Table 7 Most frequent SIFR Annotator errors for the death certificate coding task at CLEF eHealth 2017
From: SIFR annotator: ontology-based semantic annotation of French biomedical text and clinical notes
Error | Example | Percent |
|---|---|---|
Formulation different from synonym labels for expected concept | “arrêt respiratoire” (R09.2) not identified in “arrêt cardio respiratoire” or “détresse cardiorespiratoire.” | 79% |
Morphosyntactic variation | “Arrêt respiratoire” (R09.2) not identified in text “arrët réspiratore” due to incorrect diacritic. | 16% |
Annotation with a more general code (higher in the concept hierarchy) | “coma d’origine indéterminée et arrêt respiratoire progressif” matched with a more specific code, while the gold standard expects “arrêt respiratoire” (R09.2) | 2.5% |
Correct annotation dependent of detecting implicit semantic information | Code I10 “hypertension essentielle (primitive)” is hard to identify from “TC suite à une chute avec épilepsie séquellaire et tr cognitifs” as expected in the gold standard. Code R68.8 “autres symptômes et signes généraux précisés” was not identified within the text “atteinte polyviscérale diffuse.” | 2.5% |