Elucidation of qualitative patient-reported outcome (PRO) data using latent dirichlet allocation (LDA) in patients (pts) undergoing radical cystectomy.
e18245 Background: As PROs become commonplace in clinical trials and routine care, we must identify efficient analytic methods to better understand and ultimately integrate this information into clinical decision-making. While PROs are often captured via qualitative methods to provide context and pt voice to traditional objective and/or quantitative assessments, the current analytic standard is resource intensive and impractical for timely delivery of this information to clinicians and trialists. To address this problem, we propose the use of LDA, a natural language processing technique that automates the distillation of vast amounts of free-text data into underlying topics. Methods: As part of a prospective longitudinal trial of pts ( N = 200; M age = 63.8; 20% female; 46%) undergoing radical cystectomy (52% stage T2-T4) and urinary diversion (49% continent) for bladder cancer (ClinicalTrials.gov identifier NCT00745355), pts completed 30 minute interviews at baseline and 6-months post-surgery where they were asked to characterize their quality of life with respect to goals or milestones they would like to reach, problems they would like to solve, situations they would like to prevent, things they would like to remain the same, and commitments they would be willing to let go. LDA was used to extract latent topics and themes, stratified by time. Results: LDA extracted 7 latent topics and showed shifting patient priorities. At baseline, pt primary concerns were related to surgery and recovery. Six months post-surgery, pt goals transitioned toward themes related to concerns about cancer recurrence, regaining a sense of normalcy, return to work, enjoyment of life and increased appreciation of friends and family. Conclusions: LDA was used as an automated tool to illustrate a shift in pt goals and concerns as they transition into survivorship. This novel data analytic technique offers the possibility of significantly reducing the resources required to summarize qualitative PRO data. As such, this information may now be more readily available for immediate inclusion in the decision-making process as part of routine care and clinical trials. Clinical trial information: NCT00745355.