Lee Minji K, Schalet Benjamin D, Cella David, Yost Kathleen J, Dueck Amylou C, Novotny Paul J, Sloan Jeff A
Robert D. and Patricia E. Kern Center for the Science of Health Care Delivery, Mayo Clinic, 200 First St SW, Rochester, MN, 55906, USA.
Department of Medical Social Sciences, Northwestern University Feinberg School of Medicine, 625 Michigan Ave, 27th Floor, Chical, IL, 60611, USA.
J Patient Rep Outcomes. 2020 Dec 10;4(1):106. doi: 10.1186/s41687-020-00271-0.
Researchers and clinicians studying symptoms experienced by people with cancer must choose from various scales. It would be useful to know how the scores on one measure translate to another.
Using item response theory (IRT) with the single-group design, in which the same sample answers all measures, we produced crosswalk tables linking five 0-10 numeric rating scale (NRS) and 15 items from Patient-Reported Outcomes version of the Common Terminology Criteria for Adverse Events (PRO-CTCAE, scored on a 1-5 scale) to the T-Score metric of six different scales from the NIH Patient reported Outcomes Measurement Information System (PROMIS®). The constructs, for which we conducted linking, include emotional distress-anxiety, emotional distress-depression, fatigue, sleep disturbance, pain intensity, and pain interference. We tested the IRT linking assumption of construct similarity between measures by comparing item content and testing unidimensionality of item sets comprising each construct. We also investigated the correlation of the measures to be linked and, by inspecting standardized mean differences, whether the linkage is invariant across age and gender subgroups. For measures that satisfied the assumptions, we conducted linking.
In general, an NRS score of 0 corresponded to about 38.2 on the PROMIS T-Score scale (mean = 50; SD = 10); whereas an NRS score of 10 corresponded to a PROMIS T-Score of approximately 72.7. Similarly, the lowest/best score of 1 on PRO-CTCAE corresponded to 39.8 on T-score scale and the highest/worst score of 5 corresponded to 72.0.
We produced robust linking between single item symptom measures and PROMIS short forms.
研究癌症患者症状的研究人员和临床医生必须从各种量表中进行选择。了解一种测量方法的分数如何转换为另一种测量方法会很有用。
使用单组设计的项目反应理论（IRT），即同一样本回答所有测量方法，我们生成了交叉表，将五个0至10的数字评定量表（NRS）和患者报告的不良事件通用术语标准（PRO-CTCAE，评分范围为1至5）中的15个项目与美国国立卫生研究院患者报告结果测量信息系统（PROMIS®）的六种不同量表的T分数指标相联系。我们进行联系的构念包括情绪困扰-焦虑、情绪困扰-抑郁、疲劳、睡眠障碍、疼痛强度和疼痛干扰。我们通过比较项目内容并测试构成每个构念的项目集的单维性，来检验IRT关于测量方法之间构念相似性的联系假设。我们还研究了要联系的测量方法之间的相关性，并通过检查标准化平均差异，探究这种联系在年龄和性别亚组中是否不变。对于满足假设的测量方法，我们进行了联系。
一般来说，NRS评分为0对应于PROMIS T分数量表上约38.2分（均值 = 50；标准差 = 10）；而NRS评分为10对应于PROMIS T分数约72.7分。同样，PRO-CTCAE的最低/最佳评分为1对应于T分数量表上的39.8分，最高/最差评分为5对应于72.0分。
我们在单项症状测量方法和PROMIS简表之间建立了可靠的联系。
