Department of Epidemiology and Biostatistics, School of Public Health, Peking University, Beijing, China; Peking University Medical Informatics Center, Beijing, China.
Peking University Medical Informatics Center, Beijing, China.
J Affect Disord. 2024 Feb 1;346:167-173. doi: 10.1016/j.jad.2023.11.014. Epub 2023 Nov 8.
The optimal multimorbidity measures for predicting disability trajectories are not universally agreed upon. We developed a multimorbidity index among middle-aged and older community-dwelling Chinese adults and compare its predictive ability of disability trajectories with other multimorbidity measures.
This study included 17,649 participants aged ≥50 years from the China Health and Retirement Longitudinal Survey 2011-2018. Two disability trajectory groups were estimated using the total disability score differences calculated between each follow-up visit and baseline. A weighted index was constructed using logistic regression models for disability trajectories based on the training set (70 %). The index and the condition count were used, along with the pattern identified by the latent class analysis to measure multimorbidity at baseline. Logistic regression models were used in the training set to examine associations between each multimorbidity measure and disability trajectories. C-statistics, integrated discrimination improvements, and net reclassification indices were applied to compare the performance of different multimorbidity measures in predicting disability trajectories in the testing set (30 %).
In the newly developed multimorbidity index, the weights of the chronic conditions varied from 1.04 to 2.55. The multimorbidity index had a higher predictive performance than the condition count. The condition count performed better than the multimorbidity pattern in predicting disability trajectories.
Self-reported chronic conditions.
The multimorbidity index may be considered an ideal measurement in predicting disability trajectories among middle-aged and older community-dwelling Chinese adults. The condition count is also suggested due to its simplicity and superior predictive performance.
预测残疾轨迹的最佳多重疾病衡量标准尚未达成共识。我们开发了一种适用于中年和老年社区居民的中国成年人的多重疾病指数，并将其预测残疾轨迹的能力与其他多重疾病衡量标准进行了比较。
这项研究包括了来自中国健康与退休纵向研究 2011-2018 年的 17649 名年龄在 50 岁及以上的参与者。通过在每次随访和基线之间计算总残疾评分差异来估计两种残疾轨迹组。使用基于训练集（70%）的残疾轨迹的逻辑回归模型构建加权指数。在训练集中，使用指数和疾病计数以及潜在类别分析确定的模式来衡量基线时的多重疾病。使用逻辑回归模型在训练集中检验每个多重疾病衡量标准与残疾轨迹之间的关联。C 统计量、综合鉴别改善和净重新分类指数被用于比较不同多重疾病衡量标准在预测测试集（30%）中残疾轨迹的性能。
在新开发的多重疾病指数中，慢性病的权重从 1.04 到 2.55 不等。该多重疾病指数的预测性能高于疾病计数。疾病计数在预测残疾轨迹方面优于多重疾病模式。
自我报告的慢性疾病。
多重疾病指数可被认为是预测中年和老年社区居民中国成年人残疾轨迹的理想衡量标准。由于其简单性和优越的预测性能，也建议使用疾病计数。
