Evaluating the use of red flags by online symptom checkers.

Abstract

Background: Online Symptom Checkers (OSCs) are digital health tools providing triage, diagnostic, and self-care advice based on user reported symptoms. Amidst global trends of increasing demand and workforce shortages, OSCs have the potential to alleviate primary care workload. However, their ability to seek red flag symptoms, a critical marker of a safe consultation in primary care, remains unexplored. Using clinical vignettes, this study evaluates OSCs' performance in seeking red flag symptoms compared to Primary Care Physicians (PCPs).

Methods: Four OSCs (Ada, Babylon, Symptomate, Healthily) were evaluated using 51 clinical vignettes. Two standard setters used guidelines to determine which vignettes required emergency triage and identified the relevant red flags symptoms for the remaining vignettes. Two laypersons entered data from vignettes into OSCs and outputs were collected following a standardised form. The same vignettes were independently assessed by PCPs to compare triage accuracy and red flag identification. Summary statistics and 95% confidence intervals were calculated using Wilson Score intervals, and Fisher's exact test was used to compare performance between OSCs and PCPs.

Results: Of the 51 clinical vignettes, standard setters determined 14 to require emergency triage and the remaining 37 vignettes suitable for primary care triage. Of the primary care triaged vignettes, standard setters identified a total of 77 relevant red flag symptoms to be sought. Of the 14 emergency vignettes, PCPs correctly triaged 85.7% (95% CI: 74.3-92.6%) of cases compared to OSCs 76.9% (95% CI: 59.3-87.9%), with no statistically significant difference (p = 0.299). Specificity, the proportion of correctly triaged primary care vignettes, PCPs performed significantly better compared to OSCs, 91.9% (95%CI 78.9-97.0%) vs. 83.3% (95%CI 68.1-91.9%), p = 0.024.

Conclusions: OSCs demonstrated comparable ability to appropriately triage clinical vignettes requiring emergency triage as PCPs, however, were less specific, triaging more primary care vignettes as emergency. OSCs do not seek the majority of red flags. This raises concerns about their safety and effectiveness in primary care. OSCs developers should focus on improving OSCs' red flag coverage to ensure safe integration into primary care settings.