腕部加速度计时间相关性分析可预测养老院居民死亡与住院风险
本研究为ReMInDAR试验的子研究,利用213名养老院居民(中位年龄85岁)的腕部加速度计数据,评估活动信号时间相关性等指标与12个月内死亡、住院及认知下降的关联。结果显示,时间相关性越高(活动越不随机),死亡和住院风险越低,但与认知下降无关。包含时间相关性的预测模型对死亡和住院的AUC分别为0.70和0.64。
要点
- 腕部加速度计的时间相关性越高,养老院居民12个月内死亡和住院风险越低。
- 时间相关性对认知下降无显著预测价值(标准化OR=1,95%CI 0.74-1.35,p=0.98)。
- 包含时间相关性的模型对死亡预测AUC为0.70,对住院为0.64。
核心结论
- 腕部加速度计信号的时间相关性(经去趋势波动分析测量)与养老院居民12个月内死亡和住院风险独立相关,但与认知下降无关。
- 数字生物标志物可能作为客观、准确、低成本的患者指标,用于风险分层和临床规划。
研究设计与样本
- 本研究为随机对照试验(评估养老院持续药师服务效果)的二次分析,属于ReMInDAR试验子研究。
- 纳入213名居住在养老院的参与者,中位年龄85岁,具有可用的加速度计记录。
- 比较三种指标:衰弱指数、每日轻活动时间、以及通过去趋势波动分析测量的加速度计信号时间相关性。
- 采用逻辑回归评估基线加速度计指标与12个月内不良事件(死亡、认知下降、住院)的关联。
主要发现
- 时间相关性较高(活动更不随机)的个体12个月内死亡风险较低(标准化OR=0.49;95%CI 0.34-0.70;p<0.001)。
- 时间相关性较高的个体12个月内住院风险较低(标准化OR=0.57;95%CI 0.42-0.77;p<0.001)。
- 时间相关性与认知下降无显著关联(标准化OR=1;95%CI 0.74-1.35;p=0.98)。
- 包含时间相关性的预测模型对死亡的AUC为0.70(95%CI 0.60-0.80),对住院的AUC为0.64(95%CI 0.54-0.72)。
临床怎么用
- 在养老院老年综合评估中,可考虑使用腕部加速度计获取的时间相关性指标辅助识别死亡和住院高风险居民。
- 该指标不适用于预测认知下降,认知评估仍需采用专门的认知筛查工具。
- 数字生物标志物可作为客观、低成本的补充工具,支持风险分层和照护计划制定。
局限性
- 本研究为二次分析,原始试验目的为评估药师服务效果,非专门设计验证加速度计指标。
- 仅报告了死亡和住院的预测模型AUC,认知下降未报告AUC。
- 摘要未报告加速度计佩戴时长、数据采集具体参数及模型校准度等信息。
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: Objective measures for screening, prioritizing, and planning care for frail individuals are essential for appropriate aged care provision. This study evaluates metrics derived from actigraphy measures (captured by wrist accelerometer) as a digital biomarker to identify frail individuals at risk of adverse outcomes, including death, hospitalization, and cognitive decline. METHODS: This was a secondary study using data from a randomized controlled trial assessing the effectiveness of an ongoing pharmacist service in residential aged care facilities. Three metrics are studied and compared: the Frailty Index, the daily time spent in light time activity, and the temporal correlation of the actigraphy signal, measured by detrended fluctuation analysis. The association between actigraphy-derived metrics at baseline and adverse events within 12 months (death, cognitive decline, and hospitalizations) was assessed using logistic regression. RESULTS: Actigraphy records were available for 213 participants living in aged-care, median age of 85 years. Individuals with higher temporal correlation (activity is less random) were at lower risk of death (Standardized OR: 0.49; 95% CI 0.34, 0.7, p < 0.001) and hospitalization (Standardized OR: 0.57; 95% CI 0.42, 0.77, p < 0.001) in 12 months, but there was no difference in cognitive decline (Standardized OR: 1; 95% CI 0.74, 1.35, p = 0.98). The predictive model that included temporal correlation had an area under the curve of 0.70 (CI 0.60-0.80) for death and 0.64 (CI 0.54-0.72) for hospitalization. CONCLUSION: Temporal correlation of the actigraphy signal from aged care residents was strongly associated with death and hospitalization, but not cognitive decline. Digital biomarkers may have a place as an objective, accurate, and low-cost patient metric to support risk stratification and clinical planning.
引用
- Andrade AQ, Lim R, Kelly TL, et al. Wrist accelerometer temporal analysis as a prognostic tool for aged care residents: A sub-study of the ReMInDAR trial. J Am Geriatr Soc. 2022. PMID: 36524585. doi:10.1111/jgs.18181
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。