Latent profile analysis-based development and validation of a model for differentiating subjective cognitive decline profiles in older hemodialysis patients: a cross-sectional study

Subjective cognitive decline is often discussed as a potential early marker of dementia. Although subjective cognitive decline has been proposed as a key early warning sign of preclinical Alzheimer’s disease, it is important to recognize that subjective cognitive decline as a subjective perception is strongly influenced by psychological state, social environment, and physical health. Older maintenance hemodialysis patients are at high risk for subjective cognitive decline due to uremic toxin accumulation, dialysis-related brain injury, and multiple comorbidities. Traditional total-score-based assessments fail to capture the intrinsic heterogeneity among Subjective cognitive decline subgroups. Latent profile analysis, a person-oriented approach, can identify latent subgroups with similar multidimensional characteristics. Given the insidious nature of subjective cognitive decline and the differential characteristics across profiles, constructing profile-specific classification models is crucial for timely risk screening. To explore the latent profiles of subjective cognitive decline in older maintenance hemodialysis patients, analyze characteristic differences and associated factors among these profiles, and develop and evaluate a classification model for differentiating subjective cognitive decline profiles in this population. Using convenience sampling, 300 older maintenance hemodialysis patients from Li Huili Hospital of Ningbo Medical Center, Zhejiang Province, China, between May and November 2025 were recruited. Participants completed a general information questionnaire, the subjective cognitive decline Questionnaire-9, the Pittsburgh Sleep Quality Index, and the Short Physical Performance Battery. Latent profile analysis was performed based on the subjective cognitive decline Questionnaire responses. Univariate analysis and multinomial logistic regression were used to construct the classification model. The model’s performance was comprehensively evaluated across three dimensions: authenticity, reliability, and discriminatory ability. Two hundred seventy-five valid questionnaires were collected and latent profile analysis identified three distinct subjective cognitive decline profiles: Mild Cognitive Decline Group (C1, 40.00%), Moderate Cognitive Decline Group (C2, 38.18%), and Severe Cognitive Decline-Physical Dysfunction Group (C3, 21.82%). The classification model for subjective cognitive decline in older maintenance hemodialysis patients, constructed using multivariate regression analysis, demonstrated a good fit in the Hosmer–Lemeshow test ( P > 0.05). Furthermore, it exhibited outstanding discriminatory performance in terms of validity, reliability, and discriminatory ability. Subjective cognitive decline among older maintenance hemodialysis patients is prevalent at a moderate-to-high level and demonstrates significant heterogeneity, classifiable via latent profile analysis into three distinct profiles. The classification model incorporating dialysis vintage, monthly social activity participation hours, serum albumin, the Pittsburgh Sleep Quality Index, and the Short Physical Performance Battery scores demonstrates significant discriminatory efficacy within the derivation sample, offering a preliminary foundation for risk stratification in this high-risk population that warrants external validation.