FEDERATED LEARNING–DRIVEN REAL-TIME DISEASE SURVEILLANCE FOR SMART HOSPITALS USING MULTI-SOURCE HETEROGENEOUS HEALTHCARE DATA

Abstract

Federated learning is increasingly promoted as a privacy-preserving way to use multi-source heterogeneous healthcare data for real-time disease surveillance in smart hospitals, but little is known about how its capabilities, data integration, governance, and organizational conditions jointly shape perceived surveillance performance. This study therefore aimed to quantify how federated learning capability, multi-source data integration, privacy and security assurance, and organizational readiness contribute to perceived surveillance effectiveness across cloud and enterprise smart hospital cases. A quantitative cross-sectional, case-based survey design was used, with 228 usable responses (87.7 percent response rate from 260 questionnaires) from clinicians, infection prevention staff, health informatics personnel, and IT/security professionals working in digitally advanced hospitals. Key variables included federated learning capability, multi-source integration, privacy and security assurance, organizational readiness, and perceived surveillance effectiveness, all measured on five-point Likert scales. The analysis plan comprised descriptive statistics, reliability testing, Pearson correlations, and multiple regression. Mean scores indicated generally high maturity (FLC 3.78, MSI 3.92, PSA 4.06, OR 3.84, SE 3.95), with strong internal consistency (Cronbach’s alpha 0.83–0.88). Correlations between surveillance effectiveness and predictors were all positive and significant, highest for organizational readiness (r = 0.72) and federated learning capability (r = 0.65). The regression model was significant and explained 61.3 percent of the variance in surveillance effectiveness (R² = 0.613), with organizational readiness (β = 0.29), federated learning capability (β = 0.24), privacy and security assurance (β = 0.21), and multi-source integration (β = 0.18) all making significant contributions. The findings imply that effective federated learning driven surveillance in smart hospitals depends not only on advanced models and integrated data, but also on robust privacy safeguards and high organizational readiness to act on analytic insights.