The National Cancer Database (NCDB) is often used for comparative effectiveness research, and an ostensible strength is the inclusion of radiation therapy (RT) data. This project seeks to quantify internally inconsistent RT data using lymph node positive (N+) uterine cancer as a test clinical scenario. All subjects in the NCDB between 1998-2012 with clinical (c) or pathologic (p) N+ invasive uterine cancer treated with hysterectomy within 6 months of diagnosis (dx) followed by adjuvant RT were identified. T4 disease was excluded to limit analysis to stage IIIC1-IIIC2. Variables reviewed to identify coding inconsistent with accepted norms of adjuvant RT include site, modality, dose, fractions, timing, and duration. Chi squared test was used to correlate inconsistent coding with facility-specific variables available within the NCDB. 14,480 women met inclusion criteria, but 182 were excluded from further analysis because of inconsistent coding suggesting N0 and M1 stage. Overall 2,288 (16.0%) of the 14,298 remaining subjects had ≥1 of the following inconsistencies: site (n=142); modality (n=91); dose (n=1,625), fractions (n=1,031), timing/duration (n=699). 553 (24.2%) of these 2,288 subjects, however, had inconsistencies limited to variables which may result from an incomplete course of RT. Therefore, 1,735 (12.1%) subjects had fundamental inconsistencies in RT coding. Many of these 'fundamental' inconsistencies were likely errors of magnitude (extra or missing 0), wrong site, dose recorded was too high to be reasonable in an adjuvant setting, time duration that was inconsistent with the recorded therapy, or record of a fraction size that is never used clinically. The following variables were associated with higher rates of inconsistent coding: non-comprehensive community cancer programs (p=0.0007); South Central, South Atlantic, and Pacific locations (p<0.0001); reporting hospital ≥50 miles from patient’s residence (15.5% for <50 miles vs 52.9% for ≥100 miles, p<0.0001); and when all or a portion of RT was not administered at the reporting facility (15.8% for all RT at same facility vs 21.9% for regional or boost elsewhere, p=0.0038). 7,781 (54.4%) subjects had unknown or blank data in ≥1 variable(s) analyzed. While the majority of RT data within the NCDB appears to be appropriate for the clinical scenario, a significant portion of RT data within the NCDB for N+ uterine cancer is internally inconsistent. Additionally, much of the RT data is missing or not adequately specified. Data auditing is indicated for work using the NCDB to examine the effects of RT, particularly on the specific effects of dose, fractionation, and timing. These findings may also be used to direct improvements within the NCDB coding methodology to allow for more accurate reflection of the treatment received.