Establishing a denominator for palliative care quality metrics for patients with advanced cancer.

Abstract

Listen

Translate article

104 Background: Measurement and feedback of palliative care quality metrics for patients with advanced cancer is an important aspect of supporting a shared mental model for the early integration of palliative care within the oncology clinic. Accurately identifying the subset of advanced cancer patients within a population of cancer patients is critical to the development of quality improvement activities; however, there is not a methodology of identifying these patients in real time. Methods: We evaluated two main approaches to identifying patients with advanced cancer among active cancer patients defined as patients who were seen by an oncologist at least twice in the last 6 months and who had received chemotherapy or radiation at least once in the last two years. These approaches included: 1) Pattern matching of words indicating advanced cancer (e.g. metastatic, advanced) in oncology notes, radiology imaging and problem lists and 2) ICD-10 codes. To determine the final set of ICD-10 codes for the second approach, we used a conceptual model of the meaning of advanced cancer (evidence of distant metastasis and/or poor prognostic cancer with > 50% mortality rate at 5 years) and iterative chart review. In order to test our final definitions, we randomly selected 588 charts of patients with active cancer who see one of 64 oncologists in our health system. These charts were abstracted by an oncologist to establish a gold standard for advanced cancer. We evaluated the sensitivity and specificity of our approaches to identify advanced cancer patients compared to this gold standard. Results: We found that the methods used to identify patients using pattern matching had a specificity of 76% and a sensitivity of 80%. Using our final ICD-10 algorithm we achieved a specificity of 93% and a sensitivity of 68%. We improved our sensitivity to 74% while maintaining our specificity at 92% when we excluded oncologists who predominantly see hematological malignancies. Conclusions: We achieved high specificity and reasonable sensitivity for an advanced cancer quality metric denominator using an ICD-10 algorithm within an academic oncology practice. This will help inform quality improvement efforts locally and beyond.