Do SF-36 summary component scores accurately summarize subscale scores?
Do SF-36 summary component scores accurately summarize subscale scores?
复制标题
DOI：
10.1023/a:1012552211996
复制
发表时间：
2001-01-01
影响因子：
3.5
通讯作者：
Sullivan, M
中科院分区：
文献类型：
--
作者：
Taft, C;Karlsson, J;Sullivan, M
标准评分算法最近可用于将来自8个SF-36子量表的分数聚合为两个不同的高阶总结分数：物理成分摘要（PCS）和心理成分摘要（MCS）。然而，最近的研究表明，PCS和MCS分数并不是独立的，可能在某种程度上测量的是相同的结构。本文的目的是检验和说明(1)SF-36子量表与PCS/MCS分数之间的关系，(2)PCS和MCS分数之间的关系，以及(3)它们对研究结果的解释意义。模拟分析说明了评分算法的各个方面对子量表轮廓和总结成分得分之间潜在差异的贡献。使用瑞典SF-36标准数据库，进行相关分析和回归分析，以估计两个组成部分之间的关系，以及各分量表对组成部分的相对贡献。确定并解释了子量表剖面和成分得分之间的差异。PCS和MCS评分在各自的评分上限区间存在显著相关（r = -0.74, -0.67），表明各成分不是独立的。回归分析显示，在这些范围内，PCS主要测量心理健康（57%的方差），MCS主要测量身体健康（65%的方差）。讨论了研究结果的意义。结论是，目前的PCS/MCS评分程序不准确地总结了子量表的得分，因此应进行修订。在此之前，应该谨慎地解释组件分数，并且只能与概要分数相结合。
Standard scoring algorithms were recently made available for aggregating scores from the eight SF-36 subscales in two distinct, higher-order summary scores: Physical Component Summary (PCS) and Mental Component Summary (MCS). Recent studies have suggested, however, that PCS and MCS scores are not independent and may in part be measuring the same constructs. The aims of this paper were to examine and illustrate (1) relationships between SF-36 subscale and PCS/MCS scores, (2) relationships between PCS and MCS scores, and (3) their implications for interpreting research findings. Simulation analyses were conducted to illustrate the contributions of various aspects of the scoring algorithm to potential discrepancies between subscale profile and summary component scores. Using the Swedish SF-36 normative database, correlation and regression analyses were performed to estimate the relationship between the two components, as well as the relative contributions of the subscales to the components. Discrepancies between subscale profile and component scores were identified and explained. Significant correlations (r = -0.74, -0.67) were found between PCS and MCS scores at their respective upper scoring intervals, indicating that the components are not independent. Regression analyses revealed that in these ranges PCS primarily measures aspects of mental health (57% of variance) and MCS measures physical health (65% of variance). Implications of the findings were discussed. It was concluded that the current PCS/MCS scoring procedure inaccurately summarizes subscale profile scores and should therefore be revised. Until then, component scores should be interpreted with caution and only in combination with profile scores.