Development and Validation of an Algorithm to Identify Patients with Multiple Myeloma Using Administrative Claims...

Area of Science:

• Health Informatics
• Oncology
• Epidemiology

Background:

• Accurate identification of multiple myeloma (MM) patients in administrative claims is crucial for research.
• Prior algorithms for MM patient identification have limitations.

Purpose of the Study:

• To develop and validate novel algorithms for identifying multiple myeloma (MM) patients using administrative claims data.
• To improve upon the accuracy of existing MM patient identification methods.

Main Methods:

• Constructed two files using MarketScan EMR and claims data (2000-2014) to select MM cases and controls.
• Developed and validated 20 algorithms, focusing on MM diagnosis codes and timing relative to other events.
• Calculated sensitivity, specificity, and positive predictive value (PPV) for algorithm performance.

Main Results:

• Validated three claims-based algorithms for MM patient identification.
• Achieved approximately 10% improvement in PPV (87-94%) compared to prior work (81%) and the baseline algorithm (76%).
• Confirmed the necessity of MM diagnoses both before and after specific tests.

Conclusions:

• The validated algorithms offer improved accuracy for identifying MM patients in administrative claims.
• These algorithms can be utilized in future research requiring MM patient cohorts.
• Findings underscore the importance of considering the temporal relationship of diagnoses and tests.