With the deepening of population aging, the rising prevalence of multimorbidity and disability poses a significant threat to the quality of life among middle-aged and older adults. Existing research, both domestic and international, rarely explores influencing factors of disability specifically in older adults with multimorbidity and seldom analyzes the impact of variable interactions on disability.

This study aimed to investigate factors influencing disability status in older adults with multimorbidity, reveal the contribution of various variables and their intrinsic relationships to disability, and provide empirical support for developing targeted prevention and control strategies.

From November 2023 to December 2024, based on the health ecological model theory and data from the China Health and Retirement Longitudinal Study (CHARLS 2020) , an indicator system of influencing factors was constructed. Kruskal-Wallis tests, multinomial Logistic regression, and Decision tree models were employed to analyze the impact of variables on disability status.

The study included 9 124 middle-aged and elderly individuals with chronic disease comorbidity. Among them, 5 355 (58.7%) had no impairment in physical functioning, 3 442 (37.7%) had mild impairment, and 327 (3.6%) had moderate or greater impairment. Multinomial Logistic regression analysis showed age <80 years, male, primary education, mild/moderate/severe depression, self-rated health as very poor/poor/fair, visual impairment, no history of falls, no history of hip fractures, fewer sites of bodily pain, current or former smoker, alcohol consumption, no exercise or low exercise intensity, minimal night-time sleep, and low/moderate personal annual income are factors associated with preserved physical functioning in older adults with chronic disease comorbidity (P<0.05) ; Self-rated health as very poor, abnormal vision, no history of falls, no history of hip fractures, lack of exercise, and non-participation in social activities were factors associated with moderate or greater impairment of physical functioning in older adults with chronic disease comorbidity (P<0.05) . The decision tree model identified self-rated health as the primary influencing factor for disability status, followed by depression, number of pain sites, physical activity level, age, and history of falls (P<0.05) . Model comparison showed that the area under the curve (AUC) for multinomial Logistic regression was higher than that for the Decision tree model across all disability levels: no impairment [0.808 (95%CI=0.800-0.817) vs. 0.768 (95%CI=0.759-0.778) ], mild impairment [0.773 (95%CI=0.764-0.783) vs. 0.734 (95%CI=0.724-0.745) ], and moderate-to-severe impairment [0.891 (95%CI=0.875-0.907) vs. 0.833 (95%CI=0.812-0.854) ].

Age, gender, education level, depression, self-rated health, vision, fall history, hip fracture history, number of bodily pain sites, smoking status, alcohol consumption, physical activity level, nighttime sleep duration, social participation, and personal annual income significantly influence disability status in older adults with multimorbidity. Relevant authorities should implement effective preventive measures across these dimensions to reduce the disability rate. While the logistic regression model demonstrates slightly better predictive performance, both models offer distinct advantages: logistic regression clearly demonstrates the dependence between independent and dependent variables, while the decision tree elucidates the impact of variable interactions. Both models can be utilized for supplementary analysis in future research.

随着人口老龄化进程加深，慢性病共病率与失能率上升趋势明显，威胁着中老年群体的生命质量，国内外研究较少测评慢性病共病中老年人失能状况的影响因素且鲜少分析变量交互关系对因变量的影响。

探讨慢性病共病中老年人失能状况的影响因素，揭示各变量及其内在联系对失能的贡献程度，为制定针对性预防和控制策略提供实证支持。

2023年11月—2024年12月，基于健康生态学模型理论与中国健康与养老追踪调查2020年数据（CHARLS 2020），构建慢性病共病中老年人失能状况影响因素指标体系并运用Kruskal-Wallis检验、多元Logistic回归、决策树模型分析各变量对失能状况的影响。

研究共纳入9 124名慢性病共病中老年人，其中身体失能状况未受损5 355名（58.7%）、轻度受损3 442名（37.7%）、中度及以上受损327名（3.6%）。多元Logistic回归结果显示年龄<80岁、男性、受初等教育、轻度/中度/重度抑郁、自评健康很不好/不好/一般、视力不正常、未有过摔倒、未有过髋部骨折、身体疼痛部位数量较少、吸过烟且仍在吸烟或已戒烟、饮酒、不运动或运动强度较低、夜间睡眠很少、个人年收入很少/中等是慢性病共病中老年人身体失能状况未受损的影响因素（P<0.05）；自评健康很不好、视力不正常、未有过摔倒、未有过髋部骨折、不运动、不参加社会活动是慢性病共病中老年人身体能力中度及以上受损的影响因素（P<0.05）。决策树模型得出自评健康为失能状况的主要影响因素，其次是抑郁、疼痛部位数、运动、年龄、摔倒（P<0.05）；经比较，未受损、轻度受损、中度及以上受损下Logistic回归受试者工作特征曲线下面积（AUC）分别为0.808（95%CI=0.800~0.817）、0.773（95%CI=0.764~0.783）、0.891（95%CI=0.875~0.907），大于决策树模型对应AUC分别为0.768（95%CI=0.759~0.778）、0.734（95%CI=0.724~0.745）、0.833（95%CI=0.812~0.854）。

研究表明年龄、性别、受教育程度、抑郁情况、自评健康、视力、摔倒、髋部骨折、身体疼痛部位数量、吸烟、饮酒、运动、夜间睡眠时数、社会活动参与、个人年收入均对慢性病共病中老年人失能状况有显著影响。因此，有关部门需针对不同影响维度层面采取有效防范措施，降低慢性病共病中老年人失能率。此外，本研究数据显示Logistic回归预测效果略优于决策树模型，但两者各具优势，Logistic回归可清晰展示自变量与因变量间依存关系，决策树能说明变量间交互关系对因变量的影响，后续可利用两个模型进行辅助分析。