Poster presented at the International Society for Pharmacoeconomics and Outcomes Research (ISPOR) Conference 2022
Objectives: To explore strategies to identify patient cohorts and multi-agent chemotherapy regimens from administrative claims data.
Methods: We designed a study to compare the safety and effectiveness of granulocyte colony stimulating factor (G-CSF) products (i.e. filgrastim, pegfilgrastim) with their biosimilars. Using the Biologics and Biosimilars Collective Intelligence Consortium’s Distributed Research Network, we included adults aged >=20 years who, in 2015-2019, per insurance claims, who received any G-CSF originator product or biosimilar as febrile neutropenia (FN) prophylaxis following the first cycle of high or intermediate FN risk chemotherapy. We developed de novo code to identify patient cohorts.
Results: Our initial criteria of locating cancer cases first, then subsequent chemotherapy exposure, resulted in fewer cases than expected, as codes in claims to identify cancer cases may include people who were screened for cancer but did not have cancer. Changing the coding order from chemotherapy then cancer, instead of the reverse, increased the sample size from 2,814 to 3,645. Additionally, we looked back 183 days before the index date of G-CSF receipt and excluded patients with a cancer diagnosis that differed from the cancers of interest for this study, including early-stage cancers without any G-CSF or chemotherapy exposure. When these criteria were lifted, 77 additional patients with GCSF and chemotherapy were included. Finally, we incorporated a 5-day window pre- and post-GCSF to identify all chemotherapy drugs administered, excluding prednisone and methotrexate as patients may fill oral prescriptions days to weeks prior to infusion. These changes increased the number of patients with chemotherapy exposures of interest to 6,010.
Conclusions: Given the inherent imprecision of timing, administrative coding, and cross-contamination of medications with disparate indications, it is important to consider how these variables interact in order to design robust realworld studies. Changing these parameters in the current study increased the eligible patient cohort >50%.
