Time varying association between deprivation, ethnicity and SARS-CoV-2 infections in England: a space-time study

Tullia Padellini, Radka Jersakova, Peter J. Diggle, Chris Holmes, Ruairidh King, Brieuc Lehmann, Ann‐Marie Mallon, George Nicholson, Sylvia Richardson, Marta BlangiardomedRxiv 2021 · DOI

Abstract

Early evidence from the COVID-19 pandemic highlighted significant disparities in infection rates among different socio-economic and ethnic groups in England. However, these associations were often treated as static snapshots. This study employs a Bayesian spatio-temporal model to analyze how the impact of deprivation and ethnicity on SARS-CoV-2 infection rates evolved between June 2020 and September 2021. By accounting for ascertainment bias and spatial correlation, the research reveals that while deprivation was a primary driver early in the pandemic, ethnic composition—particularly South Asian heritage—became a more prominent risk factor during the Alpha and Delta variant surges.

Introduction

The COVID-19 pandemic did not affect all populations equally. In England, it became rapidly apparent that individuals from socio-economically deprived backgrounds and ethnic minority communities faced a disproportionate burden of disease. Most initial studies provided only temporary views of these inequalities, failing to account for the dynamic nature of the epidemic. As lockdown policies shifted, new variants emerged, and vaccination programs rolled out, the underlying drivers of infection risk likely changed.

This paper addresses the need for a longitudinal perspective, seeking to disentangle the time-varying effects of deprivation and ethnicity. A significant challenge in such analyses is "ascertainment bias"—the reality that raw test positivity data is influenced by healthcare-seeking behavior and testing availability. By combining routine surveillance data with randomized community testing, this study provides a more robust estimate of true prevalence and its socio-demographic drivers.

Methodology

The researchers utilized a Bayesian spatio-temporal regression model, implemented via the R-INLA package, to analyze data at the Lower Tier Local Authority (LTLA) level across England.

Outcome Metrics

The study modeled two primary outcomes:

  1. Weekly Test Positivity Rate: Derived from Pillar 2 (community) testing data provided by the PHE Second Generation Surveillance System (SGSS).
  2. Debiased Prevalence Proportion: A novel metric that adjusts Pillar 2 data using randomized surveillance data from the REal-time Assessment of Community Transmission (REACT) study. This correction accounts for the fact that symptomatic individuals or those in certain occupations are more likely to seek tests.

Model Covariates and Structure

The model adjusted for several key factors, including the 2019 Index of Multiple Deprivation (IMD), ethnic composition (based on the 2011 UK Census), and weekly second-dose vaccine uptake. The model also accounted for urban/rural status and spatial correlation between neighboring LTLAs. Crucially, the model allowed coefficients for deprivation and ethnicity to vary by week, capturing the evolving relationship between these factors and infection rates.

Results

The analysis revealed that the "driver" of inequality was not constant throughout the study period.

The Shift from Deprivation to Ethnicity

Socio-economic deprivation showed a dominant association with infection rates through October 2020. However, as the pandemic progressed into the second wave (December 2020–January 2021), the influence of ethnicity became more pronounced. During the peak of the Alpha variant surge, areas with high BAME populations but low deprivation saw a 19% increase in test positivity, suggesting the variant spread effectively through specific community networks regardless of economic status.

Ethnic Disparities and Variant Surges

The study found significant heterogeneity among ethnic groups. LTLAs with higher proportions of South Asian residents were significantly more affected during the second wave and the subsequent Delta variant surge in May 2021. Interestingly, areas with larger Black populations did not show a similarly increased risk during this specific timeframe (June 2020–September 2021), highlighting that the drivers of risk for different minority groups may vary across different phases of the pandemic.

Quantifying the Risk

When comparing the most deprived and ethnically diverse areas to the least, the weekly positivity rate was approximately 13% higher (rising from 2.97% to 3.35%), and the prevalence was 10% higher. During the initial Delta surge in May 2021, the Odds Ratio (OR) for the most deprived and diverse areas peaked at 1.23.

Discussion

A key finding of this research is that the observed inequalities were consistent across both raw test positivity and debiased prevalence. This suggests that the disparities are driven by differences in actual disease exposure or susceptibility—likely due to factors such as housing density, multi-generational households, or employment in essential sectors—rather than differences in testing behavior.

Limitations

The study is subject to "ecological bias," as it uses area-level data rather than individual-level data. While the associations at the LTLA level are strong, they do not definitively prove individual-level risk. Furthermore, the reliance on 2011 Census data for ethnicity may not perfectly reflect the demographic landscape of 2020, and the "Other BAME" category remains a broad umbrella that may obscure specific nuances.

Conclusion

This spatio-temporal analysis demonstrates that the impact of social determinants on COVID-19 risk is highly dynamic. Public health interventions must be as flexible as the epidemic itself, recognizing that the communities most at risk can change as new variants emerge and social conditions evolve. Future research should focus on the specific mechanisms of exposure, such as occupational risks and public transport usage, to better protect vulnerable populations in future health crises.

Summary updated

BibTeX
@misc{padellini2021time,
  author = {Tullia Padellini and Radka Jersakova and Peter J. Diggle and Chris Holmes and Ruairidh King and Brieuc Lehmann and Ann‐Marie Mallon and George Nicholson and Sylvia Richardson and Marta Blangiardo},
  title = {{Time varying association between deprivation, ethnicity and SARS-CoV-2 infections in England: a space-time study}},
  year = {2021},
  doi = {10.1101/2021.11.09.21266054},
  howpublished = {medRxiv},
}
Figure