To systematically investigate the independent factors influencing the life expectancy of patients with dementia enrolled in the Chengdu Long-Term Care Insurance (hereinafter referred to as “LTCI”) system, and to provide high-quality evidence for implementing precision care, developing a risk stratification management system, and optimizing LTCI policies.

A retrospective cohort study design was employed. A total of 10 108 patients with dementia enrolled in Chengdu LTCI from January 2018 to April 2022 were included and followed up until July 2025. Baseline demographic characteristics, comorbidities, adverse events, and assessment data on cognitive function, activities of daily living, and neuropsychiatric behaviors were extracted. Using Python software, descriptive analysis, univariate tests, multicollinearity tests, multivariate logistic regression, Cox proportional hazards regression, and model performance evaluation were conducted.

Among the 10 108 patients, 5 435(53.8%) died. Multivariate logistic regression indicated that age (OR=1.64, 95%CI：1.56-1.72), male gender (OR=1.63, 95%CI：1.47-1.80), mild dysphagia (OR=4.01, 95%CI：3.30-4.88), and >2 comorbidities (OR=1.50, 95%CI：1.31-1.73) were independent risk factors for mortality, while moderate cognitive impairment (OR=0.29, 95%CI：0.18-0.46) and higher education level (OR=0.77, 95%CI：0.64-0.91) were independent protective factors (all P<0.05). Multivariate Cox regression showed that gender (male: HR=1.412, 95%CI:1.336-1.491), age (HR=1.039, 95%CI:1.036-1.042), number of comorbidities (HR=1.226, 95%CI:1.143-1.316), adverse events (HR=1.293, 95%CI:1.141-1.466), and swallowing function (HR=1.092, 95%CI:1.059-1.126) were risk factors for survival, whereas better cognitive function (HR=0.399, 95%CI:0.357-0.446), activities of daily living (HR=0.948, 95%CI:0.911-0.985), higher education level (HR=0.964, 95%CI:0.944-0.985), and history of surgery (HR=0.789, 95%CI:0.661-0.940) were protective factors (all P<0.05). The AUC of the logistic regression model was 0.7315, and the Cox regression model showed good calibration and discrimination (24-month calibration slope=1.084) with strong stability.

The life expectancy of patients with dementia under long-term care is influenced by multidimensional factors. Abnormal swallowing function, advanced age, and male gender are core risk factors, while family care, higher education level, and better functional status are key protective factors. The constructed Cox regression model can be effectively used for clinical risk assessment, and targeted interventions are expected to improve patient prognosis.