| Abstract: |
As we begin to leverage Big Data in health care settings and particularly in assessing patient-reported outcomes, there is a need for novel analytics to address unique challenges. One such challenge is in coding transcribed interview data, typically free-text entries of statements made by interviewees during face-to-face interviews. Conventional coding of such qualitative data into themes is labor-intensive and prone to inconsistencies. Latent Dirichlet Allocation (LDA) may offer statistical rigor in summarizing patients' concerns and coping strategies in a life-threatening illness. We aim to apply LDA to interview data collected as part of a prospective, longitudinal study of QOL in patients undergoing radical cystectomy and urinary diversion for bladder cancer. LDA showed that, prior to surgery, patients' priorities were primarily in cancer surgery and recovery. Six months after the surgery, however, their goals shifted to a desire to spend more time with family, resume work, and enjoy life to its fullest extent. Novel analytics such as LDA offer the possibility of summarizing personal goals in real time without the need for conventional fixed-length measures and qualitative data coding. © 2018 IEEE. |
