A machine learning-guided strategy for replacement of peripheral venous catheters in the prevention of hospital-acquired bacteremia: A comparative cost-effectiveness analysis

Abstract

Machine learning (ML) models for early identification of patients at risk of hospital-acquired bacteremia (HAB) may guide a preventive strategy of routine replacement of peripheral venous catheters (PVCs). The feasibility and cost-effectiveness of such an approach have not been evaluated. In this study, we developed ML models for the preventive strategy of ML-guided routine replacement of PVCs and evaluated this strategy according to number needed to harm and incremental cost-effectiveness for preventing a HAB-related death. The ML-guided strategy of early identification of patients at risk of HAB can lead to a reduction in the labor-intensive intervention of routine PVC replacement. However, this is at the expense of increased costs for establishing and maintaining ML model guidance, as well as increased costs for patients erroneously categorized as low-risk for HAB by the ML model.