Specialty vs. Primary Palliative Care in Randomized Clinical Trials: A Systematic Review (FR420B)

Abstract

Objectives Describe the general process used to develop and implement an EHR-based mortality risk stratification model. Assess the impact and implications of implementation an EHR-based mortality risk stratification model on clinical, quality, and financial metrics. Background. The frequency and timing of palliative care consultation is highly variable among inpatients with life-limiting illnesses despite evidence of its benefits. Objectives. To develop an EHR-based risk stratification model and evaluate its impact as a trigger for inpatient palliative care consultation. Methods. We performed a retrospective cohort study at three urban hospitals among 65,045 admissions in 2016. Using a randomly split sample and a machine learning approach, we developed and validated an EHR-based model (Palliative Connect) to predict risk of death within six months of admission. We then determined a risk threshold of $30% based on expert chart review and prospectively piloted Palliative Connect in a 4month pre-post study of triggered palliative care consultation on a general medicine service at a large academic hospital. Primary clinicians could decline the consult. We performed an intention-to-treat analysis to evaluate impact on care delivery, quality metrics, and costs. Original Research Background. While outpatient palliative care (PC) began primarily in cancer centers, outpatient PC increasingly serves patients with a wide range of diagnoses. Research Objectives. Compare characteristics of patients with cancer and non-cancer diagnoses referred to clinic-based PC, and the care they receive. Methods. Data were extracted from the Palliative Care Quality Network database regarding 3,569 patients seen by 27 clinic-based PC teams between 01/ 15/2016 and 07/17/2018. Results. Overall 79.3% (n¼2,766) of all patients referred to outpatient PC had cancer. Compared to patients with non-cancer diagnoses, patients with