(M) Development of a Tool for Visualizing Patient Journeys Using Disease Blogs on The Internet

Publication Number: 671

Background: All services should exist for customers, and understanding customer needs is essential. Customer journey maps are commonly used to visualize their experiences in the marketing field. Following this system, it is quite meaningful to have a tool for patients in the healthcare field to help understanding emotional distress with treatment and making a solution for it. Therefore, we attempted to develop a patient journey map (PJM) that visualizes patients' actual experiences, and this is expected to promote the provision of patient-centric solutions.

Objectives: To develop the PJM for visualizing a patient journey using disease blogs obtained from the Internet and identifying unsatisfactory events for patients by applying natural language processing (NLP) technique.

Methods: Using NLP and graph database theory approach, we first tried to visualize medical paths. For medical terms that appear multiple times (e.g., hospital admission), we adopted only the first appearance of each term as a node. In order to evaluate the validity of our approach, we calculated the patient satisfaction index (PSI) of sentences which contain the term using the prediction model presented previously (Matsuda, et al. ICPE 2020) and verified the results.

Results: We analyzed 48 disease blogs written by patients with breast cancer (20-34 years old: n = 10, 35-49 years old: n = 30, 50-64 years old: n = 8). Using the PJM, we succeeded in visualizing a series of patient’s treatment experiences as a path which consists of each treatment experience as a node, and the path lined up in a chronological order. By displaying each patient’s path, we recognized various treatment experience patterns. Furthermore, we were able to quantify the patient satisfaction in each node; A palliative care (mean PSI: 0.11), discharge (0.06), the last chemotherapy (0.05), clinical trial (0.04), and rehabilitation (0.03) showed the highest scores. On the other hand, the nodes showing the lowest PSIs included swelling (-0.19), subjective symptoms (-0.19), morphine (-0.19), uncomfortable feelings (-0.17), and life expectancy (-0.15). Box-plotting enabled us to compare the level of patient satisfaction for different stages along the patient journey (e.g., "examination," "diagnosis," "prognosis"). Overall, the tool enabled us to trace patients' experiences with a reasonable PSI score reflected at each point.

Conclusions: PJM using disease blogs focuses on identifying the patient's disease transition and emotions to comprehend the patient's issues and anxieties from the patient's perspective. By quantitatively assessing patient satisfaction in the patient journey, it would promote understanding patient-level treatment concerns and developing solutions.