The indicated sensitivity, specificity and accuracy was evaluated based on symptomatic cases. These values may not be transferred for the purposes of identifying a current COVID-19 infection, especially for a presymptomatic or asymptomatic disease presentation. The COVID-19 chatbot does not represent an in-vitro-diagnostic device (e.g. PCR-Tests).
Enter symptoms, answer questions, and receive a score indicating the COVID-19 risk.Start Chatbot
An artificial intelligence based first-line defence against COVID-19: digitally screening citizens for risks via a chatbotDownload Study →
Symptoma correctly classifies 29 of 30 symptomatic COVID-19 cases as COVID-19 risk (96.6% sensitivity). Of 1,112 British Medical Journal (BMJ) control cases (non-COVID-19), only 41 are classified as potential COVID-19 cases by Symptoma, with only seven of these ranking COVID-19 higher than the correct diagnoses. These seven cases relate to diseases that present similarly to COVID-19, however, have far lower incidence rates and, therefore, are deemed less likely, e.g. Severe Acute Respiratory Syndrome (SARS-CoV) or the Avian influenza A (H5N1) virus infection (bird flu). The results are summarized in the table below.
|Flagged as COVID-19 risk
|Not flagged as COVID-19 risk
|29 True Positives
|1 False Negatives
|41 False Positives
|1,071 True Negatives
Differentiation of symptomatic COVID-19 cases vs non-COVID-19 cases by Symptoma, four other questionnaire-based approaches and a Symptoma variant (n = 394).More Details →
Identification of symptomatic COVID-19 cases with regards to the number of query terms entered. On the x-axis, the search rank of the query in Symptoma is given against the y-axis where each panel considers a different number of symptoms in the query. Only reported COVID-19 symptoms are considered. Points are jittered vertically for clarity only.More Details →
Although symptoms are very similar, Symptoma seems to correctly differentiate COVID-19 from other diseases, e.g. the common cold, hay fever and the flue (see Fig 2b).More Details →
A benchmark of online COVID-19 symptom checkersDownload Study →
A total of 50 COVID-19 cases were extracted by three trained medical doctors from the literature and are listed in S3 Table. Each case describes one patient’s medical situation, i.e. symptoms experienced and COVID-19 contacts. Extreme edge cases of COVID-19 such as patients with several severe comorbidities were not included in this study.
We used a total of 460 clinical cases to evaluate the performance of the COVID-19 symptom checkers. Each case lists both, symptoms and the correct diagnosis, alongside the age and sex of the patient when available. Details of the two case sets used are given below and in Table 2.
COVID-19 cases allow us to evaluate the sensitivity of symptom checkers. To also evaluate the specificity, 410 control cases from the British Medical Journal (BMJ) were sourced [6,7]. To allow a fair assessment, we only used cases containing at least one of the COVID-19 symptoms (see S4 Table) reported by the WHO .
Classifying non-relevant cases (e.g. a fracture) would overestimate the symptom checkers’ specificity. Furthermore, these patients would not consult an online COVID-19 symptom checker. None of these 410 BMJ cases has COVID-19 listed as the diagnosis as the cases where collected before the COVID-19 outbreak.
Symptoma shows the overall highest accuracy rate for COVID-19. Even SF-COS and SF-DIST (two alternative algorithms developed by Symptoma within a few hours each) perform higher than almost all other solutions.More Details →
Score your COVID-19 risk. Free. Anonymous. Fast. Accurate. Safe. Validated.
Screen citizens safely online and gather insights to act upon.
Screen patients to prioritize your resources.
Spread the word.
Integrate our chatbot into your website(s) for free.
Access our anonymous real world database for COVID-19.