Investigating the Psychometric Properties of the Suicide Stroop Task
Investigating the Psychometric Properties of the Suicide Stroop Task
复制标题
DOI：
10.1037/pas0000723
复制
发表时间：
2019-08-01
影响因子：
3.6
通讯作者：
Cha, Christine B.
中科院分区：
文献类型：
--
作者：
Wilson, Kelly M.;Millner, Alexander J.;Cha, Christine B.
行为测量越来越多地用于评估自杀想法和行为。一些措施，如自杀Stroop任务，在文献中产生了混合的结果。这些行为测量的一个未充分研究的特征是它们的心理测量特性，这可能会影响检测显著效果和再现性的概率。在同类最大的调查中，我们测试了自杀Stroop任务的内部一致性和并行有效性，目前的形式，从七个独立的研究（N = 875参与者，64%女性，年龄12至81岁）。结果表明，最常见的自杀Stroop评分方法，干扰分数，产生了不可接受的低内部一致性（rs = -.09-.13），并未能证明并行效度。内部一致性系数的平均反应时间（RT），以每种刺激类型的范围从rs = 0.93 - 0.94。所有自杀相关干扰的评分方法都显示出较差的分类准确性（AUC = 0.52 - 0.56），这表明评分在区分自杀未遂者和非自杀未遂者的能力方面接近偶然。在平均RT的情况下，尽管我们的可靠性结果非常好，但我们没有发现并行有效性的证据，强调可靠性并不能保证测量在临床上有用。这些结果进行了讨论的背景下，更广泛的影响测试和报告心理测量性能的行为措施，在心理健康研究。
Behavioral measures are increasingly used to assess suicidal thoughts and behaviors. Some measures, such as the Suicide Stroop Task, have yielded mixed findings in the literature. An understudied feature of these behavioral measures has been their psychometric properties, which may affect the probability of detecting significant effects and reproducibility. In the largest investigation of its kind, we tested the internal consistency and concurrent validity of the Suicide Stroop Task in its current form, drawing from seven separate studies (N = 875 participants, 64% female, aged 12 to 81 years). Results indicated that the most common Suicide Stroop scoring approach, interference scores, yielded unacceptably low internal consistency (rs = -.09-.13) and failed to demonstrate concurrent validity. Internal consistency coefficients for mean reaction times (RTs) to each stimulus type ranged from rs = .93-.94. All scoring approaches for suicide-related interference demonstrated poor classification accuracy (AUCs = .52-.56) indicating that scores performed near chance in their ability to classify suicide attempters from nonattempters. In the case of mean RTs, we did not find evidence for concurrent validity despite our excellent reliability findings, highlighting that reliability does not guarantee a measure is clinically useful. These results are discussed in the context of the wider implications for testing and reporting psychometric properties of behavioral measures in mental health research.