Cognitive impairment is common among peritoneal dialysis (PD) patients and is shaped by sociodemographic, psychosocial, and behavioral determinants. Traditional variable-centered analyses often overlook clinically meaningful subgroups with distinct constellations of risk and protective factors. This study aimed to identify multidimensional patient profiles in PD using cluster-driven analysis, compare cognitive performance across profiles, and characterize within-profile cognitive risk pathways using decision tree modeling to inform targeted screening and intervention strategies.

This cross-sectional study enrolled 333 PD patients from a tertiary hospital in China. Seven variables—age, education, health literacy, self-efficacy, family support, sleep duration, and physical activity—were assessed. K-means clustering identified distinct profiles. Cognitive function was measured with the Montreal Cognitive Assessment (MoCA). Non-parametric tests compared MoCA scores across profiles, multiple linear regression identified independent predictors, and decision tree modeling explored risk stratification within the highest-risk profile.

Three clinically interpretable profiles were identified. The youngest profile, with the lowest health literacy, self-efficacy, and family support but highest physical activity, had the lowest MoCA scores. Education (β = 1.25, p < 0.001) and family support (β = 0.10, p = 0.0046) were independent predictors of cognition. Within the high-risk profile, decision tree analysis revealed that lower education, compounded by low self-efficacy and shorter sleep duration, defined subgroups at greatest risk, while higher education with strong psychosocial resources conferred resilience.

Cluster-driven profiling combined with decision tree modeling identified heterogeneous cognitive risk pathways in PD. This framework offers practical utility for targeted cognitive screening, optimized resource allocation, and individualized intervention strategies in PD care.