The Mid-Atlantic Oncology Program's comparison of two data collection methods.

Abstract

The Mid-Atlantic Oncology Program (MAOP) compared clinical trial data collected by circuit riding data managers (CDMs) from the MAOP statistical center with data collected by local data managers (LDMs) from clinics and practices, the latter being the standard data capture method in cancer clinical trials. LDMs and CDMs filled out identical study forms, using the same patient charts, for randomly selected patients on MAOP protocols. All coded answers on the forms were compared by one of the authors (DJ) and discrepant items were resolved in a blinded manner by the local MAOP physician acting as the principal investigator (PI). Thirty-three patient charts were reviewed with 53 pairs of forms completed and 1023 pairs of codes compared. A total of 129 (13%) pairs of codes were considered discrepant. Of the 100 discrepancies resolved (29 items were answered differently by the CDM, LDM, and PI) the PI's answers matched 66 of 100 codes as recorded by the CDMs and 34 of 100 codes as recorded by the LDMs. This results in chi 2 = 9.61 (p less than 0.005), demonstrating a significant difference between the frequency with which the PI's answers matched data collected by the CDMs and LDMs. It was also determined that CDMs consistently coded toxicities as more severe than did LDMs and were more often correct. Given the results of this study, CDMs should be considered an acceptable alternative to LDMs in the context of regional programs.