Clustering of Social Determinants of Health and Their Association With Adverse Cardiovascular Outcomes in Atrial Fibrillation

Abstract

Background

Individuals with atrial fibrillation remain at high risk for major adverse cardiovascular events despite advances in treatment. Social determinants of health (SDOH) are associated with cardiovascular outcomes, yet how these factors cluster to create distinct vulnerability profiles remains poorly understood.

Methods

Using UK Biobank data, we examined 15 SDOH indicators across economic, psychological, and neighborhood domains among 3842 participants with atrial fibrillation (35.1% female). Multigroup latent class analysis evaluated sex‐based differences in vulnerability pattern distribution. Cox proportional hazards models assessed the association between identified SDOH clusters and composite cardiovascular outcomes comprising major adverse cardiovascular events and all‐cause mortality.

Results

Five distinct SDOH vulnerability clusters were identified: (1) low vulnerability across all domains, (2) primarily economic vulnerability, (3) primarily neighborhood‐related vulnerability, (4) economic and neighborhood vulnerability with favorable psychological conditions, and (5) high vulnerability across all domains. Men were more represented in advantaged profiles than women (Cluster 1: 37.0% versus 26.6%; Cluster 4: 25.3% versus 14.1%), and women showed the greatest representation in the highest vulnerability cluster (Cluster 5: 35.1% versus 25.2%). Compared with Cluster 1, Cluster 2 had a 30% higher risk (hazard ratio [HR], 1.30 [95% CI, 1.07–1.58]), Cluster 4 had 37% increased risk (HR, 1.37 [95% CI, 1.13–1.68]), and Cluster 5 more than twice the risk (HR, 2.05 [95% CI, 1.67–2.52]).

Conclusions

Five distinct SDOH vulnerability patterns were identified among individuals with atrial fibrillation, with economic determinants as pivotal drivers of cardiovascular risk. These findings support incorporating comprehensive SDOH assessment into atrial fibrillation clinical management.

Nonstandard Abbreviations and Acronyms

LCA

latent class analysis

MACE

major adverse cardiovascular events

SDOH

social determinants of health

Research Perspective.

What Is New?

• In a UK Biobank prospective cohort of 3842 individuals with atrial fibrillation, multigroup latent class analysis identified 5 distinct social determinants of health vulnerability profiles spanning low, economic, neighborhood, combined, and high‐vulnerability patterns, with economic disadvantage emerging as the predominant driver of elevated cardiovascular risk.

• Sex differences manifested as distributional rather than structural variations: women were overrepresented in higher vulnerability clusters, and the social determinants of health–outcome association followed a graduated risk gradient in men but a threshold pattern in women, in whom only the highest combined‐vulnerability profile reached statistical significance.

What Question Should Be Addressed Next?

• Future research should investigate whether longitudinal trajectories of social determinants of health and sex‐tailored interventions targeting economic vulnerability—including financial support, medication‐cost assistance, and broader structural and policy reforms— an mitigate cardiovascular event rates among individuals with atrial fibrillation across diverse health care systems.

Despite widespread use of anticoagulation therapy and improvement in both screening and treatment strategies, individuals with atrial fibrillation (AF) ¹ still experience high rates of major adverse cardiovascular events (MACE) such as stroke, transient ischemic attacks, arterial thromboembolic events, myocardial infarctions, and cardiovascular mortality. ² , ³ , ⁴ Although traditional clinical risk factors for MACE in participants with AF are well established, growing evidence suggests that social determinants of health (SDOH) play a crucial role in cardiovascular outcomes. ⁵ , ⁶ SDOH, defined as the conditions in which people are born, grow, live, work, and age, ⁷ , ⁸ have been shown to influence various aspects of AF care, including treatment adherence, health care use, and clinical outcomes. ⁹

Although there is strong evidence supporting SDOH determinants (eg, income, education, social support) as upstream drivers, ¹⁰ , ¹¹ , ¹² , ¹³ less attention has been paid to how SDOH intersect to create distinct patterns of vulnerability. In fact, SDOH factors rarely manifest in isolation. Instead, they tend to demonstrate co‐occurrence patterns and potentially interact in ways that may amplify their impact on health outcomes. ¹⁴ , ¹⁵ This phenomenon is particularly relevant for individuals with chronic disease, who frequently face multiple social and economic challenges that may influence their capacity for effective disease self‐management and health care engagement. ⁵ , ¹⁶

In the present study, we used multigroup latent class analysis (LCA) models ¹⁷ to identify distinct patterns of SDOH vulnerability among individuals with AF and examine their associations with MACE outcomes and sex differences. This method is particularly relevant for studying the impact of SDOH, as they tend to cluster and interact in complex ways that may not be apparent when examined individually. This approach is therefore useful for revealing underrecognized mechanisms that might be targeted with precision‐based social intervention strategies for reducing cardiovascular risk.

METHODS

Data Availability

The data used in this work were obtained from the UK Biobank (Application Number: 45551), which is a large‐scale public health research database. The data that support the findings are available from the corresponding author upon reasonable request.

Source and Study Population

We conducted a prospective cohort study using data from the UK Biobank, a population‐based cohort of ∼500 000 participants aged 40 to 69 years, recruited between 2006 and 2010 across the United Kingdom. This study is reported in accordance with the Strengthening the Reporting of Observational Studies in Epidemiology statement for observational research. ¹⁸ Participants were evaluated at dedicated centers across the United Kingdom, where both self‐reported and electronic health record data were collected and subsequently monitored during follow‐up. ¹⁹ For this analysis, participants with a diagnosis of AF were identified through hospital admission records (International Classification of Diseases, Tenth Revision [ICD‐10] code I48) and self‐reported medical history at baseline. To maintain temporal proximity between SDOH assessments at enrollment and AF diagnosis, we restricted inclusion to participants diagnosed with AF within 2 years before or after enrollment. The UK Biobank study received approval from the Northwest Multi‐Centre Research Ethics Committee (11/NW/0382) (https://www.ukbiobank.ac.uk/learn‐more‐about‐uk‐biobank/about‐us/ethics). Participants provided informed consent at baseline.

The selection of the study population is summarized in Figure 1. Briefly, of the ~500 000 UK Biobank participants, participants were identified with an AF diagnosis (ICD‐10: I48) or self‐reported history. After applying the ±2‐year enrollment window and excluding those with outcome events before AF diagnosis, 3842 participants were included in the final study sample.

Figure 1. Consolidated Standards of Reporting Trials participant flow diagram.

Selection of the study population from the UK Biobank cohort. Of ∼500 000 UK Biobank participants, individuals with an atrial fibrillation diagnosis (International Classification of Diseases, Tenth Revision [ICD‐10] code I48) or self‐reported history were identified. AF indicates atrial fibrillation; and UKBB, UK Biobank.

Definition of SDOH Variables

SDOH variables selected for the LCA to identify clusters were based on availability in the UK Biobank database. ²⁰ , ²¹ Fifteen SDOH were included and categorized across 3 domains: (1) economic factors including household income, highest level of education attained, employment status (employed versus unemployed or retired) and accommodation ownership; (2) psychological factors (ie, living alone, social isolation, social support, emotional distress, and social inactivity; and (3) neighborhood characteristics (ie, area education quality level, material deprivation, housing quality, crime rate, greenspace remoteness, and blue space remoteness). Each SDOH indicator was dichotomized as 0 (favorable/reference) or 1 (disadvantaged); detailed cutoff thresholds and operational definitions for all 15 indicators are provided in Tables S1–S3. For missing values, LCA applies full‐information maximum likelihood estimation, which derives parameter estimates while retaining all available data and performs well under the missing at random assumption. SDOH variables had missing rates ranging from <1% to ∼8% across indicators. ²²

Covariates

Baseline demographic, clinical, and lifestyle variables were included as covariates. Demographic characteristics in this study refer to age, sex, and self‐reported race. ²³ , ²⁴ , ²⁵ Clinical factors included a history of heart failure, hypertension, diabetes, chronic kidney disease stages 3 or more, chronic obstructive pulmonary disease, anticoagulant use, body mass index, waist circumference, and high‐density lipoprotein cholesterol. In addition, lifestyle variables included smoking history and alcohol consumption.

Outcomes

The primary outcome was a composite of first occurrence of MACE (ie, nonfatal myocardial infarction, nonfatal stroke, and cardiovascular death) and all‐cause mortality. Secondary outcomes included the individual components of MACE, and noncardiovascular mortality, separately. Time‐to‐event was defined as the interval between AF diagnosis and the first occurrence of the outcome. To ensure appropriate temporal sequencing and minimize immortal time bias, participants who experienced outcome events at or before the AF diagnosis were excluded. Participants were followed from the date of AF diagnosis until the first occurrence of MACE, death, loss to follow‐up, or the study end date (December 31, 2022).

Statistical Analysis

LCA was employed to identify distinct patterns of SDOH vulnerability among AF participants. ¹⁷ Observed SDOH indicators were used to determine the most meaningful clusters. ²⁶ Rather than generating phenotypes, LCA identifies homogeneous subgroups. ²⁷ Given established evidence demonstrating significant sex‐based variations in SDOH distributions, health behaviors, and cardiovascular risk profiles, ²⁸ , ²⁹ , ³⁰ we employed multigroup LCA with self‐reported biological sex as the grouping variable. Formal a priori power calculations are not standard for LCA, as the technique is exploratory and data driven. However, the adequacy of our sample size (N=3842) was considered in the context of published simulation studies, which generally recommend a minimum of 500 participants for stable 5‐class LCA solutions and a minimum class size of ∼5% of the total sample. ³¹

Multigroup LCA with sex as the grouping variable examined sex‐specific distributional differences while testing measurement invariance across groups. Model selection incorporated multiple statistical criteria including Bayesian information criterion, with optimal class enumeration determined through systematic evaluation of 2 to 6 class solutions. Detailed LCA methodology, including model selection criteria, measurement invariance testing procedures, and technical specifications are provided in Data S1. Cox proportional hazards regression models evaluated associations between identified vulnerability clusters and the composite cardiovascular outcomes. Models were adjusted for covariates. Proportional hazards assumptions were verified using scaled Schoenfeld residuals. We conceptualize the identified SDOH clusters as upstream composite social exposure profiles rather than as simple confounders. Adjusted hazard ratios (HRs) from the Cox models should therefore be interpreted as the association of these upstream profiles with cardiovascular outcomes after adjustment for demographic, clinical, behavioral, and metabolic covariates. The analysis is not designed to disentangle direct effects from pathways mediated through clinical or metabolic factors such as obesity, waist circumference, or dyslipidemia, which are treated as covariates rather than mediators.

Statistical analyses were conducted using RStudio, poLCA statistical software package version 1.6.0.1, ²⁸ glca statistical software package version 1.4.2. ²⁹ Statistical significance was defined as P<0.05, with results presented as hazard ratios and 95% CIs.

RESULTS

Study Population Characteristics

The study cohort comprised 3842 participants with AF, of whom 35.1% were female, with a mean age of 62.5 years (SD=6.1) and a median follow‐up of 11.2 years (interquartile range, 9.5–12.6) (Table). Men had a higher prevalence of diabetes; no significant differences were found in other comorbidities. Lifestyle factors also differed by sex, with men reporting higher rates of smoking and alcohol consumption. A higher proportion of men experienced the composite outcome (29.1%) compared with women (21.3%).

Table 1.

Percentage Distributions of Characteristics and Outcomes of Participants by Latent Class

Variables	Overall (N=3842)	Cluster 1 (N=972)	Cluster 2 (N=1023)	Cluster 3 (N=475)	Cluster 4 (N=831)	Cluster 5 (N=541)
Demographics (%)
Age (≥65 y)	44.8	30	59.1	28.8	55.2	42.5
Male sex	64.8	72.5	58.6	69.5	62.5	62.3
Racial or ethnic minority groups*	3.3	1.6	1.8	7.2	2.9	6.5
Comorbidities (%)
Heart failure	3.0	2	2.4	4	3.2	5
Hypertension	25.7	17.5	28.4	19.6	30.2	33.8
Vascular diseases	3.0	1.9	3.2	1.9	2.8	6.1
Diabetes	5.9	3.9	5.5	6.3	6.5	9.1
Chronic obstructive pulmonary disease	5.5	1.6	4.7	3.4	6.7	14
Chronic kidney disease	3.4	1.9	3.7	2.1	4.3	5.2
Dyslipidemia	43.7	43.6	41.8	41.8	44.4	47.7
Abdominal obesity†	65.8	62.6	62.9	63.9	68.3	73.9
Overweight	41.5	44.7	42.0	42.7	42.3	33.4
Obesity	36.3	31.6	32.7	31.1	40.1	46.9
Lifestyle factors (%)
Smoking	54.7	51	50.5	52.2	57.2	66
Alcohol	47.7	59.9	47.2	51.2	40.2	35.5

Cluster description: Cluster 1: Low vulnerable across all domains; Cluster 2: Primarily economic vulnerability; Cluster 3: Predominantly neighborhood‐related vulnerability; Cluster 4: Economic and neighborhood vulnerability but favorable psychological conditions; Cluster 5: High vulnerability across all domains.

^*Racial or ethnic minority groups includes the following self‐reported races or ethnicities: Asian or Asian British, Black or Black British, Mixed, and Other ethnic group.

^† Body mass index categories: overweight=25.0–29.9 kg/m²; obesity=≥30 kg/m². Abdominal obesity defined as waist‐to‐hip ratio >0.90 in men and >0.85 in women.

Sex differences in distribution across SDOH domains were observed. Female participants demonstrated a higher prevalence of economic disadvantage compared with their male counterparts (low income: 71.3% versus 59.2%; unemployment: 70.3% versus 60.2%) and psychological vulnerability indicators (living alone: 24.2% versus 19.0%; emotional distress: 53.3% versus 49.4%). Conversely, male participants exhibited greater social isolation (23.7% versus 14.3%) and reduced social support (29.0% versus 23.9%) compared with female participants. Neighborhood‐level determinants demonstrated minimal sex difference, suggesting structural environmental factors operate more uniformly across sex groups.

Identification of Latent Classes of SDOH

Measurement invariance testing favored the partially invariant model, which showed superior fit (Table S2). This observation supported the hypothesis that equivalent vulnerability constructs existed across sexes while accounting for significant differences in distribution across clusters. We selected a 5‐class approach because it appeared to fit the data best.

The distribution of SDOH indicators for each cluster is shown in Figure 2. Specifically, cluster 1 (“low vulnerable across all domains” n=972) showed the most favorable profile with no low‐income participants and relatively low rates of psychological and neighborhood deleterious factors. Cluster 2 (“primarily economic vulnerability” n=1023) presented a unique pattern of lower household income (100%) and high unemployment (80.8%), but with relatively better performance in neighborhood factors and psychological stress levels. Cluster 3 (“primarily neighborhood‐related vulnerability” n=475) was distinguished by major neighborhood disadvantages including high material deprivation (92.6%) and higher crime rates (86.9%), despite moderate economic factors related vulnerabilities. Cluster 4 (“economic and neighborhood vulnerability but favorable psychological conditions” n=831) demonstrated high rates of poverty (92.7%) and poor neighborhood conditions, but relatively better psychological wellbeing with the lowest social isolation (12.3%). Finally, cluster 5 (“high vulnerability across all domains” n=541) emerged as the most disadvantaged group, characterized by higher rates of low income (99.6%), the highest rates of emotional distress (70.1%), social isolation (living alone 61.2%), and severe material deprivation (99.1%), despite having the highest rate of home ownership (61.9%).

Figure 2. Distribution of latent class and conditional probability from single LCA.

Heatmap displaying the prevalence (%) of each of the 15 social determinants of health indicators within each latent class, identified by partially invariant multigroup LCA among 3842 participants with atrial fibrillation. Indicators are grouped by domain (economic, psychological, neighborhood). LCA indicates latent class analysis.

Notable sex differences in vulnerability patterns were identified (Figure S2, Table S3). Male participants demonstrated significantly higher representation in advantaged social profiles than female participants (Cluster 1: 28.3% versus 19.7%), suggesting greater likelihood of favorable SDOH. Economic‐only vulnerability (Cluster 2) showed the opposite trend, with a higher percentage of women (31.4%) compared with men (24.1%). Of note, this class accounts for the largest proportion of women. Cluster 3 had the lowest representation in the total population, though slightly including more men (13.3%) than women (10.7%). Cluster 4 and Cluster 5 had relatively similar distributions between men and women, though women were slightly more represented in Cluster 4 (23.1% versus 20.8%) and men slightly more in Cluster 5 (15.1% versus 13.5%).

Association Between Clusters and Outcomes

Over a median follow‐up of 11.2 years (interquartile range, 9.5–12.6), 1012 composite outcome events occurred among 3842 participants (total: 38741 person‐years; crude incidence rate: 26.1 per 1000 person‐years). Crude incidence rates increased with SDOH vulnerability burden, ranging from 17.4 per 1000 person‐years in Cluster 1 (low vulnerability) to 43.0 per 1000 person‐years in Cluster 5 (high vulnerability). LCA was further performed separately within each sex as a sensitivity analysis.

The association between the composite outcome and SDOH clusters were assessed. In a Cox proportional hazards analysis fully adjusted for clinical and behavioral variables, using Cluster 1 as comparison group, participants in Cluster 2 showed a 30% higher risk of the composite outcome (HR, 1.30 [95% CI, 1.07–1.58]), whereas Cluster 4 demonstrated a 37% increased risk (HR, 1.37 [95% CI, 1.13–1.68]). The most vulnerable group (Cluster 5) exhibited the strongest association, with more than doubled risk of MACE (HR, 2.05 [95% CI, 1.67–2.52]). However, Cluster 3 showed no significant difference in outcome risk compared with Cluster 1 (HR, 1.04 [95% CI, 0.81–1.34]). Kaplan–Meier survival curves for the composite outcome, stratified by SDOH cluster, are presented in Figure 3.

Figure 3. Kaplan–Meier curve: composite outcome (MACE+all‐cause mortality).

Event‐free survival from the date of atrial fibrillation diagnosis through December 31, 2022, for the composite outcome of MACE and all‐cause mortality. The 5 curves correspond to the SDOH vulnerability clusters identified by multigroup latent class analysis. MACE indicates major adverse cardiovascular events; and SDOH, social determinants of health.

Sex‐stratified analysis revealed distinct SDOH‐outcome association patterns (Figure 4). Female participants demonstrated a threshold effect pattern, with increased cardiovascular risk only for Cluster 5 (HR_female, 1.84 [95% CI, 1.22–2.76]). In contrast, male participants exhibited a graduated risk pattern with significant elevation across vulnerability configurations: Cluster 2 (HR_male, 1.37 [95% CI, 1.09–1.71]) and Cluster 4 (HR_male, 1.36 [95% CI, 1.08–1.71]) demonstrated comparable risk magnitudes, and Cluster 5 yielded the highest risk (HR_male, 2.14 [95% CI, 1.69–2.72]).

Figure 4. Forest plot for identified SDOH cluster associations with composite cardiovascular outcome by sex.

Forest plot of adjusted HRs and 95% CIs from sex‐stratified Cox proportional hazards models, with Cluster 1 (low vulnerability) as the reference. HR indicates hazard ratio; and SDOH, social determinants of health.

Several sensitivity analyses were conducted to assess the robustness of the main findings. First, LCA performed separately within each sex yielded 4 classes among men and 3 among women; notably, the sex‐specific LCA did not yield a low‐vulnerability class for women. This pattern reflects how the classes grouped under our specific modeling choices and should not be read as evidence of a universally elevated SDOH burden in women; further work in independent cohorts is needed to confirm the observation. The SDOH profiles from these sex‐specific models closely resembled those derived from the main multigroup model, supporting the validity of the partially invariant multigroup LCA solution (Figures S1–S3). Second, analyses of individual outcome components—cardiovascular mortality, nonfatal stroke, nonfatal myocardial infarction, and noncardiovascular mortality—demonstrated consistent risk hierarchies across clusters, confirming the robustness of the vulnerability‐outcome associations observed for the composite end point (Figures S3–S6).

DISCUSSION

This study of 3842 individuals with AF applied multigroup LCA to identify SDOH clusters and their associations with MACE and all cause death. Three main findings emerged. First, SDOH cluster into 5 distinct vulnerability patterns among individuals with AF, spanning low, economic, neighborhood, combined economic and neighborhood, and high vulnerability across all domains. Second, although the qualitative structure of these clusters was consistent across sexes, their distribution differed: women were overrepresented in higher vulnerability clusters, and sex‐specific LCA did not identify a typical low‐vulnerability class among women, suggesting that SDOH burden is universally elevated in women with AF. Third, the association between SDOH vulnerability and adverse outcomes was more pervasive among men, who showed a graduated risk pattern across multiple vulnerability classes, whereas among women only the highest combined vulnerability class reached significance, consistent with a threshold effect in which co‐occurring burdens across domains are required to elevate cardiovascular risk in women.

Although our findings highlight sex differences in the clustering of social determinants in individuals with AF, economic disadvantage represents the predominant social factor associated with cardiovascular outcomes after adjusting demographic and clinical covariates. Although various SDOH contribute to cardiovascular risk, our results point to the substantial impact of economic vulnerability. This is evidenced by the LCA where all classes with severe economic disadvantage (Clusters 2, 4, and 5) showed significantly increased composite outcome risk. These findings align with previous research highlighting the fundamental role of economic status in cardiovascular health. ³⁰ , ³² For instance, findings from the PURE (Prospective Urban Rural Epidemiologic) study, a large international prospective cohort study, showed that participants with low income and low education are up to 2 times more likely to experience MACE and 2‐ to 3‐fold more likely to experience cardiovascular disease‐associated mortality, relative to those with favorable socioeconomic profiles. ³³ The impact of economic vulnerability on cardiovascular outcomes likely operates through multiple pathways. First, economic constraints may affect medication adherence, particularly for expensive anticoagulants and antiarrhythmics that are crucial for stroke prevention in individuals with AF. ⁹ Second, despite universal health care coverage in the United Kingdom, individuals' economic barriers still affect health care use through indirect mechanisms such as transportation costs to appointments, inability to take time off from hourly paid work, digital exclusion limiting access to remote monitoring services, and challenges navigating complex health care systems without adequate support. ³⁴ Third, economic hardship may lead to chronic stress and adoption of adverse health behaviors, such as smoking and poor dietary choices, which were more prevalent in our economically vulnerable groups. ³⁰ Additionally, economic status is associated with education, health literacy, and self‐management capabilities, which are essential for successful AF care. ⁹ , ³⁵ , ³⁶

Interestingly, our finding that isolated neighborhood disadvantage (Cluster 3) was not linked to cardiovascular outcomes differs from what has been reported in some established social epidemiology studies. ³² For instance, a study from United States found that residents of disadvantaged neighborhoods had a higher risk of coronary heart disease, even after adjusting for personal income, education, and occupation, ³⁷ and another Canadian study demonstrated associations between neighborhood deprivation and cardiovascular mortality. ³⁴ Several factors might explain the apparent discrepancy between our findings and previous research. First, participants with better economic resources might be able to overcome neighborhood limitations through better access to transportation and health care services outside their immediate area. ³⁸ Second, they may have greater capacity to modify their home environment or more likely to engage in healthier behaviors to mitigate neighborhood disadvantages. ¹⁶ Third, the protective effects of economic stability might buffer against the negative impacts of poor neighborhood conditions. ³⁹ , ⁴⁰ These findings may suggest that while neighborhood factors contribute to cardiovascular risk, their impact might be less direct or more modifiable than individual‐level economic factors among individuals with AF.

Multigroup LCA revealed vulnerability constructs maintain identical structural characteristics across sexes while accommodating significant distributional prevalence differences. This methodological achievement enables meaningful cross‐group comparisons while ensuring construct validity, essential prerequisites for evidence‐based intervention development and health policy formulation. The finding that sex‐based heterogeneity manifests primarily through distributional variations provides important insights regarding SDOH operational mechanisms. It aligns with previous view that different patterns exist in lifestyle, health beliefs, and behaviors among men and women. ⁴¹ These findings underscore the importance of considering sex‐specific factors when designing health interventions and support systems. ⁴² For instance, interventions aimed at enhancing resilience and health managing might need to be tailored differently for men and women to be effective.

Strengths and Limitations

This study is novel as it applies an intersectional and multidimensional‐based approach to explore the interaction between SDOH, sex and cardiovascular outcomes in a large sample of participants with AF. This study highlights the ways in which SDOH may exert detrimental effects on outcomes that differ by sex. Partial measurement invariance supports comparability of the latent vulnerability constructs across sexes but does not establish strict item‐level equivalence. The social meaning and lived experience of individual indicators (eg, living alone, unemployment) may differ between men and women, and cross‐sex comparisons in the present study should therefore be interpreted at the level of latent constructs rather than as evidence of identical item‐level meaning across groups.

However, our findings should be interpreted with caution considering several limitations. First, SDOH were measured at a single time point (baseline enrollment), which limits our ability to capture changes in social circumstances over time and to understand the cumulative effects of long‐term social disadvantage on cardiovascular risk. Second, the direction of causality cannot be definitively established, as declining health might lead to changes in social and economic circumstances (such as unemployment or social isolation), ⁴³ potentially creating bidirectional relationships between SDOH and health outcomes. Third, although we adjusted for multiple clinical factors, unmeasured confounding may still exist. For example, health literacy and health insurance usage were not directly measured but could be associated with both SDOH patterns and cardiovascular outcomes. ³⁵ In addition, as our study was conducted in the United Kingdom with a universally inclusive health care system, our findings may not be generalizable to individuals with AF in different health care systems or social contexts. Furthermore, UK Biobank participants are known to be generally healthier and more educated than the UK population, with implicit selection cohort biases. ⁴⁴ Future research would be enhanced by longitudinal assessment of SDOH patterns, more detailed measurement of social circumstances, and validation in diverse populations.

CONCLUSIONS

In conclusion, this study demonstrates that SDOH cluster into 5 distinct vulnerability profiles among individuals with AF, with distributional differences by sex. The association between SDOH vulnerability and adverse cardiovascular outcomes was more pervasive among men, who exhibited elevated risk across multiple vulnerability classes, whereas women showed significant risk elevation only at the highest combined vulnerability threshold. These findings highlight the need for sex‐informed SDOH screening in AF care, with economic vulnerability representing a priority target for intervention across both sexes.

Sources of Funding

This project was funded by a Canadian Institute of Health Research grant # PJT‐178237.

Disclosures

None.

Supporting information

Tables S1–S3

Figures S1–S6

References 26, 45–47