Using the Electronic Health Record to Identify a Cohort of Seriously Ill Patients from a Primary Care Population Across Three Academic Health Systems (FR441C)

Abstract

•Identify a seriously ill patient population across primary care clinics from three academic health systems using data available in the electronic medical record.•Describe how validity of this serious illness definition was evaluated. Automated identification of a serious illness population is needed to implement palliative care and advance care planning interventions. We aimed to develop a way to identify a serious illness (SI) population within primary care (PC) that is consistent across three University of California (UC) health systems. We included patients who were 18 or older, had 2 or more PC visits in the last 12 months, and had either advanced cancer, advanced heart failure, advanced COPD, end-stage liver disease, CKD requiring dialysis, ALS, or were a vulnerable elder with serious illness using previously published electronic medical record (EMR) data elements that were identified through iterative chart abstraction. For validation, charts were reviewed to ensure that patients met one or more of the following criteria: 1) ≤2 year prognosis, 2) worsening decision making capacity, 3) worsening functional status, 4) high burden of disease causing suffering due to symptoms or excessive healthcare utilization. To evaluate the final criteria, we conducted implicit chart abstraction for 306 patients across three health systems. We also checked the consistency of the population across the health systems by looking at the percent of patients who meet the SI definition among all PC patients. After application of the EMR elements across the three health systems, chart abstraction of 306 patients revealed that 301 (98%) met the pre-specified criteria. Lessons learned included that source of ICD code is critical and that billing codes can be linked with other available data elements such as utilization, specialty visits, chemotherapy use and clincal data to identify a SI population. Patients who met final criteria for SI represented 4%-7% of the PC population depending on if the PC clinic was geographically located near the academic medical center. An EMR data algorithm can identify a SI primary care patient population that is consistent across three academic health systems.