Assessment of knowledge on COVID-19 and estimation of the factors influencing the knowledge

ABSTRACT

RationaleKnowledge and awareness of COVID-19 likely results in better compliance with the guidelines set forth by local governments and may help prevent community spread of the disease. We designed a population-based survey to estimate the difference in knowledge among different groups of individuals based on sociodemographic diversities. The secondary objective was to identify significant determinants of knowledge on COVID-19. Methods: The survey questionnaire was built in Redcap. Knowledge-related questions were based on 34 metrics, including risk factors for severe infection, disease spread, prevention, treatment, and general information. Most of the questions were dichotomous ('Yes' or 'No' response). We used social media platforms, Studyfinder, and Researchmatch to send the survey link. Five sociodemographic determinants were considered potential predictors of knowledge, including age, race, education level, gender, and Healthcare Worker. Participants were also asked about their preferred source of information on COVID-19. The data was collected between June to November 2020. The 'Factor analysis' function in 'SPSS' was used to convert the 34-knowledge metrics into a single standardized scale of '0-10' for further analyses. Generalized Linear Model was built to measure the degree of association among the predictors and the knowledge score, based on 'model effect'. We further used the Machine learning tool 'Neural network' to generate a rank list of the knowledge determinants. Results: 1139 participants (male female = 13) from all 50 US states participated in the survey. The generalized linear model demonstrated that the chosen determinants could precisely predict the knowledge score (χ2=93.68,p<0.001). All predictors, except healthcare workers, had significant associations with the knowledge score. Race had the highest association followed by Education, Gender, and Age (χ2=69.29,p<0.001;χ2=15.35,p<0.001;χ2=12.34,p=0.002 and χ2=4.04,p=0.044, respectively). Neural network reproduced the exact rank list with the normalized importance for Race, Education, Gender, and Age, which were 100%, 24.6%, 20%, 15.8%, respectively. Participants belonging to the 'White' race had a significantly higher score (7.71±1.13) compared to the 'Black' race (6.77±1.69) (p<0.001), while 'Female' participants performed better (7.60±1.39) compared to 'Male' (7.20±1.61) (p<0.001). Younger participants (18-44 years) had statistically significantly lower knowledge compared to older age groups (>60years) (p<0.05). Among available sources of information on COVID-19, 'City/State websites' was the most popular (67.9% favored 'Yes' vs. 'No'), followed closely by 'Television' (67.8%) and 'CDC website' (62.2%). Conclusions: 'Race' was the strongest sociodemographic predictor of COVID-related knowledge, and white, older, female participants with higher education (Masters/Ph.D.) demonstrated better knowledge than the rest of the population.