Construct Validity of Medicare Chemotherapy Claims

Abstract

The elderly are under represented in clinical trials of cancer therapy and the elderly who are enrolled may be unrepresentative. To assess whether Medicare claims data might be used to understand the benefits and tolerance of chemotherapy in the general elderly population, the construct validity of Medicare 5FU claims for elderly colon cancer patients within the SEER-Medicare data set was determined. In this validation study of Medicare chemotherapy claims from the linked the SEER-Medicare data set, the patterns of 5FU chemotherapy claims were evaluated for an incident cohort of elderly colon cancer patients (n = 15,039) during the 13 months following their diagnosis. Patterns of Medicare National Claims History (NCH) 5FU claims were evaluated with respect to prespecified patient-level disease and demographic factors from the data set. Twenty-two percent of patients had at least one detectable 5FU claim during the observation period. Among those patients, the median dose of 5FU was 1000 mg, the median interval between 5FU claims was 7 days, and the median number of claims during this period was 24. Multivariate regression revealed expected associations between demographic and disease factors and the likelihood of having a Medicare NCH 5FU claim. With increasing cancer stage, patients' likelihood of having a 5FU claim increased. Younger patients, married patients, white patients, patients with low comorbidity, and patients living in urban and less impoverished regions were each more likely to have 5FU claims. Because their pattern is consistent with the standard of medical care and with previously described associations with disease and demographic factors, it was concluded that Medicare NCH claims for 5FU administration in the SEER-Medicare data set exhibit construct validity. Criterion validation studies with an external gold standard should be pursued to determine the sensitivity and specificity of chemotherapy codes in the Medicare NCH files.