Department of Medicine, Stanford University School of Medicine, Palo Alto, California, USA.
J Rheumatol. 2011 Aug;38(8):1759-64. doi: 10.3899/jrheum.110402.
Use of item response theory (IRT) and, subsequently, computerized adaptive testing (CAT), under the umbrella of the NIH-PROMIS initiative (National Institutes of Health-Patient-Reported Outcomes Measurement Information System), to bring strong new assets to the development of more sensitive, more widely applicable, and more efficiently administered patient-reported outcome (PRO) instruments. We present data on current progress in 3 crucial areas: floor and ceiling effects, responsiveness to change, and interactive computer-based administration over the Internet.
We examined nearly 1000 patients with rheumatoid arthritis and related diseases in a series of studies including a one-year longitudinal examination of detection of change; compared responsiveness of the Legacy SF-36 and HAQ-DI instruments with IRT-based instruments; performed a randomized head-to-head trial of 4 modes of item administration; and simulated the effect of lack of floor and ceiling items upon statistical power and sample sizes.
IRT-based PROMIS instruments are more sensitive to change, resulting in the potential to reduce sample size requirements substantially by up to a factor of 4. The modes of administration tested did not differ from each other in any instance by more than one-tenth of a standard deviation. Floor and ceiling effects greatly reduce the number of available subjects, particularly at the ceiling.
Failure to adequately address floor and ceiling effects, which determine the range of an instrument, can result in suboptimal assessment of many patients. Improved items, improved instruments, and computer-based administration improve PRO assessment and represent a fundamental advance in clinical outcomes research.
利用项目反应理论（IRT）和随后的计算机自适应测试（CAT），在 NIH-PROMIS 计划（美国国立卫生研究院-患者报告结局测量信息系统）的框架下，为开发更敏感、更广泛适用和更高效管理的患者报告结局（PRO）工具带来强大的新资产。我们介绍了在以下 3 个关键领域的当前进展数据：地板效应和天花板效应、对变化的反应能力以及基于互联网的交互式计算机管理。
我们在一系列研究中检查了近 1000 名类风湿关节炎和相关疾病患者，其中包括一项为期一年的变化检测的纵向研究；比较了基于 IRT 的工具与 Legacy SF-36 和 HAQ-DI 工具的反应能力；进行了 4 种项目管理模式的随机头对头试验；并模拟了缺乏地板和天花板项目对统计功效和样本量的影响。
基于 IRT 的 PROMIS 工具对变化更敏感，这有可能将样本量要求大大减少多达 4 倍。在任何情况下，所测试的管理模式之间的差异都不超过十分之一的标准差。地板和天花板效应大大减少了可用受试者的数量，尤其是在天花板处。
未能充分解决决定工具范围的地板和天花板效应，可能导致许多患者的评估不理想。改进的项目、改进的工具和基于计算机的管理改进了 PRO 评估，代表了临床结果研究的重大进展。
