Department of Medical Social Sciences, Northwestern University Feinberg School of Medicine, Chicago, USA.
Centre for Patient Reported Outcomes Research, Department of Applied Health Sciences, University of Birmingham, Edgbaston Birmingham, UK, Edgbaston, Birmingham, UK.
Qual Life Res. 2024 Dec;33(12):3223-3228. doi: 10.1007/s11136-024-03788-9. Epub 2024 Sep 28.
Interpretation of patient-reported outcome (PRO) scores has been supported by identifying score thresholds or ranges that indicate clinical importance. There has been a recent focus on the estimation of meaningful within patient change (MWPC). While much attention has been focused on anchor-based methods, some researchers prefer that a lower bound to these estimates should exceed a change score that could be observed due to measurement error alone as a safeguard against misclassifying individual patients as changed when they have not. The standard error of measurement (SEM) is often used as the lower bound of anchor estimates. Here, we argue that the SEM is not an the best lower bound for MWPCs. Instead, statistically significant individual change as calculated by the reliable change index (RCI) should be used as the lower bound. Our argument is based on two points. First, conceptually, the SEM does not provide specific enough information to serve as a lower bound for MWPCs, which should be based on the level of observed score change that is unlikely to be due to chance alone. Second, the SEM is not appropriate for direct application to observed scores, and requires a multiplier when examining observed change instead of true change. We conclude with recommendations for using the RCI with a thoughtful range of p-values in combination with anchor estimates.
患者报告结局（PRO）评分的解释得到了确定评分阈值或范围的支持，这些范围表明了临床重要性。最近人们关注的重点是估计有意义的患者内变化（MWPC）。虽然锚定方法得到了广泛关注，但一些研究人员希望这些估计的下限应该超过由于测量误差单独导致的变化得分，以防止将没有变化的个体患者错误地归类为有变化。测量误差（SEM）通常用作锚定估计的下限。在这里，我们认为 SEM 不是 MWPC 的最佳下限。相反，应该使用可靠变化指数（RCI）计算的统计学上显著的个体变化作为下限。我们的论点基于两点。首先，从概念上讲，SEM 不能提供足够具体的信息作为 MWPC 的下限，MWPC 的下限应该基于不太可能仅仅由于机会导致的观察到的评分变化水平。其次，SEM 不适合直接应用于观察到的分数，并且在检查观察到的变化而不是真实变化时需要一个乘数。我们的结论是，建议使用 RCI 并结合深思熟虑的 p 值范围，以及锚定估计。
