Quality of Cancer Recurrence Data in the National Cancer Database: A Reappraisal of Reporting Readiness.

BACKGROUND

This study evaluated the quality of cancer recurrence data in the National Cancer Database (NCDB) to determine if missingness and reporting consistency have improved enough to support national research.

METHODS

This multi-methods study included NCDB analyses and a cancer registry staff survey. Trends in recurrence data missingness from 2004 to 2021 and multivariable analyses of factors associated with missingness from 2017 to 2021 were evaluated for 4,568,927 patients with non-metastatic cancer. A survey of cancer registry staff at Commission on Cancer-accredited hospitals investigated challenges with recurrence data abstraction.

RESULTS

From 2004 to 2021, recurrence data missingness decreased from 15.7 to 8.4% for breast, 19.8 to 9.3% for colon, 20.5 to 7.4% for lung, 17.6 to 6.6% for melanoma, 29.3 to 9.0% for pancreas, and 18.5 to 9.2% for thyroid cancers. Driving distance ≥100 miles (odds ratio [OR] 1.96, 95% confidence interval [CI] 1.90-2.02) and Southern geographic region (OR 2.86, 95% CI 2.80-2.93) were associated with increased data missingness. Of 565 completed surveys (39.1% response rate), the most common challenges identified were inadequate physician documentation of no evidence of disease (67.8%) and inadequate documentation of recurrence (50.5%). High variability was noted in the interpretation of registry rules specific to the assignment of cancer recurrence or new primary cancer, with discordant assignment occurring 25.5-40.8% of the time.

CONCLUSION

Despite overall low rates of recurrence data missingness in the NCDB, data quality concerns remain related to inadequate clinical documentation and discrepancies with abstracting practices. Multi-organizational efforts are underway to improve the abstraction of high-quality recurrence data to support outcomes research.