Keywords: clinical data, data validation, electronic health records, oncology
Overall survival (OS) is an important outcome measure in oncology, but it may require large sample sizes and long follow-up time to provide meaningful insights. As a supplement to OS, outcomes based on disease progression and tumor response typically require less follow-up time and are commonly captured in oncology clinical trials to assess treatment efficacy. The standard methods to collect these outcomes in clinical trials, however, may not be feasible using electronic health records (EHR), but such data present an opportunity to answer research questions at a scale and recency not available from clinical trials, while also reflecting treatment patterns and populations seen in routine clinical practice. As we develop real-world oncology outcome measures, it is critical to understand their quality and the appropriate statistical methods required to derive valid insights. We illustrate this with a case study, discuss statistical considerations unique to working with EHR data, and identify methodological gaps that present opportunities for future development.