临床预测模型:新预测因子的预测增量值

Clinical prediction models:incremental value of new predictors

目的 新的生物标志物的发现和开发一直以来都是充满前景的领域,这些研究为预测模型提供了新的预测因子,可以极大改善对疾病风险的预测.为了对新的预测因子的临床效用得到有意义的结论,必须采取适当的统计方法.本文介绍了一些传统的和新颖的指标,包括模型拟合,预测因子效应量,模型区分度改进,以及基于风险重分类表来计算的统计量,并且结合案例说明如何对这些统计量进行计算.最后,我们提出了一些对于新预测因子增量值类研究的注意事项,供读者参考.

The discovery and development of new biomarkers is promising and will provide new predictors to clinical prediction models and improve the disease risk prediction.In order to get meaningful conclusions about the clinical usefulness of new predictors,appropriate statistical measures are needed.In this article,we introduce several traditional and novel measures,including model fitting,effect size,improvement in discrimination,and statistics based on reclassification tables.We also provide example calculation of these measures.At last,we emphasize some considerations in performing incremental value analysis for new risk predictors.