加权多病共存指数与中国中老年人新发衰弱风险的关联——基于CHARLS的纵向研究

Association of Weighted Multimorbidity Index with New-Onset Frailty Risk in Middle-Aged and Older Adults—A Longitudinal Study Based on CHARLS

摘要: 背景：多病共存是中老年人群常见的健康问题。传统研究多采用简单疾病计数评估多病负担，但该方法默认不同慢性疾病对健康结局的影响相同，可能忽视疾病组成对衰弱风险的异质性。本研究基于中国健康与养老追踪调查(China Health and Retirement Longitudinal Study, CHARLS)，评估加权多病共存指数(weighted multimorbidity index, WMI)与中国中老年人新发衰弱风险之间的纵向关联，并比较其与简单疾病计数的预测表现。方法：本研究使用CHARLS 2011~2020年随访数据，纳入基线年龄 ≥ 50岁、基线无衰弱且具有完整暴露、协变量及随访结局信息的6659名参与者。多病共存基于8种慢性疾病定义，包括高血压、糖尿病、癌症、慢性肺病、心脏病、卒中、关节炎以及精神/情绪疾病。8种疾病同时进入Cox比例风险模型，并通过重复5折交叉拟合估计β系数以构建WMI；简单疾病计数作为传统比较指标。主要结局为随访期间新发衰弱，采用基于累积缺陷模型构建的衰弱指数定义，FI ≥ 0.25判定为衰弱。采用Kaplan-Meier曲线、限制性立方样条和Cox比例风险模型评估WMI与新发衰弱的关联，并通过C-index和赤池信息准则(AIC)比较模型区分能力和拟合表现。结果：最终纳入的6659名参与者平均年龄为60.8岁，男性占56.7%，基线FI均值为0.14。平均随访5.6年期间，共2841名参与者发生新发衰弱，发生率为42.7%。随着WMI分位组升高，参与者年龄、慢性病负担、基线FI及新发衰弱比例总体升高；新发衰弱发生率由Q1组的33.0%升至Q4组的57.4%。Kaplan-Meier曲线显示，高WMI组无衰弱生存概率更低(log-rank P < 0.001)。限制性立方样条显示WMI与新发衰弱风险之间存在非线性剂量–反应关系(总体关联P < 0.001，非线性检验P < 0.001)。在完全调整模型中，WMI每增加1个标准差，新发衰弱风险增加43% (HR = 1.43, 95% CI: 1.38~1.48, P < 0.001)。简单疾病计数同样与新发衰弱风险升高相关(HR = 1.39, 95% CI: 1.34~1.45, P < 0.001)，但其模型区分能力和拟合表现略弱于WMI。结论：在CHARLS中国中老年人群中，较高的加权多病共存负担与更高的新发衰弱风险显著相关。与简单疾病计数相比，WMI的模型区分能力和拟合表现略优，提示疾病组成可能在疾病数量之外提供额外的风险信息，但其临床增量价值仍需进一步验证。

Abstract: Background: Multimorbidity is a common health problem among middle-aged and older adults. Traditional studies often use simple disease counts to assess the burden of multiple diseases, but this method assumes that different chronic diseases have the same impact on health outcomes, potentially neglecting the heterogeneity of disease composition in relation to frailty risk. This study, based on the China Health and Retirement Longitudinal Study (CHARLS), assesses the longitudinal association between the weighted multimorbidity index (WMI) and the risk of new-onset frailty in middle-aged and older Chinese adults, and compares its predictive performance with that of the simple disease count. Methods: This study used CHARLS follow-up data from 2011 to 2020, including 6659 participants with a baseline age ≥ 50 years, no baseline frailty, and complete exposure, covariate, and follow-up outcome information. Multimorbidity was defined based on eight chronic diseases, including hypertension, diabetes, cancer, chronic lung disease, heart disease, stroke, arthritis, and mental/mood disorders. All eight diseases were simultaneously included in a Cox proportional hazards model, and the beta coefficient was estimated using repeated 5-fold cross-fitting to construct the WMI; simple disease count was used as a conventional comparative indicator. The primary outcome was new-onset frailty during follow-up, defined using a frailty index based on a cumulative deficiency model, with an FI ≥ 0.25 indicating frailty. The association between WMI and new-onset frailty was assessed using Kaplan-Meier curves, restricted cubic splines, and Cox proportional hazards models. The model discrimination ability and fit performance were compared using the C-index and the Akaike Information Criterion (AIC). Results: The final 6659 participants had a mean age of 60.8 years, 56.7% were male, and the mean baseline FI was 0.14. During the mean follow-up of 5.6 years, 2841 participants experienced new-onset frailty, with an incidence rate of 42.7%. As the WMI quantile increased, participants’ age, chronic disease burden, baseline fibrillation index (FI), and the overall proportion of new-onset frailty increased; the incidence of new-onset frailty rose from 33.0% in the Q1 group to 57.4% in the Q4 group. Kaplan-Meier curves showed that the probability of frailty-free survival was lower in the high WMI group (log-rank P < 0.001). Restricted cubic splines showed a non-linear dose-response relationship between WMI and the risk of new-onset frailty (overall association P < 0.001, non-linearity test P < 0.001). In the fully adjusted model, for every 1 standard deviation increase in WMI, the risk of new-onset frailty increased by 43% (HR = 1.43, 95% CI: 1.38~1.48, P < 0.001). Simple disease count was also associated with an increased risk of new-onset frailty (HR = 1.39, 95% CI: 1.34~1.45, P < 0.001), but its model discrimination and fit were slightly weaker than WMI. Conclusion: Among middle-aged and older Chinese adults in CHARLS, a higher weighted multimorbidity burden was significantly associated with a higher risk of new-onset frailty. Compared with simple disease counts, the WMI model showed slightly better discriminative ability and fit, suggesting that disease composition may provide additional risk information beyond the number of diseases, but its incremental clinical value still needs further validation.