Cancer Recurrence: An Important but Missing Variable in National Cancer Registries

Abstract

Background

Cancer recurrence is a critically important outcome to patients and providers. However, no publicly available cancer registry data contain recurrence information. The National Cancer Data Base (NCDB) collects recurrence data; however, this information is not provided to researchers because of completeness and accuracy concerns. Our objective was to examine completeness of cancer recurrence information in the NCDB.

Methods

Stage I–III thyroid/colon/melanoma/pancreas/breast cancers diagnosed in 2002–2005 were identified. Recurrence status, recurrence type, and recurrence date were evaluated for data completeness. Patient, tumor, and hospital factors were examined using generalized linear mixed models. Pseudo-R ² statistics estimated the relative contribution of patient and hospital factors.

Results

Of 702,144 patients with thyroid/colon/melanoma/pancreas/breast cancers treated in 1405 hospitals, recurrence information was incomplete in 21.5/24.0/20.2/34.8/18.2 % of patients, respectively. On average, hospitals had incomplete recurrence information on 56.7–66.7 % of their patients. Patients with incomplete information had more comorbidities, a higher cancer stage, non-private insurance, and lived farther from the hospital. Hospitals with the poorest collection were larger tertiary hospitals serving higher-income patients. However, these patients and hospital factors explained less than 3 %, while unexplained hospital variation accounted for the largest part of the observed variation (%ΔR ² = 84 %).

Conclusions

The majority of hospitals report incomplete recurrence information for more than half of their patients. The presence of incomplete recurrence information was largely dependent on undefined hospital factors, rather than patient or tumor characteristics. Attempts to improve cancer recurrence information should focus on hospital operational and process factors surrounding how the hospital tumor registries collect recurrence data.

Appendices

Appendix A

Variable Codes That Define Incomplete Information

• Date of recurrence is coded as being entirely unknown (DATE OF FIRST RECURRENCE, NAACCR Item #1860 = 99999999) and the recurrence flag variable indicates there is a recurrence, but the date is unknown (RECURRENCE DATE-1ST FLAG, NAACCR Item #1861 = 12).

• Date of recurrence is coded as being entirely unknown (DATE OF FIRST RECURRENCE NAACCR Item #1860 = 99999999) and the recurrence flag variable indicates it is unknown if the patient was ever disease-free or had first recurrence (RECURRENCE DATE-1ST FLAG, NAACCR Item #1861 = 10).

• Disease has recurred, but the type of recurrence is unknown (TYPE OF FIRST RECURRENCE, NAACCR Item#1880 = 88).

• It is unknown whether the disease has recurred or if the patient was ever disease-free (TYPE OF FIRST RECURRENCE, NAACCR Item#1880 = 99).

• The variable to indicate whether there was or there was not a recurrence is left entirely blank (TYPE OF FIRST RECURRENCE, NAACCR Item#1880 = .)

Appendix B

See Table 5.

Table 5 Characteristics of EXCELLENT, MIDDLE, and BAD collection hospitals