Free-Text Responses in a Nationally Representative Experimental Survey about End-of-Life Care Choices:...

Area of Science:

• Medical Ethics
• Gerontology
• Health Services Research

Background:

• Understanding surrogate decision-making in end-of-life care for incapacitated patients is limited.
• Surrogates' choices significantly impact patient outcomes and quality of life during terminal illness.

Purpose of the Study:

• To identify key themes in surrogate decision-making for end-of-life care using qualitative analysis.
• To evaluate the efficacy of an advanced artificial intelligence (AI) chatbot in performing qualitative analysis of free-text survey responses.

Main Methods:

• Qualitative analysis of 3931 free-text responses from a national survey on end-of-life care decisions.
• Initial manual coding of 200 responses to develop a codebook, followed by AI analysis using ChatGPT-4o.
• Validation of AI-generated themes and keyword counts against manual coding and qualitative software (NVivo).

Main Results:

• Key themes included prioritizing comfort care for dementia patients, especially in advanced stages, and weighing perceived quality of life.
• Surrogates' decisions were often based on personal experiences or values, not always aligning with patients' stated wishes.
• AI analysis demonstrated comparable keyword counts to qualitative software and generated themes consistent with human analysis, significantly reducing initial coding time.

Conclusions:

• End-of-life decisions by surrogates are influenced by dementia stage, quality of life perceptions, and personal experiences over patient preferences.
• AI chatbots offer a promising tool to augment qualitative research capabilities, particularly for analyzing large datasets in health services research.