Self-care–based clinical and metabolic profiles among adults with type 2 diabetes: a cross-sectional cluster analysis

Type 2 diabetes mellitus is a heterogeneous condition influenced by clinical, metabolic, and behavioral factors. Traditional approaches to diabetes management often rely on single clinical indicators and may overlook meaningful patterns in patients’ self-care behaviors. This study aimed to identify distinct patient profiles by integrating clinical characteristics, metabolic parameters, and self-care behaviors among adults with type 2 diabetes. A multicenter cross-sectional study was conducted among 303 adults with type 2 diabetes attending outpatient clinics in northern Italy. Sociodemographic, clinical, and laboratory data were collected together with self-care behaviors assessed using the Self-Care of Diabetes Inventory. Continuous variables were standardized prior to analysis. Dimensionality reduction was performed using t-distributed stochastic neighbor embedding, followed by hierarchical agglomerative clustering using Ward’s method. The five-cluster solution was retained through joint evaluation of average silhouette width, the Calinski–Harabasz index, dendrogram structure, cluster-size balance, resampling stability, and clinical interpretability. Differences between clusters were examined using chi-squared tests, Fisher’s exact tests, and Kruskal–Wallis tests with post hoc comparisons. A five-cluster solution comprising clinically distinguishable profiles was retained. The partition showed moderate resampling stability (mean adjusted Rand index = 0.593). The clusters differed significantly in age, duration of diabetes, multimorbidity burden, glycemic control, lipid profile, renal function, insulin therapy use, and self-care behaviors ( p < 0.05). Profiles ranged from older adults with low self-care management and self-efficacy to working-age individuals with poor glycemic control and to older adults showing high self-care engagement despite suboptimal metabolic outcomes. Adults with type 2 diabetes present heterogeneous combinations of clinical status and self-care behaviors. Person-centered clustering approaches may help characterize clinically interpretable patient profiles and inform the future development of tailored educational and clinical strategies for more personalized diabetes care. However, the exploratory profiles require external validation before clinical implementation.