Background: Critically ill patients are at particular risk of developing bloodstream infection. Such infections are associated with the development of sepsis, leading to a marked increase in mortality rate. Early detection of the causative organism and appropriate antibiotic treatment are therefore critical for optimum outcome of patients with nosocomial infection. Current infection diagnosis is based on standard blood culture techniques. However, microbiological culture has a number of limitations, not least that it takes several days to confirm infection and is therefore not useful in directing the early treatment with antibiotics. New techniques based on the detection of pathogen DNA using real-time polymerase chain reaction (PCR) technology have the potential to address these limitations but their clinical utility is still to be proved. Objectives: Develop and evaluate novel PCR-based approaches to bloodstream infection diagnosis in critical illness based on detection and identification of bacterial and fungal DNA in blood. Methods: A range of commercial and 'in-house' PCR-based assays for detection of bacterial and fungal DNA were developed and/or optimised for use in clinical blood samples. These included LightCycler SeptiFast, a CE-marked multi-pathogen assay for common bloodstream pathogens, BactScreen and GramScreen, broad spectrum bacterial assays based on 16S rRNA gene and real-time PCR assays developed to detect a range of clinically important fungal pathogens. Novel approaches to speciation of pathogen DNA using melting temperature (Tm) profiling and high resolution melting analysis (HRMA) were developed. Clinical evaluation of assays was either on blinded clinical isolates or blood samples from critically ill patients with clinical suspicion of bloodstream infection against conventional microbiological culture. Several techniques aimed at improving extraction of pathogen DNA from blood were also investigated. Results: The CE-marked commercial assay SeptiFast showed analytical sensitivity and specificity of 79% and 83% respectively. Concordance with positive culture results was good but high levels of 'false positives' were detected possibly attributed to detection of free pathogen DNA not associated with viable pathogens. The predictive value of a negative SeptiFast test was 98% suggesting that absence of pathogen DNA is a strong indicator of absence of infection. Further studies were aimed at detailed optimisation and validation of 16S rRNA gene real-time PCR assays for bacterial DNA. BactScreen and GramScreen were able to detect a broad range of clinically important bacteria down to <50 CFU/ml blood. A preliminary comparative evaluation against SeptiFast showed BactScreen gave excellent concordance with blood culture results with minimal false positive results compared to SeptiFast. Efficient extraction of pathogen DNA was shown to be a key factor in determining analytical sensitivity and several protocols were evaluated. Low cost approaches to speciation of bacterial DNA were developed by combining broad range real-time PCR with HRMA. A novel HRMA method based on Tm profiling was shown to identify 89% and 96% of blinded clinical isolates at species or genus level respectively. Real-time PCR/HRMA approaches were also successfully developed for detection and identification of fungal pathogens including a range of Candida and Aspergillus species associated with bloodstream fungal infection. Conclusions: These studies have highlighted some of the key factors that need to be considered when developing and validating PCR based assays for pathogen DNA detection in blood. A set of novel tools have been developed for rapid detection and identification of bacterial and fungal pathogens that could address the challenges of infection diagnosis based on pathogen DNA detection. Further work is required, not least in development of more efficient pathogen DNA extraction and detailed clinical validation but the tools described here have the potential to provide cost effective solutions to aid infection diagnosis that would be complementary to current culture-based methods. The provision of time critical information could have a positive impact on clinical decision-making leading to more effective management and treatment of patients with suspected bloodstream infection.