Association between the risk perception of contracting COVID-19 and sociodemographic characteristics in a Peruvian population

Participants

The population consisted of residents from four cities in Peru (Callao, La Libertad, Lambayeque, and Lima Metropolitana). The study included residents of the selected cities who had access to an internet-connected device. Participants under 18 years old and those who were unavailable to complete the questionnaire were excluded. The sample size was determined in 492 participants considering that it was expected to find moderate to small size effects (f = 0.15) and a statistical power of 80%. However, data was collected from 1044 participants due to the fact possibility of having significant losses in data due to non-complete answers and also to increase the statistical power. Participants taking part in the online survey were chosen using a convenience sampling method.

Study design and instruments

This was an analytical and cross-sectional study in which participants were surveyed using a self-administered questionnaire developed in Google Forms.³¹ The questionnaire was distributed via email and social media from October to December, 2021.

The online form consisted of three sections. The first section included the informed consent. The second section collected information on sociodemographic characteristics (sex, age in terms of the complete years from your birthday, marital status, highest level of education, occupation, personal and family history of COVID-19 in the household). The third section had the scale of risk perception of contracting the COVID-19 infection, which was designed based on the Health Belief Model (HBM) and validated in Brazil.¹¹ The scale had 24 questions that assessed five factors: a) perceived susceptibility, corresponding to knowledge and belief about the possibility of getting coronavirus; b) perceived severity, asking about personal beliefs about how the individual would experience the disease process and the intensity of symptoms; c) perceived benefits, corresponding to the effectiveness of adopted behavioral mechanisms to prevent infection; d) perceived barriers, aiming to understand the difficulties in adhering to protective and preventive measures for coronavirus transmission, and e) health motivation, seeking to improve overall health. Responses were rated on an analog scale from 0 to 100, where zero represented “not at all” and 100 represented “extremely high.”

Before the validation process, a back-translation was performed. Firstly, the questionnaire was translated from Portuguese to Spanish by a certified translator, and then vice versa. Subsequently, a pilot test was conducted with 30 participants to evaluate the comprehension of the questions and response options. In this activity, it was identified that participants had difficulty rating the probability of risk (response option) with a value greater than 100. This is why the range of response options was modified from the original version, which ranged from 0 to 120, to a range from 0 to 100.

The content of the instrument went through a process of expert judgment validation. In this process, health professionals who are methodologists, members of the COVID-19 command, and epidemiologists.

In order to explore the reliability of the instrument, a split-half correlation procedure was conducted, correlating all possible halves of items. Based on the obtained correlations, a data distribution was obtained for each subscale of the instrument, and the 2.5th, 50th, and 97.5th percentiles are reported. Split-half correlations with a score of 0.70 or higher are considered adequate.

Reliability analysis

Table 1 shows the results of the reliability analysis conducted for the scales of perception of COVID-19 contagion risk and its factors. Specifically, we can observe that the median split-half correlations for the general risk perception scale (0.91), susceptibility scale (0.83), and severity scale (0.88) are considerably higher than 0.70, indicating that these scales have very good levels of reliability. On the other hand, the severity, benefit, and barrier scales have low levels of reliability as the median split-half correlations are below 0.70 (0.67, 0.53, 0.57, respectively). Regarding the descriptive statistics of the subscales, it can be observed that the highest average perception of risk is found in the barrier subscale (M = 42.34, SD = 18.03), while the lowest average perception of risk is observed in the benefit perception (M = 28.90, SD = 15.68), followed by susceptibility perception (M = 28.94, SD = 19.50).

Statistical analysis

To ascertain the differences in susceptibility to COVID-19 based on sex, marital status, education level, occupation, and family history, a combination of Student’s t-tests and Analysis of Variance (ANOVA) was utilized, contingent on the number of categories within the variable under consideration. When statistically significant disparities were identified using ANOVA, a post hoc Tukey analysis was implemented to elucidate the nature of these differences. Consistently, the assumptions of parametric analyses were validated by constructing half-normal simulation plots for each model to assess the distribution of residuals and homoscedasticity. As indicated by Moral et al.,¹³ Conventional tests for normality and homoscedasticity, such as Shapiro-Wilk and Bartlett, can be sensitive to large sample sizes. A more robust method to verify the adherence to these assumptions is to construct plots that juxtapose residuals with a half-normal simulation, while also illustrating 95% confidence bands to determine if the observed distribution aligns with outcomes where the assumptions are satisfied. Visual representations of these analyses can be viewed in Supplemental Material, confirming that in all instances the assumptions of parametric analysis were met. Lastly, to investigate the correlation between age and susceptibility to COVID-19, a Spearman correlation coefficient was computed, given that age did not adhere to a normal distribution. All analyses were executed using R software v4.2.1.¹⁴

Potential biases

Due to the absence of an available sampling frame and the utilization of convenience sampling, the sample obtained in this study was not representative of the population across the four Peruvian cities, limiting the ability to generalize the findings. Nonetheless, the number of participants in this study, nearly twice the minimum sample size, allows for statistically robust results. On another note, as the survey was conducted virtually, the results may be subjected to biases, such as social desirability bias—the tendency for respondents to provide socially acceptable responses rather than responding truthfully.

Lastly, given that the participants had to respond numerically to the risk perception questionnaire, potential difficulty in providing responses may have arisen, as some individuals have limitations in quantitatively evaluating their perceptions, or confusion surrounding the interpretation of the questions. For this reason, clear instructions were provided in the questionnaire, and each question reiterated the numerical scale that could be used for responses. In addition, rigorous quality control was conducted on the responses to select the questionnaires that were considered for statistical analysis.

Ethical considerations

The study was approved on May 24, 2021 by the Institutional Ethics Committee of Norbert Wiener University (file number 560-2021). All participants provided informed consent before responding to the questionnaire. The data collected through the survey in each city were coded, excluding identifying information, ensuring the confidentiality of the information and exclusive access to the data by the researchers.

Sociodemographic data

A total of 1044 participants answered, of which 223 were excluded for quality control purposes because they provided letters or words instead of indicating a number between 0 and 100 in their responses.³³ Out of the 821 participants, 53.71% (n = 441) were women, with an average age of 28.29 years (SD = 11.56). Regarding marital status, 76.49% (n = 628) reported being single, 20.95% (n = 172) reported being married, while 2.56% (n = 21) reported being separated or divorced. In terms of participants’ education level, it was observed that 73.33% (n = 602) had a higher education level, while 26.31% (n = 216) had completed secondary education. In regard to the occupation of the participants, 45.55% (n = 374) reported being self-employed, 42.39% (n = 348) reported being employed, 11.33% (n = 93) reported being homemakers, and the remaining 0.73% (n = 6) indicated being retired. Finally, with regard to the history of COVID-19 among family members or individuals living in the same household, at the time of the survey, 40.44% (n = 332) of the participants reported that no family member or household member had contracted COVID-19, 28.01% (n = 230) reported that one family member had been infected, while the remaining 31.55% (n = 259) reported that several members of their family and household had been infected with COVID-19.

In Table 2, the Pearson correlations for the different scales of risk perception can be observed. In this table, it can be seen that the general scale of risk perception maintains a high and significant correlation with all the subscales (0.72 < r’s < 0.80). On the other hand, the different subscales show moderate to strong correlations if (0.32 < r’s < 0.60).

Analysis of risk perception of contracting COVID-19

In Table 3, the Student’s t-tests for the means of risk perception of contracting the COVID-19 infection can be observed according to the reported sex of the participants. Specifically, it can be indicated that none of the subscales showed statistically significant differences (p’s > 0.5). These results indicate that the risk perception of contracting the COVID-19 infection is the same for men and women.

In Table 4, the Analysis of Variance (ANOVA) comparisons of the means of risk perception of contracting the COVID-19 infection can be observed according to the marital status reported by the participants. Particularly, it can be observed that none of the subscales showed statistically significant differences (p’s > 0.5). These results indicate that the risk perception of contracting the COVID-19 infection is the same for married, single, and separated participants.

In Table 5, we can observe the results of the Analysis of Variance (ANOVA) for the means of risk perception of contracting the COVID-19 infection according to the participants’ educational level. For this sociodemographic variable, significant differences were found in the benefit subscale based on educational level (F = 5.75, df = 2,818, p = 0.003). According to the post-hoc Tukey analysis, participants with a secondary education level have a significantly higher benefit score than those with a higher education level (p = 0.004). No other significant differences were observed in any of the other subscales of risk perception of contracting the COVID-19 infection.

In Table 6, we can see the results of the Analysis of Variance (ANOVA) for the mean scores of risk perception of contracting the COVID-19 infection according to participants’ occupation. For this variable, significant differences were observed in the susceptibility subscale (F = 7.47, df = 3,817, p < 0.001). According to the post-hoc Tukey analysis, participants who are employed reported higher levels of susceptibility compared to homemakers (p = 0.008). Similarly, participants who are employed reported higher levels of susceptibility compared to self-employed participants (p < 0.001). No significant differences were found in the other subscales of the instrument.

Table 7 shows the results of the ANOVA for the mean scores of risk perception of contracting the COVID-19 infection according to participants’ family history of COVID-19. Significant differences were observed in the general perceived risk scale (F = 3.06, df = 2,818, p = 0.047) and the susceptibility scale (F = 9.66, df = 2,818, p < 0.001) for this sociodemographic variable. Following post-hoc Tukey analysis, it was found that participants with multiple infected family members had higher levels of perceived risk on the general perceived risk scale compared to participants with no infected family members (p = 0.04). In terms of the susceptibility scale, participants with no infected family members reported lower levels of susceptibility compared to those with at least one infected family member (p = 0.03), and those with multiple infected family members (p < 0.001). No other significant differences were observed in any of the other subscales of risk perception of contracting the COVID-19 infection.

Finally, regarding the correlation between age and perceived risk, Table 8 reveals that only the susceptibility scale shows a positive, weak, and significant relationship with age (Rho = 0.11, p = 0.002). This indicates that older participants in the sample reported higher susceptibility scores compared to younger participants. None of the other subscales showed significant associations.

Implications

Overall, the findings of this study, in line with the literature, highlight the impact of sociodemographic variables on susceptibility perception and perception of benefits. This aspect can be useful in targeting public health interventions aimed at calibrating risk perception in the population to promote compliance with preventive measures, as preventive behavior is only evident when the event is perceived as highly contagious or dangerous.²⁶ While risk perception influences the adoption of preventive behaviors during a pandemic,^23,29,30 especially the affective dimension, this construct alone is not sufficient to promote the adoption of these behaviors.²⁷ Therefore, to reduce the incidence of the disease, it is necessary to consider the sociodemographic characteristics of the population, complemented by the issuance of prevention guidelines, and thus ensure adequate vaccination coverage.^31,32

Another implication of this study stems from the psychometric analysis conducted, which demonstrated that the scale developed in Brazil to measure the perception of risk behaviors for COVID-19 contagion based on the health belief model¹² exhibits adequate psychometric properties in the Peruvian population. This suggests that this scale could be used in other Latin American countries with a previous a validation process.

Limitations

The first limitation is that data collection was based on self-reporting, which could increase the likelihood of common method variance. However, these results represent an initial approach to the evaluated phenomenon and should be verified using complementary methodologies to control this effect. The second limitation is related to the convenience sampling method employed, which allowed quick data collection and access to remote areas of Peru during the pandemic but diminished the representativeness of the sample. The third limitation is the unequal composition of the sample according to sociodemographic characteristics, which could impact the detection of differences. Nevertheless, the statistical tests employed enabled the identification of statistically significant differences. Despite these limitations, this study reports findings from residents of different cities in Peru.