Prodinger B, Hammond A, Tennant A, Prior Y, Tyson S
Swiss Paraplegic Research, Nottwil, Switzerland.
Department of Health Sciences and Health Policy, University of Lucerne, Lucerne, Switzerland.
BMC Musculoskelet Disord. 2019 Jan 25;20(1):41. doi: 10.1186/s12891-019-2414-6.
Limitations in upper limb functioning are common in Musculoskeletal disorders and the Disabilities of the Arm, Shoulder and Hand scale (DASH) has gained widespread use in this context. However, various concerns have been raised about its construct validity and so this study seeks to examine this and other psychometric aspects of both the DASH and QuickDASH from a modern test theory perspective.
Participants in the study were eligible if they had a confirmed diagnosis of Rheumatoid Arthritis (RA). They were mailed a questionnaire booklet which included the DASH. Construct validity was examined by fit to the Rasch measurement model. The degree of precision of both the DASH and QuickDASH were considered through their Standard Error of Measurement (SEM).
Three hundred and thirty-seven subjects with confirmed RA took part, with a mean age of 62.0 years (SD12.1); 73.6% (n = 252) were female. The median standardized score on the DASH was 33 (IQR 17.5-55.0). Significant misfit of the DASH and QuickDASH was observed but, after accommodating local dependency among items in a two-testlet solution, satisfactory fit was obtained, supporting the unidimensionality of the total sets and the sufficiency of the raw (ordinal or standardized) scores.
Having accommodated local response dependency in the DASH and QuickDASH item sets, their total scores are shown to be valid, given they satisfy the Rasch model assumptions. The Rasch transformation should be used whenever all items are used to calculate a change score, or to apply parametric statistics within an RA population.
Most previous modern psychometric analyses of both the DASH and QuickDASH have failed to fully address the effect of a breach of the local independence assumption upon construct validity. Accommodating this problem by creating 'super items' or testlets, removes this effect and shows that both versions of the scale are valid and unidimensional, as applied with a bi-factor equivalent solution to an RA population. The Standard Error of Measurement of a scale can be biased by failing to take into account the local dependency in the data which inflates reliability and thus making the SEM appear better (i.e. smaller) than the true value without bias.
上肢功能受限在肌肉骨骼疾病中很常见，手臂、肩部和手部功能障碍量表（DASH）在此背景下得到了广泛应用。然而，人们对其结构效度提出了各种担忧，因此本研究旨在从现代测试理论的角度检验DASH和快速DASH（QuickDASH）的这一及其他心理测量学方面。
如果参与者被确诊患有类风湿性关节炎（RA），则符合本研究的条件。他们收到了一本包含DASH的问卷手册。通过拟合拉施测量模型来检验结构效度。通过测量标准误（SEM）来考虑DASH和QuickDASH的精确程度。
337名确诊为RA的受试者参与了研究，平均年龄为62.0岁（标准差12.1）；73.6%（n = 252）为女性。DASH的标准化分数中位数为33（四分位距17.5 - 55.0）。观察到DASH和QuickDASH存在显著的不拟合，但在采用双子测验解决方案来处理项目间的局部依赖性后，获得了满意的拟合度，支持了总分集的单维度性以及原始（有序或标准化）分数的充分性。
在DASH和QuickDASH项目集中考虑了局部反应依赖性后，它们的总分被证明是有效的，前提是它们满足拉施模型假设。每当使用所有项目来计算变化分数，或在类风湿性关节炎人群中应用参数统计时，都应使用拉施变换。
以前对DASH和QuickDASH进行的大多数现代心理测量分析都未能充分解决违反局部独立性假设对结构效度的影响。通过创建“超级项目”或子测验来解决这个问题，消除了这种影响，并表明这两个量表版本在应用于类风湿性关节炎人群的双因素等效解决方案中是有效且单维度的。如果不考虑数据中的局部依赖性，量表的测量标准误可能会产生偏差，这会夸大信度，从而使测量标准误看起来比无偏差的真实值更好（即更小）。
