Responsiveness to risk explains large variation in COVID-19 mortality across countries
This article has been Reviewed by the following groups
Listed in
- Evaluated articles (ScreenIT)
Abstract
Background
Health outcomes from the COVID-19 pandemic vary widely across countries – by 2022 New Zealand suffered ∼10 total deaths per million people, whereas the U.K. and U.S. have had over 2000. Differences in infection fatality rates are insufficient to explain such vastly divergent outcomes. We propose that endogenous behavioral responses to risk shape countries’ epidemic trajectories, and that differences in responsiveness to risk are a primary driver of variation in epidemic scale and resultant mortality.
Methods
We develop several testable predictions based on the proposed endogenous risk response mechanism. We test these using a simple modified SEIR model incorporating this mechanism, which we estimate for 131 countries (5.96 billion people) using data on daily reported SARS-CoV-2 infections and COVID-19 deaths. We further examine associations between COVID-19 deaths and several observed and model-estimated country characteristics using linear regression.
Findings
We find empirical support for all predictions tested: 1) endogenous risk response substantially improves an SEIR model’s fit to data (mean absolute errors normalized by mean=66% across countries, vs. 551% without endogenous risk response); 2) R e converges to ∼1 across countries in both empirical data and model estimates with endogenous risk response (but not without it); 3) most cross-country variation in death rates cannot be explained by intuitively important factors like hospital capacity or policy response stringency; and 4) responsiveness to risk, which governs the sensitivity of the endogenous risk response, correlates strongly with death rates (R 2 =0.75) and is the strongest explanatory factor for cross-country variation therein.
Interpretation
Countries converge to policy measures consistent with R e ∼1 (or exponentially growing outbreaks will compel them to increase restrictions). Responsiveness to risk, i.e. how readily a country adopts the required measures, shapes long-term cases and deaths. With greater responsiveness, many countries could considerably improve pandemic outcomes without imposing more restrictive control policies.
Funding
None.
Research in context
Evidence before this study
While numerous studies have examined drivers of variation in COVID-19 infection fatality rates (e.g. age, comorbidities), relatively few have sought to explain cross-country variation in overall mortality outcomes. Mortality outcomes are shaped primarily by variation in infection rates and only secondarily by IFR variation, yet the drivers of cross-country variation in infection rates are poorly understood as well. There is ample evidence that policy responses such as mask mandates, quarantines, and other non-pharmaceutical interventions (NPIs) can influence transmission, yet across countries the association of policy responses with long-term infection rates remains weak.
Added value of this study
We propose a simple and intuitive theoretical mechanism – endogenous behavioral response to risk – to explain variation in cases and deaths across countries. While simple, the implications of this mechanism are rarely examined; a recent review found only 1 of 61 models in the CDC COVID-19 Forecast Hub incorporates endogenous risk response. We demonstrate how it can explain several empirical regularities, including multiple outbreak waves and convergence in effective reproduction number, which existing models largely do not explain. Our empirical estimates of responsiveness to risk also show that it drives wide cross-country variation in mortality outcomes, which otherwise remains unexplained.
Implications of all the available evidence
Our results highlight the key role of responsiveness to risk in shaping COVID-19 mortality outcomes. They suggest that adopting similar policy responses but with greater responsiveness could improve outcomes and reduce mortality. The responsiveness mechanism also resolves the apparent paradox that despite clear evidence of proximal NPI effectiveness, their impact on infection and death outcomes in cross-country comparisons is rather weak. Our results highlight the need to understand better the determinants of differences in responsiveness across countries, and how to improve it.
Article activity feed
-
-
-
SciScore for 10.1101/2020.12.11.20247924: (What is this?)
Please note, not all rigor criteria are appropriate for all manuscripts.
Table 1: Rigor
NIH rigor criteria are not applicable to paper type.Table 2: Resources
No key resources detected.
Results from OddPub: Thank you for sharing your code and data.
Results from LimitationRecognizer: We detected the following sentences addressing limitations in the study:Our simple model was designed primarily for investigating the tradeoffs between contact reduction and deaths and includes several limitations. One set of extensions may model the epidemic in more detail. For example, one could more explicitly model under-reporting processes, use testing data to inform the analysis, account for changing fatality rates over time, and couple model parameters across nations in a hierarchical Bayesian …
SciScore for 10.1101/2020.12.11.20247924: (What is this?)
Please note, not all rigor criteria are appropriate for all manuscripts.
Table 1: Rigor
NIH rigor criteria are not applicable to paper type.Table 2: Resources
No key resources detected.
Results from OddPub: Thank you for sharing your code and data.
Results from LimitationRecognizer: We detected the following sentences addressing limitations in the study:Our simple model was designed primarily for investigating the tradeoffs between contact reduction and deaths and includes several limitations. One set of extensions may model the epidemic in more detail. For example, one could more explicitly model under-reporting processes, use testing data to inform the analysis, account for changing fatality rates over time, and couple model parameters across nations in a hierarchical Bayesian framework. A second set of extensions could focus on better representation of various policies and economic costs. We do not differentiate between policies used to reduce contact rates (e.g. contact tracing, isolation of vulnerable groups, school closures, masks, etc.). Some policies are more efficient in reducing contacts with less overall disruption to social life and economic activity (Benzell et al. 2020, Han et al. 2020). With effective policy sets, some nations may better maximize economic activity given the limited contact ‘budget’ they have based on Re∼1 condition. Another valuable extension would be to use economic output data to more directly test the predicted (lack of) tradeoff between saving lives and maintaining economic activities. A third line of further research could explore the intriguing possibility that better economic outcomes may actually coincide with fewer deaths (Correia et al. 1918, Hasell 2020). Our analysis hints at this possibility, as we find a negative correlation between deaths and normalized contacts. Part of this ne...
Results from TrialIdentifier: No clinical trial numbers were referenced.
Results from Barzooka: We did not find any issues relating to the usage of bar graphs.
Results from JetFighter: We did not find any issues relating to colormaps.
Results from rtransparent:- Thank you for including a conflict of interest statement. Authors are encouraged to include this statement when submitting to a journal.
- Thank you for including a funding statement. Authors are encouraged to include this statement when submitting to a journal.
- No protocol registration statement was detected.
-