Williams H C, Burney P G, Strachan D, Hay R J
St John's Institute of Dermatology, London, U.K.
Br J Dermatol. 1994 Sep;131(3):397-405. doi: 10.1111/j.1365-2133.1994.tb08531.x.
The level of agreement between 14 physicians as to what constitutes a case of atopic dermatitis was tested on 15 selected patients with a range of diagnoses. Between-observer agreement was good, with a mean pair agreement index (P0) of 0.94, and a chance corrected index (kappa) of 0.78. Between-observer agreement in the recording of 18 separate physical signs of atopic dermatitis was then tested by asking the 14 physicians to note the presence or absence of each sign in a different group of patients to those seen in the first part of the exercise. Substantial between-observer agreement (kappa > 0.61) was only present for truncal dermatitis. Most signs showed only fair to moderate agreement (kappa 0.21-0.60), and some signs, such as keratosis pilaris, xerosis, orbital pigmentation, fine hair, and extensor dermatitis, showed poor agreement (kappa 0.01-0.20). The findings were similar when the responses of two independent observers from the national study outlined in Paper I were compared for each sign. Within-observer variation for the recording of physical signs was substantially better than between-observer variation. Physicians interested in atopic dermatitis agree reasonably well on what constitutes a typical case of atopic dermatitis. Between-observer variation with regard to some physical signs of atopic dermatitis is of a magnitude which argues against their continued use in clinical and epidemiological studies.
在15名患有一系列诊断疾病的选定患者身上，测试了14名医生对于什么构成特应性皮炎病例的一致程度。观察者间的一致性良好，平均配对一致性指数（P0）为0.94，机遇校正指数（kappa）为0.78。然后，通过要求这14名医生在与第一部分研究中所观察患者不同的另一组患者中记录特应性皮炎的18种不同体征的存在或不存在，来测试观察者间在记录这些体征方面的一致性。观察者间的实质性一致性（kappa>0.61）仅在躯干性皮炎中存在。大多数体征仅显示出一般到中等程度的一致性（kappa 0.21 - 0.60），而一些体征，如毛发角化病、皮肤干燥、眶周色素沉着、细毛和伸侧皮炎，显示出较差的一致性（kappa 0.01 - 0.20）。当比较论文I中概述的全国性研究中两名独立观察者对每种体征的反应时，结果相似。体征记录的观察者内变异明显优于观察者间变异。对特应性皮炎感兴趣的医生在什么构成典型的特应性皮炎病例上意见相当一致。特应性皮炎某些体征的观察者间变异程度表明，这些体征不宜继续用于临床和流行病学研究。
