Table 2.
Validation of chemotherapy episodes compared to chart review.
| Treatment regimen | Without information, n | Positive predictive valuea,b of regimen type, n/N (%) | Accuracyc of treatment cycle number, % | Mean absolute error | Root mean square error |
| FULVd | 30 | 67/70 (95) | 94 | 0.1 | 0.4 |
| FOLFOXe | 8 | 92/92 (100) | 87 | 0.3 | 0.6 |
| FOLFIRIf | 21 | 79/79 (100) | 89 | 0.4 | 1.4 |
| Capecitabine monotherapy | 65 | 35/35 (100) | 73 | 0.7 | 1.5 |
aFor each regimen, 100 cases were randomly sampled and reviewed. The information for chemotherapy was not available in the discharge summary in 30, 8, 21, and 65 cases with FULV, FOLFOX, FOLFIRI, and capecitabine monotherapy, respectively.
bPositive predictive value for the matched cases for the type of regimen in the manual comparison of generated episode records with the contents of the clinical notes.
cThe percentage of matched cases for the number of treatment cycles in the manual comparison of generated episode records with the contents of the clinical notes.
dFULV: fluorouracil and leucovorin.
eFOLFOX: fluorouracil, leucovorin, and oxaliplatin.
fFOLFIRI: fluorouracil, leucovorin, and irinotecan.