Objectives 1. Identify predictors of referral to an inpatient palliative care consult service in an academic medical center. 2. Identify key concepts of how to use administrative data to evaluate outcomes of consults. Background. There is still no consensus on what triggers should be used to initiate inpatient palliative care consults. Research objectives. Identify predictors of referral to an inpatient palliative care service (PCS) in an academic medical center. Methods. We conducted logistic regression analyses using administrative data for all hospitalizations between 2005e2008 to examine patient and physician factors associated with PCS consults for patients who survived (n 1⁄4 477) or died (n 1⁄4 282). Patient factors included age, gender, race, comorbidities, diagnosis of cancer or metastatic cancer, prior hospitalizations, length of stay (LOS) category, and discharge status. LOS and discharge were included as proxies for severity of illness and complexity of care. Physician factors included a propensity score for referral (coded as terciles for analysis) computed using age, gender, years of employment and hospital department. Results. For survivors, significant predictors (31 days: OR 1⁄4 14.6; CI 1⁄4 10.2-20.8), discharge status (hospice: OR 1⁄4 19.5; CI 1⁄4 13.9-27.5; skilled nurse facility: OR 1⁄4 2.3; CI 1⁄4 1.7-3.1), physician propensity (upper tercile: OR 1⁄4 2.4; CI 1⁄4 1.83.1) metastatic cancer diagnosis (OR 1⁄4 1.9; CI 1⁄4 1.2-2.9), and comorbidities (OR 1⁄4 1.2; CI 1⁄4 1.1-1.3). For decedents, significant predictors included LOS (7-30 days: OR 1⁄4 3.4; CI 1⁄4 2.4-4.7; > 31 days: OR 1⁄4 6.9; CI 1⁄4 4.4-10.7), metastatic cancer diagnosis (OR 1⁄4 2.7; CI 1⁄4 1.4-5.3), and physician propensity (upper tercile: OR 1⁄4 1.4; CI 1⁄4 1.0-1.9). Conclusion. Palliative care referrals were mostly likely to be initiated for survivors who were expected to die or for all patients with long LOS. Implications for research, policy, or practice. These results suggest examining LOS as a primary trigger for consults as this variable is objective and may serve as a proxy for severity of illness and complex care.