Harbst K B, Ottenbacher K J, Harris S R
Department of Physical Therapy, University of Wisconsin-La Crosse 54601.
Phys Ther. 1991 Feb;71(2):107-15. doi: 10.1093/ptj/71.2.107.
Increased emphasis on the use of single-subject designs in physical therapy research suggests the need to examine whether therapists can meaningfully interpret the results of such research as part of the clinical decision-making process. With this goal in mind, the interrater reliability of therapists to make visual judgments from graphed data that included a trend line was examined. Thirty therapists were presented with 24 graphs of single-subject data from AB (baseline-treatment) designs. Each graph included a trend line calculated using the split-middle method of trend estimation. The trend line was computed using the baseline data and then extended into the treatment phase to "predict" patient performance. The analysis, using intraclass correlation coefficients (ICCs), revealed low interrater agreement, with ICC values ranging from .37 to .55 for the entire sample. Evidence is presented that the statistical backgrounds of some raters positively influenced interrater reliability. No statistically significant relationship was found between interrater agreement and visual components of the graphed data, such as changes in slope or variability.
物理治疗研究中对单病例设计使用的日益重视表明，有必要审视治疗师能否将此类研究结果作为临床决策过程的一部分进行有意义的解读。出于这一目的，研究人员考察了治疗师根据包含趋势线的图表数据做出视觉判断时的评分者间信度。向30名治疗师展示了24张来自AB（基线-治疗）设计的单病例数据图表。每张图表都包含一条使用趋势估计的中分法计算得出的趋势线。趋势线是根据基线数据计算得出的，然后延伸至治疗阶段以“预测”患者的表现。使用组内相关系数（ICC）进行的分析显示，评分者间一致性较低，整个样本的ICC值在0.37至0.55之间。有证据表明，一些评分者的统计学背景对评分者间信度产生了积极影响。在评分者间一致性与图表数据的视觉成分（如斜率变化或变异性）之间未发现具有统计学意义的关系。
