Table 3:
Reportable Cancer Case Detection Algorithm Test Results
| Algorithm | Number of features | Accuracy (%) | Cancer precision (%) | Cancer recall (%) | Cancer F-measure (%) | Training test skew (%) |
|---|---|---|---|---|---|---|
| Source only (∼CRCP-DUAL) | 3 | 79.92 | 68.96 | 93.87 | 79.51 | 0.86 |
| Prefix feature only | 1847 | 77.38 | 78.31 | 62.90 | 69.77 | 7.26 |
| CUI name only | 1307 | 82.87 | 87.60 | 68.39 | 76.81 | 3.02 |
| ICD-9 codes (cancer) only | 541 | 83.80 | 80.39 | 80.65 | 80.52 | −10.15 |
| ICD-9 codes (all) only | 4921 | 78.18 | 72.48 | 76.45 | 74.41 | 7.89 |
| CUI name and prefix feature | 3153 | 83.94 | 86.82 | 72.26 | 78.87 | 5.67 |
| CUI name and ICD-9 codes (cancer) | 1847 | 87.15 | 84.30 | 84.84 | 84.57 | 1.30 |
| CUI name and ICD-9 codes (all) | 6227 | 82.73 | 79.29 | 79.03 | 79.16 | 10.35 |
| Source, CUI name, and ICD-9 codes (cancer) | 1849 | 86.61 | 84.09 | 83.55 | 83.82 | 2.82 |
| Prefix feature, CUI name, and ICD-9 codes (cancer) | 3693 | 87.15 | 84.08 | 85.16 | 84.62 | 4.04 |
| Source, prefix feature, and ICD-9 codes (cancer) | 2389 | 85.41 | 82.32 | 82.58 | 82.45 | 3.86 |
| Source, prefix feature, CUI name, and ICD-9 codes (cancer) | 3695 | 87.02 | 85.86 | 82.26 | 84.02 | 4.63 |
| Source, prefix feature, CUI name, and ICD-9 codes (all) | 8075 | 83.27 | 82.01 | 76.45 | 79.13 | 11.70 |
CUI, Concept Unit Identifier; ICD-9, International Classification of Disease – 9. Bolded text indicates the feature set that was ultimately selected for implementation.