127 Background: Hospice care can improve quality of life (QOL) for patients with advanced cancer. Studies suggest that hospice services are offered late in the course of terminal illness; limiting patient comfort and QOL. Limited literature exists on barriers to timely hospice referral. We examined demographic and clinical characteristics associated with late hospice referral. We hypothesized that age, malignancy type, and performance status (PS) may predict late referral to hospice. Methods: Patients were age 18+ cared for by oncologists of Baystate Medical Practices, referred to Baystate Hospice from 1/1/08 to 12/31/2010. Demographics and diagnosis (e.g., breast, lung, GI) were abstracted from the electronic medical record. Referral interval was calculated as date of death minus date of referral. Late referrals (“cases”) were classified as interval quintiles 1 and 2; early referrals (“controls”) quintiles 4 and 5. Fisher’s exact test and Wilcoxon Rank Sum identified demographic and clinical characteristics associated with late referral. Multivariable logistic regression was used to identify independent predictors of late referral. Results: 209 patients met inclusion criteria. The median referral interval was 29 days (IQR 12, 57). In the multivariable model, ECOG score (OR 1.51 95% CI 1.09 2.08); dementia (OR 0.08, 95% CI 0.01, 0.82); CNS cancer (OR 0.14, 95% CI 0.03, 0.73); and unclassified cancer (OR 0.32, 95% CI 0.14, 0.73) independently predicted late referral. Age, race, and sex were not associated in uni- or multivariable comparisons. Conclusions: In this study of 209 patients with advanced cancer enrolled to hospice, malignancy type, dementia, and performance status are associated with late hospice referral. We believe it is important to track timing of referral and look for ways to reduce barriers.