Complex Network Analysis Reveals Associations Between Physical Activity and Knee Arthroplasty Burden Across Brazilian State Capitals (2009–2023)

Background: Total knee arthroplasty (TKA) is one of the most frequently performed procedures for advanced knee osteoarthritis. Demographic, cardiometabolic, behavioral, and socioeconomic factors may be associated with TKA demand and hospital outcomes, but these indicators are often investigated separately. Objective: To investigate the structural relationships between population-level indicators and major TKA hospital outcomes across Brazilian state capitals using complex network analysis. Methods: An ecological study was conducted using data from the Brazilian Unified Health System (SUS) and the Surveillance System for Risk and Protective Factors for Chronic Diseases by Telephone Survey (VIGITEL) between 2009 and 2023. Variables were residualized for state-capital and calendar-year effects, and undirected weighted correlation networks were constructed using FDR-retained Pearson correlations. Targeted eigenvector centrality was used to characterize the structural position of variables relative to hospital outcomes. Results: In the primary hospitalization network, physical activity frequency of 3–4 days/week showed the highest targeted eigenvector centrality (0.466) and the only FDR-retained direct association with hospitalization rates. In the primary cost network, hypertension (0.484), diabetes mellitus (0.457), and mean age (0.443) showed the highest centralities, while mean age showed the only FDR-retained direct association with cost. However, these direct associations were retained after FDR correction in only 58.7% and 52.3% of cluster-bootstrap replicates, respectively, and their 95% bootstrap confidence intervals included zero. The hospitalization association was also sensitive to alternative representations of exercise-frequency composition. Centrality rankings also showed only moderate resampling stability. No direct FDR-retained association was observed for hospital length of stay. Conclusions: TKA-related hospital outcomes were characterized by distinct population-level correlation structures, although the bootstrap analysis indicated uncertainty in individual associations and centrality rankings. Network centrality reflects structural position rather than causal influence or independent association with hospital outcomes. Undirected weighted correlation networks may complement conventional epidemiological approaches by characterizing interdependencies and generating hypotheses for future longitudinal and individual-level studies.