PS1-16: Quality Assurance in the VDW TUMOR File

Abstract

Background: The VDW TUMOR database is constructed using nationally standardized variable definitions. The North American Association of Central Cancer Registries (NAACCR) is a professional agency for establishing data standards. The database is NAACCR compatible with respect to variable definitions and record layout. Many of the CRN sites are also Surveillance, Epidemiology, and End Results (SEER) Program sites. SEER collects data from specific geographic regions representing 26% of US population and is demographically diverse. Access to SEER data provides a mature data source and familiar variable list for use in cancer research. Non SEER site can access comparable data using either hospital based or State based data collection systems. Methods: The TUMOR content quality team was formed to examine data quality to report back to the VDW operations committee. Frequency of values in selected fields can be used to indicate data quality. The team has decided to test for non standard values, number of unknowns (stage, anatomic site others), sex specific cross tabs (i.e., males with female GYN cancer, females with prostate cancer) and histology frequencies by year to detect outliers. In addition to identifying areas for quality improvement, the TUMOR content team will evaluate the need for additional references to be appended to the VDW data area and suggest future directions to improve the utility of the TUMOR database. The team will also review the need for additional data processing tools (i.e. macros) to facilitate ease of content use. Results: Data from 9 sites have tumor data posted to the CRN web counter. This represents 725,000 tumors. Over 77,000 un-staged cancers (10.6%) were identified using the CRN web based cancer counter. The next step is to determine the proportion for which un-staged is an appropriate designation by examining the tumor behavior and other clinical parameters. This process will be used to examine additional variables and provide the TUMOR quality team with guidance on which specific content areas require attention. This work will allow the TUMOR quality team to identify areas for data improvement without distributing many work-plans to each participating sites. Conclusion: Examining and documenting the findings of the TUMOR content team will provide a mechanism to establish best practices for use of the TUMOR data. Expanded knowledge of current status will facilitate future additions to the content.