OBJECTIVES: Administrative databases collect electronic health records data that can have a secondary use for pharmaco-epidemiological research and support generation of real-world evidence. In oncology, overall survival and progression-free survival are often used as endpoints to evaluate treatment effectiveness, but administrative databases generally don’t record progression markers (e.g. repeat stage recording, laboratory measures, response to treatment). As such proxy measures are required. Here, we outline our methodology for developing proxy measures for a study of melanoma using Hospital Episode Statistics in England.

METHODS: Stages of diagnoses and effectiveness endpoints were defined with clinical experts’ input from diagnoses coded using the International Classification of Diseases version 10 (ICD-10 diagnostic) and procedures and treatment administration coded using the Office of Population Censuses Surveys Classification of Surgical Operations and Procedures, 4th Revision (OPCS-4).

RESULTS: Using ICD-10 diagnosis codes for melanoma diagnosis and presence of metastases lymph node removal procedure codes and groups of chemotherapy administration codes), it was possible to determine stage 0, stage I/II, III and IV Stages of melanoma, and resectable versus unresectable melanoma. The following proxy-effectiveness endpoints were also definable: time-to-next treatment (time from start of systemic treatment to date of initiation of next OPCS treatment combination), progression-free survival (reaching disease next stage, the presence of a new metastasis, or death; a sensitivity analysis also used start of a new OPCS treatment combination as an indicator of progression), recurrence-free survival (with recurrence being defined as new metastasis or a new lymph node removal surgery or a new cutaneous melanoma diagnosis >30 days after last recorded diagnosis) and overall survival (using absence of a record of death).

CONCLUSIONS: Derivation of oncology effectiveness endpoints using diagnoses, treatments and procedures recorded in HES is possible; nevertheless, data granularity remains lower than data that would be collected via primary data collection methods such as chart review.