Metrics used in comparison of condition-suggestion accuracy
| Abbrev. | Full name | Description |
| M1 (%) | M1 (Matching-1) accuracy | % of cases where the top-1 condition-suggestion matches the gold-standard main diagnosis.7 |
| M3 (%) | M3 (Matching-3) accuracy | % of cases where the top-3 condition-suggestions contain the gold-standard main diagnosis.7 |
| M5 (%) | M5 (Matching-5) accuracy | % of cases where the top-5 condition-suggestions contain the gold-standard main diagnosis |
| COMP (%) | Comprehensiveness | Ratio of the (number of gold standard differentials matched by the suggested differentials) to the (number of gold standard differentials for the vignette), expressed as a mean across all vignettes.13 |
| RELE (%) | Relevance | Ratio of the (number of the suggested differentials that match with any of the gold standard differentials for the vignette) to the(number of differentials provided by the tested-GP or the symptom assessment app for the vignette), expressed as a mean across all vignettes.13 |