To examine the developmental trajectory and associated factors of frailty in Chinese older people using the data of four national waves of China Health and Retirement Longitudinal Study (CHARLS) .

The data of this study obtained from four national waves〔2011 (the baseline survey), and 2013, 2015 and 2018 (follow-up surveys) 〕 of CHARLS, which was initially conducted in 2011, and was followed by tracking once every 2 to 3 years with multi-stage PPS sampling for middle-aged and elderly groups in 28 provincial administrative regions of China, covering 150 counties and 450 villages. The surveyees were coded, and matched, then 2 267 cases (≥60 years old) involved in the four waves of surveys were selected as the sample. Frailty was assessed by the frailty index (FI). Mplus was used to construct three types of unconditional latent growth models, and the optimal fitting model was selected to determine the developmental trajectory of frailty of Chinese older people, and was used to develop the conditional latent growth model. The effects of time-invariant factors (gender, education level) and time-varying factors (physical activity, smoking, alcohol consumption, sleep) on frailty were examined.

The latent growth model with undefined curve fit the data better, and was selected as the optimal model to determine the frailty development trajectory. The results of χ² (3) =36.16, CFI=0.992, TLI=0.984, RMSEA=0.070, SRMR=0.022, indicating that the frailty prevalence in older adults showed a trend of curvilinear increase. The values of intercept (initial level), slope (growth), and the variation of them of the model were significantly higher than 0 (P<0.01), indicating that there were significant individual differences in the initial level and growth rate of frailty. Gender and education level were negatively associated with the initial level of frailty (β=-0.113, -0.173, P<0.01). They were also negatively associated with the growth of frailty (β=-0.181, -0.151, P<0.01). Compared with men, women had higher initial level and faster growth rate of frailty (P<0.05). Compared to those with higher education level, those with lower education level had higher initial level and faster growth rate of frailty (P<0.05). Physical activity and sleep were negatively associated with frailty in all waves of surveys (P<0.05). Smoking was positively associated with frailty in 2011, 2015, 2018 waves of surveys (P<0.05). Alcohol consumption was positively associated with frailty in 2013 and 2015 waves of surveys (P<0.05) .

The frailty in Chinese older people showed a trajectory of curvilinear increase, and its initial level and growth rate had significant individual differences. Comparatively speaking, being female and having lower education level were associated with increased risk of having frailty. Moderate- and high-level physical activity and adequate sleep were associated with decreased risk of having frailty or alleviating frailty. Long-term smoking and drinking too much could exacerbate frailty.

基于中国健康与养老追踪调查（CHARLS）4期追踪数据，考察我国老年人衰弱的发展轨迹及影响因素。

本研究资料来源于CHARLS 2011年、2013年、2015年、2018年数据，该调查于2011年开展，随后每2~3年追踪1次，采用多阶段PPS抽样，对我国28个省级行政区的中老年人群进行家户调查，范围覆盖150个县级单位、450个村级单位。其中2011年为全国基线调查，2013年、2015年和2018年为全国追踪调查。将4期数据通过个人编码匹配合并，最终形成4期调查均参与的2 267例60岁及以上老年人作为研究样本。采用衰弱指数（FI）对老年人的衰弱状况进行评估。利用Mplus工具构建3类无条件潜变量增长模型，选取最优的拟合模型确定我国老年人衰弱的发展轨迹，并在最优模型的基础上构建条件潜变量增长模型，考察随时间恒定因素（性别、教育程度）和随时间变化因素（体力活动、吸烟、饮酒、睡眠）对老年人衰弱的影响。

不定义曲线潜变量增长模型能更好地拟合数据，为我国老年人衰弱发展轨迹的最优模型，其中χ²（3）=36.16，比较拟合指数（CFI）=0.992，非规范拟合指数（TLI）=0.984，近似均方根误差（RMSEA）=0.070，标准化均方根残差（SRMR）=0.022，表明我国老年人衰弱水平呈曲线增长的趋势；模型截距、斜率以及截距的变异和斜率的变异均显著>0（P<0.01），表明老年人衰弱的初始水平和增长速度均存在显著的个体差异。恒定因素性别、教育程度对模型截距（性别：β=-0.113，P<0.01；教育程度：β=-0.173，P<0.01）和斜率（性别：β=-0.181，P<0.01；教育程度：β=-0.151，P<0.01）均有显著的负向预测作用，相比于男性和教育程度高的老年人，女性、低教育程度老年人衰弱的初始水平更高，增长速度更快。时间变化因素中体力活动、睡眠在4期调查中均对老年人衰弱具有显著的负向影响（P<0.05）；吸烟和饮酒分别在2011年、2015年、2018年和2013年、2015年调查中对老年人的衰弱有显著的正向影响（P<0.05）。

我国老年人的衰弱呈曲线增长的发展轨迹，初始水平和增长速度均存在显著的个体差异，女性、低教育程度能预测衰弱的发展，中高体力活动、睡眠充足有助于衰弱水平的降低，长期吸烟、饮酒过多则会使衰弱恶化。