Table 3.
Recall and Precision of Natural Language Processing (NLP) Algorithms
| NLP of Clinical Narratives only | NLP of Semi-Structured Reports only | NLP of All Notes | Adjudicated Reference Standard | |
|---|---|---|---|---|
|
| ||||
| Unique colonoscopy tests | ||||
| Discovered (n) | 157 | 105 | 185 | 190 |
| Correctly identified (n) | 146 | 105 | 174 | -- |
| Recall | 77% | 55% | 92% | -- |
| Precision | 93% | 100% | 94% | -- |
| F-measure | 84% | 71% | 93% | -- |
|
| ||||
| Unique FSIG tests | ||||
| Discovered (n) | 9 | 6 | 11 | 10 |
| Correctly identified (n) | 8 | 6 | 10 | -- |
| Recall | 80% | 60% | 100% | -- |
| Precision | 89% | 100% | 91% | -- |
| F-measure | 84% | 75% | 95% | -- |
|
| ||||
| Unique DCBE tests | ||||
| Discovered (n) | 6 | 17 | 19 | 19 |
| Correctly identified (n) | 6 | 17 | 19 | -- |
| Recall | 32% | 89% | 100% | -- |
| Precision | 100% | 100% | 100% | -- |
| F-measure | 48% | 94% | 100% | -- |
|
| ||||
| Unique FOBT tests | ||||
| Discovered (n) | 11 | 42 | 48 | 46 |
| Correctly identified (n) | 9 | 40 | 44 | -- |
| Recall | 20% | 87% | 96% | -- |
| Precision | 82% | 95% | 92% | -- |
| F-measure | 32% | 91% | 94% | -- |
CRC = colorectal cancer; FSIG = flexible sigmoidoscopy; DCBE = double contrast barium enema; FOBT = fecal occult blood testing