Background 

Negative mood is an important risk factor for poor clinical outcomes among individuals with musculoskeletal pain. Screening for negative mood can aid in identifying those who may need additional psychological interventions. Limitations of current negative mood screening tools include (1) high response burden, (2) a focus on single dimensions of negative mood, (3) poor precision for identifying individuals with low or high negative mood levels, and/or (4) design not specific for use in populations with orthopaedic conditions and musculoskeletal pain.

Questions/purposes 

(1) Can item response theory methods be used to construct screening tools for negative mood (such as depression, anxiety, and anger) in patients undergoing physical therapy for orthopaedic conditions? (2) Do these tools demonstrate reliability and construct validity when used in a clinical setting?

Methods 

This was a cross-sectional study involving outpatients having physical therapy in tertiary-care settings. A total of 431 outpatients with neck (n = 93), shoulder (n = 108), low back (n = 119), or knee (n = 111) conditions were enrolled between December 2014 and December 2015, with 24% (103 of 431) seeking care after orthopaedic surgery. Participants completed three validated psychological questionnaires measuring negative mood, resulting in 39 candidate items for item response theory analysis. Factor analysis was used to identify the dimensions (factors) assessed by the candidate items and select items that loaded on the main factor of interest (negative mood), establishing a unidimensional item set. Unidimensionality of an item set suggests they are assessing one main factor or trait, allowing unbiased score estimates. The identified items were assessed for their fit to the graded item response theory model. This model allows for items to vary by the level of difficulty they represent and by their ability to discriminate between patients at different levels of the trait being assessed, in this case, negative mood. Finally, a hierarchical bifactor model where multiple subfactors are allowed to load on an overall factor was used to confirm that the items identified as representing a unidimensional item set explained the large majority of variance of the overall factor, providing additional support for essential unidimensionality. Using the final item bank, we constructed a computer adaptive test administration mode, and reduced item sets were selected to create short forms including items with the highest information (reliability) at targeted score levels of the trait being measured, while also considering clinical content.

Results 

We identified a 12-item bank for assessment of negative mood; eight-item and four-item short-form versions were developed to reduce administrative burden. Computer adaptive test administration used a mean ± SD of 8 ± 1 items. The item bank’s reliability (0 = no reliability; 1 = perfect reliability) was 0.89 for the computer adaptive test administration, 0.86 for the eight-item short form, and 0.71 for the four-item short form. Reliability values equal to or greater than 0.7 are considered acceptable for group level measures. Construct validity sufficient for clinical practice was supported by more severe negative mood scores among individuals with a previous episode of pain in the involved anatomical region, pain and activity limitations during the past 3 months, a work-related injury, education less than a college degree, and income less than or equal to USD 50,000.

Conclusion 

These newly derived tools include short-form and computer adaptive test options for reliable and valid negative mood assessment in outpatient orthopaedic populations. Future research should determine the responsiveness of these measures to change and establish score thresholds for clinical decision-making.

Clinical Relevance 

Orthopaedic providers can use these tools to inform prognosis, establish clinical benchmarks, and identify patients who may benefit from psychological and/or behavioral treatments.

中文翻译：

---

为骨科肌肉骨骼疼痛患者开发可靠有效的消极情绪筛查工具

背景 

消极情绪是肌肉骨骼疼痛患者临床结果不佳的一个重要危险因素。筛查负面情绪可以帮助识别那些可能需要额外心理干预的人。当前负面情绪筛查工具的局限性包括（1）高反应负担，（2）关注负面情绪的单一维度，（3）识别具有低或高负面情绪水平的个体的精度较差，和/或（4）设计不适用于患有骨科疾病和肌肉骨骼疼痛的人群。

问题/目的 

（1）项目反应理论方法能否用于构建骨科物理治疗患者负面情绪（如抑郁、焦虑、愤怒）的筛查工具？(2) 这些工具在临床环境中使用时是否表现出可靠性和结构有效性？

方法 

这是一项横断面研究，涉及在三级医疗机构接受物理治疗的门诊患者。2014 年 12 月至 2015 年 12 月期间，共有 431 名患有颈部（n = 93）、肩部（n = 108）、腰部（n = 119）或膝盖（n = 111）疾病的门诊患者入组，其中 24%（103第 431 页）在骨科手术后寻求护理。参与者完成了三份经过验证的心理问卷来测量负面情绪，从而产生了 39 个用于项目反应理论分析的候选项目。通过因子分析确定候选项目评估的维度（因子），并选择加载主要感兴趣因子（负面情绪）的项目，建立一维项目集。项目集的一维性表明他们正在评估一个主要因素或特征，从而允许无偏的分数估计。评估所确定的项目是否适合分级项目反应理论模型。该模型允许项目根据它们所代表的难度级别以及它们区分处于不同级别的被评估特征（在本例中为消极情绪）的患者的能力而变化。最后，使用允许多个子因子加载到整体因子上的分层双因子模型来确认被识别为代表一维项目集的项目解释了整体因子的大部分方差，为基本的一维性提供了额外的支持。使用最终的项目库，我们构建了计算机自适应测试管理模式，并选择减少的项目集来创建简短的表格，其中包括在被测量特征的目标分数水平上具有最高信息（可靠性）的项目，同时还考虑临床内容。

结果 

我们确定了一个包含 12 个项目的库来评估负面情绪；开发了八项和四项简式版本，以减轻行政负担。计算机自适应测试管理使用 8 ± 1 个项目的平均值 ± SD。计算机自适应测试管理的题库可靠性（0 = 无可靠性；1 = 完全可靠性）为 0.89，八题简式为 0.86，四题简式为 0.71。对于组级测量，等于或大于 0.7 的可靠性值被认为是可接受的。足以用于临床实践的构建效度得到了以下因素的支持：曾在相关解剖区域出现过疼痛、过去 3 个月内疼痛和活动受限、工伤、受教育程度低于大学学历的个体中更严重的负面情绪评分，且收入低于或等于 50,000 美元。

结论 

这些新衍生的工具包括简短的和计算机自适应测试选项，用于对门诊骨科人群进行可靠且有效的负面情绪评估。未来的研究应该确定这些措施对变化的响应能力，并建立临床决策的评分阈值。

临床相关性 

骨科医生可以使用这些工具来告知预后、建立临床基准并确定可能受益于心理和/或行为治疗的患者。