The Longevity-Frailty Hypothesis: Evidence from COVID-19 Death Rates in Europe

Abstract

By the end of spring (31 May), the COVID-19 death rate was remarkably unevenly distributed across the countries in Europe. While the risk of COVID-19 mortality is known to increase with age, age-specific COVID-19 death rates across Europe were similarly unevenly distributed. To explain these mortality distributions, we present a simple model where more favorable survival environments promote longevity and the accumulation of health frailty among the elderly while less favorable survival environments induce a mortality selection process that results in lower health frailty. Because the age-related conditions of frailty render the elderly less resistant to SARS-CoV-2, pre-existing survival environments may be non-obviously positively related to the COVID-19 death rate. To quantify the survival environment parameter of our model, we leveraged historic cohort- and period-based age-specific probabilities of death and life expectancies at age 65 across Europe. All variables are significantly correlated with indicators of frailty like elderly dependence on others for personal and household care for a subset of European countries. With respect to COVID-19 death rates, we find significant positive relationships between our survival indicators and COVID-19 death rates across Europe, a result that is robust to statistical control for the capacity of a healthcare system to treat and survive infected persons, the timing and stringency of non-pharmaceutical interventions, population density, age structure, case rates and the volume of inbound international travelers, among other factors. To address possible concerns over reporting heterogeneity across countries, we show that results are robust to the substitution of our response variable for a measure of cumulative excess mortality. Also consistent with the intuition of our model, we also show a strong negative association between age-specific COVID-19 death rates and pre-existing all-cause age-specific mortality rates for a subset of European countries. Overall, results support the notion that variation in pre-existing frailty, resulting from heterogeneous survival environments, partially accounts for striking differences in COVID-19 death during the first wave of the pandemic.

Keywords: COVID-19; frailty; longevity.

Figure A1

Correlation matrix for indicators of survival environment and elderly frailty. NOTE: Survival environment indicator data are derived from the Human Mortality Database (HMD) (https://www.mortality.org/ (accessed on 15 July 2020)) and elderly frailty indicator data are from European Commission Eurostat database (https://ec.europa.eu/eurostat/data/database (accessed on 15 July 2020)). The data underlying the correlations presented here can be found in Appendix A Table A1.

Figure 1

Uneven Distribution of Wave I (ending 31 May 2020) COVID-19 mortality risk across the countries of Europe. NOTE: The countries of Europe are arranged in descending order by reported COVID-19 deaths per million, as of 31 May 2020. COVID-19 death data by country are from Johns Hopkins Coronavirus Resource Center (JHU CSSE) (https://github.com/CSSEGISandData/COVID-19 Unified-Dataset (accessed on 15 June 2020)).

Figure 2

Heterogeneous Wave I (ending 31 May 2020) COVID-19 mortality risk by age group across select European countries. NOTE: COVID-19 death and population data (closest to 31 May 2020) by age and country are from National Institute for Demographic Studies (INED) (https://dccovid.site.ined.fr/en/data/ (accessed on 15 June 2020)).

Figure 3

Spatial distribution of Wave I (ending 31 May 2020) COVID-19 mortality risk and relative survival environment. NOTE: COVID-19 death data by country are from Johns Hopkins Coronavirus Resource Center (JHU CSSE) (https://github.com/CSSEGISandData/COVID-19 Unified-Dataset (accessed on 15 June 2020)), and Survival Index data are derived from the Human Mortality Database (HMD) (https://www.mortality.org/ (accessed on 15 June 2020)).

Figure 4

Model predicted cumulative COVID-19 death. Rate at various percentiles of survival environment indicators. NOTE: Model predicted cumulative death rates are generated via the Equation (12) model estimates presented in columns (1)–(4) of Table 1, where all other covariates are evaluated at their mean.

Figure 5

Model predicted cumulative COVID-19 death rate by week of Wave I for survival index terciles. NOTE: Model predicted cumulative death rates by week and survival index tercile are generated via Equation (14) model estimates, where all other covariates are evaluated at their mean.