Automating the surveillance of healthcare-associated infections

Abstract

Healthcare-associated infections (HAI) are among the most common complications of medical care, affecting one in twenty-five hospitalized patients on any given day. Surveillance of HAI by systematically assessing patients for the development of an infection is a key component of successful infection prevention programs, both to measure the effects of an intervention and to assess quality of care. Traditional methods of surveillance, manual review of patients’ medical records, are time-consuming and prone to error however. Developments in healthcare information technology and the increasing demands for reliable HAI surveillance have driven the development and implementation of automated surveillance systems that employ routine care data stored in electronic health records to identify (probable) cases of HAI. These systems can increase efficiency, and hence capacity, of surveillance efforts and improve consistency of case-finding. As is illustrated for drain-related meningitis in neurosurgical patients, an automated surveillance approach can reliably identify patients affected by HAI and achieve up to eighty percent workload reduction. This system uses clinical routine care data, e.g. antibiotic use, laboratory testing and microbiology culture results in multivariable regression models to identify high-risk patients requiring manual confirmation of infection. Across the board, use of clinical routine care is preferable over administrative data such as diagnosis codes; despite their ease of accessibility and standardization, this latter data source often has subpar sensitivity and mediocre positive predictive value for HAI. Importantly, successful automation of HAI surveillance depends on adequate documentation in medical records in a format that allows for data extraction. In addition, thoughtful development and validation of systems is necessary to ensure their reliability over time and in different hospitals. In line with increasing use of HAI surveillance as means to assess quality of care, ensuring comparability of HAI rates is becoming paramount. In particular in the United States, concerns with current manual surveillance methods and complexity of case-definitions have motivated revisions of surveillance protocols. For ventilator-associated pneumonia, the revised definitions no longer contain subjective components and have been made amenable to electronic implementation. Despite these efforts, however, we found that in the current specification such modifications do not guarantee consistency of case identification. In addition, when modifying definitions the nature of the events detected will change and these new events may not necessarily lend themselves well to prevention programs. When aiming to compare HAI rates across hospitals, another aspect that requires consideration is adequate correction for differences in severity of underlying disease. Current adjustment methods are fairly crude as for most device-associated infections this limited to stratifying by type of hospital or intensive care unit. Clinical routine care data can also be used to further refine this risk stratification. In summary, automated surveillance of healthcare-associated infections using data extracted from electronic medical records is reliable and feasible. These systems increase capacity of surveillance efforts by identifying patients at high risk of infection and increasing consistency of case-finding. The use of clinical routine care data is preferable over the use of administrative data, and can also be used to refine corrections for severity of underlying disease.