“I Want to Do It on My Own”: A Multimethod Study Exploring When and Why Hospitalists Opt for Palliative Care to Support Goals-of-Care Discussions

Background: Timely goals-of-care discussions (GOCDs) are critical for hospitalized patients with serious illness, yet clinicians face barriers in identifying appropriate patients and initiating these conversations, prompting mortality prediction models and triggered palliative care consults. Limited palliative care capacity impedes the scalability of automatic consultations. We developed a machine-learning-based workflow that prompted hospitalists to initiate GOCDs, with palliative care consultation available through an opt-in mechanism rather than a default trigger. Our study evaluated when, how often, and why hospitalists engage with palliative care in this system. Methods: We conducted a multimethod study involving a quantitative data analysis of message responses ( N = 1263) and semi-structured qualitative interviews with hospitalists who engaged with the system ( n = 9). Quantitative data were analyzed using mixed-effects logistic regression to identify encounter and patient characteristics associated with selecting a palliative care consult versus choosing to initiate a GOCD themselves. Qualitative data were analyzed using a qualitative descriptive approach by coding transcripts and generating themes. Outcomes: Hospitalists elected to initiate GOCDs themselves more often than opting for palliative care to have the conversation (67.9% vs. 32.1%). The odds of involving palliative care varied significantly by hospitalist and hospital but did not differ across patient demographics and other clinical factors. In interviews, hospitalists described a preference to manage GOCDs themselves, driven by perceived professional responsibility, efficiency, and personal meaning, but endorsed palliative care involvement for complex clinical or family situations and during periods of high workload. Conclusion: In a hospitalist-as-default workflow, hospitalists often initiated GOCDs themselves, driven by professional responsibility, while reserving palliative care for complex cases and high workloads. Paired with observed variability in palliative care engagement by hospitalist and hospital, this suggests that selectively triggered consults based on hospitalist preferences, workload, case complexity, and hospital norms can balance professional ownership with sustainable use of palliative care resources.