In this study, we sought to determine the accuracy with which the International Classification of Diseases, 9th Edition, Clinical Modification (ICD-9-CM) diagnosis code for "secondary neoplasm of brain and spinal cord" in health insurance claims measures clinically evident central nervous system (CNS) metastases in patients with non-small cell lung cancer (NSCLC). For 241 consecutive patients with newly diagnosed NSCLC, we compared ICD-9-CM "secondary neoplasm" codes indicating tumor spread to the CNS from institutional billing records to gold-standard chart review to determine: (1) sensitivity, specificity and positive predictive value (PPV) of the site-specific secondary neoplasm code and (2) the accuracy in time of its appearance within billing records compared with the gold standard date of CNS relapse. The occurrence of at least one ICD-9-CM code for brain metastasis (Algorithm 1) had a sensitivity of 100% (95% CI: 100-100%) and PPV of 91% (95% CI: 87-94%). By requiring >or= 2 codes (Algorithm 2) or >or= 3 codes (Algorithm 3) for the diagnosis of brain metastasis in claims, specificity and PPV improved, while sensitivity did not drop substantially. The claims-based date of diagnosis was also accurate, with 92% of dates falling within 30 days of the gold standard. ICD-9-CM codes in institutional billing claims reliably documented NSCLC metastases to the CNS. These results suggest that Medicare claims data may be used to evaluate clinical and epidemiological issues related to brain metastases in elderly cancer patients.