Facebook and the University of Maryland (UMD) are analyzing a spike in people reporting symptoms similar to those of COVID-19, such as having experienced fever, along with cough, shortness of breath, or difficulty breathing, based on the COVID-19 Symptom Survey data. The change in data coincides with the release of a new version of the survey questionnaire that went into effect on May 20. These methodological updates are common and intended to improve the survey’s accuracy or increase the number of respondents. In this case, the revision of the survey was also intended to help align the survey conducted in the United States by Carnegie Mellon University with the survey run globally by the UMD.
To test the effects of the changes to the survey and in accordance with standard procedure, 20% of respondents received the old version and 80% received the new version of the survey. Having analyzed the data, our partners saw that the data after May 20 was likely impacted by the survey questionnaire changes, and therefore estimates of COVID-19-like symptoms should not be directly compared to the data from previous weeks.
Specifically, this issue impacted the results in Brazil, where we have seen a spike in people reporting symptoms similar to those of COVID-19. Because we determined that the spike was affected by the questionnaire change, the data after May 20 cannot be compared to previous dates.
It is important to note that the new version of the survey questionnaire was communicated by UMD on its website (see What’s New section) and also sent to researchers using the data that are subscribed via email. Moving forward, however, we will communicate any survey questionnaire improvements more clearly on our Data for Good website and make them visible to people who see our data visualizations.