Discussion

The perceived risks of illness have long been associated with health behavior; however, much of the literature on risk perceptions refers only to the probability of contracting a disease. ^{Reference Ferrer and Klein5} This study evaluated perceptions of 3 types of risks: infection, hospitalization, and death. A large proportion of individuals infected with COVID-19 are either asymptomatic (approximately 40%) or have mild symptoms; thus, the probability of infection may provide only limited information about risk perceptions that affect health behaviors. ²⁷ The degree of association between these 3 risk measures was relatively weak, therefore, they cannot be used as proxies.

Some authors have argued that a more precise measure of risk of illness involves the probability of infection and a measure of the disease’s expected severity. ^{Reference Leppin and Aro28} This new measure involves multiplying the perceived risk of infection by the probability of hospitalization or death (both provide measures of the expected severity of illness). Such a measure can be interpreted as the perceived joint probability of infection and the adverse health outcomes (hospitalization or death). This interpretation is based on the formula for conditional probability: P(B∣A) = P(A∩B)/P(A), in which A refers to infection, and B refers to the adverse health outcome (hospitalization or death), ‘∣’ denotes conditional probability, and ‘Ո’ denotes joint probability. Thus, the joint probability of infection and an adverse health outcome is P(A∩B) = P(B∣A)*P(A). ^{Reference Hogg, Tanis and Zimmerman29} Note that the question about the risk of infection corresponds to the unconditional probability P(A), while the questions about the risk of hospitalization and death correspond to conditional probabilities (P(B∣A)). More work would validate these risk perception measures and relate them to more formal conceptual constructs, and individuals’ expectations of the costs of a disease. Consistent with previous COVID-19 studies, each risk perception measure is discussed separately below.

The estimated median values of the risk of infection reported in Ecuador (30%) and Kenya (20%) were similar to those in the United States (32%) but were significantly higher than those in Indonesia (10%): countries where comparable studies had been conducted at approximately the same time using comparable risk elicitation methods. These values are significantly higher than the official rates of infection. The cumulative risks of infection reported by December 12, 2020, were lower than the perceived risks. Ecuador reported 200000 confirmed cases in a population of 17 million, and Kenya reported 19000 cases in a population of 51 million. Similarly, median perceived risks of death in Ecuador (20%) and Kenya (12%) were significatively higher than the reported fatality ratios of 6.9% and 1.7%, respectively (also reported by December 12, 2020). However, it is important to note that official statistics may differ from actual statistics because of limited testing capacity, for example. Also, in both countries, the number of cases estimated using several epidemiological models is significantly higher than the number of confirmed cases. ^{Reference Giattino30}

In short, there is a large difference between the risks of infection and death that governments and researchers have estimated in both countries, and the greater risks individuals in the population perceived. These large perceived risks of infection and adverse health outcomes did not translate into total compliance with health prevention measures. For example, mask use reported by the Institute for Health Metrics and Evaluation (IHME) was estimated at approximately 70% in Ecuador, and 50% in Kenya during the time of data collection (both values are closer to the maximum values reported throughout the pandemic). ³¹

In both countries, risk perceptions tended to be homogenous across age groups, as respondents’ ages were not associated with any of the perceived risks. The lack of association between age and hospitalization and death risks could be problematic as, older individuals were more susceptible to the disease at the beginning of the pandemic; however, it is important to note that perceptions of the risks were high overall and the sample tended to be younger, which may have hindered the ability to detect a relation. This finding may also reflect the lack of communication about these risks to the population. While the risk of infection can be reduced by individual behavioral changes, the probability of being hospitalized or dying after contracting the disease depends less upon individual behavior. This finding contrasts with studies from Indonesia and the United States that reported higher perceived risks of infection among individuals in their 20s to 40s relative to older individuals. ^{Reference Harapan, Anwar and Nainu15,Reference Masters, Shih and Bukoff16}

A consistently negative and significant association between income and the risk of death was found in both countries. This may be related to perceptions of the quality of services available to poorer individuals relative to wealthier individuals. However, the estimated effects were very small. In Ecuador, a $100 increase in income was associated with only a 0.42% reduction in the perceived risk of death, while in Kenya, a $100 increase in income was associated with a 0.83% reduction in the perceived risk of dying. Larger, significant, and consistent associations were found between gender and risk perceptions in both countries. Relative to men, women’s perceived risk of dying was 2.9% and 4.7% higher in Kenya and Ecuador, respectively. Similarly, women’s perceived risk of contracting the disease was 1.6% and 4.6% lower than men in Kenya and Ecuador, respectively (significant only in Ecuador). Previous studies that have evaluated COVID-19-related risk perceptions found no differences between genders or did not consider this variable relevant. ^{Reference Masters, Shih and Bukoff16} At the beginning of the pandemic, little was known about differences in the associations between gender, infection, and mortality. ^{Reference Harapan, Anwar and Nainu15} A recent study with data from 17 countries reported a higher infection rate among women (+15%), but higher fatality rates among men (+50%). ^{Reference Seeland, Coluzzi and Simmaco32} It is possible that women’s final exposure was greater than expected initially, given their role as primary homemakers in charge of procuring food for the household and caring for the sick. Some research has also shown that women in developing countries were less likely to have jobs that could be conducted remotely. ^{Reference Beltrán-Castillo and Corti-Velazquez33}

Large differences in risk perceptions between locations and regions were found in both countries, but they were not always consistent. For example, while the perceived risk of dying was higher among rural residents in Ecuador (+3.1%), it was lower in Kenya (−2.3%) compared to urban residents, ceteris paribus. These findings suggest that individuals are aware of potential regional differences in the incidence of the disease and the importance of context (e.g., health infrastructure) in health outcomes.