基于纵向衰弱指数的中国老年人死亡风险动态统计预测模型
本研究利用中国老年健康影响因素跟踪调查(CLHLS)数据,构建并验证了基于纵向重复测量的衰弱指数(FI)的动态死亡风险预测模型。训练集和三个验证集分别纳入9748、7459、9093和6368人,模型AUC范围为0.64至0.84,校准曲线显示良好预测能力。
要点
- 基于35项健康缺陷构建衰弱指数,联合模型纳入纵向FI及年龄、居住地、性别等协变量。
- 动态模型可随FI更新而更新死亡风险预测,AUC为0.64至0.84。
- 模型适用于65岁以上中国老年人死亡风险评估。
核心结论
- 开发的动态预测模型能够随衰弱指数测量更新而更新死亡风险预测,可用于评估65岁以上个体的死亡风险。
研究设计与样本
- 前瞻性队列研究,使用CLHLS数据,训练集9748人,三个验证集分别为7459、9093和6368人。
- 使用35项健康缺陷构建衰弱指数,基于每轮CLHLS的重复测量构建纵向FI。
- 采用联合模型构建动态预测模型,同时考虑基线协变量和纵向FI。
- 使用时间依赖ROC曲线下面积(AUC)和校准曲线评估预测性能。
主要发现
- 线性混合效应模型使用时间、性别、居住地(城市、城镇或农村)、独居、吸烟和饮酒计算个体特异性纵向FI。
- 动态预测模型基于纵向FI、年龄、居住地、性别及FI-年龄交互项构建。
- AUC范围为0.64至0.84,AUC和校准曲线均显示良好预测能力。
关键数值与推荐
- 训练集和验证集1至3分别纳入9748、7459、9093和6368人。
- 使用35项健康缺陷构建衰弱指数。
- AUC范围为0.64至0.84。
临床怎么用
- 可利用纵向衰弱指数动态更新65岁以上老年人的死亡风险预测,辅助临床风险评估。
以上解读依据论文摘要(全文未开放获取或许可不允许翻译转载)。
原文
BACKGROUND: Frailty is a common characteristic of older people with the ageing process. We aimed to develop and validate a dynamic statistical prediction model to calculate the risk of death in people aged ≥65 years, using a longitudinal frailty index (FI). METHODS: One training dataset and three validation datasets from the Chinese Longitudinal Healthy Longevity Survey (CLHLS) were used in our study. The training dataset and validation datasets 1 to 3 included data from 9,748, 7,459, 9,093 and 6,368 individuals, respectively. We used 35 health deficits to construct the FI and a longitudinal FI based on repeated measurement of FI at every wave of the CLHLS. A joint model was used to build a dynamic prediction model considering both baseline covariates and the longitudinal FI. Areas under time-dependent receiver operating characteristic curves (AUCs) and calibration curves were employed to assess the predictive performance of the model. RESULTS: A linear mixed-effects model used time, sex, residence (city, town, or rural), living alone, smoking and alcohol consumption to calculate a subject-specific longitudinal FI. The dynamic prediction model was built using the longitudinal FI, age, residence, sex and an FI-age interaction term. The AUCs ranged from 0.64 to 0.84, and both the AUCs and the calibration curves showed good predictive ability. CONCLUSIONS: We developed a dynamic prediction model that was able to update predictions of the risk of death as updated measurements of FI became available. This model could be used to estimate the risk of death in individuals aged >65 years.
引用
- Chen Q, Tang B, Zhai Y, et al. Dynamic statistical model for predicting the risk of death among older Chinese people, using longitudinal repeated measures of the frailty index: a prospective cohort study. Age Ageing. 2020. PMID: 32365173. doi:10.1093/ageing/afaa056
供执业医师工作参考,不替代临床判断;数值与分级以指南原文为准。