Identifying newly approved medications in Medicare claims data: a case study using tocilizumab

Area of Science:

• Health Informatics
• Pharmacovigilance
• Biostatistics

Background:

• Parenterally administered medications in U.S. Medicare claims often use non-specific drug codes post-licensure.
• Accurate identification of these medications is crucial for patient safety and treatment effectiveness research.
• Existing methods for identifying medications before specific code assignment are not well-documented.

Purpose of the Study:

• To develop and describe a generalized algorithm for identifying parenteral therapies in Medicare claims using non-specific drug codes.
• To evaluate the algorithm's effectiveness in identifying tocilizumab (TCZ), a biologic agent approved in 2010.
• To assess the algorithm's performance using rheumatoid arthritis patient data.

Main Methods:

• Utilized Medicare data from 2008-2010 for a rheumatoid arthritis patient cohort.
• Developed an algorithm classifying non-specific drug codes based on ICD-9 codes, dose, procedure codes, and reimbursement data.
• Assessed algorithm performance by linking to an external arthritis registry for validation.

Main Results:

• The algorithm identified 9762 claims potentially representing TCZ from over 472,000 claims with non-specific codes.
• Classification accuracy was high: 74.3% by unit price, 4.4% by unique doses, and 21.3% by diagnosis and price deviation.
• Demonstrated strong performance: 94% sensitivity, 100% specificity, and 97% positive predictive value (PPV).

Conclusions:

• Claims-based algorithms can reliably identify newly approved, parenterally administered biologic medications in Medicare data.
• This approach is effective even before specific drug codes are available.
• The developed algorithm offers a robust method for pharmacovigilance and research involving new parenteral therapies.