JMIR Publications

ABSTRACT

Background:

Italy has experienced very severe consequences in terms of hospitalizations and deaths during the COVID-19 pandemic. Online decision support systems and self-triage applications have been used in several settings to supplement recommendations from health authorities to prevent and manage COVID-19. A digital Italian health tech startup developed a non-commercial online decision support system to assist individuals in managing their potential exposure to COVID-19 and interpret their symptoms, with a chat user interface, available since the early phases of the pandemic in Italy.

Objective:

To compare the trend of sessions in this online support decision system with that of COVID-19 cases reported by the national health surveillance system in Italy, from February 2020 to March 2021.

Methods:

We analyzed the number of sessions by users with a COVID-19 positive contact and by users with symptoms compatible with COVID-19, with the number of cases reported by the National surveillance system. To calculate the distance between the time series, we used the Dynamic Time Warping algorithm. We also applied Symbolic Aggregate approXimation (SAX) encoding to the time series in one-week periods and we calculated the Hamming distance between the SAX strings. We shifted time series of sessions from the online decision support system one week ahead and we measured the improvement in Hamming distance to verify the hypothesis that sessions in the online decision support systems anticipate the trends in cases reported to the official surveillance system.

Results:

We analyzed a total of 75,557 sessions in the online decision support system. Among them, 65,207 were sessions by users with symptoms, while 19,062 were by contacts with individuals with COVID-19. The highest number of sessions in the online decision support system was recorded in the early phases of the pandemic. A second peak was observed in October 2020 and a third peak was observed in March 2021, in parallel with the surge of reported cases. Peaks in sessions of the online decision support system preceded the surge of COVID-19 notified cases by approximately one week. The distance between sessions by users with COVID 19 contacts and reported cases calculated by dynamic time warping was 61.23 while the distance with sessions by users with symptoms was 93.72. As the time series of users with a COVID 19 contact was more consistent with the trend of confirmed cases, we applied Symbolic Aggregate approXimation encoding and we measured the Hamming distance between these two time series. After applying a one-week shift, the Hamming distance between the time series of sessions by users with a COVID-19 contact and reported cases improved from 0.49 to 0.46. We repeated the analysis restricting the time window to the time period between July and December 2020. The corresponding Hamming distance was 0.16 before shifting the time series, and improved to 0.08 after the time shift.

Conclusions:

Temporal trends in the number of sessions of an online COVID-19 online decision support system may precede the trend of reported COVID-19 cases obtained through traditional surveillance. The trends of sessions by users with a contact with COVID-19 cases may better predict reported cases of COVID-19 than sessions by users with symptoms. Data from online decision support systems may represent a useful information source to supplement traditional surveillance and to support the identification of early warning signals in the COVID-19 pandemic.