International journal of obstetric anesthesia · 2026
The use of structured prompt to improve the readability, accuracy, and quality of ChatGPT responses on labor epidural analgesia: a brief report.
СтатьяНазвание и аннотация: Перевод в очереди
Обработка статьи · этапы, ошибки и управление
Загрузка состояния…
Ошибки и ограничения
История · было → стало
Состояние обновляется каждые 5 секунд, пока блок открыт. Процент показывается только там, где известен объём работы. Пауза останавливает продолжение этапа; текущий запрос к источнику может завершиться.
Tan C W, Sng R J K, Chan J C Y, Chan J J I, Nagarajan S, Sng B L
Аннотация
BACKGROUND: Artificial intelligence chatbot responses to frequently asked questions on labor epidural analgesia reported poor readability with limited consideration on personalized context. We investigated whether a structured prompt based on Context, Role, Action, Format, Tone (CRAFT) framework could improve the readability and accuracy of these chatbot-generated responses.
METHODS: Twenty-nine questions on labor epidural analgesia covering patient preferences and clinical scenarios were evaluated in ChatGPT-5.2. Two sets of responses were produced: one generated with direct questions and the other using CRAFT-based prompt prior to the questions. Readability metrics were applied to each response and averaged, while accuracy was calculated as the proportion of correct statements. Quality was assessed using Patient Education Materials Assessment Tool for Print, with four obstetric and gynecological anesthesiologists independently evaluating understandability and actionability on top of accuracy.
RESULTS: Responses generated using CRAFT-based prompt had significantly better readability than those with direct questions only. The direct-question responses ranged from slightly difficult to professional (9th grade to college level), whereas the CRAFT-based prompt responses ranged from fairly easy to slightly difficult (6th to 9th grade). Accuracy was significantly higher in CRAFT-based prompt group than direct-question group (99.8% ± 0.2% vs. 99.0% ± 0.4%, P = 0.029). No significant differences were observed in understandability or actionability scores.
CONCLUSIONS: Incorporating CRAFT-based prompts could generate responses with high accuracy and quality on understandability and actionability. Patient education could involve customized prompts to improve readability and accuracy. This approach could thus improve access to reliable clinical information and support shared decision-making.
METHODS: Twenty-nine questions on labor epidural analgesia covering patient preferences and clinical scenarios were evaluated in ChatGPT-5.2. Two sets of responses were produced: one generated with direct questions and the other using CRAFT-based prompt prior to the questions. Readability metrics were applied to each response and averaged, while accuracy was calculated as the proportion of correct statements. Quality was assessed using Patient Education Materials Assessment Tool for Print, with four obstetric and gynecological anesthesiologists independently evaluating understandability and actionability on top of accuracy.
RESULTS: Responses generated using CRAFT-based prompt had significantly better readability than those with direct questions only. The direct-question responses ranged from slightly difficult to professional (9th grade to college level), whereas the CRAFT-based prompt responses ranged from fairly easy to slightly difficult (6th to 9th grade). Accuracy was significantly higher in CRAFT-based prompt group than direct-question group (99.8% ± 0.2% vs. 99.0% ± 0.4%, P = 0.029). No significant differences were observed in understandability or actionability scores.
CONCLUSIONS: Incorporating CRAFT-based prompts could generate responses with high accuracy and quality on understandability and actionability. Patient education could involve customized prompts to improve readability and accuracy. This approach could thus improve access to reliable clinical information and support shared decision-making.