A Validated Electronic Medical Record-Based Algorithm to Identify Hospitalized Patients with Serious Illness

Population-based methods to identify patients with serious illness are necessary to provide equitable and efficient access to palliative care services. Create a validated algorithm embedded in the electronic medical record (EMR) to identify hospitalized patients with serious illness. An initial algorithm, developed from literature review and clinical experience, was twice adjusted based on gaps identified from chart review. Each iteration was validated by comparing the algorithm's results for a subset of patients (approximately 10% of the populations screened in and screened out on a given day) with the expert consensus of two independent palliative care physicians. The final algorithm was run daily for nine months to screen all hospitalized adults at our academic medical center in the United States. Compared with the gold standard of expert consensus, the final algorithm for identifying hospitalized patients with serious illness was found to have a sensitivity of 89%, specificity of 82%, positive predictive value of 80%, and negative predictive value of 90%. At our hospital, an average of 284 patients a day (54%) screened positive for at least one criterion, with an average of 38 patients newly screening positive daily. Data from the EMR can identify hospitalized patients with serious illness who may benefit from palliative care services, an important first step in moving to a system in which palliative care is provided proactively and systematically to all who could benefit.