Validation of an Administrative Claims-based Line of Therapy Algorithm for Women with Ovarian Cancer Using Medical...

New maintenance therapies to treat advanced ovarian cancer have added complexity to identifying lines of therapy (LOTs) for real-world evidence (RWE) studies. This study evaluated the performance of a claims-based algorithm that identifies LOTs among patients with ovarian cancer using medical chart review validation.

The algorithm was developed previously utilizing the Optum Research Database (ORD), a US database that contains administrative claims data. To validate the algorithm, LOT results generated using claims data vs chart data were compared at the patient level by calculating the percent agreement between total number of active and maintenance LOTs, type of therapy (neoadjuvant vs adjuvant classification), and type of regimen (individual drugs). Patients with a diagnosis of ovarian cancer who initiated chemotherapy between December 1, 2014, and September 15, 2017, were included in the study. We report descriptive statistics, the percentage correspondence between medical records and claims data, and kappa statistics to measure the magnitude of agreement.

A total of 294 patients were included in the analysis; 164 received only chemotherapy and no maintenance, 77 received bevacizumab, and 53 patients received poly (ADP-ribose) polymerase inhibitors (PARPi). Mean age was 64.9 years, and 47.3% had stage III cancer. The algorithm demonstrated substantial agreement between claims and medical records for total number of lines of active and maintenance therapy (weighted kappa 0.65 and 0.62 p < 0.0001). There was moderate-to-substantial agreement for neoadjuvant and adjuvant therapy (kappa 0.56 and 0.62 p < 0.0001). The algorithm performed best at identifying early treatment with a regimen match of 82% and 88% agreement for first-line active and first-line maintenance, respectively.

We validated an administrative claims-based algorithm that demonstrates strong concordance with medical records for identifying LOT among patients with ovarian cancer. The algorithm can be applied in future studies to analyze treatment patterns and outcomes.