• A C Berry, B D Cash, B Wang, M S Mulekar, A B Van Haneghan, K Yuquimpo, A Swaney, M C Marshall, and W K Green.
• Division of Gastroenterology,Larkin Community Hospital,Miami, FL,USA.
• Epidemiol. Infect. 2019 Jan 1; 147: e104.

AbstractWe sought to address the prior limitations of symptom checker accuracy by analysing the diagnostic and triage feasibility of online symptom checkers using a consecutive series of real-life emergency department (ED) patient encounters, and addressing a complex patient population - those with hepatitis C or HIV. We aimed to study the diagnostic and triage accuracy of these symptom checkers in relation to an emergency room physician-determined diagnosis. An ED retrospective analysis was performed on 8363 consecutive adult patients. Eligible patients included: 90 HIV, 67 hepatitis C, 11 both HIV and hepatitis C. Five online symptom checkers were utilised for diagnosis (Mayo Clinic, WebMD, Symptomate, Symcat, Isabel), three with triage capabilities. Symptom checker output was compared with ED physician-determined diagnosis data in regards to diagnostic accuracy and differential diagnosis listing, along with triage advice. All symptom checkers, whether for combined HIV and hepatitis C, HIV alone or hepatitis C alone had poor diagnostic accuracy in regards to Top1 (<20%), Top3 (<35%), Top10 (<40%), Listed at All (<45%). Significant variations existed for each individual symptom checker, as some appeared more accurate for listing the diagnosis in the top of the differential, vs. others more apt to list the diagnosis at all. In regards to ED triage data, a significantly higher percentage of hepatitis C patients (59.7%; 40/67) were found to have an initial diagnosis with emergent criteria than HIV patients (35.6%; 32/90). Symptom checker diagnostic capabilities are quite inferior to physician diagnostic capabilities. Complex patients such as those with HIV or hepatitis C may carry a more specific differential diagnosis, warranting symptom checkers to have diagnostic algorithms accounting for such complexity. Symptom checkers carry the potential for real-time epidemiologic monitoring of patient symptoms, as symptom entries and subsequent symptom checker diagnosis could allow health officials a means to track illnesses in specific patient populations and geographic regions. In order to do this, accurate and reliable symptom checkers are warranted.

16 keywords...

hide…