Schousboe John T, Langsetmo Lisa, Taylor Brent C, Ensrud Kristine E
Park Nicollet Clinic and HealthPartners Institute, HealthPartners, Minneapolis, MN, USA; Division of Health Policy and Management, University of Minnesota, Minneapolis, MN, USA.
Division of Epidemiology and Community Health, University of Minnesota, Minneapolis, MN, USA.
J Clin Densitom. 2017 Jul-Sep;20(3):280-290. doi: 10.1016/j.jocd.2017.06.012. Epub 2017 Jul 13.
Fractures are binary events (they either occur or they do not), and predicting whether fractures may occur involves assigning probabilities of one or more of those events occurring over time to populations and to individuals. Fracture risk prediction has become central to the management of osteoporosis and fracture prevention in clinical practice, and the ultimate clinical usefulness of the prediction tools used to estimate these risks depends, at a minimum, on the validity and accuracy of those tools. In this paper, we will describe how fracture prediction models are developed and validated, and how their performance characteristics are assessed. We will provide a checklist by which clinicians, clinical researchers, and reviewers of journal submissions can judge whether a fracture prediction tool meets basic requirements of good performance. We will further describe how the incremental predictive value of additional diagnostic tools, such as bone mass measurement technologies, is assessed.
骨折是二元事件（要么发生，要么不发生），预测骨折是否可能发生涉及为群体和个体确定这些事件中一个或多个在一段时间内发生的概率。骨折风险预测已成为临床实践中骨质疏松症管理和骨折预防的核心，用于估计这些风险的预测工具的最终临床实用性至少取决于这些工具的有效性和准确性。在本文中，我们将描述骨折预测模型是如何开发和验证的，以及如何评估其性能特征。我们将提供一份清单，临床医生、临床研究人员和期刊投稿评审人员可据此判断骨折预测工具是否满足良好性能的基本要求。我们还将进一步描述如何评估其他诊断工具（如骨量测量技术）的增量预测价值。
