Improving Palliative Measurement Application with Computer-Assisted-Abstraction Study

Background: VA needs to incorporate the Veteran and family voice in measuring performance, and it may improve care to involve Veterans and families more deeply in improvement. This is especially true for palliative and end of life care, given the 30% of lifetime Medicare costs in the last year of life, and the divergent perspectives that Americans have expressed regarding end of life care. We have identified major gaps in VA performance, but work is needed to prioritize indicators from the Veteran and family perspective and to improve measure feasibility. The latter will foster better VA quality and experience of care and facilitate monitoring the potential impact of paid, non-VA care on seriously ill Veterans. Aims: The Improving Palliative Measurement Application with Computer-Assisted-Abstraction Study (ImPACS), will prioritize measures and operationalize a subset of higher, intermediate, and lower feasibility process and utilization measures for palliative and end of life cancer care. We aim to: Aim 1: Solicit priorities from two Delphi panels - one of Veterans and families and a second of experts regarding process and healthcare use quality measures for advanced cancer care, including which of the 42 Cancer Quality ASSIST measures to extract, and conduct interviews regarding how to integrate Veterans and families in VA measurement and improvement for palliative and end of life cancer care, and Aim 2: Extract high priority process and utilization measures using natural language processing of VA charts and administrative data and in a sample drawn from Stanford and Dana Farber’s Healthcare's Epic systems, focusing on the domains of pain and opioids, mental health, and goals of care communication, and Aim 3: Examine associations of the extracted measures with Veteran characteristics, focusing on disparities in the care of rural and nonwhite Veterans and palliative care use. Methods: For Aim 1, we will recruit two Delphi panels - one of Veterans, family members, and a second of expert stakeholders. Veteran-family members will have experience with cancer. Experts will have expertise in the methods and application of quality measures. Purposive sampling will focus on critical attributes (e.g., race) that may affect priorities. Panelists will rate and rank measures within tiers of high, intermediate, and low feasibility, informed by reviews of the evidence for intervention and impact of performance gaps on patients and caregivers. We will also interview Veterans, family members and VA leaders at high and low performing VA facilities based on the Bereaved Family Survey of end of life experience, to see how deeper Veterans-family involvement might strengthen quality and experience of end of life care. For Aim 2, we will operationalize a subset of prioritized process measures from Aim 1 including Cancer Quality ASSIST measures using natural language processing of text notes and machine learning. In Aim 3, we will characterize variation in measures with Veteran characteristics focusing on known disparities among rural and nonwhite Veterans and palliative care services use. Impact: We will produce Veteran and family-informed priorities for a balanced measure set for palliative and end of life cancer care. We will inform how Veterans and families might be more deeply engaged in fostering improved quality and experience and fostering a learning healthcare system. Finally, we will extract prioritized measures using state of the art methods to improve their feasibility, characterize variation at the Veteran level focused on known disparities and palliative care potential to mitigate them.