Case Study

Delphi COVID-19 Symptoms Survey Analysis

Challenge: The University of Maryland and Carnegie Mellon University have disseminated global and U.S. COVID-19 survey data collected by Facebook. Since the inception of the pandemic, these two universities have provided data products to help inform public health policies by U.S. and international public health organizations. Summit served as part of the research team, helping both universities produce data products for public consumption.

Solution: Summit supported the universities by automating the production of weekly and monthly survey results tables, exploring possible bias correction methods, and conducting ad hoc analyses for both the global and U.S. surveys. We produced contingency tables, which aggregated the data by time period (weeks or months), geography (nations or regions), and various demographic characteristics. In addition, we provided an automated solution that read in and cleaned the survey microdata, performed the calculations, and exported the files each week. These files were posted on the University of Maryland’s server for researchers, public health organizations, and ministries of health to use.  

Result: The Delphi survey data exhibited deviations in vaccine uptake, vaccine acceptance, and other measures compared to CDC benchmarks. Summit conducted an exploratory analysis to determine whether the deviations could be mitigated. Using the demographic variables available in the survey, including race, gender, education, and age, Summit conducted post-stratification, Generalized Regression (GREG) estimation, inverse propensity score weighting, and raking. We identified promising methods that adjusted a portion of the deviation and provided recommendations for further exploration.