Assessing Completeness of Cancer Treatment Data from an Academic Medical Center's Tumor Registry Through Comparison to the Central Registry.

Background: Researchers often rely on hospital tumor registry data to provide comprehensive cancer therapy information. The purpose of this study was to determine the completeness of treatment information found in the abstracted records of patients seen at an academic medical center located in a rural Midwestern state.

Approach: The cohort included 846 Iowa residents diagnosed with a single malignant tumor of the female breast, colon/rectum, lung, pancreas, or prostate in 2017-2018 with an abstract recorded by the academic medical center and at least 1 other hospital. Treatment/no treatment agreement between the academic medical center's abstract and the central registry's consolidated abstract was examined for the following summary variables of the North American Association of Central Cancer Registries (NAACCR): surgery of the primary site, chemotherapy, radiation therapy, immunotherapy, and hormone therapy. Treatment summary variables from the academic medical center abstract that agreed with the corresponding variables from the central registry abstract were classified as concordant. The proportion of concordance for each treatment modality was the outcome measure, and 95% confidence intervals were calculated with the Agresti-Coull method. Concordance was also examined at the specific treatment level.

Results: There was high concordance between the treatment information recorded in the academic medical center and the central registry records. The average proportion of treatment/no treatment agreement across all treatment modalities and cancer sites was 0.97 (SD, 0.02). Concordance remained high even when examining specific treatments (average concordance, 0.95; SD, 0.04). The lowest treatment/no treatment concordance proportion was 0.92 (95% CI, 0.86-0.96) for chemotherapeutic treatment of pancreatic cancer. We also found that the academic medical center's summary variables captured most treatments given at other facilities, ranging from 74.4% capture of immunotherapy to 88.2% capture of surgery of the primary site.

Conclusions: These results indicate that NAACCR-formatted, summary variables from the academic medical center's tumor registry are likely to provide comprehensive treatment information for those individuals diagnosed or treated in this setting. Analyses of either the academic medical record registry records or consolidated records from the central registry should yield similar results. Future research should establish whether similar findings are obtained at other medical centers.