Skip to content

Socioeconomic status → cardiovascular disease

Background

Socioeconomic status (SES) is one of the most robust predictors of cardiovascular disease (CVD) across populations and settings. Lower SES is associated with higher CVD incidence and mortality, and this relationship persists after adjustment for conventional risk factors — suggesting both direct pathways (e.g., material deprivation, chronic stress) and indirect pathways through behavioral intermediaries.

This DAG represents a domain-level causal structure intended as a starting point for study-specific derivation. Researchers should add design-specific nodes (selection, measurement, timing) when deriving a study-specific DAG.

Causal structure

dag { SES [exposure, pos="0,1"] CVD [outcome, pos="4,1"] Age [pos="2,0"] Smoking [pos="1,2"] PhysicalActivity [pos="3,2"] StressPathways [pos="2,1"] SES -> CVD SES -> Smoking SES -> PhysicalActivity SES -> StressPathways Smoking -> CVD PhysicalActivity -> CVD StressPathways -> CVD Age -> SES Age -> CVD Age -> Smoking }

The key structural features are:

  • SES → CVD direct path: material deprivation, neighborhood environment, healthcare access
  • SES → Smoking → CVD: behavioral intermediary; note this is a mediator not a confounder if the research question is total effect of SES
  • SES → PhysicalActivity → CVD: behavioral intermediary
  • Age as a common cause of SES (life course accumulation) and CVD

Assumptions

Included:

  • Smoking and physical activity are on the causal pathway from SES to CVD (mediators), but they are also influenced by age independently. The adjustment decision depends on the estimand: total effect vs. direct effect of SES.
  • Age is placed as a cause of SES to capture life course SES accumulation. In cross-sectional data, this arrow is often omitted.

Intentionally omitted:

  • Race/ethnicity: a major modifier and upstream cause of SES in US contexts. Including it would require a separate DAG or a context-specific variant.
  • Diet, alcohol: downstream of SES and on the causal pathway; omitted for parsimony.
  • Healthcare access: mediator; can be added as explicit node if of interest.
  • Sex/gender: similar to race/ethnicity — important modifier, context-dependent.

Identification strategy

If the estimand is the total effect of SES on CVD:

  • Adjustment set: {Age} only
  • Smoking and PhysicalActivity are mediators — conditioning on them blocks part of the causal path

If the estimand is the direct effect of SES on CVD (not through behavioral intermediaries):

  • Adjustment set: {Age, Smoking, PhysicalActivity, StressPathways}
  • Note: adjusting for mediators introduces potential collider bias if the mediators share unmeasured common causes with CVD

In both cases, backdoor criterion is satisfied with the listed adjustment sets.

Known variants

US context with race/ethnicity: Add Race as a cause of SES and CVD. This changes the adjustment set and requires careful consideration of whether Race is a cause or a proxy for structural racism.

Life course approach: Split SES into childhood SES and adult SES, with ChildhoodSES → AdultSES and both affecting CVD. Changes interpretation substantially.

Mediator-as-confounder: If Smoking has independent causes that also affect CVD (unmeasured confounders), adjusting for Smoking in a mediation analysis introduces collider bias.

Open questions

  • The arrow Age → SES is contested. In some frameworks, Age is a proxy for cohort effects, not a cause.
  • The StressPathways node is highly aggregate. Disaggregating into allostatic load, cortisol dysregulation, and inflammatory markers may be warranted for mechanistic research.
  • Evidence level for SES → CVD direct path is moderate: while observational evidence is robust, the mechanism is not fully established.