Risk prediction model for early detection of urinary tract infection in a hospital setting in Australia.

Background: Hospital-acquired complications have detrimental effects on patient outcomes and recovery, with increased morbidity and mortality burdens, and hospital efficiency. The Australian Commission on Safety and Quality in Healthcare has identified 16 high-priority complications, including healthcare-associated infections, as potential targets of clinical risk mitigation strategies. Within the North Metropolitan Health Service in Western Australia, the prevalence of urinary tract infections (UTIs) was recognised as one of the most ubiquitous hospital-acquired complications and thus, there was desire to find new and innovative ways to enhance the existing infection prevention and control practices.

Objective: To develop a risk prediction model for early identification of inpatients at risk of acquiring a UTI, to support clinical processes to facilitate targeted intervention strategies.

Method: Prognostic modelling techniques were employed using retrospective hospital separation data encompassing patient and local health service factors.

Results: The risk prediction model, developed from approximately 350 variables, used just 9 factors: 2 patient characteristics (age, gender), 4 clinical factors (paraplegia, dementia, prostate hyperplasia, neurosurgeon care), and 3 process measures (hospital stay duration, long theatre time, intensive care unit stay). It predicted UTI risk with 91% sensitivity, 86% specificity, and 95% discrimination (area under the curve). Real-time use in ward settings suggested it could help reduce hospital-acquired urinary tract infections (HAUTIs).

Conclusion: Predictive modelling techniques can identify patients at risk of developing a HAUTI with high sensitivity and specificity. The resulting model can be used as a real-time clinical decision-making tool to guide proactive interventions and help reduce the prevalence of UTIs among hospital inpatients.Implications for health information management practice:The development and successful validation of a real-time predictive model for HAUTIs demonstrates how health information managers can leverage routinely collected data to support proactive clinical risk mitigation. Integrating such models into electronic health record systems can enhance patient safety, improve clinical workflows, and inform targeted infection control interventions across hospital settings.