隨機化研究設計及其衍生樣本流失之處理策略

Managing Attrition in Randomized Research Design

隨機控制試驗是當前評價實證護理之介入措施成效的黃金準則。隨機化可降低選擇偏誤對內在效度的威脅，排除外在變數的影響，強化臨床試驗結果的因果推論。然而，隨機化可能衍生樣本流失的問題、降低試驗的統計效力，嚴重威脅內在效度，損壞資料結構，使研究結果推論產生改變或被扭曲。幸而，可運用減少樣本流失、重獲樣本及統計分析等處理策略來因應樣本流失的影響，進而提高研究的效度與推論性。研究文章亦應具體說明有關樣本流失的研究設計與處理策略，以及完整的樣本流失分析及質化資料。本文期能提供讀者做為實證護理之隨機化設計與執行的參酌。

A randomized clinical trial is nowadays the gold standard for evaluating the effectiveness of interventions in evidence-based nursing. Randomization can reduce the threat of selection bias to internal validity and remove the influence of extraneous variables; therefore, it strengthens causal inference from results of clinical trials. However, randomization may result in sample loss and reduction of statistical power that leads to seriously threaten the internal validity, corrupt the data set, and distort the casual inference. Fortunately, to deal with such issues, the strategies of reducing attrition, recapturing samples, and statistical methods could be used. This article provides a reference in designing and implementing randomization on evidence-based nursing.