Web-based self-reporting of symptoms is a growing field and has been used to improve survival in oncology [1,2]; it can be used as a participatory surveillance tool for coronavirus disease (COVID-19) or other influenza-like illnesses as well [3,4]. We thought of applying the same technology to optimize patient triage for COVID-19 patients in France and alleviate the burden on emergency call centers. A self-assessment and participatory surveillance website [5] was developed and launched during the growing phase of the COVID-19 epidemic in France in March 2020. Our objective was to determine if self-reported symptoms could help monitor outbreak dynamics in France. We report here the analysis of the first 13 days of web application usage.

Users were recruited via national media campaigns in France, including social media, radio, and magazine campaigns, between March 17-29, 2020. Participants were recruited through the maladiecoronavirus.fr website [5]. Respondents provided information on sociodemographic data, zip code, and coexisting disorders anonymously. They were asked about nine symptoms associated with possible COVID-19 infection—fever (body temperature >37.7°C), unusual cough, shortness of breath, sore throat, muscle aches, diarrhea, anorexia, and asthenia. Anosmia, a loss of the sense of smell, was added on March 21, 2020. Following symptom reporting, a notification was sent, recommending the user either to stay at home and use the website again in case of evolving symptomatology (self-monitoring), or to contact a general practitioner (GP), or to call an emergency number if they reported dyspnea or anorexia. A questionnaire was built according to Chinese reports and French experience [6]. The website was not considered a medical device by regulatory authorities since no tracking was performed and data were anonymous. We compared the distribution of web-based, self-reported symptoms to that of hospitalized COVID-19 patients according to Ministry of Health reports. Spearman correlation coefficients were used for statistical analysis.

Between March 17-29, 2020, the website was accessed 4,126,789 times; 3,799,535 electronic questionnaires were filled out; and 2,477,174 users had at least one out of the nine symptoms included in the questionnaire (Figure 1). In total, 1,322,361 (34.8%) participants reported no symptoms. The remaining patients (median age 37 years; range 15-99 years) were directed to self-monitoring (858,878, 22.6%), GP visit or teleconsultation (1,033,922, 27.2%), or an emergency phone call (584,374, 15.4%).

Of all symptomatic patients, anosmia was reported by 17.1% (325,910/1,903,741), fever was reported by 33.5% (828,952/2,477,174), and cough by 61.2% (1,515,557/2,477,174). Emergency warning signs (dyspnea or complete anorexia) were reported by 39.1% of participants with anosmia (n=127,586) versus 22.7% of participants without anosmia (n=1,597,289; P<.001; Table 1). Anosmia and fever and/or cough were correlated with COVID-19–related hospitalizations (Spearman correlation coefficients=0.87 and 0.82, respectively; P<.001 for both; Figure 2).

This study suggests that self-reported symptoms of COVID-19 are correlated with COVID-19–related hospitalizations and that anosmia may be strongly associated with COVID-19. This could be explained by a greater specificity compared to other reported symptoms that could result from other respiratory viruses [7,8]. Limitations include lack of medical assessment and virological confirmation of COVID-19 and comparison of application-retrieved data with a distinct set of data from the Ministry of Health reports on hospitalizations. Self-checking application data could be a relevant tool to monitor the dynamics of an outbreak and thus can be a real-time health system response to the epidemic.