The Use of Natural Language Processing to Interpret Unstructured Patient Feedback on Health Services: Scoping Review

Area of Science:

• Health Informatics
• Natural Language Processing
• Patient Experience Research

Background:

• Unstructured patient feedback (UPF) offers direct patient insights, amplified by online healthcare platforms.
• Natural language processing (NLP) techniques like sentiment analysis and topic modeling are increasingly used to analyze UPF.
• The scope and clinical relevance of NLP in analyzing UPF within healthcare remain unclear.

Purpose of the Study:

• To conduct a scoping review on the application of NLP techniques for interpreting UPF in healthcare.
• To identify healthcare settings, purposes, and reported clinical impacts of NLP in UPF analysis.
• To synthesize current research on NLP's role in understanding patient experiences.

Main Methods:

• Searched MEDLINE, Embase, CINAHL, Cochrane, and Google Scholar in February 2024 without date restrictions.
• Included English-language studies using NLP on UPF in identifiable healthcare settings or involving providers.
• Excluded studies with solely human coding or NLP on structured/non-patient feedback; data synthesized narratively.

Main Results:

• 52 studies met inclusion criteria, predominantly from secondary care (n=33).
• Common NLP techniques included sentiment analysis (n=32), topic modeling (n=15), and text classification (n=7).
• Limited association was found between NLP insights and traditional quality metrics, with few reported clinical impacts.

Conclusions:

• NLP is applied to UPF mainly to identify factors supporting positive patient experiences.
• Despite academic interest, evidence of NLP implementation in clinical settings is scarce.
• Future research should focus on NLP's ability to capture healthcare nuances, align with quality metrics, and influence clinician behavior.