Assessing resource use in oncology patients: a comparison of analyses based on claims data and medical chart review.

Background: Provider claims data are used to make medical analyses and decisions, but such databases typically lack important clinical information.
Objective: To compare the patterns of use of filgrastim in analyses of a claims database and a medical chart review.
Methods: We extracted data from the Medicare 5% sample claims database for the years 1996 through 1998 and from a medical chart review of the Oncology Practice Pattern Survey (OPPS) for the same period to determine the patterns of use of filgrastim in patients with non-Hodgkin.s lymphoma treated with cyclophosphamide, doxorubicin, vincristine, and prednisone. The analyses were restricted to the first 3 cycles of the chemotherapy.
Results: A total of 414 patients in the Medicare database were treated with 1,360 cycles of chemotherapy. The mean duration of filgrastim use in these patients was 6.6 days. In the OPPS database, 80 patients were treated with 152 cycles of chemotherapy, with a mean duration of filgrastim use of 9.3 days.
Conclusion: The mean duration of filgrastim use in the OPPS database was greater than that in the Medicare database and was closer to that shown in clinical trials to produce optimal results (approximately 11 days). The actual use of resources may be underestimated in claims databases, owing to their limitations and potential for bias.