Nicholas A. Berry, from the Mayo Clinic in Scottsdale, Arizona, and colleagues entered clinical data from 15 rash vignettes into eight online symptom checkers that were capable of diagnosis (not just triage), allow user input of skin rash data, and handle pediatric and adult ages.
The researchers found that online symptom checkers correctly diagnosed skin rashes in only 37 of 120 entries (30.8 percent). Symptom checkers generated a mean total of 13.6 differential diagnoses in those 37 scenarios. For correct diagnosis via symptom checker, the mean number of data entry terms needed to maintain that correct diagnosis was 6.1 following one-by-one random elimination. Variation existed in each symptom checker’s accuracies, ranging from 13.3 to 73.3 percent. Symptom checkers were most often correct for diagnosis of atopic dermatitis (pediatric; 75 percent) and shingles (62.5 percent). No symptom checker was correct for diagnosis of dermatitis herpetiformis. Only one symptom checker correctly diagnosed erythema nodosum, hidradenitis suppurativa, lichen sclerosus, or perioral dermatitis.
“For diagnosing skin rashes, symptom checkers have poor accuracy due to limitations on asking about pertinent medical history and insufficient diagnoses in their algorithms,” the authors write.