key: cord-0860764-xqlrfh18 authors: John, J. title: Reopen or redistribute? Modeling years of life lost due to Covid-19, socioeconomic status, and non-pharmaceutical interventions date: 2021-04-26 journal: nan DOI: 10.1101/2021.04.23.21256005 sha: 63600e482cbbb9a65a93adc353fc95494c9e5745 doc_id: 860764 cord_uid: xqlrfh18 Research in the current pandemic has put a sharp focus on the health burden of Covid-19, thereby largely neglecting the cost to life from the socioeconomic consequences of its containment. The paper develops a model for assessing their proportionality. It compares the years of lost life (YLL) due to Covid-19 and the socioeconomic consequences of its containment. The model reconciles the normative life table approach with de facto socioeconomic realities by correcting YLL estimates for socioeconomic differences in life expectancy. It thereby aims to improve on the attribution of YLL due to immediate and fundamental sources of inequalities in life expectancy. The application of the approach to the pandemic suggests that the socioeconomic consequences of containment measures come with a much higher life tag than the disease itself and therefore need urgent attention, especially in poorer and more unequal societies. Avoiding 3 million additional cases of extreme poverty may come with a similar life tag as protecting 1 million people from dying from Covid-19. individual's life expectancy. 1 Accordingly, it is possibly to assess the damage due to NPIs in years of life lost (YLL). The approach complements but is distinct to other perspectives on the pandemic such as the burden of disease and value of life. The model rather contributes to the discourse on the relationship between health inequality and social justice. 14, 15 The paper targets some of the key conceptual difficulties when attributing YLL to individual causes. Because any such assessment can only be plausible estimates at best, the efforts at quantifying the model are primarily for purposes of illustration and need to be complemented with context-specific data to be able to inform policy. The model starts from the basic assumption that proportionality can be expressed as a correspondence of YLL due to Covid-19 and the YLL due to the socioeconomic damage from the NPIs. YLL refer to the gap between the age of death and the age to which a person could have lived. No single methodology for estimating YLL exists, but it is common practice to use life tables that either assume an ideal life expectancy in a counterfactual disease and poverty free world or draw on hazard ratios within the age bracket of the birth cohort. 16, 17 As a result life tables rather state an aspiration than provide information about the actual number of years an individual would have lived in the absence of a specific cause of death. While there is nothing wrong about such a normative approach, it entails problems of correctly attributing YLL to individual causes. The question therefore is to what extent these YLL can be attributed to the immediate cause of death or in fact reflect more fundamental socioeconomic differences in life expectancy. International differences in average life expectancy amount to about 30 years between the poorest and the richest countries. 18 Accordingly, the WHO's international life tables state an average of more than 50 YLL for a death in a low income country compared with less than 20 YLL in a high income country. Those numbers roughly halve when correcting for the average life expectancy in the income group. National life tables such as those of UN's World Population Prospects take into 1 If not stated otherwise all references to life expectancy concern averages of both sexes. 3 account international differences in life expectancy but fall short of accounting for systematic differences in life expectancy among socioeconomic groups within a society. By common ways of measurement socioeconomic differences in life expectancy in high income countries usually amount to 5-10 years between groups with a low and high SES. 19 These differences may reach up to 15 or 20 years in poorer and more unequal societies and when using more fine-grained measures. 20, 21 The common tripartite division in low, mid, and high SES means that a person that permanently falls into a lower status group, for example, as a consequence of losing 40% of their income due to unemployment or a transition from full-to part-time work, or as a consequence of a forgone 3-4 year higher education block (e.g. secondary or tertiary education). 22 However, it would almost certainly be an overestimate to infer that such a decline directly translates into a reduction of the individual life span as suggested by the systematic socioeconomic differences in life expectancy in a society because only a part of those differences are in fact malleable. To ascertain YLL more precisely it is necessary to disentangle the determinants of the socioeconomic gap in life expectancy. Unfortunately, key determinants tend to overlap and only unfold their effects indirectly and in the long run, leaving life expectancy overdetermined. It is therefore difficult to specify the relative causal influence of fundamental sources such as the genetic disposition 23 and SES 24, 25 , the mechanisms through which they work such as health behaviors [26] [27] [28] and morbidities (e.g. chronic diseases) 29, 30 , and the immediate causes of death. Temporal and causal complexity as well as a reliable data complicate measurement immensely. 31, 32 Lately however, a number of studies made headway into ascertaining the individual contribution of factors such as income and education in the socioeconomic gap in life expectancy. In the European mean low income explains around 10-20% of an average 5-year gap in life expectancy between educational groups. 33 For disability-adjusted life expectancy it is around 20% of a 8,5-year gap in life expectancy between educational groups. 34 Educational and occupational status also account for around 20% of the 10-year gap in life expectancy between SES groups. 35, 36 In sum, income and educational status may each account for about a fifth (~20%) of the socioeconomic gap in life expectancy. The findings do not travel easily from the European high income countries to the rest of the world. In poorer countries morbidity and mortality are generally higher but health behaviors account for a smaller share of the socioeconomic differences in life expectancy. 30, 37 This plausibly leaves a larger share for socioeconomic factors. Education tends to entail a higher income premiums, 38 but like 4 . CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted April 26, 2021. ; https://doi.org/10.1101/2021.04.23.21256005 doi: medRxiv preprint health services is often not universally supplied and depends on personal income. The Socio-Demographic Index of the Global Disease Burden Project accounts for 85% of international differences in average healthy life expectancy by building the geometric mean of lagged per capita income, education of the population aged 15+, and the fertility rate of women aged 25+ (as a proxy for the standing of women in society). 39 Against this background it seems plausible that in middle and low income countries, factors such as income and education may each account for 30% and more of the socioeconomic differences in life expectancy. In other words, the high, mid, and low income countries have higher factors of socioeconomic determination of the life expectancy. Based on these findings it is possible to construct a rough estimate of YLL due to permanent unemployment, reduction of working hours, and forgone education that entail a loss in SES group (from high to mid or mid to low). The YLL vary with the nominal socioeconomic gap in life expectancy (5y, 10y, 15y) and the degree of socioeconomic determination (SOD: 20%, 30%, 40%), depending on the overall level of income and development of a society. The two main components of the socioeconomic damage are due to income loss (YLLI) and forgone education (YLLE): . CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. The copyright holder for this preprint this version posted April 26, 2021. ; https://doi.org/10.1101/2021.04.23.21256005 doi: medRxiv preprint current losses in lifetime earnings may amount to 5% on average (or about one eights of a decline in SES group). For many low to middle income countries this is possibly an underestimate, given that school closures were on average longer, potentially entail higher dropout rates, and would have held a higher income premium. 42 Amid these uncertainties, the article suggest correcting the aspirational YLL from the life tables for socioeconomic differences in life expectancy. In other words, to drop the assumption of a povertyfree world. This may improve YLL estimates in two ways: a more accurate attribution of YLL to its fundamental and immediate causes; and a potentially more precise estimate of the actual YLL of an individual without giving up on the normative problem of accepting a lower than ideal life expectancy. To that end, country-specific findings on socioeconomic differences in life expectancy should be combined with data on seroprevalence, hospitalization rates, and deaths among groups with a low SES compared to the general population. Studies consistently find that people with a low SES are significantly more affected than people with a high SES. In high income countries around half of all Covid-19 deaths occur among people with a low SES. In the US, people with below median income account for two thirds of deaths (the lowest tertile for about half). 50 The poorest quintile suffers from a third more co-morbidities and 6 . CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. The copyright holder for this preprint this version posted April 26, 2021. ; https://doi.org/10.1101/2021.04.23.21256005 doi: medRxiv preprint twice the case count and death rate. 51 In Scotland, the lowest quintile may account for half of all ICU admissions. 52 Also, in the otherwise more equitable Germany and Sweden people with low SES account for 40-60% of hospitalizations and death. 53, 54 Undertaking similar estimate for low and middle income countries faces severe data challenges. The only large cross-country comparison to date, using national UN WPP life tables for remaining life expectancy at the exact age of deaths, estimates mean YLL per capita at 13 YLL for high income countries and 19 YLL for middle and low income countries. 552 Their estimates are used in the following examples. The profile of Covid deaths in low and middle income countries is more ambiguous. Similar to all-cause mortality Covid deaths tend to be younger. While this reduces case mortality, high levels of inequality and poverty come with additional risks for the younger. In other words, fewer people die but those that die lose more life years. Socioeconomic factors associated with a low SES such as poor living conditions and work in the informal economy, which often accounts for 50-90% of economic activity, seem to play a key role in the spread and higher mortality in the younger age brackets. [56] [57] [58] It is therefore reasonable to assume that the share of people with a low SES is even higher in low and middle income countries. . CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. The copyright holder for this preprint this version posted April 26, 2021. ; levels of poverty and large lifespan inequalities, the aspirational 19 YLL from the life tables may have to be reduced by 10-12 YLL to 7-9 YLL due to Covid. 3 In sum, when accounting for socioeconomic differences in life expectancy YLL attributable to Covid may be between 6 The loss of life years in these scenarios is then juxtaposed with the socioeconomic damage in the pandemic to ascertain proportionality. The main idea is to calculate the number of people affected by the socioeconomic damage that would entail the equivalent amount of YLL. . CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. The copyright holder for this preprint this version posted April 26, 2021. ; https://doi.org/10.1101/2021.04.23.21256005 doi: medRxiv preprint N g =n e +n i +n p n i ≈3.492.000 .000 ¹ ; n e ≈1.500 .000 .000 ² ; n p ≈640.000 .000 ³ 1 ILO; 2 UNESCO; 3. World Bank. Learning loss is divided into two subgroups. Those students suffering the average learning and subsequent income loss from school closures YLLe2 and the worst hit students that forgo a 3-4 year higher learning block YLLe1 (i.e. additional dropouts due lack of funding or qualification for higher education). Because students with a high SES may have more capacities to compensate for learning losses it is assumed that two thirds of the students worldwide (=0.9 billion) suffer the average learning and subsequent income loss due to school closures. The resulting value is subtracted from the overall years of life lost due to Covid (YLLcov). The remaining damage is then distributed among the students with a learning block loss (YLLE1), people with an income loss YLLI, and those that fall into extreme poverty as a result YLLP. Each group carries a weighted burden that reflects group size and the social gradient (α,ꞵ,γ). Income: Education: Poverty: 9 . CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. The copyright holder for this preprint this version posted April 26, 2021. ; x p = γ⋅(YLL cov −YLL e 2 ) YLL p The standard model reflects specifications for the global average (8 YLL per Covid death, a 7,5- year socioeconomic gap in life expectancy, and a SOD of 0.3). The tables read as follows. Each row provides the total number of workers, poor, and students for which the socioeconomic damage would have to become permanent for the YLL to be equivalent of those attributable to Covid-19. Lines are gray for negative values, i.e. when the average socioeconomic damage from school closures is higher than the YLL due to Covid. A separate column provides the common percentage share, which by definition is identical for all groups. . CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. The copyright holder for this preprint this version posted April 26, 2021. ; https://doi.org/10.1101/2021.04.23.21256005 doi: medRxiv preprint the analysis. Dropping YLLE2 from the calculation results in an increase of the percentage shares in the previously disproportionate scenarios W1-3 to 0.6-1.1%. The absolute numbers for proportional socioeconomic damage thus remain relatively small in the first three scenarios. That illustrates the potential harm to life from only comparatively minor socioeconomic consequences. Subsequent scenarios roughly represent a shift by one scenario compared to the model including school closures. The percentage shares in W4-6 increase from 1.4-5.8% to 3.3-7.7%. In the worst case scenario 270 million people suffer from income loss, 50 million fall into extreme poverty, and 116 million forgo a higher learning block. The results highlight the urgent need for compensating existing and avoiding further socioeconomic damage in the pandemic. Avoiding a globally speaking relatively minor number 5 million people with income loss, 1 million additional people in poverty, and 2 million with a higher learning loss saves a similar amount of life years as saving more than 1 million people from dying from Covid-19. Current predictions of the long-term socioeconomic consequences in the pandemic are uncertain but illustrate the effort that would be required. According to the ILO, the equivalent to 255 million full time jobs was lost in 2020, including around 114 full time jobs and two to three hundred million working hour reductions. 2 Depending on which scenario becomes reality over the next years, 60-90% of those job losses would have to be saved from becoming permanent for the loss in life years not to exceed the YLL due to the pandemic. Looking at educational loss one year into the pandemic, UNESCO estimates that more than half of all school children continue to face learning disruptions and 24 million (1.6%) students are at risk of dropping out completely, not including those that will not qualify for higher education. 59 Educational loss is thus similar to levels estimated for scenario W4. Furthermore, around 100-150 million people fell into extreme poverty due to the pandemic. 4,5 The number almost doubles to 228 million for the higher poverty threshold of $3. 20 . CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. The copyright holder for this preprint this version posted April 26, 2021. ; https://doi.org/10.1101/2021.04.23.21256005 doi: medRxiv preprint $1.90 per day. Even in the worst case scenario these losses must not become permanent for more than 20-40% of that recently fell into (extreme) poverty. For a second set of results the standard model specifications are adapted to reflect different country conditions (see Table 6 ). The three columns on the left assume conditions more similar to a typical high income country. The SOD is at 20%, the average per capital YLL due to Covid are at 6 The different model specifications obtain two main results. One at the cross-country level and one at the within-country level. First, the differences in parameters tend to largely even out across the different model specifications. The first three scenarios remain disproportionate in low, middle, and high income countries. The results also largely hold when dropping the average damage from school closures (YLLE2) ( Table 7) . Larger differences only occur in scenario W4-6. At the extreme ends of the model specifications, the proportionality of the socioeconomic damage differs by the factor 3 (2, . CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. The copyright holder for this preprint this version posted April 26, 2021. ; https://doi.org/10.1101/2021.04.23.21256005 doi: medRxiv preprint twice the socioeconomic damage is proportional than in the least egalitarian ones. In middle and low income these differences are less pronounced but remain significant. The article set out to narrow in on the difficult question of proportionality between the health and socioeconomic damage in the pandemic. YLL were suggested as a common measure of comparison for the two dimensions. Correctly attributing YLL to Covid-19 and SES, requires disentangling the fundamental and immediate sources of life expectancy. In order to use the life table approach that is common in YLL estimates, the article made an attempt at dropping the inherent assumption of a poverty-free world (or society). The discussion concluded that up to half of the average per capita YLL from the life tables may in fact be socioeconomically determined. Because SES is associated with morbidity and mortality which in turn is similar for Covid-19 and all-cause mortality, the approach may yield analytic benefits beyond the current pandemic. 60, 61 Ecological data on the SES of the population of interest may proxy for a lack of individual-level data on the prevalence of morbidity and other risk factors. The application to the pandemic highlights the difficult trade-offs involved in the short-and longterm protection of health. While NPIs target immediate health concerns, the long-term socioeconomic damage is likely to entail a steep cost to life that requires the same if not more attention in the intermediate aftermath of the pandemic. Avoiding 3 million additional cases of extreme poverty comes with a similar life tag as protecting 1 million people from dying from Covid-19. In countries that lack the necessary resources to compensate for the socioeconomic damage in the pandemic, good communication strategies may therefore be more appropriate than more drastic NPIs such as business and school closures. 7 likely to be disproportionate even in high income countries. At the very least they carry a significant burden of proof regarding their suitability and necessity. Interestingly, the question of proportionality has otherwise been rather similar across different income groups, largely because the social gradient and the associated loss of life is steeper for both Covid-19 and the NPIs. Sill, levels of within-country inequalities may be a key concern regarding the proportionality of the NPIs. This is especially true because a wider socioeconomic gap in life expectancy signals a weaker social safety net that could compensate for losses. . CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted April 26, 2021. ; https://doi.org/10.1101/2021.04.23.21256005 doi: medRxiv preprint Impacts of COVID-19 on global poverty, food security, and diets: Insights from global model scenario analysis What Have We Learnt?: Overview of Findings from a Survey of Ministries of Education on National Responses to COVID-19 Ranking the effectiveness of worldwide COVID-19 government interventions Inferring the effectiveness of government interventions against COVID-19 Demographic and public health characteristics explain large part of variability in COVID-19 mortality across countries Wie gefährlich ist COVID-19? Die subjektive Risikoeinschätzung einer lebensbedrohlichen COVID-19-Erkrankung im Frühjahr und Frühsommer 2020 in Deutschland Risk perceptions of COVID-19 around the world COVID-19 fears Why health equity? The concern for equity in health Valuing the years of life lost due to COVID-19: the differences and pitfalls Reflection on modern methods: years of life lost due to premature mortality-a versatile and comprehensive measure for monitoring noncommunicable disease mortality The health gap: the challenge of an unequal world Inequalities in longevity by education in OECD countries: Insights from new OECD estimates Income inequality, mortality, and self rated health: meta-analysis of multilevel studies Messung des sozioökonomischen Status in der Studie zur Gesundheit Erwachsener in Deutschland (DEGS1) Genetics and health inequalities: hypotheses and controversies Social epidemiology Social Conditions As Fundamental Causes of Disease Health behaviours as explanations for educational level differences in cardiovascular and all-cause mortality: a follow-up of 60 000 men and women over 23 years The contribution of health behaviors to socioeconomic inequalities in health: A systematic review Association of socioeconomic position with health behaviors and mortality Inequalities in non-communicable diseases and effective responses Socioeconomic status and non-communicable disease behavioural risk factors in low-income and lower-middle-income countries: a systematic review The great escape: health, wealth, and the origins of inequality The Health Effects of Income Inequality: Averages and Disparities Determinants of inequalities in life expectancy: an international comparative study of eight risk factors Determinants of educational inequalities in disability-free life expectancy between ages 35 and 80 in Europe Reducing socio-economic inequalities in all-cause mortality: a counterfactual mediation approach Socioeconomic status and the 25 × 25 risk factors as determinants of premature mortality: a multicohort study and meta-analysis of 1·7 million men and women Association of Socioeconomic Status with Overall and Cause Specific Mortality in the Republic of Seychelles: Results from a Cohort Study in the African Region The knowledge capital of nations: Education and the economics of growth Global age-sex-specific fertility, mortality, healthy life expectancy (HALE), and population estimates in 204 countries and territories, 1950-2019: a comprehensive demographic analysis for the Global Burden of Disease Study The long-term distributional and welfare effects of Covid-19 school closures Folgekosten ausbleibenden Lernens: Was wir über die Corona-bedingten Schulschließungen aus der Forschung lernen können. München: ifo Institut What Have We Learnt? Overview of findings from a survey of ministries of education on national responses to COVID-19 Education: From disruption to recovery COVID-19-exploring the implications of long-term condition type and extent of multimorbidity on years of life lost: a modelling study COVID-19: How Many Years of Life Lost? medRxiv 2020 Estimating (quality-adjusted) life-year losses associated with deaths: With application to COVID-19 The Mean Unfulfilled Lifespan (MUL): A New Indicator of a Disease Impact on the Individual Lifespan Years of life lost estimates cannot always be taken at face value: Response to "COVID-19 -exploring the implications of long-term condition type and extent of multimorbidity on years of life lost: a modelling study Impact of COVID-19 on Care-Home Mortality and Life Expectancy in Scotland Social determinants of mortality from COVID-19: A simulation study using NHANES Social inequality and the syndemic of chronic disease and COVID-19: county-level analysis in the USA Socio-economic deprivation and the risk of death after ICU admission with COVID-19: The poor relation Higher risk of COVID-19 hospitalization for unemployed: an analysis of health insurance data from 1.28 million insured individuals in Germany A population-based cohort study of socio-demographic risk factors for COVID-19 deaths in Sweden Years of life lost to COVID-19 in 81 countries Seroprevalence of SARS-CoV-2 in slums versus non-slums in Mumbai, India The Younger Age Profile of Covid-19 Deaths in Developing Countries. IDB Effect of socioeconomic inequalities and vulnerabilities on health-system preparedness and response to COVID-19 in Brazil: a comprehensive analysis One year into COVID-19 education disruption: Where do we stand? UNESCO Factors associated with deaths due to COVID-19 versus other causes: population-based cohort analysis of UK primary care data and linked national death registrations within the OpenSAFELY platform Use of "normal" risk to improve understanding of dangers of covid-19 Fear, lockdown, and diversion: Comparing drivers of pandemic economic decline 2020