A Fuzzy Logic-Based Optimization Model for Efficient and Equitable Resource Allocation in Diabetes Mellitus Care: Analysis of SUS Data (2015–2024)

Abstract

Objective: To develop and validate a fuzzy logic-based optimization model for identifying effective resource allocation strategies for diabetes mellitus (DM) care within the Brazilian Unified Health System (SUS). Methods: Retrospective cross-sectional study utilizing DATASUS, SIH-SUS, and Hiperdia registries (January 2015 to December 2024) across 5,570 Brazilian municipalities. We constructed a hierarchical Mamdani-type fuzzy inference system incorporating epidemiological, economic, clinical, and structural indicators. The model was calibrated using historical data, validated through technical, empirical, and expert-based evaluation, and embedded within a multi-objective optimization framework to assess alternative investment scenarios. Results: The integrated dataset comprised 8,347,219 DM-related hospitalizations. The fuzzy system demonstrated 97.3% coverage and outperformed conventional approaches with a mean absolute percentage error of 12.4% in expenditure predictions. Under baseline conditions, the model recommended increasing primary care investments from 31.2% to 42.7%, while reducing tertiary hospital care from 38.4% to 28.9%. These reallocations predicted an 8.4% improvement in glycemic control, a 12.7% reduction in hospitalizations, and a 6.2% decrease in mortality over five years. The analysis identified 847 priority municipalities requiring targeted intervention. Conclusion: Fuzzy logic-based optimization demonstrates substantial potential for enhancing diabetes care efficiency through strategic reallocation prioritizing primary care expansion and equity-focused interventions in underserved regions. Limitations include retrospective data utilization, inability to conduct prospective validation, and non-capture of private healthcare services.