Cowen J S, Kelley M A
Division of Pulmonary and Critical Care Medicine, University of Pennsylvania Medical Center, Philadelphia.
Crit Care Clin. 1994 Jan;10(1):53-72.
Scoring systems used to predict clinical outcomes for critically ill patients have been refined in the past decade, yet even the most recently developed systems contain flaws that limit their application. In general, prediction rules are derived by defining an association between a number of clinical variables and a particular outcome in a reference patient population. By systematically examining the qualities of the independent variables and the size and scope of the derivation data set, potential sources of error and bias can be identified. Existing and future predictive systems must be validated on large groups of patients and continuously updated to keep pace with new approaches to the practice of critical care medicine.
在过去十年中，用于预测重症患者临床结局的评分系统已得到完善，但即使是最新开发的系统也存在一些缺陷，限制了它们的应用。一般来说，预测规则是通过定义一组临床变量与参考患者群体中特定结局之间的关联来推导的。通过系统地检查自变量的质量以及推导数据集的大小和范围，可以识别潜在的误差和偏差来源。现有的和未来的预测系统必须在大量患者群体上进行验证，并不断更新，以跟上危重症医学实践的新方法。
