Derivation and validation of a risk model for emergency department palliative care needs assessment using the Screen for Palliative and End-of-Life Care Needs in the Emergency Department (SPEED).

22 Background: A key setting for the provision of palliative care is the emergency department (ED) where important decisions regarding treatment and next site of care are determined; however identifying patients who would benefit from a palliative care consult is an ongoing challenge. The (SPEED) is a 5-question tool that assesses unmet palliative care needs. Methods: We performed a retrospective derivation and temporal validation of a risk model for a palliative care event (PCE) among cancer patients with an ED visit and subsequent hospital admission using data available upon arrival, including data from the SPEED tool. A PCE was defined as a palliative care consult, discharge to hospice, or in-hospital death. We developed a multivariate logistic regression model to predict PCEs. We assessed model performance using a receiver operating characteristic curve and visual inspection of quintile plots. Results: Eleven factors were identified as predictive of a PCE, including SPEED score, proxy SPEED informer, age, EMS arrival, emergent or immediate ED acuity, the number of ED visits within the last 90 days, metastatic cancer, cardiac arrhythmias, coagulopathy, depression and weight loss. In validation, the risk model had an area under the curve of 0.72 and calibration showed an underestimation of risk in the second and third quintiles. Conclusions: A risk model based on SPEED score has been successfully derived, but needs a larger dataset for proper validation. If the predictive ability of the model is confirmed, a risk model can efficiently identify cancer patients arriving to the ED who may benefit from early initiation of a palliative care consult.