96 Background: Administration of chemotherapy close to death is widely recognized as poor quality care. The goals of this study were to describe the processes of how chemotherapy is discontinued and examine their relationships with timing before death, hospice referrals, and terminal hospitalizations. Methods: We reviewed electronic health records of a prospective cohort of 151 patients with newly diagnosed metastatic non-small cell lung cancer (NSCLC) who participated in a randomized trial of early palliative care. Chemotherapy treatment during the final regimen was qualitatively analyzed, identifying categories of discontinuation processes. We then quantitatively investigated predictors and outcomes of the discontinuation process categories. Results: 144 patients had died, with 81 and 48 receiving intravenous (IV) and oral chemotherapy as final regimens, respectively. Five discontinuation processes were identified: definitive decisions; deferred decisions (breaks); disruptions for radiation therapy; disruptions due to hospitalizations; and no decisions. For patients receiving IV chemotherapy, definitive decisions were least frequent (19.7%), while disruptions due to hospitalizations were most frequent (27.2%). The different processes occurred at significantly different times before death; and often hospice referrals occurred months after chemotherapy discontinuation. Deferred decisions, or “breaks,” occurred at the longest interval before death, but definitive decisions resulted in greatest hospice utilization. Among patients receiving oral chemotherapy, 83.3% were switched from IV to oral as their final regimen, sometimes concurrent with or even after hospice referral. Conclusions: Patients with metastatic NSCLC stop their final chemotherapy regimen via different processes, which significantly vary in time before death and in subsequent end-of-life care.