Unravelling COVID-19 social mortality patterns using conditional inference tree algorithm

Abstract

Abstract Background The COVID-19 pandemic led to significant excess mortality in 2020 in Belgium. By using microlevel cause-specific mortality data for the total Belgian adult population in 2020, three outcomes were considered in this study aiming at predicting sociodemographic (SD) and socioeconomic (SE) patterns of 1) COVID-19 specific death compared to survival; 2) all other causes of death (OCOD) compared to survival; 3) COVID-19 specific deaths compared to all OCOD. Methods Multivariable logistic regression models were fitted for the three outcomes. In addition, we computed conditional inference tree (CIT) algorithms, complementing regression models, to identify and rank the strongest SD and SE predictors of the three outcomes. Results Older people, males, people living in collectivities, first-generation migrants, and deprived SE groups experienced higher odds of dying from COVID-19 compared to survival; living in collectivities was identified by the CIT as the strongest predictor followed by age and sex. Education emerged as one of the strongest predictors for people not living in collectivities. Overall, similar patterns were observed for all OCOD except for first- and second-generation migrants having lower odds of all OCOD compared to survival; age group was identified by the CIT as the strongest predictor. Older people, males, people living in collectivities, first- and second-generation migrants, and people with lower levels of education had higher odds of COVID-19 death compared to all OCOD; living in collectivities was identified by the CIT as the strongest predictor followed by age, sex, and migration status. Education and income emerged as the strongest predictors among people not living in collectivities. Conclusions This study identified important SD and SE disparities in COVID-19 mortality underlying the importance of implementing preventive measures, particularly within the most vulnerable populations, in infectious disease pandemic preparedness. Key messages • Living in collectivities appeared to be the strongest predictor of COVID-19 death across all age groups, highlighting the importance of preventive measures in reducing the transmission of the virus. • Poor education plays a crucial role in predicting COVID-19 death among people not living in collectivities, highlighting the importance of targeting preventive measures towards low educated groups.