Validation of Claims-Based Algorithms to Classify Thoracic Radiation Therapy Courses.

Area of Science:

• Oncology
• Health Informatics
• Medical Imaging

Background:

• Administrative data offer insights into healthcare utilization but lack clinical detail for radiation therapy (RT).
• Identifying specific RT sites and intents (e.g., thoracic RT [TRT], curative-intent RT) is crucial for outcome analysis.
• Claims-based algorithms require validation to accurately capture RT characteristics.

Purpose of the Study:

• To validate claims-based algorithms for accurately identifying thoracic radiation therapy (TRT).
• To assess the accuracy of algorithms in distinguishing curative-intent radiation therapy (RT) within administrative databases.

Main Methods:

• Analyzed 3,846 RT episodes from patients with lung cancer and RT CPT codes.
• Developed and tested a priori algorithms using CPT codes, treatment counts, and diagnosis codes.
• Calculated positive predictive value (PPV) stratified by RT modality (3DCRT, IMRT, SBRT).

Main Results:

• The primary TRT algorithm achieved high PPVs: 97% for IMRT, 99% for SBRT, and 87% for 3DCRT.
• Curative-intent RT identification yielded PPVs of 87% for IMRT, 90% for SBRT, and 55% for 3DCRT.
• Algorithm performance was highest when exclusive thoracic malignancy diagnosis codes were included.

Conclusions:

• Clinically informed algorithms can reliably identify TRT in claims data, particularly for IMRT and SBRT.
• These validated algorithms can be utilized in claims databases to evaluate RT toxicity and effectiveness.
• External validation across diverse datasets is recommended to confirm generalizability.