Utility of the SEER-Medicare data to identify chemotherapy use

Abstract

Background: Medicare claims include codes for chemotherapy administration and specific drugs given, and researchers are increasingly using these data to measure the use of chemotherapy. However, the validity and completeness of these data as a source of information has not been established.

Objectives: This analysis is intended to assess the utility of the Medicare claims to capture chemotherapy use.

Methods: Persons with breast, colorectal, and ovarian cancer were identified from the linked SEER-Medicare data. Their Medicare claims were reviewed to determine if there were any bills for chemotherapy, and if so, if there were claims for specific agents. This information was compared with data on the first course of treatment obtained from hospitals and treating physicians by the SEER registries through an NCI-supported Patterns of Care Studies (POC). Agreement was measured using kappa statistics. The sensitivity of the Medicare claims to capture chemotherapy, as reported from the POC data, was also measured. An additional comparison assessed the agreement between the two data sources concerning which specific drugs had been given.

Results: For all of the cancers, there was a high level of agreement between the Medicare claims and the POC data regarding whether or not the patient had received chemotherapy (kappa >or=0.73). The sensitivity of the Medicare data to determine if a person had received chemotherapy was high (>or=88%). In cases where the Medicare claim included a code for a specific drug, there high agreement between Medicare and POC about the specific drug given in breast and colorectal cancers, although the agreement was lower for ovarian cancers. The sensitivity of the Medicare claims to identify specific agents varies by cancer type.

Conclusions: The Medicare claims can be used to identify which persons are receiving chemotherapy. The utility of Medicare data to measure treatment with specific agents varies by cancer type and specific agent. For some cancers, it is possible to use these claims to assess use of specific drugs, while for other drugs the data are limited.