The determinants of COVID-related perceptions

Published: 22-02-2021| Version 2 | DOI: 10.17632/892g33m7hh.2
Contributors:
Pritish Mondal,

Description

We used an online survey to estimate the impact of various sociodemographic factors on the COVID-related perceptions measured by stress, knowledge, and adoption of preventive behaviors. The working hypothesis was that the sociodemographic factors such as age, race, education level, family income, gender would influence the perception of COVID-19, and prediction models built on those factors would precisely forecast preventive behaviors. The survey questionnaire was built and sent out via red cap. The primary outcome of the study was COVID-related perceptions. The study participants were divided into categories based on several sociodemographic factors. Fifty US states were reorganized into nine regions based on the census bureau's recommendation. We also asked whether the participant was a healthcare worker (HCW) (yes vs. no), had a family member with COVID-19 (yes vs. no). We also divided the survey respondents into pre vs. post-vaccine launch, through November 1st, when Pfizer announced a successful vaccine trial and a new wave of pandemic hit the USA. A new variable was computed to estimate perceived threat based on age, HCW, chances of having severe COVID infection, state of COVID in his area. We used 'factor analyses' to abbreviate the metrics on stress, knowledge, preventive behavior, and perceived risk into normalized combined nominal scores (0-10), respectively. Prediction models for COVID-related perceptions were built using a generalized linear model. Variables such as age, race, education level, family income, gender, US region, HCW, healthcare access, and vaccine launch period, perceived threat were considered as potential predictors. Three thousand seven hundred thirty-four US-based adults responded to the survey link and were included in the study. The participants were predominantly white and females. Gender and age were the highest predictors of stress, while participants with better education and income performed better with knowledge related questions. Females, participants with higher education, and African-Americans or Hispanics had a better attitude to preventive behaviors than the remaining population. The GLM prediction models performed better compared to null-models for all three metrics of COVID-related perceptions.

Files

Steps to reproduce

The basic structure in the excel document is unchanged from the original downloaded CSV file from the Redcap. The following variables were re-coded or added to facilitate study analyses. A. Recoding: All the re-coded variables were marked in green. 1) Child, hosp_rate, hosp_rate, death_rate, incubation, corona, animal, screening, climate, sunlight, nasal_spray, hot_beverages, baby_wipes: we changed the scoring compared to the codebook. We scored response 0 in the codebook as 1, and response 1 in the codebook as 0. 2) vaccine_pcv_flu: response 1-3 in the codebook were incorrect response (scored 0), and response 4 in the codebook was correct response (scored 1) 3) mask_gloves_2: We swapped 0 and 1, and 1 in our datasheet should represent a correct response now. 4) social_distance_feet, isolation: response 1,3,4 in the codebook were incorrect, thus scored 0, and response 2 was correct and scored 1. 5) corona live: response 1,3 in the codebook were incorrect and scored 0, and response 2 scored 1. 6) spread: response 3,4 in the codebook were incorrect and scored 0, and response 1,2 scored 1. 7) chloroquin: response 1,2 codebook were incorrect and scored 0, response 3 codebook was partially correct and scored 0.5, response 4,5 in codebook was correct and scored 1. 8) hospital, vent, administration: Response 1 to 5 in the codebook was upside down; now they are scored from 5 to 1, respectively. B. Addition: Few other variables were computed and highlighted in yellow. For example, we used factor analyses to consolidate stress, preventive behavior, perceived threat and knowledge scores into normalized nominal scores. C. Reorganize categorization: We regrouped Racial distribution. All the other races apart from whites, African-American, Asian, Hispanic were categorized as others. Educational categories were categorized too. Middle school and high school were categorized as <high school.