Table 4. Comparison of the precision of different BERT (Bidirectional Encoder Representations from Transformers) models.
| Model size and model name | All codes, 95% CI | Top 80% codes, 95% CI | ||
|---|---|---|---|---|
| Multilabel | Top-5 | Multilabel | Top-5 | |
| Base | ||||
| NorDeClin-BERT-base-NorICDa | 0.65 (0.64-0.66) | 0.80 (0.79-0.81) | 0.71 (0.70-0.73) | 0.89 (0.88-0.90) |
| SweDeClin-BERT-SweICD | 0.38 (0.36-0.40) | 0.61 (0.60-0.62) | 0.46 (0.44-0.48) | 0.69 (0.67-0.70) |
| SweDeClin-BERT-NorICD | 0.58 (0.56-0.59) | 0.77 (0.76-0.78) | 0.66 (0.65-0.68) | 0.87 (0.86-0.88) |
| ScandiBERT-NorICD | 0.57 (0.55-0.58) | 0.77 (0.76-0.78) | 0.67 (0.66-0.69) | 0.87 (0.87-0.88) |
| NorBERT3-base-NorICD | 0.63 (0.61-0.64) | 0.79 (0.78-0.80) | 0.69 (0.68-0.70) | 0.88 (0.88-0.89) |
| Large | ||||
| NorDeClin-BERT-large-NorICD | 0.66 (0.65-0.68)b | 0.82 (0.81-0.82)b | 0.72 (0.71-0.74) | 0.90 (0.90-0.91)b |
| NorBERT3-large-NorICD | 0.65 (0.64-0.67) | 0.80 (0.79-0.81) | 0.73 (0.72-0.74)b | 0.89 (0.89-0.90) |
ICD-10: International Statistical Classification of Diseases, Tenth Revision.
Highest score for each scenario.