目的：基于美国国家健康和营养检查调查（NHANES）数据，评估整合炎症、肌肉及共病状态的残余易
感性指数（residual vulnerability index，RVI）评分与成年人功能障碍（functional impairment，FI）风险的关系，探
究两者关联的稳健性与临床获益。方法：选取NHANES 2017～2018周期中2 478例≥20岁的成年受试者开
展横断面研究。RVI评分由反映系统性炎症的超敏C反应蛋白（high-sensitivity C-reactive protein，hs-CRP）、
反映肌肉储备的肌肉指数（四肢瘦体重／体重指数）及共病数量经Z分数标准化后构建。FI判定基于身体功
能调查问卷。采用随机森林（RF）算法筛选关键预测特征，并构建多因素加权Logistic回归模型分析RVI评分
与FI风险的独立关联，计算比值比（OR）及95%置信区间（CI）。采用限制性立方样条（RCS）模型拟合剂量-
反应关系，并通过亚组分析、E值（E-value）敏感性分析及临床决策曲线分析（DCA）验证结果的稳健性与临床
价值。结果：功能障碍组的RVI评分显著高于对照组（P＜0.001）。RF分析显示，RVI评分的预测重要性位列
第三，优于抑郁评分及睡眠时长。多因素Logistic回归分析显示，在调整混杂因素后，RVI评分每增加1个标
准差（SD），FI发生风险显著增加[OR=1.912 (95% CI 1.548～2.361)，P＜0.001]。按四分位数分组分析显示，
最高分位组较最低分位组的FI风险显著升高[OR=6.204 (95% CI 3.013～12.777)，P=0.003]，且呈显著线性趋
势（P for trend＜0.001）。RCS 模型提示 RVI 评分与 FI 风险之间存在显著非线性正相关关系（非线性 P＜
0.001）。亚组分析显示，该关联在不同特征人群中总体保持一致，在超重／肥胖人群中更为显著（P=0.002）。
敏感性分析显示，E-value点估计值为2.115。DCA结果表明，纳入RVI评分的预测模型在不同阈值概率范围内
的临床净获益均优于基础模型。结论：RVI评分是成年人FI的独立非线性危险因素。
Based on data from the National Health and Nutrition Examination Survey (NHANES),
to evaluate the association between a residual vulnerability index (RVI) score—integrating inflammation, muscle
reserve, and comorbidity status—and the risk of functional impairment (FI) in adults, and to examine the
robustness of this association and its clinical utility. Methods: A cross-sectional study was conducted using 2,478
adult participants aged ≥20 years from the NHANES 2017～2018 cycle. The RVI score was constructed by
Z-score standardization of high-sensitivity C-reactive protein (hs-CRP, reflecting systemic inflammation), muscle
index (appendicular lean mass divided by body mass index, reflecting muscle reserve), and the number of
comorbidities. FI was defined based on the physical functioning questionnaire. The random forest (RF) algorithm
was used to screen key predictive features. A multivariable weighted logistic regression model was built to
analyze the independent association between the RVI score and FI risk, with odds ratios (ORs) and 95%
confidence intervals (CIs) calculated. Restricted cubic spline (RCS) models were fitted to model the
dose-response relationship. Subgroup analyses, E-value sensitivity analysis, and decision curve analysis (DCA)
were performed to verify the robustness and clinical value of the results. Results: The RVI score was
significantly higher in the FI group than in the non-FI group (P<0.001). RF analysis showed that the RVI score
ranked third in predictive importance, surpassing depression scores and sleep duration. Multivariable logistic
regression analysis revealed that, after adjusting for confounders, each 1-standard-deviation (SD) increase in the
RVI score was significantly associated with an increased risk of FI [OR=1.912(95% CI 1.548～2.361), P<0.001].
Quartile-based analysis showed that the highest quartile group had a significantly higher risk of FI compared with
the lowest quartile group [OR=6.204(95% CI 3.013～12.777), P=0.003], with a significant linear trend (P for
trend<0.001). The RCS model indicated a significant nonlinear positive association between the RVI score and FI
risk (nonlinearity P<0.001). Subgroup analyses demonstrated that this association remained generally consistent
across different population subgroups and was more pronounced among overweight/obese individuals (P=0.002
for interaction). Sensitivity analysis yielded an E-value point estimate of 2.115. DCA results showed that the prediction model
incorporating the RVI score provided superior clinical net benefit over the base model across a range of threshold probabilities.
Conclusion: The RVI score is an independent nonlinear risk factor for FI in adults.
PDF(1068 KB)
