Skip to main content

Table 4 Comparison of the annotations of texts for 1,000 randomly sampled MIMIC-III patient visits before and after expansion, and their associated performance with respect to how predictive semantic similarity scores calculated from the annotations were of shared first diagnosis

From: Improved characterisation of clinical text through ontology-based vocabulary expansion

Investigation Annotations MAP MRR
Unexpanded 1,380,216 0.88 0.947
Expanded 2,088,765 0.913 0.986