Does socioeconomic and environmental burden affect vulnerability to extreme air pollution and heat? A case-crossover study of mortality in California

The study investigates whether socioeconomic and socioenvironmental burden modifies the mortality risk associated with extremes of heat and fine particulate air pollution (PM2.5), alone and in combination, in California. Climate change is increasing heat waves and affecting air quality, with established links between high temperatures, PM2.5, and increased cardiorespiratory mortality. Communities with lower socioeconomic status often face higher exposures and vulnerabilities to these hazards. The authors hypothesized that neighborhood-level socioeconomic (SDI) and socioenvironmental (CES) burden, as well as individual-level SES (education), could modify the mortality effects of extreme heat and PM2.5, particularly when both exposures co-occur.

Prior work shows separate associations of heat and PM2.5 with mortality and suggests potential interactions when co-occurring. Evidence on SES as an effect modifier is mixed: some studies report larger air pollution effects in lower-SES areas or among individuals with lower SES, while others find null results—often depending on whether SES is measured at the individual vs. area level. There is limited prior research on SES modification of the combined effects of heat and PM2.5. The authors’ earlier California study reported higher mortality risk from co-occurring extreme heat and PM2.5 than from either exposure alone. The current study extends this by evaluating modification by neighborhood socioenvironmental (CES), socioeconomic (SDI), and individual education. Background literature also notes environmental justice contexts, cumulative exposures, and mechanisms (higher exposure and higher vulnerability due to preexisting conditions) that could produce effect modification.

Design: Time-stratified case-crossover study of deaths in California from 2014-2019. For each decedent, the death date was the case day; control days were same day-of-week within the same month and year. This design controls for individual time-invariant factors, long-term trends, seasonality, and day of week. Population: 1,580,799 deaths obtained; excluded 51,557 missing residential census tract and 14,950 with out-of-state tracts, yielding 1,514,292 deaths. Outcomes: all-cause; cardiovascular (ICD-10 I10–I70); respiratory non-malignant (J00–J99). Demographics from death certificates included age, sex, race/ethnicity, and education. Exposure assessment: Daily PM2.5 estimated for census tract centroids using inverse distance-squared weighting of up to four monitors within 50 km (EPA AQS). Maximum daily temperature and relative humidity from a 4-km gridded reanalysis (University of Idaho). Exposures assigned by residential census tract for lag 0 (main) and lag 1 (secondary). Extreme exposure defined as above the 90th, 95th, or 97th percentile. For temperature, percentiles were tract-specific across 2014–2019; for PM2.5, percentiles were statewide across all tracts. Socioenvironmental/socioeconomic burden: Census tract-level CalEnviroScreen (CES, 0–100; higher indicates greater cumulative burden) combining socioeconomic and environmental components; Social Deprivation Index (SDI, 1–100; higher indicates greater deprivation). Healthy Places Index (HPI) used in sensitivity analyses. Individual-level SES proxied by educational attainment (with/without high school diploma) for decedents aged 25+. Exposure categorization: Four mutually exclusive daily categories: (i) extreme PM2.5 only; (ii) extreme heat only; (iii) both extreme heat and extreme PM2.5; (iv) neither (referent). All extremes within a model used the same percentile threshold. Statistical analysis: Conditional logistic regression adjusted for relative humidity (natural cubic spline, 3 df). Product terms tested effect measure modification by CES or SDI (quartiles for all-cause; quartiles for CV at 90th percentile; binary high/low for respiratory at 90th percentile). Main models used lag 0 exposure. Secondary analyses: effect modification by individual education; lag 1 exposure. Sensitivity analyses: CES socioeconomic and environmental components; HPI; continuous temperature and PM2.5 (summer months) with interaction, and product terms with CES/SDI quartiles. Analyses conducted in R. Observations missing exposure or humidity were excluded.

- Sample: 1,514,292 all-cause deaths (33% cardiovascular; 9% respiratory). Mean age at death 74.0 years. - Exposure distributions: Average max temperature on extreme heat-only days was 34.6°C; on co-exposure days 36.7°C. Average PM2.5 on extreme PM2.5-only days was 26.7 µg/m3; on co-exposure days 24.2 µg/m3. - Disparities in exposure frequency: Days with extreme PM2.5 were more common in higher-burden neighborhoods (e.g., 6% vs 12% of days in low vs high CES; 8% vs 11% in low vs high SDI for extreme PM2.5 at the 90th percentile threshold). - Mortality associations: Extreme heat alone, extreme PM2.5 alone, and especially their co-occurrence were each associated with increased odds of all-cause, cardiovascular, and respiratory mortality, consistent with prior findings; co-exposure had the strongest odds ratios. - Effect modification: No consistent evidence that neighborhood CES or SDI modified the mortality effects of extreme heat, extreme PM2.5, or their co-occurrence, across all-cause, cardiovascular, or respiratory outcomes. No modification by individual education. - Isolated signals: Some statistically significant positive modification in SDI quartile 3 for certain outcomes (e.g., all-cause and CV mortality for extreme heat alone; all-cause mortality for extreme PM2.5 alone) at the 90th percentile, but not consistent across quartiles or higher thresholds (95th/97th). - Sensitivity analyses: Results remained largely null for modification when using lag 1 exposure, CES component scores, HPI, or continuous exposures. Slight indication of higher PM2.5-associated mortality odds ratios in SDI quartiles 2–3 versus quartile 1 for continuous PM2.5, but not in quartile 4 and no modification for continuous heat or the heat–PM2.5 interaction.

The study addressed whether socioeconomic or socioenvironmental burden modifies mortality risks from extreme heat and PM2.5, alone and combined. While co-exposure increased mortality risk more than either exposure alone, the analysis did not find consistent modification by neighborhood CES or SDI, nor by individual education, across all-cause, cardiovascular, or respiratory mortality. These findings suggest that, within California during 2014–2019, increased vulnerability due to neighborhood-level socioenvironmental or socioeconomic burden did not systematically amplify short-term mortality risks from extreme heat and PM2.5 beyond the observed main effects. However, higher-burden neighborhoods experienced more frequent extreme PM2.5 days, highlighting environmental injustice in exposure frequency. Potential explanations for null modification include limitations of area-level indices for short-term susceptibility, crude individual SES measure (education dichotomy), unmeasured factors like air conditioning access and usage, and power constraints for detecting interaction effects with relatively rare extreme-exposure days. The results align with some literature reporting null SES modification at area level, but contrast with other studies showing larger effects in lower-SES settings, underscoring contextual variability and measurement differences.

Co-exposure to extreme heat and PM2.5 was associated with increased mortality risk and was stronger than either exposure alone. Nevertheless, there was no consistent evidence that neighborhood socioenvironmental (CES) or socioeconomic (SDI) burden, or individual education, modified these mortality associations in California (2014–2019). High-burden neighborhoods did experience more days with extreme PM2.5. Future work should examine finer-grained and up-to-date SES and environmental metrics, incorporate individual-level vulnerability factors (e.g., comorbidities, AC access), study broader geographies or longer periods to increase power, and further investigate modification of joint exposure effects.

- Exposure assessment at residential census tract using monitor-based interpolation and gridded meteorology may introduce exposure measurement error; non-residential exposures (work, transit) not captured. - Lack of data on air conditioning (AC) access/use, an important modifier of heat and PM2.5 exposure and vulnerability. - Exclusion of decedents without residential tract (e.g., people experiencing homelessness) may omit a highly vulnerable group. - Individual SES measured only by dichotomized educational attainment, a crude proxy that may not capture SES complexity. - Focus on extreme exposures limits exposed-day prevalence (e.g., ≤10% at 90th percentile), reducing power for interaction tests. - Generalizability may be limited outside California or to markedly different climates or sociodemographic contexts. - Area-level burden indices (CES, SDI, HPI) based on multi-year aggregates may be suboptimal for short-term effect modification analyses.