Herrmann N, Cayten C G, Senior J, Staroscik R, Walsh S, Woll M
Health Serv Res. 1980 Summer;15(2):127-43.
The collection of data by abstraction from patient records is a widely used method of research, evaluation, and registry. Since valid conclusions depend on the accuracy of the abstracted data, it is essential to examine the abstracting procedures. In this paper, we report on a study of patient data abstracted from emergency department records by nurses trained by project personnel. Twenty-five charts were selected at each of five hospitals. To test interobserver reliability, the nurses were asked to abstract all of the charts at each hospital; to test intraobserver reliability, four of the nurses each reabstracted a set of charts. The results show that even with highly trained, well motivated abstractors, there are considerable differences in the accuracy with which the variables are abstracted. Disposition from the hospital, quantitative vital signs, and blood gas values tend to be abstracted with higher reliability; whereas variables requiring judgment, such as character of vital signs or history of disease, tend to have low reliability. To improve the quality of abstracted data, we propose improved retrieval methods for hospital records, monitoring of data collection procedures, cooperation of all medical personnel providing the raw data, and careful selection of variables.
通过从患者记录中提取数据进行收集是一种广泛用于研究、评估和登记的方法。由于有效的结论取决于提取数据的准确性，因此检查提取程序至关重要。在本文中，我们报告了一项关于由项目人员培训的护士从急诊科记录中提取的患者数据的研究。在五家医院中，每家医院选取了25份病历。为了测试观察者间信度，要求护士提取每家医院的所有病历；为了测试观察者内信度，四名护士每人重新提取了一组病历。结果表明，即使是训练有素、积极性高的提取人员，在提取变量的准确性方面也存在相当大的差异。出院情况、定量生命体征和血气值的提取可靠性往往较高；而需要判断的变量，如生命体征特征或疾病史，可靠性往往较低。为了提高提取数据的质量，我们建议改进医院记录的检索方法、监测数据收集程序、所有提供原始数据的医务人员进行合作以及仔细选择变量。
