Effect modification and interaction between ethnicity and socioeconomic factors in severe COVID-19: analyses of linked national data for Scotland

Objective Minority ethnic groups disproportionately experienced adverse COVID-19 outcomes, partly a consequence of disproportionate exposure to socioeconomic disadvantage and high-risk occupations. We examined whether minority ethnic groups were also disproportionately vulnerable to the consequences of socioeconomic disadvantage and high-risk occupations in Scotland.

Design We investigated effect modification and interaction between area deprivation, education and occupational risk and ethnicity (assessed as both a binary white vs non-white variable and a multi-category variable) in relation to severe COVID-19 (hospitalisation or death). We used electronic health records linked to the 2011 census and Cox proportional hazards models, adjusting for age, sex and health board. We were principally concerned with additive interactions as a measure of vulnerability, estimated as the relative excess risk due to interaction (RERI).

Results Analyses considered 3 730 837 individuals aged ≥16 years (with narrower age ranges for analyses focused on education and occupation). Severe COVID-19 risk was typically higher for minority ethnic groups and disadvantaged socioeconomic groups, but additive interactions were not consistent. For example, non-white ethnicity and highest deprivation level experienced elevated risk ((HR=2.7, 95% CI: 2.4, 3.2) compared with the white least deprived group. Additive interaction was not present (RERI=−0.1, 95% CI: −0.4, 0.2), this risk being less than the sum of risks of white ethnicity/highest deprivation level (HR=2.4, 95% CI: 2.3, 2.5) and non-white ethnicity/lowest deprivation level (1.4, 95% CI: 1.2, 1.7). Similarly, non-white ethnicity/no degree education (HR=2.5, 95% CI: 2.2, 2.7; RERI=−0.1, 95% CI: −0.4, 0.2) and non-white ethnicity/high-risk occupation (RERI=0.3, 95% CI: −0.2, 0.8) did not experience greater than additive risk. No clear evidence of effect modification was identified when using the multicategory ethnicity variable or on the multiplicative scale either.

Conclusion We found no definitive evidence that minority ethnic groups were more vulnerable to the effect of social disadvantage on the risk of severe COVID-19.