Comparison of Diagnostic Accuracy in Vascular Neurology Between Neurology Residents and a Neurology Differential Diagnosis App: A Multi-Center Cross-Sectional Observational Study

Abstract

Background: Diagnostic errors in neurological diagnosis are a source of preventable harm. Software tools like Differential Diagnosis (DDx) apps in neurology that hold the potential to mitigate this harm are conspicuously lacking. Materials and Methods: A multicenter cross-sectional observational study was designed to compare the diagnostic accuracy of a Neurology DDx App (Neurology Dx) with neurology residents by using vascular neurology clinical vignettes. The study was conducted at 7 leading neurology institutes in India. Study participants comprised of 100 neurology residents from the participating institutes. Measurements: Detecting diagnostic accuracy of residents and App measured as a proportion of correctly identified high likely gold standard DDx was prespecified as the main outcome. Proportions of correctly identified first high likely, first 3 high likely, first 5 high likely, and combined moderate plus high likely gold standard differentials by residents and App were secondary outcomes. Results: 1,000 vignettes were attempted by residents. Frequency of gold standard, high likely differentials correctly identified by residents was 27% compared to 72% by App (absolute difference 45%, 95% CI 35.7-52.8). When high and moderate likely differentials were combined, residents scored 17% compared to 57% by App (absolute difference 40%, 95% CI 33.8-50.0). Residents correctly identified first high likely gold standard differential as their first high likely differential in 34% compared to 18% by App (absolute difference 16%, 95% CI 1.2-25.4). Conclusion: App with predefined knowledge base can complement clinical reasoning of neurology residents. Portability and functionality of such Apps may further strengthen this symbiosis between humans and algorithms (CTRI/2017/06/008838).

Ähnliche Arbeiten