Validation of a Claims-Based Algorithm for Specialist Palliative Care Delivery in Metastatic Cancer

The lack of valid methods to identify specialist palliative care (PC) delivery in population-level data impedes comprehensive understanding of its use.

To develop and validate a claims-based algorithm to identify receipt of specialist PC in Medicare beneficiaries with metastatic cancer.

We developed a claims-based algorithm to identify specialist PC, using a physician billing claim from a known PC clinician as the gold standard, retaining candidate variables with a positive predictive value (PPV) >60%. We evaluated algorithm performance and conducted simulation to measure bias resulting from algorithm use when examining the association between specialist PC and outcomes.

We identified 1,384,750 claims from 68,121 patients. The prevalence of specialist PC was 3.8% on the claim-level and 26.8% on the patient-level. The provider specialty code for "hospice and palliative care" (PPV 80.4%) and the diagnosis code Z51.5 for "encounter for palliative care" (PPV 67.5%) were included, where claims were counted as specialist PC if they had either variable. PPV and sensitivity of the algorithm were 68.0% and 83.0% respectively on a claim-level, and 78.4% and 88.3% respectively on a patient-level. Percent bias differed by outcome (hospice 4.2%, hospice enrollment ≥3 days 5.3%, intensive care unit use in the last 30 days of life -1.7%, chemotherapy use in the last 14 days of life -1.3%).

A simple algorithm can identify receipt of specialist PC care in Medicare claims for patients with metastatic cancer with reasonable accuracy. Algorithm use results in potentially acceptable amounts of bias, depending on study aims.