Abstract
Introduction The Covid-19 pandemic has driven widespread and rapid application of non-pharmaceutical interventions (NPIs), including travel restrictions, bans on gathering, and community-wide lockdowns. To date, decision-makers have not explicitly weighed the predictable secondary health impacts that may result from responses to Covid-19.
Methods This analysis applies two pooled effects hazard ratios from published meta-analyses on unemployment to quantify the impacts of Covid-19 unemployment in the United States on suicides and deaths from all causes.
Results Applying observational epidemiological data to the number of the population recently unemployed in the United States predicts approximately 9,000 annual deaths from suicide 61,000 annual deaths from all causes.
Discussion Indirect health impacts of societal responses to Covid-19 are identifiable and quantifiable. These health impacts may have a longer duration than those of the Covid-19 pandemic itself. Decision-makers can include indirect health impacts in policy making calculi for Covid-19 mitigation and suppression strategies. A full accounting of trade-offs is likely to reveal multiple ethical challenges requiring transparent public discussion and solutions.
Introduction
Limiting death and mitigating demand on health care systems have been the primary aims of policy actions to reduce transmission of the Covid-19 virus. The spectre of millions dead and overwhelmed tertiary health care (e.g. ventilator rationing) drove widespread and rapid application of rarely used policies to limit mobility and economic activity. These policies include the deployment of non-pharmaceutical interventions (NPIs), including travel restrictions, bans on gathering, and community-wide lockdowns.(1) The simultaneous and widespread application of NPI’s in a globalized world is an event without precedent. In the face of uncertainty, decisions to employ these strategies were made quickly, without scientific peer review of the models informing decisions and without a transparent analysis of trade-offs.
The authors of the Imperial College of London modeling report, one of those credited with influencing policy, wrote a warning: “We do not consider the ethical or economic implications of either strategy here … Suppression … carries with it enormous social and economic costs which may themselves have significant impact on health and well-being in the short and longer-term”(2) The US CDC’s guidelines for responding to pandemics also explicitly call for weighing any community mitigation actions against other human needs.(3)
Behavioral responses to Covid-19 combined with the application of NPIs are reasonably likely to produce an number of secondary impacts that are relevant to health, including fear and anxiety, social isolation, business closures, unemployment, bankruptcies, foreclosures, homelessness, loss of heath insurance, deferred or delayed health care, and suicide.(4) A number of these effects are already manifest.
Available economic and risk assessment tools might help evaluate and quantify these effects and secondary health impacts.(5) As the epidemic evolves, both real-time measurement and modeled prediction of these secondary outcomes might be useful in balancing the health impacts of the response to Covid-19. Furthermore, enumeration and quantification of the health consequences of pandemic countermeasures may inform the adoption of policy countermeasures, such as income food and housing supports and expansion of health and mental health services.
One highly visible and early consequence of the pandemic response has been widespread unemployment. Unemployment results both from individual-level changes in consumer behavior as well as governmental restrictions on mobility and economic activities. As of April 30th 2020, in the United States, thirty million Americans have become newly unemployed due to the combined public, private, and governmental responses to the Covid-19 pandemic. A similar situation exists in many other countries.
A substantial literature exists on the associations among unemployment and poor health. A review of studies conducted following the great recession of 2007 found that unemployment may contribute to hunger, the loss of housing, declines in birthweight, increased symptom reporting, substance abuse, disability, medication use, and hospital visits, and untimely deaths due to suicide.(6)
Several studies on unemployment have produced quantitative estimates of health risk. For example, on study of suicide conducted in the aftermath of the great recession, observed that a 1% increase in the unemployment rate translated into a nearly 1% increase in the suicide rate in the US.(7)
This analysis predicts unemployment health impacts using risk estimates from published meta-analysis of observational studies and current estimates of unemployment, population size, and baseline diseases and mortality burdens. The approach is simple and transparent with acknowledged limitations and uncertainties.
Methodology
Data on unemployment came from the US Department of Labor (Unemployment Insurance Weekly Claims Report). Data accessed on May 7th, 2020 indicated that 33.5 million US workers had become unemployed since the beginning of the Covid-19 epidemic. (URL: https://www.dol.gov)
Data on the composition of the current labor force came from the Bureau of Labor Statistics. (URL: https://www.bls.gov/cps/cpsaat03.htm) Data accessed was for calendar year 2019.
Data on all cause age specific death rates and age-adjusted suicide rates came from the US National Bureau of Health Statistics. Gender specific rates were averaged.(URL: https://www.cdc.gov/nchs/products/databriefs/db355.htm)
Estimates of suicide risk attributable to unemployment came from a meta-analysis of 16 studies.(8) Studies included in the meta-analysis were published after 1980, had a longitudinal cohort design, included data on duration of unemployment. Included studies came from Sweden, Finland, and Denmark, which are all countries with strong social safety nets and longitudinal employment and mortality registers. The overall pooled relative risk of suicide associated with long-term unemployment (average follow-up time 7.8 years) compared to those currently employed was 1.70 (95% CI 1.22 to 2.18). The authors found non-linear effects with the suicide hazard greater for unemployment < 5 years versus >5 years. Predictions applied in this analysis use estimates based on the 1-5 year follow up period (2.50; 95% CI 1.83 to 3.17). The publication did not provide ge-specific risk estimates.
Estimates of age-specific excess all-cause mortality risk due to unemployment came from a published meta-analysis of 40 studies.(9) Studies were published between 1980 and 2008, came primarily from OECD countries, involved 20 million people, were of observational designs and had longitudinal follow up for up to 10 years. Overall pooled estimates of the hazard ratio in unemployed working aged people relative to the entire population were 1.63 (95% CI 1.49 to 1.79). This estimate represents the average effect over 10 years adjusted for age and other covariates, which varied by study. Estimates used for this analysis were age-group specific and adjusted for health risk covariates. The authors also estimated pooled hazard ratios stratified by duration of follow up but did not publish separate estimates for age and duration.
Results
Applying age-specific hazard ratios from Milner et al. (2013) to the newly unemployed population distributed based on the composition of the current labor force and age-specific suicide rates, produces the results in the table below (Table 1). The central estimate based on the current number of the newly unemployed in the US population is approximately 9000 excess suicide deaths per year.
Applying age-specific hazard ratios from Roelfs et al. (2011) to the newly unemployed population distributed based on the composition of the current labor force and age specific death rates, produces the results in the table below (Table 2). The central estimate based on the current number of the newly unemployed in the US population is approximately 60,000 excess deaths per year.
Discussion
This analysis demonstrates the feasibility of quantifying indirect health impacts of societal responses to Covid-19. The risk assessment methodology, though simple, provides an evidence-based prediction of the impact of current unemployment since Covid-19 on excess suicide and pre-mature death.
This analysis is not a full accounting of all of the health impacts of either unemployment or the overall Covid-19 response. The choice of unemployment as a measure of health exposure and all-cause and suicide mortality as the measures of outcome is based on convenience. Unemployment is an early published indicator of the economic effects of Covid-19 responses. The consequences of unemployment on mortality and suicide are well studied and the subject of published meta-anlysis. Additional data exists predict impact on unemployment on other outcomes, including health symptoms, disease conditions, and biological markers. (10–12)
This analysis would be improved with consideration of uncertainties. For example, unemployment following Covid-19 may be short lived or prolonged. Some share of the predicted health effects might be mitigated by national policy countermeasures; however, many studies occurred in countries where social safety nets are already robust.
Unemployment is typically accompanied by other economic factors, such as the loss of income and health insurance which are both associated with increased mortality. These effects are not estimated separately because of limited data on precise income and insurance effects and because doing so may double count adverse health impacts.
The effects of poverty on mortality are consistent across well designed longitudinal cohorts. In the lower third of the income distribution each additional $10 000 of income reduces mortality risk by >50%.(13) Lack of health insurance is an independent predictor of mortality with smaller but consistent and statistically significant effects.(14)
Others have correlated recessions with health improvements. Economists observe that overall mortality rates decline, including due to declines in traffic and possibly improvement in behavioral risk factors. Health impacts due to unemployment are distinct from and are experienced by a different group from the general population. Society has not applied recessions as a tool to improve health.
The application of NPIs raises difficult ethical questions that are not addressed here. For example, school closure might disproportionately impact lower income families who have less access to other educational resources.(15) In the case of Covid-19 specifically, the prevention of deaths among the elderly population might come at the expense of delayed deaths in the working age population resulting from secondary unemployment. Those most able to tolerate lockdowns, such as the retired and those working in the knowledge economy, might have more political influence than those whose day to day livelihood requires a physical presence at work.
In the face of uncertainty and threat, precautionary and decisive action is reasonable. Yet, finding ways to account for and strike a balance among the benefits and costs of alternative courses of action is wise if the goal is optimizing health.(16) The Community Mitigation Guidelines to Prevent Pandemic Influenza--United States recognizes these principles. “To be effective, these measures must be implemented early and strategically targeted, layered, and tailored to pandemic severity, and their public health benefits must be balanced against economic and social costs.”Given the scope of social disruption and the complexity of decision-making, decisions may benefit from being more transparent and participatory.
More monitoring and evaluation may support more infomed trade-offs both now and in the case of future pandemics. This may include studies on the cost-effectiveness of individual NPIs for mitigating hospitalizations and deaths.
Monitoring can also include short term health outcomes and risk factors. Alongside, Covid-19 infections and mortality, we mght monitor suicide crisis calls, markers of glycemic control and blood pressure, ambulatory care sensitive conditions like heart failure and asthma.
Unanticipated effects of societal Covid-19 responses, are also coming into view. Migrant workers are stranded between their former jobs and their homes. Schools in low resource countries may be closing long-term. People are becoming fearful, with implications for social trust. All of these issues merit attention and some value. Sadly, such effects may create new health inequalities and political divisions.
The Chilean economist Manfed Max-Neef astutely pointed out that society often does something to satisfy one need but, tragically, in doing so, undermines others.(17) Optimal health requires the fulfillment of multiple human needs.
Public health planning and action to contain and mitigate Covid-19 might benefit from a more holistic policy framework, greater transparency and public participation. In places where Covid-19 response has led to significant economic and social disruption, policy leaders might develop indicators and tools to monitor these effects. Longer term income support, universal health insurance coverage and other policy interventions may be warranted to mitigate the indirect health impacts of Covid-19 response.
Data Availability
All data used in this analysis is publicly available without a fee.
Conflicts of interest
None
Funding
None
Acknowledgments
Jennifer Hughes for her review of this manuscript.