Predictive modeling of disinfectant efficacy: Integrating bacterial inactivation kinetics with human health risk assessment.

Existing regulatory frameworks for disinfectant efficacy rely on fixed pass/fail thresholds under predefined conditions, which limits their relevance to variable exposure conditions and prevents integration with toxicological risk assessment. To address these limitations, we developed a Gaussian Process (GP) modeling framework to characterize inactivation under variable exposure conditions. We then integrated safety risk assessment through the Hazard Quotient (HQ) analysis based on toxicological reference values. Specifically, the GP model was fitted to an experimental dataset obtained from microbial inactivation experiments conducted using benzalkonium chloride against Staphylococcus aureus and Escherichia coli across varying temperatures and exposure times. During GP model fitting, a Tobit-type likelihood formulation was used within the Bayesian framework to handle experimental observations at the analytical limit of detection. To account for potentially complex inactivation behavior across these variable conditions, several GP kernel structures were compared, and the Matérn plus Rational Quadratic configuration was identified as optimal. From the fitted GP model, concentration was identified as the dominant determinant of efficacy, whereas the inactivation response showed non-linear structure that conventional parametric models may fail to capture. By integrating the fitted GP model with multi-route HQ analysis, we quantitatively derived route-specific usage limits. Notably, the inhalation-based limit was far more restrictive than the oral and dermal limits, highlighting the limitations of conventional uniform guidelines. This integrated framework provides a robust, data-driven methodology for establishing safety-critical usage limits, bridging the gap between probabilistic microbiology and multi-route risk assessment to support evidence-based disinfection management.