OBJECTIVES: Accurately determining line of therapy (LOT) is important for characterizing patients’ treatment history in the real-world setting. This study aims to describe an innovative algorithm of identifying LOT in metastatic colorectal cancer (mCRC) patients using the real-world data in China.

METHODS: We developed an innovative algorithm to determine LOT in mCRC patients by using both the structured prescription information (including systemic anticancer therapy (SACT) drug name, start and end dates) and unstructured medical notes (including tumor progression etc.) from authorized electronic medical records in China. We have filed patent application for this innovative algorithm. The first SACT given after the first metastatic diagnosis was considered as the initiation of the 1st line (1L). For those without metastatic diagnosis, who underwent 1 radical surgery, the date of first SACT administered after 6 months from the surgery was considered as the start date of the 1L. Successive SACTs administered within 9 days were considered as the same regimen. Tumor progression identified from medical notes by natural language processing (NLP) or switching to subsequent regimen (a gap of >25 days) whichever occurs first were used to identify line advancement. The accuracy of this algorithm was further validated by manual chart review of 50 patients randomly selected from the study population.

RESULTS: An accuracy rate of 95.2% was achieved by comparing our results to the results of manual chart review of 50 randomly selected patients.

CONCLUSIONS: In additional to prescription information, our algorithm also used tumor progression to improve the precision of identifying LOT, which resulted in better efficiency and reduced costs compared to manual review. However, our algorithm is developed for colorectal cancer and the accuracy is contingent on the data quality. LOT algorithm in other tumors needs further investigation.