Importance of social inequalities to contact patterns, vaccine uptake, and epidemic dynamics

The spread of infectious diseases is profoundly shaped by individuals' socio-demographic and economic characteristics, influencing both exposure levels and disease severity. Pandemics disproportionately affect certain populations due to variations in exposure (social interactions, mobility, compliance with NPIs) and vulnerability (preexisting health conditions, access to healthcare). While the importance of socioeconomic inequalities is widely acknowledged, traditional epidemiological models often overlook these factors, typically stratifying only by age. This study addresses this gap by investigating how individual characteristics influence two key epidemic-relevant behaviors: contact patterns and vaccination uptake. The research leverages data from the MASZK study conducted in Hungary during the COVID-19 pandemic. This extensive dataset captures face-to-face interactions, vaccination status, and other relevant behavioral and attitudinal data across a representative sample of the Hungarian population, collected monthly from April 2020 to June 2022. The study aims to identify the most influential individual characteristics differentiating epidemic-relevant behaviors and subsequently incorporates these differences into a data-driven epidemiological model to analyze their impact on real-world epidemic outcomes, focusing specifically on the fourth wave of the COVID-19 pandemic in Hungary. This approach offers valuable insights into the unequal burden of disease and informs strategies for more effective public health interventions.

Existing literature extensively documents the disproportionate impact of pandemics on specific population groups, highlighting health inequalities stemming from differences in exposure to pathogens and disparities in disease severity and fatality. These disparities are often linked to socioeconomic factors influencing social interactions, mobility, and compliance with non-pharmaceutical interventions (NPIs). However, traditional epidemiological models frequently neglect these socioeconomic heterogeneities, primarily stratifying populations by age. Recent research has advocated for integrating social aspects into infectious disease modeling to improve our understanding of how socioeconomic disparities translate into heterogeneous disease spread. This study builds upon this call, seeking to elucidate the mechanisms connecting individual characteristics to epidemic-relevant behaviors and ultimately to epidemiological outcomes.

The study uses data from the MASZK study, a large-scale survey conducted in Hungary during the COVID-19 pandemic. The data comprises 26 monthly cross-sectional surveys (April 2020 – June 2022), each including approximately 1000 respondents, providing a nationally representative sample. The data includes information on face-to-face contact patterns (interactions lasting at least 15 minutes within 2 meters), categorized as work and community contacts (excluding household contacts). Socio-demographic variables considered include education level, employment status, income level, gender, settlement type, chronic and acute disease status, and smoking habits. The data is aggregated into six periods encompassing four epidemic waves and two interim periods to reflect the course of the pandemic in Hungary. Initially, a negative binomial regression model is used to identify the key socio-economic dimensions that, in interaction with age, most significantly influence contact numbers. The average marginal effect (AME) is computed, and its maximum confidence level across age groups is determined for each variable, providing a measure of the variable's overall influence on contact patterns. Similarly, a logistic regression is used to analyze the dimensions impacting vaccination uptake. The findings from the regression analyses are then incorporated into an extended SEIR compartmental model. This model extends a standard SEIR model by incorporating decoupled age contact matrices, stratifying contacts not only by age but also by the identified socio-economic dimensions (employment, education, settlement, and income). The model is used to simulate the spread of infection, calculating attack rates across different age and socioeconomic groups. Finally, the extended SEIR model is applied to simulate the fourth wave of the COVID-19 pandemic in Hungary, incorporating both contact patterns and vaccination uptake differences across employment and income levels. The model is calibrated using Approximate Bayesian Computation (ABC) on daily death data. Bootstrapping is employed to assess uncertainty in the estimation of contacts and contact matrices.

The analysis reveals that employment, education, and income are the most significant predictors of contact numbers, with higher socioeconomic groups exhibiting higher contact rates. High- and mid-high-educated individuals showed greater adaptability in their contact patterns, reducing contacts during waves and increasing them during interim periods. Employed individuals consistently maintained more contacts than the unemployed, both within and outside the workplace. Age contact matrices, decoupled by education and employment, demonstrate substantial differences in contact patterns beyond age alone. The extended SEIR model, incorporating these differences, predicts significantly different attack rates across socioeconomic groups, highlighting the unequal burden of the epidemic. Employed and wealthier individuals, along with those in the capital, exhibit higher attack rates. Regarding vaccination uptake, income emerges as the most significant determinant, with higher socioeconomic groups exhibiting higher vaccination rates. The extended SEIR model, accounting for vaccination uptake variations, shows that the benefits of vaccination are also socioeconomically stratified, with groups having higher vaccination rates experiencing the greatest reduction in attack rates. Simulations of the fourth wave in Hungary demonstrate higher infection rates among employed individuals and higher mortality rates among unemployed, low-income, and older individuals. However, further stratification by age within employment and income subgroups reveals nuanced patterns, highlighting the complex interplay between age and socioeconomic factors.

The findings underscore the significant influence of socioeconomic factors on both contact patterns and vaccination uptake, shaping the transmission dynamics of infectious diseases. The results challenge assumptions of homogeneous mixing in traditional epidemiological models, demonstrating the need to incorporate socioeconomic heterogeneity. The study's data-driven approach provides insights into the mechanisms underlying unequal epidemiological outcomes, revealing how socioeconomic disparities translate into differential exposure and vulnerability to infection. The model's ability to capture these heterogeneities allows for a more accurate prediction of disease spread and a better understanding of the unequal burden of the epidemic across different socioeconomic groups. The observed variations in contact patterns highlight the importance of considering social and behavioral adaptations to public health measures. The integration of vaccination uptake in the model further reveals the stratified effects of vaccination campaigns. The results strongly support the need for tailored public health strategies that address the unique needs and characteristics of different socioeconomic groups.

This study demonstrates the critical role of socioeconomic inequalities in shaping COVID-19 transmission dynamics. The developed data-driven epidemiological model provides a more nuanced and accurate representation of disease spread, highlighting the limitations of traditional age-stratified models. The findings emphasize the need for public health strategies that account for these inequalities to effectively mitigate the unequal burden of epidemics. Future research could investigate the causal pathways linking socioeconomic factors to disease outcomes and explore the development of more sophisticated models incorporating a wider range of social determinants and dynamic interactions between behaviors and policies.

The study's reliance on survey data introduces potential limitations. Contact patterns are only characterized by the participants' characteristics, lacking information on the contacted individuals' attributes. The extended SEIR model, therefore, only accounts for the respondent's socioeconomic status. Data gaps in the survey regarding the initial distribution of individuals across compartments and socioeconomic groups led to the assumption of proportional distribution, which may not accurately reflect the reality of the pandemic's initiation. The model also simplifies complex interactions between vaccination and contact patterns, focusing primarily on the stratified effects of each aspect separately. The study focuses on Hungary during the COVID-19 pandemic, and its findings might not be generalizable to other contexts.