The development and application of statistical models to detect outbreaks of infection in the healthcare setting using local and national surveillance systems

Abstract

Background: Surveillance has been described as an essential component of an effective infection prevention and control programme. With increasing concerns over the rising levels of antimicrobial resistance there has been emphasis on the requirement to improve infectious disease surveillance and outbreak detection. The availability of healthcare information in electronic formats provides an opportunity to perform surveillance and outbreak detection at the hospital level. Aim: To determine if data available from national and local microbiology surveillance systems can be utilised for outbreak detection of infectious diseases within the acute healthcare setting. Methods: A critical analysis of a national microbiological surveillance system is performed prior to the application of methods for automated outbreak detection. At the local level, an epidemiological analysis is performed to ascertain the levels of antimicrobial resistance and identify trends before carrying out outbreak detection for multidrug-resistant organisms (MDRO). Several methods for outbreak detection are investigated, including exceedance detection algorithms, cumulative sum methods and variable life adjusted display charts. Results: Results from a comprehensive systematic review found that the employment of systems utilising electronic data sources for healthcare-associated infection surveillance is feasible. Evaluation of a national microbiological surveillance system identified several issues with using data reported through a voluntary system for outbreak detection at the hospital level. After identification of a hospital laboratory exhibiting consistent and timely reporting it became apparent that outbreak detection methods could be applied to data available through the national system. At the local level, MDRO were identified through the application of algorithms to electronically stored microbiology data. The selected outbreak detection methods were successfully applied to local level data, identifying several potential MDRO outbreak situations. Conclusions: This thesis demonstrates that there is potential for the implementation of automated systems for hospital level outbreak detection using both national and local microbiological data sources.