Automated surveillance of healthcare-associated infections: Towards large-scale implementation

Healthcare-associated infections (HAIs) are among the most common complications of healthcare and affect approximately 8% of hospital patients in Europe. HAIs include surgical site infections (SSI), bloodstream infections, urinary tract infections, and respiratory tract infections. HAIs increase hospital stays, mortality, and healthcare costs, but up to half may be preventable through effective infection prevention. Surveillance is essential because it enables healthcare professionals to monitor infection trends and take timely preventive measures.
Traditionally, HAI surveillance relies heavily on manual patient-record review, which is time-consuming and prone to errors. Automated surveillance (AS) can improve efficiency and standardization. In semi-automated systems, algorithms identify high-risk patients for manual review, whereas fully automated systems determine whether an infection has occurred without manual assessment. Electronic health records make large-scale AS increasingly feasible, although differences in definitions, data infrastructures, and clinical practices remain important challenges.
This thesis evaluated the potential for large-scale implementation of AS, focusing on hospital-onset bacteraemia (HOB) and SSI. HOB is particularly suitable for automation because it can be objectively identified using microbiological data, while SSI is a widely used infection for surveillance but is transitioning from manual to automated methods.
Chapter 2 demonstrated that an international consensus definition for HOB could be applied consistently across four Dutch hospitals despite differences in data infrastructures. The approach enabled monitoring of HOB trends by hospital, ward, micro-organism, and time period. The study also showed that assumptions within the consensus definition can influence results, highlighting the importance of standardized methods.
Chapter 3 focused on risk adjustment to enable fair comparisons between hospitals. Patient case-mix characteristics and blood-culture variables were associated with HOB rates. The use of standardized infection ratios (SIRs) and funnel plots was proposed as a framework for interpreting differences between hospitals and over time while accounting for variations in patient populations and diagnostic practices.
Chapter 4 investigated whether national antimicrobial resistance surveillance data (ISIS-AR) could support a centrally coordinated national HOB surveillance system. The data were suitable for detecting pathogen-related HOB and required only little additional work from hospitals. However, improvements in reporting hospital admission dates are needed, and local diagnostic and care practices must be considered when interpreting results.
Chapter 5 evaluated an implementation strategy for large-scale semi-automated SSI surveillance. Successful implementation required clear roles and responsibilities, strong project management, sufficient knowledge of AS, stakeholder engagement, training, and collaboration between hospitals.
Overall, the thesis shows that automated HAI surveillance is feasible and has potential to improve infection monitoring on a large scale. Standardized definitions, appropriate risk adjustment, reliable data sources, and understanding of local practices are essential. Successful implementation also requires strong leadership, stakeholder involvement, knowledge sharing, and support from coordinating organizations. Further research should examine the preventability and clinical relevance of HOB and its potential for national trend monitoring.