OBJECTIVES: Despite using fully adjudicated claims, analyzing biologics treatment patterns requires reliable data cleaning and imputation methods for both pharmacy and professional claims. Self-injected agents with longer dosing intervals present unique challenges for data analysis. We developed methods to improve interpretability while minimizing data loss using TNF blocker claims. METHODS: A large health plan claims database was used to obtain 3,725 Psoriasis and/or Psoriatic Arthritis subjects initiating adalimumab (ADL) or etanercept (ETN) between January 1, 2003 and March 31, 2009; and were enrolled for 360 days pre- and ≥180 days post-index (first TNF claim). Patients were excluded if they had other inflammatory disorders pre- or post-index or received any biologic pre-index. We reviewed patients’ drug dispensing histories, established acceptable ranges for key claim values, and developed imputation rules that leveraged allowed reimbursements and dispensed quantities when other values were discrepant or missing. For professional claims, we divided total doses across weeks between fills to obtain average weekly dose. RESULTS: A total of 89.2% of 46,206 ETN and 94.6% of the 3,470 ADL claims were from pharmacies. 9.7% subjects with ≤0 reimbursed amounts on index were excluded, as were 0.9% with extreme quantity values (>48/>16 for ETN pharmacy/professional claims, >12 for ADL) and 1.0% patients with extreme weekly dose values in any claim (>250mg for ETN, >200mg for ADL). 8.8% subjects had >1 same-day claims, which were sorted in descending order of charge, allowed, and paid, with the top claim retained. After these steps, 3065 (82.3%) subjects remained in the cohort. CONCLUSION: Analyzing TNF blocker treatment patterns from claims requires adjustments for weekly dosing schedules and for professional claims that reflect dispensing of supplies for home injection. However, the limited dosing schedules for ADL and ETN allow for data cleaning strategies that address these challenges while retaining the vast majority of data.