Free-Text Responses in a Nationally Representative Experimental Survey about End-of-Life Care Choices: ChatGPT-4o-Assisted Qualitative Analytical Study
Bibliographic record
Abstract
Background: Little is known about how surrogates make end-of-life care choices for patients who lack the ability to make decisions for themselves. Objective: The study aims (1) to identify key themes that emerged from participants' free-text responses to a large nationally representative vignette survey about surrogate decision-making in end-of-life care and (2) to determine if an advanced artificial intelligence (AI) chatbot could assist us in accurately and efficiently performing qualitative analyses. Methods: Our dataset included 3931 free-text responses from a nationally representative survey of 6109 individuals. In this qualitative study, we first familiarized ourselves with the free-text responses and hand-coded the first 200 responses until we reached saturation. We then created a codebook, initial themes, subthemes, and illustrative quotes. Subsequently, we prompted ChatGPT-4o to analyze the entire dataset of 3931 responses and identify frequent keywords and generate themes and quotable quotes. We validated responses by comparing the AI's keyword counts to qualitative software (NVivo, Lumivero) counts and cross-validating AI-generated quotes with the original transcripts. Results: We identified several key themes: surrogates more often chose comfort care for care recipients with dementia, particularly at advanced stages. They also strongly weighed the patients' perceived quality of life and functional status. Many reported making surrogate decisions based on their own lived experiences or values, rather than making decisions aligned with the patients' previously stated wishes. There was no significant difference between the AI and qualitative software's keyword counts. The most frequent keywords included "life" (2051/81,713, 2.51%), "quality" (903/81,713, 1.11%), and dementia (507/81,713, 0.62%). Overall, AI-generated themes closely aligned with aforementioned human-generated themes. Manual coding of the first 200 free-text responses required 4 hours, including codebook development. In contrast, ChatGPT-4o generated themes in <10 seconds using the predefined codebook. However, dataset preparation, output verification, iterative prompting, debugging, and validation required several weeks. Conclusions: Surrogates often base end-of-life decisions on dementia stage, perceived quality of life, and their own lived experiences, rather than patient preferences. Using an AI chatbot to perform qualitative analysis on free-text responses may help extend the work of qualitatively trained investigators, especially for large datasets such as free-text responses to large surveys.
Fetched live from OpenAlex and de-inverted. Abstracts are not stored in this database: the inverted indexes are 8.6 GB of the frame’s 9.3 GB of text, and the host has 13 GB free.
How this classification was reachedexpand
Full frame distilled prediction
Teacher imitationNot calibrated prevalence, not ground truth. Human validation pending. Learned from the 10,348 direct Codex labels and 10,348 direct Gemma labels. Candidate is the union of thresholded teacher heads; consensus is their intersection. These outputs are machine_predicted_unvalidated and are not human labels or direct frontier model labels.
Codex and Gemma teacher scores by category
| Category | Codex | Gemma |
|---|---|---|
| Metaresearch | 0.001 | 0.003 |
| Meta-epidemiology (narrow) | 0.000 | 0.000 |
| Meta-epidemiology (broad) | 0.000 | 0.000 |
| Bibliometrics | 0.001 | 0.001 |
| Science and technology studies | 0.000 | 0.000 |
| Scholarly communication | 0.000 | 0.000 |
| Open science | 0.000 | 0.000 |
| Research integrity | 0.000 | 0.000 |
| Insufficient payload (model declined to judge) | 0.000 | 0.000 |
Machine scores (provisional)
The two teacher heads of the student model, read on this work. A score orders the frame for review; it never asserts a category, and the validation status ships verbatim with every row.
Baseline scores from an immature model (maturity gate not passed, 7 training rounds). Scores rank; they never assert a category.
score_only:v0-immature-baseline · verbatim from the scoring run: score_only means the number may rank works, and no category label ships from itClassification
machine, unvalidatedMachine predicted; a candidate call from one teacher head, not a consensus.
How this classification was reached, model by model and score by score, is at the end of the page under "How this classification was reached".