Choosing the right COVID-19 indicator: crude mortality, case fatality, and infection fatality rates influence policy preferences, behaviour, and understanding

The COVID-19 pandemic saw a proliferation of different mortality rate metrics used to describe its severity—crude mortality rate (CMR), case fatality rate (CFR), and infection fatality rate (IFR). This study explores how the presentation of these different rates influences public attitudes, policy preferences, and behaviors. The authors highlight the discrepancies in how these rates are presented and interpreted, referencing President Trump's statement about the US having the "number one mortality rate" which actually referred to the CFR, while CNN responded with a per capita death rate (related to CMR). The research seeks to understand the consequences of these differing presentations for public health communication and policy effectiveness. The study will examine the impact of the three mortality rates on people's understanding, attitudes, and policy preferences, as well as their compliance with preventive measures, acknowledging the complexities in calculating and understanding the IFR, especially in the context of COVID-19.

The research draws upon several bodies of literature. First, it connects to studies demonstrating how the framing of COVID-19 information shapes public attitudes, preferences, and behavior (Sotis et al., 2020; Garcia et al., 2020; Van Bavel et al., 2020). Secondly, it engages with research on health literacy, emphasizing its role in facilitating informed health decisions (Nutbeam, 2008; Paasche-Orlow et al., 2005; Paakkari and Okan, 2020). The study also references literature showing how message framing significantly impacts health-related choices, citing the examples of Bolsonaro's dismissal of COVID-19 risks and the resulting decreased compliance in Brazil (Ajzenman et al., 2020), and the impact of mistrust in scientific guidelines (Plohl and Musil, 2020). Finally, it connects to literature on the economic impact of pandemics and the valuation of human life (Rice and Cooper, 1954; Colley, 1976; Weisenthal, 1980; Viscuzi, 2008; Power, 2020).

A double-blind experiment was conducted using a sample of 1,199 US residents recruited via Amazon Mechanical Turk (MTurk). The study used the Qualtrics platform and employed a design that avoided forcing responses to ensure data quality. Participants were randomly assigned to one of three treatment groups (exposed to one of the three mortality rates) or a control group. All participants were presented with sufficient data to calculate all three mortality rates before being asked questions about their assigned rate. The experiment assessed understanding using three questions: calculating the assigned rate using real data from Singapore, identifying the correct definition of the rate, and calculating the rate for a hypothetical disease. Policy preferences and behavior were also measured, along with standard demographic information. An ordinal logit model was employed to analyze the data, controlling for various demographic and news consumption factors.

The results show that the type of COVID-19 indicator significantly impacts attitudes, policy preferences, behavior, and understanding. Compared to the control group, individuals presented with the CMR were 26.9 percentage points more likely to support a mask mandate. Presenting the CFR decreased support for banning indoor events (though this effect was influenced by news consumption). The CMR was the strongest predictor of future worries about a health crisis. Understanding of the rates was generally poor across all groups. The use of the IFR led to less compliance with preventive behaviors, potentially due to difficulty in understanding its calculation. The CMR, by providing the total population size, might have aided in better understanding of the economic implications of the pandemic. The study also found that while providing information on COVID-19 indicators can increase support for preventive policies, it doesn't guarantee increased compliance. In fact, it can even reduce the willingness to engage in preventive behaviors (e.g., using the CMR encouraged attendance at small gatherings, the IFR led to less time spent at home, and the CFR decreased mask usage). These findings suggest a level of moral hypocrisy: people may support policies but not act accordingly. The authors note potential influences of political affiliation (Republicans in the CMR group showed higher misunderstanding of rules), news consumption (via Fox News, which has a history of downplaying the virus), cultural factors (uncertainty avoidance), and risk aversion on the efficacy of different indicators.

The findings contradict previous research showing no correlation between communication strategies and support for preventive measures (Freira et al., 2021), but support research linking pandemic risk perceptions to individual behavior (Allcott et al., 2020). The study highlights the importance of clear communication and understanding of mortality rates in shaping public health policy and behavior. The difficulty in understanding the IFR and its changing estimates are highlighted as potential reasons for its ineffectiveness in influencing behavior. The stronger influence of the CMR is potentially due to its presentation of the overall population size, aiding in the understanding of the potential economic impacts of the pandemic. The observed moral hypocrisy underscores the gap between supporting policies and acting upon them, influenced by factors like political affiliation and news sources.

The study demonstrates that the choice of COVID-19 mortality indicator significantly affects public perception, policy preferences, and behaviors. While the CMR appears most effective in promoting positive actions, public understanding remains a challenge. The authors recommend policymakers provide clear definitions of indicators, prioritize the CMR for general awareness, and utilize the CFR when emphasizing the pandemic's severity. Future research should expand on the findings, examining different populations and pandemic phases.

The study's limitations include potential omitted variable bias, the use of a non-randomized sample (US residents only), and the focus on a single point in time. The findings may not generalize to other populations or pandemic phases. Future research should explore these factors to enhance the generalizability of the results. The potential for omitted variable bias and the limitations of the sample's demographic representation also warrant attention.