Educational inequalities in heat-related mortality across Belgian districts

Technical Summary: Educational Inequalities in Heat-Related Mortality Across Belgian Districts

Problem Statement
Climate change is intensifying the frequency and severity of extreme heat events, creating disproportionate health risks for vulnerable populations. While socioeconomic position is known to shape exposure and adaptive capacity, evidence regarding the specific role of educational attainment in modifying heat-related mortality remains limited, particularly at the individual level. Previous studies often rely on area-level proxies, leaving a gap in understanding how individual circumstances versus geographic context drive vulnerability. This study addresses the need to quantify educational gradients in heat vulnerability among older adults (aged 60+) across the 42 NUTS-3 districts of Belgium, a region characterized by well-documented socioeconomic and spatial health inequalities.

Methodology
The study utilizes a comprehensive dataset linking Belgian mortality records (2002–2019) with individual educational attainment derived from the 2001 census. The analysis focuses on adults aged 60 and over, restricted to a closed cohort to ensure consistent exposure classification.

• Data Integration:

  • Mortality: 1,392,134 deaths with known educational attainment were analyzed, categorized into three groups: Low (no formal/primary/lower secondary), Secondary (upper secondary/post-secondary non-tertiary), and Tertiary (Bachelor's and above).
  • Temperature: Daily maximum temperatures (1992–2022) from the Royal Meteorological Institute of Belgium were spatially joined to NUTS-3 district boundaries.
  • Contextual Variables: District-level deprivation scores from the Belgian Index of Multiple Deprivation were compiled as meta-predictors.

• Analytical Framework:
  The authors applied a two-stage distributed lag non-linear modelling (DLNM) framework:

  1. First Stage: Non-linear and lagged temperature-mortality associations were estimated separately for each of the 126 district-education combinations (42 districts × 3 education levels) using DLNMs with a maximum lag of seven days.
  2. Second Stage: A multivariate meta-regression pooled these estimates to produce population-average exposure-response functions by educational group. The framework utilized Best Linear Unbiased Predictions (BLUPs) for location-specific refinements and explicitly propagated first-stage uncertainty.

• Definitions and Metrics:

  • Extreme Heat: Defined as days exceeding the 97.5th national temperature percentile (28.3°C).
  • Minimum Mortality Temperature (MMT): A common MMT of 21.0°C (77th national percentile) was established as a shared centering point to ensure comparability across groups, despite education-specific MMTs varying from 15.8°C to 21.4°C.
  • Outcome: Age-standardised heat-related mortality rates (ASMR) per 100,000 person-years were computed using direct standardisation against the 2011 Belgian census age distribution. Uncertainty was quantified via Monte Carlo simulation (1,000 iterations).

Key Results
The analysis reveals a clear and consistent educational gradient in extreme heat vulnerability across all Belgian districts:

• Relative Risks (RR): At the 97.5th percentile (28.3°C), the RR for heat-related mortality was 1.12 (95% CI: 1.10–1.15) for the low education group, 1.11 (1.08–1.14) for secondary, and 1.10 (1.05–1.16) for tertiary. The gradient widens significantly at more extreme temperatures (99th percentile, 31.3°C), where the RR for the low education group rises to 1.28 (1.24–1.31), compared to 1.16 (1.09–1.22) for the tertiary group.
• Mortality Rates: Age-standardised heat-related mortality rates during extreme heat days were 26.5 per 100,000 person-years for the low education group, 19.4 for secondary, and 13.0 for tertiary. The confidence intervals for the low education group do not overlap with the two higher groups, indicating a statistically significant disparity.
• Spatial Consistency: Between-district heterogeneity was low ($I^2 = 8.4\%$), suggesting the educational gradient is a national phenomenon rather than being driven by specific outlier districts. However, absolute inequalities were most pronounced in the west (Tournai-Mouscron) and south (Virton, Arlon).
• Mechanism: Spline decomposition indicated that the protective effect of education becomes statistically significant only at the highest temperatures (coefficient $b_5$), implying that higher education mitigates risk specifically during extreme heat events.

Significance and Claims
The paper claims that educational attainment is a significant modifier of extreme heat vulnerability among older adults in Belgium, operating primarily through individual-level pathways (such as health literacy, thermoregulatory behavior, and access to resources) rather than geographic context alone.

• Differentiation from Demographics: The persistence of the gradient after age standardisation demonstrates that the disparity is not merely a demographic artifact of the older age structure within lower-educated groups, but reflects genuine differences in heat susceptibility.
• Policy Implications: As climate change increases the frequency of days exceeding current extreme thresholds, the unequal burden documented is likely to intensify. The authors argue that this underscores the necessity for targeted heat policies that address individual-level vulnerability, rather than relying solely on geographic or area-level interventions.
• Limitations: The authors modestly note that the study cannot fully disentangle the effects of education from frailty or underlying health status, as lower-educated individuals tend to be older and have higher baseline mortality. Additionally, the analysis is scoped to heat-related mortality due to the short lag structure of heat effects, precluding reliable estimation of cold-related mortality gradients.

In conclusion, the study provides robust evidence that in the context of a warming climate, lower educational attainment acts as a critical amplifier of heat-related mortality risk, creating a cumulative disadvantage for older, less-educated populations in Belgium.