Estimating Patient Satisfaction Through a Language Processing Model: Model Development and Evaluation

Area of Science:

• Natural Language Processing (NLP)
• Medical Informatics
• Psychological Medicine

Background:

• Patient satisfaction is vital in healthcare.
• Natural language processing (NLP) offers advanced insights from textual data.
• Patient-reported data is often limited.

Purpose of the Study:

• To develop a model quantifying patient satisfaction.
• To utilize diverse, patient-written textual data for satisfaction analysis.

Main Methods:

• A neural network-based NLP model was created.
• Data from ~20 million sentences in 56,357 Japanese disease blogs (1994-2020) were used.
• A regression approach predicted the patient satisfaction index (PSI).

Main Results:

• The model achieved excellent effectiveness in quantifying patient satisfaction (Spearman's ρ=0.832).
• Predicted and actual PSI values showed high correlation.
• PSI significantly decreased post-cancer notification, reflecting emotional impact.

Conclusions:

• The developed model quantifies patient dissatisfaction and emotional shifts.
• This approach can identify issues in routine medical practice.
• It offers a novel method for analyzing patient-reported outcomes.