Beyond Forward Chaining: Logical Perspectives on Clinical Infection Monitoring of BSIs

Healthcare-associated infections (HAIs), particularly bloodstream infections (BSIs), require timely and reliable detection. Typically, rule-based clinical decision support systems use forward chaining to propagate patient data through knowledge-based systems. While effective, this data-driven approach can become computationally costly in data-rich settings and obscures the underlying reasoning structure. We analyze infection monitoring from a computational perspective by formalizing rule-based detection as a fixed-point computation and examining its complexity. As an alternative, we consider goal-driven reasoning, which focuses inference on specific diagnostic hypotheses, and formulate BSI detection as a satisfiability problem in propositional logic. This perspective supports more efficient, interpretable, and uncertainty-aware approaches to clinical decision support.