Predictors of COVID-19 perceived susceptibility: insights from population-based self-reported survey during lockdown in the United States

Abstract

Background: The COVID-19 pandemic during lockdown has highlighted the importance of identifying individuals most at risk of infection with SARS-CoV-2, underscoring the need to assess factors contributing to susceptibility to disease. With the rapidly evolving nature of the pandemic and its new variants, there is an inadequate understanding on whether there are certain factors such as a specific symptom or collection of symptoms that combined with life-style behaviors may be useful to predict susceptibility. The study aims to explore such factors from pre-vaccination data to guide public health response to potential new waves.

Methods: An anonymous electronic survey was distributed through social media during the lockdown period in the United States from April to June 2020. Respondents were questioned regarding COVID testing, presenting symptoms, demographic information, comorbidities, and confirmation of COVID-19 test results. Stepwise logistic regression was used to identify predictors for COVID-19 perceived susceptibility. Selected classifiers were assessed for prediction performance using area under receiver operating characteristic (AUROC) curve analysis.

Results: A total of 130 participants deemed as susceptible because they self-reported their perception of having COVID-19 (but without the evidence of positive test) were compared with 130 individuals with documented negative test results. Participants had a mean age of 45 years, and 165 (63%) were female. Final multivariable model showed significant associations with perceived susceptibility for the following variables: fever (OR:33.5; 95%CI: 3.9,85.9), body ache (OR:3.0; 95%CI:1.1,6.4), contact history (OR:2.7; 95%CI:1.1,6.4), age> 50 (OR:2.7; 95%CI:1.1, 6.6) and smoking (OR:3.3; 95%CI: 1.2,9.1) after adjusting for other symptoms and presence of comorbid conditions. The AUROC ranged from poor to fair (0.65-0.76) for cluster of symptoms but improved to a good model (AUROC = 0.803) after inclusion of sociodemographic and lifestyle behaviors e.g., age and smoking tobacco.

Conclusions: Fever and body aches suggest association with perceived COVID-19 susceptibility in the presence of demographic and lifestyle behaviors. Using other constitutional and respiratory symptoms with fever and body aches, the parsimonious classifier correctly predicts 80.3% of COVID-19 perceived susceptibility. A larger cohort of respondents will be needed to study and refine classifier performance in future lockdowns and with expected surge of new variants of COVID-19 pandemic.

Keywords: COVID-19; Infections; Population survey; Predictors; Susceptibility.