Active syndromic surveillance of COVID-19 in Israel

Syndromic surveillance systems monitor disease indicators to detect emergence of diseases and track their progression. Here, we report on a rapidly deployed active syndromic surveillance system for tracking COVID-19 in Israel. The system was a novel combination of active and passive components: Ads were shown to people searching for COVID-19 symptoms on the Google search engine. Those who clicked on the ads were referred to a chat bot which helped them decide whether they needed urgent medical care. Through its conversion optimization mechanism, the ad system was guided to focus on those people who required such care. Over 6 months, the ads were shown approximately 214,000 times and clicked on 12,000 times, and 722 people were informed they needed urgent care. Click rates on ads and the fraction of people deemed to require urgent care were correlated with the hospitalization rate (R2=0.54">R2=0.54

$R^2=0.54$

and R2=0.50">R2=0.50

$R^2=0.50$

, respectively) with a lead time of 9 days. Males and younger people were more likely to use the system, and younger people were more likely to be determined to require urgent care (slope: −0.009">−0.009

$-0.009$

, P=0.01">P=0.01

$P=0.01$

). Thus, the system can assist in predicting case numbers and hospital load at a significant lead time and, simultaneously, help people determine if they need medical care.