Google and Facebook searches also collect data in different ways. The Google search is a unique question, written by CMU: “Do you know anyone in your community who is sick (fever, cough, shortness of breath or difficulty breathing) now?” The answer options are Yes, No and I’m not sure. Google will display the search box on Google-owned products, including the Google Opinion Rewards app, and on content such as news articles that are part of the Google search publisher network. You can have access to an article, for example, if you complete the search.
Facebook is acting as yet another promoter of CMU’s own research. The CMU survey is a detailed questionnaire consisting of at least a dozen questions about the participant’s age, postal code, family size, symptoms, attempts to connect with doctors or get tested for Covid-19 and interactions with people outside the country. immediate family. When the symptom search appears in someone’s Facebook news feed, and the user clicks on it, it will be taken to the CMU website, outside of Facebook.
The COVIDCast map finally appears as a large map of the United States, with five tabs to separate each data source. On the map side, there is the option to view Covid indicators by state, metropolitan area or municipality (the most granular option); and analyze the current intensity of cases or trends of intensity in the last seven days. In its current version, the map is very obviously what Adelphi might call “instant transmission”, or perhaps an almost final release; makes no predictions.
The goal is to do that eventually. “It’s useful to think of Covid-19 as a gravity pyramid,” says Rosenfeld, with people at the bottom who are not infected, people who have Covid-19, but who may not have symptoms, and others who have symptoms, but don’t ‘don’t see a doctor; even hospitalized people end up in intensive care or die of Covid-19 or related complications.
“The bottom is much more difficult to measure, but what happens at the bottom penetrates to the top. So, if you have an increase in symptoms reported in a specific region, you can expect an increase in doctor visits a few days later and perhaps predict an increase in hospitalizations after that, ”says Rosenfeld.
Tibshirani, the other team leader, says the Delphi team is hardly the first research group to use symptom research to try to identify Covid-19 outbreaks. “There are probably 15 such surveys that I could cite,” he says.
An example is Covid Near Year, a crowdsourcing symptom tracker led by John Brownstein at Boston Children’s Hospital and a team of volunteer bioinformatics from companies like Apple, Amazon and Google. If survey participants indicate that they are not feeling well, they will be asked to go through a more intense questionnaire. It will not provide diagnostics because WIREDMaryn McKenna reported, but could alert health officials about where Covid-19 could go up next.
But a big part of CMU’s strategy was to get Big Tech to implement these surveys “because it would help in creating a data source with a high sample size and that would be maintained at a high sample size in the coming months”, says Tibshirani. To date, about one million Facebook users per week have responded to the CMU survey, while nearly 600,000 Google users respond to the survey hosted by Google in a single question every day.
CMU researchers acknowledge that some data may be incomplete or biased due to the fact that participants report their symptoms. Majumder of Boston Children’s Hospital says that this type of syndromic surveillance can be “highly imperfect science”. If correction methods are not used, research-based work may result in the potential overestimation of Covid-19 cases in a given population. Even if correction methods are used, they are not perfect, she says. “In other words, people with seasonal allergies can be accidentally counted as Covid-19 simply because they reported a dry cough in their research,” she told WIRED.