant. Similarly, among PA-insufficient participants in the complex cardiometabolic group, the estimated OR of being in the moderate, moderately severe, or severe depressive symptom trajectory was also not statistically significant. Table 2Joint association between clustering of cardiometabolic risk factors, physical activity, and digital exclusion with trajectories of depressive symptomsVariablesPhysical activityNoDepressive symptoms trajectories (reference: mild)Digital exclusionNoDepressive symptoms trajectories (reference: mild)ModerateModerately SevereSevereModerateModerately SevereSevereHRSCardiometabolic latent classes C1: Healthy obesity26111.00 (ref)1.00 (ref)1.00 (ref)26111.00 (ref)1.00 (ref)1.00 (ref) C2: Obesity-hypertensionInsufficient6651.30 (1.06–1.61)1.30 (1.01–1.67)1.18 (0.81–1.73)Yes5411.50 (1.17–1.92)1.53 (1.15–2.03)1.43 (0.97–2.12)Sufficient3190.98 (0.75–1.28)1.00 (0.71–1.41)0.59 (0.31–1.12)No4430.96 (0.76–1.21)0.93 0.69–1.26)0.46 (0.24–0.91) C3: Complex cardiometabolicInsufficient6191.62 (1.27–2.08)2.16 (1.64–2.85)2.27 (1.60–3.24)Yes5312.05 (1.54–2.73)2.59 (1.89–3.54)2.55 (1.72–3.76)Sufficient1661.54 (1.04–2.29)1.47 (0.90–2.40)1.81 (0.94–3.48)No2541.17 (0.85–1.59)1.37 (0.94–2.02)2.00 (1.18–3.39)ELSACardiometabolic latent classes C1: Healthy obesity27771.00 (ref)1.00 (ref)1.00 (ref)27771.00 (ref)1.00 (ref)1.00 (ref) C2: Obesity-hypertensionInsufficient10381.22 (1.02–1.46)1.51 (1.22–1.89)1.69 (1.24–2.31)Yes8691.10 (0.90–1.34)1.43 (1.13–1.81)1.52 (1.09–2.10)Sufficient3930.76 (0.60–0.96)0.89 (0.65–1.23)0.31 (0.14–0.68)No5621.00 (0.81–1.23)1.09 0.82–1.45)0.72 (0.42–1.24) C3: Complex cardiometabolicInsufficient2581.28(0.90–1.80)2.11 (1.43–3.12)2.63 (1.58–4.37)Yes2011.52 (1.00–2.30)2.53 (1.60–4.00)2.96 (1.66–5.29)Sufficient531.12 (0.60–2.10)1.23 (0.52–2.91)1.64 (0.46–5.84)No1100.98 (0.63–1.52)1.25 (0.68–2.27)1.97 (0.87–4.44)PooledCardiometabolic latent classes C1: Healthy obesity1.00 (ref)1.00 (ref)1.00 (ref)1.00 (ref)1.00 (ref)1.00 (ref) C2: Obesity-hypertensionInsufficient1.25 (1.09–1.44)1.42 (1.20–1.67)1.44 (1.01–2.04)Yes1.27 (0.94–1.72)1.47 (1.23–1.76)1.48 (1.15–1.91)I2 = 0.0%, P = 0.651I2 = 0.0%, P = 0.379I2 = 51.4%, P = 0.151I2 = 72.7%, P = 0.056I2 = 0.0%, P = 0.720I2 = 0.0%, P = 0.815Sufficient0.86 (0.67–1.10)0.94 (0.74–1.19)0.45 (0.24–0.83)No0.98 (0.84–1.15)1.01 (0.82–1.24)0.60 (0.39–0.92)I2 = 49.0%, P = 0.161I2 = 0.0%, P = 0.626I2 = 34.8%, P = 0.215I2 = 0.0%, P = 0.798I2 = 0.0%, P = 0.453I2 = 4.4, P = 0.306 C3: Complex cardiometabolicInsufficient1.49 (1.19–1.86)2.14 (1.71–2.69)2.38 (1.78–3.18)Yes1.84 (1.38–2.44)2.57 (1.98–3.33)2.67 (1.93–3.69)I2 = 15.1%, P = 0.278I2 = 0.0%, P = 0.923I2 = 0.0%, P = 0.641I2 = 25.7%, P = 0.246I2 = 0.0%, P = 0.934I2 = 0.0%, P = 0.676Sufficient1.41 (1.01–1.96)1.41 (0.92–2.15)1.77 (0.99–3.17)No1.10 (0.85–1.42)1.33 (0.97–1.84)1.99 (1.28–3.10)I2 = 0.0%, P = 0.399I2 = 0.0%, P = 0.724I2 = 0.0%, P = 0.892I2 = 0.0%, P = 0.520I2 = 0.0%, P = 0.801I2 = 0.0%, P = 0.976Model was adjusted for age, sex, marital status, education level, smoking and drinking status, cardiovascular diseases. HRS, The Health and Retirement Study; ELSA, The English Longitudinal Study of Ageing Table 2Joint association between clustering of cardiometabolic risk factors, physical activity, and digital exclusion with trajectories of depressive symptomsVariablesPhysical activityNoDepressive symptoms trajectories (reference: mild)Digital exclusionNoDepressive symptoms trajectories (reference: mild)ModerateModerately SevereSevereModerateModerately SevereSevereHRSCardiometabolic latent classes C1: Healthy obesity26111.00 (ref)1.00 (ref)1.00 (ref)26111.00 (ref)1.00 (ref)1.00 (ref) C2: Obesity-hypertensionInsufficient6651.30 (1.06–1.61)1.30 (1.01–1.67)1.18 (0.81–1.73)Yes5411.50 (1.17–1.92)1.53 (1.15–2.03)1.43 (0.97–2.12)Sufficient3190.98 (0.75–1.28)1.00 (0.71–1.41)0.59 (0.31–1.12)No4430.96 (0.76–1.21)0.93 0.69–1.26)0.46 (0.24–0.91) C3: Complex cardiometabolicInsufficient6191.62 (1.27–2.08)2.16 (1.64–2.85)2.27 (1.60–3.24)Yes5312.05 (1.54–2.73)2.59 (1.89–3.54)2.55 (1.72–3.76)Sufficient1661.54 (1.04–2.29)1.47 (0.90–2.40)1.81 (0.94–3.48)No2541.17 (0.85–1.59)1.37 (0.94–2.02)2.00 (1.18–3.39)ELSACardiometabolic latent classes C1: Healthy obesity27771.00 (ref)1.00 (ref)1.00 (ref)27771.00 (ref)1.00 (ref)1.00 (ref) C2: Obesity-hypertensionInsufficient10381.22 (1.02–1.46)1.51 (1.22–1.89)1.69 (1.24–2.31)Yes8691.10 (0.90–1.34)1.43 (1.13–1.81)1.52 (1.09–2.10)Sufficient3930.76 (0.60–0.96)0.89 (0.65–1.23)0.31 (0.14–0.68)No5621.00 (0.81–1.23)1.09 0.82–1.45)0.72 (0.42–1.24) C3: Complex cardiometabolicInsufficient2581.28(0.90–1.80)2.11 (1.43–3.12)2.63 (1.58–4.37)Yes2011.52 (1.00–2.30)2.53 (1.60–4.00)2.96 (1.66–5.29)Sufficient531.12 (0.60–2.10)1.23 (0.52–2.91)1.64 (0.46–5.84)No1100.98 (0.63–1.52)1.25 (0.68–2.27)1.97 (0.87–4.44)PooledCardiometabolic latent classes C1: Healthy obesity1.00 (ref)1.00 (ref)1.00 (ref)1.00 (ref)1.00 (ref)1.00 (ref) C2: Obesity-hypertensionInsufficient1.25 (1.09–1.44)1.42 (1.20–1.67)1.44 (1.01–2.04)Yes1.27 (0.94–1.72)1.47 (1.23–1.76)1.48 (1.15–1.91)I2 = 0.0%, P = 0.651I2 = 0.0%, P = 0.379I2 = 51.4%, P = 0.151I2 = 72.7%, P = 0.056I2 = 0.0%, P = 0.720I2 = 0.0%, P = 0.815Sufficient0.86 (0.67–1.10)0.94 (0.74–1.19)0.45 (0.24–0.83)No0.98 (0.84–1.15)1.01 (0.82–1.24)0.60 (0.39–0.92)I2 = 49.0%, P = 0.161I2 = 0.0%, P = 0.626I2 = 34.8%, P = 0.215I2 = 0.0%, P = 0.798I2 = 0.0%, P = 0.453I2 = 4.4, P = 0.306 C3: Complex cardiometabolicInsufficient1.49 (1.19–1.86)2.14 (1.71–2.69)2.38 (1.78–3.18)Yes1.84 (1.38–2.44)2.57 (1.98–3.33)2.67 (1.93–3.69)I2 = 15.1%, P = 0.278I2 = 0.0%, P = 0.923I2 = 0.0%, P = 0.641I2 = 25.7%, P = 0.246I2 = 0.0%, P = 0.934I2 = 0.0%, P = 0.676Sufficient1.41 (1.01–1.96)1.41 (0.92–2.15)1.77 (0.99–3.17)No1.10 (0.85–1.42)1.33 (0.97–1.84)1.99 (1.28–3.10)I2 = 0.0%, P = 0.399I2 = 0.0%, P = 0.724I2 = 0.0%, P = 0.892I2 = 0.0%, P = 0.520I2 = 0.0%, P = 0.801I2 = 0.0%, P = 0.976Model was adjusted for age, sex, marital status, education level, smoking and drinking status, cardiovascular diseases. HRS, The Health and Retirement Study; ELSA, The English Longitudinal Study of Ageing Joint association between clustering of cardiometabolic risk factors, physical activity, and digital exclusion with trajectories of depressive symptoms Model was adjusted for age, sex, marital status, education level, smoking and drinking status, cardiovascular diseases. HRS, The Health and Retirement Study; ELSA, The English Longitudinal Study of Ageing Additionally, compared to participants with healthy obesity, among digitally deprived participants in the obesity-hypertension group, the significant OR of being in the moderately severe or severe depressive symptom trajectory (compared with the mild depressive symptom trajectory) was 1.51 (95% CI: 1.26–1.81, I2 = 0.0%, p = 0.479) or 1.55 (95% CI: 1.21–1.98, I2 = 0.0%, p = 0.878), respectively; among digitally deprived participants in the complex cardiometabolic group, the significant OR of being in the moderate, moderately severe, or severe depressive symptom trajectory (compared with the mild depressive symptom trajectory) was 1.72 (95% CI: 1.34–2.21, I2 = 0.0%, p = 0.472), 2.47 (95% CI: 1.88–3.25, I2 = 0.0%, p = 0.901), or 2.72 (95% CI: 1.94–3.82, I2 = 0.0%, p = 0.730), respectively. Among digitally deprived participants in the obesity-hypertension group, the significant OR of being in the moderately severe or severe depressive symptom trajectory (compared with the mild depressive symptom trajectory) was 1.51 (95% CI: 1.26–1.81, I2 = 0.0%, p = 0.479). Among digitally engaged participants in the obesity-hypertension, the significant OR of being in the severe depressive symptom trajectory (compared with the mild depressive symptom trajectory) was 0.58 (95% CI: 0.36–0.93, I2 = 26.0%, p = 0.245). Among digitally engaged participants in the complex cardiometabolic group, the significant OR of being in the severe depressive symptom trajectory was 1.86 (95% CI: 1.15–2.99, I2 = 0.0%, p = 0.860). Among digitally engaged participants in the obesity-hypertension or complex cardiometabolic group, the estimated OR of being in the moderate, moderately severe depressive symptom trajectory was not statistically significant. Mediation effects of physical activity and digital exclusion Figure 6 illustrates the mediating effects of PA and digital exclusion on the relationship between clustering of cardiometabolic risk factors and depressive symptoms trajectories. Pooled analyses showed that 9.09% (I2 = 47.9%, p = 0.166) or 7.69% (I2 = 0.0%, p = 1.000) of the effect of clustering of cardiometabolic risk factors on moderate or moderately severe depressive symptom trajectory was mediated through PA, while no significant mediating effect was observed for digital exclusion. The proportion of the association between clustering of cardiometabolic risk factors and the severe depressive symptom trajectory mediated by PA was 12.50% (I² = 0.0%, p = 0.587), whereas the proportion mediated by digital exclusion was 6.67% (I² = 70.7%, p = 0.065). The mediating effects of PA and digital exclusion on the relationship between clustering of cardiometabolic risk factors and depressive symptoms trajectories for HRS and ELSA are shown in Additional file 1: Fig. S2–S3.Fig. 6Diagram of mediation analyses. The direct and indirect effects of cardiometabolic risk factors on depressive symptoms trajectories. We first used random-effects models to pool the indirect and direct effects and then calculated the mediation proportions through the pooled indirect and direct values. MP, mediation proportion Diagram of mediation analyses. The direct and indirect effects of cardiometabolic risk factors on depressive symptoms trajectories. We first used random-effects models to pool the indirect and direct effects and then calculated the mediation proportions through the pooled indirect and direct values. MP, mediation proportion Sensitivity analyses The results of the sensitivity analysis using the waist obesity criteria specific to the U.S. population were consistent with the previous findings (Additional file 1: Table S9–S12). Similarly, the use of the inverse probability-weighted method did not materially alter the main findings (Additional file 1: Table S13–S16). In addition, when PA is treated as a three-category variable, cardiometabolic risk factors exhibit the same effect trend as in the main analysis (Additional file 1: Table S17–S20). The sensitivity analyses using the imputed data (Additional file 1: Table S21–S24; Additional file 1: Fig. S4 and Additional file 1: Fig. S6) and defining the ELSA baseline as Wave 4 (2008/2009) (Additional file 1: Table S25–S28; Additional file 1: Fig. S5 and Additional file 1: Fig. S7) both yielded results consistent with the main findings. Discussion This study found that the obesity–hypertension phenotype was associated only with higher odds of belonging to the moderately severe depressive symptom trajectory, while the complex cardiometabolic pattern was associated with increased odds across multiple depressive symptom trajectories. Sufficient PA and digital inclusion were both associated with reduced odds of more severe depressive symptom trajectories. Furthermore, sufficient PA and digital inclusion may moderate the associations of obesity-hypertension and complex cardiometabolic patterns with depressive symptom trajectories. These combined findings highlight the significant benefits of integrated strategies addressing clustering of cardiometabolic risk factors, PA, and digital inclusion in reducing the odds of depressive symptom trajectories. The present study found that obesity-hypertension was only associated with an increased odds of being moderately severe depressive symptom trajectory, and complex cardiometabolic pattern was associated with an increased odds of depressive symptoms trajectories. Previous studies have shown that managing cardiometabolic risk factors can reduce depression risk [57]. Obesity may activate the hypothalamic–pituitary–adrenal (HPA) axis by stimulating adrenal and pituitary stem cells [58], whereas hypertension induces prolonged sympathetic nervous system activation [59]. Both mechanisms contribute to chronic stress responses that disrupt neurotransmitter function [5, 60–63]. Severe depressive symptom trajectory may be accompanied by neurobiological impairments, including disrupted serotonin transmission [64], and the physiological burden of obesity-hypertension is likely insufficient to induce these severe consequences. However, the obesity-hypertension pattern was associated with higher odds of moderate depressive symptoms in the HRS cohort, but not in the ELSA cohort. This discrepancy may be attributed to the England healthcare system’s greater equity compared to the US, allowing more people to access blood pressure and related treatments [65], thereby mitigating the impact of cardiometabolic risks on depression. Individuals with complex cardiometabolic patterns, characterized by the co-occurrence of multiple cardiometabolic risk factors, may experience cumulative neurological impairments, including disrupted neural connectivity and white matter abnormalities, which adversely affect brain regions related to depression [66]. In conclusion, the control of cardiometabolic risk factors may play a vital role in the prevention and treatment of depression in middle-aged and older adults. We found that sufficient PA and digital inclusion were associated with lower odds of depressive symptoms trajectories. PA has been shown to promote the release of brain-derived neurotrophic factor (BDNF), enhance neuronal growth, survival, and plasticity, and improve depressive symptoms [24]. Additionally, PA helps regulate HPA axis function, reduce cortisol secretion, and alleviate depression-related HPA axis hyperactivity [24]. It also improves self-perception and overall body image, thereby disrupting the negative cycle between low self-esteem and depressive symptoms [24]. Furthermore, internet use may reduce depressive symptoms in middle-aged and older adults by alleviating feelings of helplessness and insecurity triggered by health issues [27], enhancing social participation, expanding social networks, and mitigating the risk of social isolation [25, 67, 68]. Therefore, promoting PA and digital inclusion may be an effective strategy for the prevention and control of depression in middle-aged and older populations. However, some pooled analyses showed high heterogeneity, which may be attributed to differences in measurement methods, as well as variations in healthcare and cultural contexts. We found that sufficient PA was associated with lower odds of depressive symptoms trajectories related to the obesity-hypertension and complex cardiometabolic pattern. Previous studies have demonstrated that PA alleviates depressive symptoms in patients with hypertension and diabetes [69, 70], which supports the findings of this study. Diabetes has been associated with lower BDNF levels, which may exacerbate depressive symptoms [71], whereas hypertension has been linked to vascular impairments that may increase vulnerability to depression [5]. PA may be associated with a reduction in these effects by upregulating BDNF [72] and promoting vascular health [73]. Additionally, PA may improve neurological function impaired by abdominal obesity and dyslipidemia, thereby lowering the odds of depression [74, 75]. Although the combined burden of multiple cardiometabolic risk factors in the complex cardiometabolic pattern may diminish the potential benefits of physical activity [76], PA remained associated with lower odds of adverse depressive symptom trajectories in this group. Furthermore, we found that digital inclusion may moderate the associations of obesity-hypertension and complex cardiometabolic patterns with depressive symptoms trajectories. Digital interventions have been shown to alleviate depressive symptoms in patients with hypertension or diabetes [77]. Digital inclusion also helps middle-aged and older adults maintain social connections, receive emotional support, and reduce social isolation [78], as well as access timely health information and medical advice [79], all of which contribute to lower odds of depressive symptoms. Therefore, maintaining adequate PA and digital inclusion may serve as a crucial strategy for preventing and managing depressive symptoms among middle-aged and older adults with cardiometabolic risk factors. This study found that the association between clustering of cardiometabolic risk factors and depressive symptoms trajectories in middle-aged and older adults was partly mediated by PA and digital exclusion. This suggests that comprehensive strategies targeting cardiometabolic risk factors while simultaneously improving PA and reducing digital exclusion may effectively lower the odds of depressive symptoms trajectories. Evidence suggests that simultaneous reduction of cardiometabolic risk factors [80], increased PA [81], and enhanced digital inclusion [82] more effectively decrease the secretion of stress hormones, such as cortisol, via modulation of the HPA axis, thereby lowering depression odds in middle-aged and older adults. Conversely, coexisting cardiometabolic risk factors and chronic diseases, such as hypertension [83], compounded by digital exclusion and social isolation [84], may undermine the beneficial effects of increased PA on depressive symptoms. Furthermore, digital exclusion can exacerbate inflammatory processes by elevating factors such as interleukin-6 [78], due to diminished social support, which in turn counteracts the positive effects of controlling cardiometabolic risk factors (e.g., abdominal obesity) [85] and engaging in PA [24] by increasing oxidative stress, thereby elevating depression odds. However, we found no evidence that digital exclusion mediated the association between clustering of cardiometabolic risk factors and the moderate or moderately severe depressive symptom trajectory. Compared with the severe depressive symptoms trajectory, individuals in the moderate or moderately severe depressive symptoms trajectory have milder symptoms and less functional impairment [86]. They can still obtain support through friends or offline social interactions, which limits the impact of digital exclusion. Given that the mediators in this study were measured at a fixed baseline time point, future research will apply longitudinal mediation analysis and the g-method to explore the effects of time-varying mediators. Therefore, public health policies aimed at preventing and managing depressive symptoms in middle-aged and older populations should emphasize integrated strategies that concurrently address cardiometabolic risk factors, PA, and digital inclusion. This study has several strengths. First, to the best of our knowledge, it is the first cohort study to examine the relationships between clustering of cardiometabolic risk factors, PA, digital exclusion, and depressive symptoms trajectories. Second, we analyzed two large, prospective cohorts from different countries, each with representative samples, and all analyses accounted for their complex survey designs. Third, LCA was utilized to identify potential clusters of cardiometabolic factors, allowing for a comprehensive assessment of cardiometabolic status. Additionally, dynamic changes in depressive symptoms over time were captured. However, this study still has several limitations. First, despite controlling for multiple covariates, residual confounding may remain (e.g., medication use, wealth/income, and comorbidities) because the existing data lacked additional relevant variables. Second, both the LCA and GBTM methods rely on statistical assumptions, and the classification of trajectories and clusters involves some degree of uncertainty. However, the results remained consistent across multiple sensitivity analyses. Third, the long follow-up period may have led to participant attrition due to death or withdrawal, resulting in a sample biased toward healthier survivors. This may lead to biased estimates of the main exposure–outcome associations. Finally, some subgroups had relatively small sample sizes, which may have limited statistical power and increased the risk of Type II errors. Further studies are needed to determine whether the associations observed in this selected sample hold in broader populations. Conclusions We found that obesity-hypertension was only associated with an increased odds of being on a moderately severe depressive symptoms trajectory, while the complex cardiometabolic pattern was associated with an increased odds of depressive symptoms trajectories. In contrast, sufficient PA and digital inclusion were associated with lower odds of belonging to these trajectories. In addition, PA and digital inclusion are associated with an attenuation of the adverse effects of the “obesity hypertension” and complex cardiometabolic pattern on depressive trajectories. The combined effects of clustering of cardiometabolic risk factors, PA, and digital inclusion underscore the substantial benefits of integrated strategies for reducing the burden of depressive symptoms among middle-aged and older adults. Supplementary Information Additional file 1. eMethods. Tables S1–S2. Physical activity definitions and MET values used in HRS and ELSA. Table S3–S4. Model fit and goodness-of-fit statistics for LCA and GBTM in HRS and ELSA. Tables S5–S8. Associations of cardiometabolic risk factor clustering, physical activity, and digital exclusion with depressive symptom trajectories under baseline models (Models 1 and 2), including stratified and joint effects. Tables S9–S12. Sensitivity analyses using U.S. criteria for abdominal obesity: associations and mediation effects of cardiometabolic risk, physical activity, and digital exclusion on depressive symptom trajectories. Tables S13–S16. Sensitivity analyses using inverse probability weighting: associations and mediation effects across cardiometabolic latent classes. Tables S17–S20. Analyses based on physical activity tertiles: associations and mediation effects involving cardiometabolic risk and digital exclusion. Tables S21–S24. Multiple imputation–based results: associations and mediation effects of cardiometabolic risk, physical activity, and digital exclusion with depressive symptom trajectories. Tables S25–S28. Analyses restricted to participants with complete baseline data (“same baseline”): associations and mediation effects. Figure S1. Study roadmap. Figures S2–S3. Mediation analysis diagrams for HRS and ELSA. Figures S4–S5. Predicted probabilities of cardiometabolic risk factors by latent classes from LCA, under multiple imputation and same-baseline scenarios. Figures S6–S7. Depressive symptom trajectories under multiple imputation and same-baseline scenarios. Additional file 1. eMethods. Tables S1–S2. Physical activity definitions and MET values used in HRS and ELSA. Table S3–S4. Model fit and goodness-of-fit statistics for LCA and GBTM in HRS and ELSA. Tables S5–S8. Associations of cardiometabolic risk factor clustering, physical activity, and digital exclusion with depressive symptom trajectories under baseline models (Models 1 and 2), including stratified and joint effects. Tables S9–S12. Sensitivity analyses using U.S. criteria for abdominal obesity: associations and mediation effects of cardiometabolic risk, physical activity, and digital exclusion on depressive symptom trajectories. Tables S13–S16. Sensitivity analyses using inverse probability weighting: associations and mediation effects across cardiometabolic latent classes. Tables S17–S20. Analyses based on physical activity tertiles: associations and mediation effects involving cardiometabolic risk and digital exclusion. Tables S21–S24. Multiple imputation–based results: associations and mediation effects of cardiometabolic risk, physical activity, and digital exclusion with depressive symptom trajectories. Tables S25–S28. Analyses restricted to participants with complete baseline data (“same baseline”): associations and mediation effects. Figure S1. Study roadmap. Figures S2–S3. Mediation analysis diagrams for HRS and ELSA. Figures S4–S5. Predicted probabilities of cardiometabolic risk factors by latent classes from LCA, under multiple imputation and same-baseline scenarios. Figures S6–S7. Depressive symptom trajectories under multiple imputation and same-baseline scenarios.