Introduction

Understanding patients? daily life experience provides valuable information to promote patient-centred healthcare in pulmonary fibrosis(PF). Patients often share needs and struggles online, which may differ from those shared with healthcare providers. 

Aim

We applied a new machine learning-driven(MLD) approach to analyze patient experience data from online communities to develop a patient disease trajectory map for PF.

Methods

MLD topic modelling was used to generate topic clusters among the posts of online PF communities, after ethics and data management plan approval. Study information was shared with the communities. No indirect and direct identifiers were collected. MLD sentimental analysis was used to label sentiments. Then, humans analyzed each cluster by reading the 50 most relevant posts. A patients? disease trajectory map was made based on these findings. Further validation is done by medical experts and patient panels.

Results

Based on 47.282 posts, 100 topic clusters were produced such as: moments of making difficult decisions, aspects of life change, and coping strategies of symptom. The patient disease trajectory map was based on these findings (Figure).

Conclusion

This machine learning-driven trajectory map provided 1) patient-oriented insights into the needs and challenges in their daily lives and 2) structured insights to guide the right service at the right moment to build patient-centred care.