以ChatGPT輔助非隨機試驗報告規範（TREND）在護理類實驗研究

Applying the Transparent Reporting of Evaluations with Nonrandomized Designs (TREND) Guidelines to Nursing Non-Experimental Studies: An Examination Using ChatGPT

研究背景：隨機對照試驗雖被視為介入研究的黃金標準，但在護理領域常因倫理與操作限制而採用類實驗設計。然而，類實驗設計易受選擇偏差影響，降低研究效度。為提升類實驗研究的透明度與可信度，非隨機試驗報告規範（TREND, Transparent Reporting ofEvaluations with Nonrandomized Designs）提供了系統化報告框架，但其在護理研究中的應用仍不普及，近十年來尚缺乏系統性探討。本研究利用ChatGPT分析2014至2023年間TREND在護理類實驗設計中的應用情況及其對研究品質的影響。
研究方法：研究分為兩階段：第一階段重現Fuller et al. (2012, 2014)的設計，擴展TREND在護理領域的應用範圍；第二階段基於EPHPP-QA工具設計符合度評分系統，量化TREND各項目的應用情況，並運用ChatGPT進行系統化文獻分析。
研究結果：2014至2023年間，類實驗設計文獻的發表數量有所波動，但引用與運用TREND的文獻比例平均僅為所有發表文獻的7.4%。引用與運用TREND的文獻主要應用於病人醫療介入研究，而未使用TREND的文獻多見於學生教育介入。運用TREND的文獻在研究品質上更為優異，特別是在雙組對照設計的嚴謹性、介入措施的詳細描述及結構化摘要上均有顯著提升。ANOVA分析結果顯示，運用TREND的文獻在各項符合度指標上的得分顯著高於未使用組別。
結論與建議：TREND在護理類實驗設計中的應用仍不普及，可能因研究者認識不足及推廣力度不夠所致。應用TREND能提升研究設計的嚴謹性與透明度。建議未來開發基於AI的TREND檢查工具，以提升文獻報告品質並促進其更廣泛應用。

Background: While randomized controlled trials (RCT) are considered the gold standard for intervention research, quasi-experimental designs are often used in the nursing field due to ethical and practical constraints. However, quasiexperimental designs are prone to selection bias, which can undermine internal validity. To address these issues, Transparent Reporting of Evaluations with Nonrandomized Designs (TREND) provides a structured framework to improve the transparency and credibility of nonrandomized studies. Despite its potential benefits, TREND has not been widely adopted in nursing research and its application over the past decade has not been systematically explored. This study utilizes ChatGPT to analyze the application of TREND in nursing-related quasi-experimental studies from 2014 to 2023 and its impact on research quality.
Methods: The study was conducted in two phases. The first phase replicated the design of Fuller et al. (2012, 2014) to expand the scope of TREND in nursing research. The second phase involved the development of a compliance scoring system based on the EPHPP-QA tool to quantify the application of TREND items and systematically analyze the literature using ChatGPT.
Results: Between 2014 and 2023, the number of quasi-experimental studies fluctuated, but the proportion of studies citing and applying TREND was only 7.4% on average. TREND was used primarily in studies focused on patient medical interventions, while studies not using TREND were more common in educational interventions for students. Studies that applied TREND showed significantly higher research quality, particularly in the rigor of two-group comparisons, detailed description of the intervention, and structured abstracts. The ANOVA results indicated that the scores of the studies using TREND were significantly higher across all compliance indicators compared to those not using TREND.
Conclusion and recommendations: The application of TREND in nursing quasi-experimental designs remains limited, likely due to researchers' lack of awareness and insufficient promotion. Using TREND can improve the rigor and transparency of research design. Future efforts should focus on developing AI-based TREND compliance tools to enhance reporting quality and promote broader adoption.