315 Background: Effective decision-making on whether to start the next line of therapy for advanced cancer requires comprehensive, tailored information aligned with needs and preferences of patients and caregivers. Our study aimed to compare patient and caregiver information needs to information that oncologists consider and share when discussing prognosis before starting a new line of therapy. Methods: Our study used Latent Dirichlet Allocation (LDA)-based topic modeling, a text mining method, to extract topics embedded in semi-structured interview transcripts from patient and caregiver (n=20) and oncologist (n=10) participants. Transcripts were collected as part of a larger study to design an interface for a prognostic tool. Patients and caregivers were asked to identify information needs when deciding to start a new line of therapy. Oncologists were asked through critical incident technique to recall a recent patient encounter and, with probing questions, describe information they consider and how they communicate expected survival time. Two members of the study team identified extracted topics using an interactive, intertopic distance map, then compared topics between patient/caregiver and oncologist datasets. Results: Results revealed 6 distinct topics from patients and caregivers and 6 distinct topics from oncologists that we grouped into 3 overall themes (Table 1). While topics across both participant groups were similar, two notable differences were found in patient and caregiver interviews: 1) the desire for information about support-related resources, which was not noted by oncologists, and 2) topics included keywords like “helpful” and “clear” to describe information needs, highlighting the importance of how information is presented by oncologists. Conclusions: LDA-based topic modeling was useful to identify topics from transcripts, highlighting information needs and what is shared when discussing prognosis and starting a new line of therapy. Limitations include use of self-reported descriptions and lack of actual conversations between patients and oncologists. However, findings show the ongoing gap in communication about support-related resources when discussing prognosis. Opportunities exist to enhance communication about prognosis and promote access to resources using decision support tools.[Table: see text]