The World Health Organisation describes antimicrobial resistance (AMR) as one of the biggest threats to global health. One of the drivers of AMR in healthcare settings is the overuse of broad-spectrum antimicrobials. For severe infections, paediatricians must prescribe antimicrobial therapy urgently; however, this limits the performance of microbiological culture tests to determine the presence of bacterial infection. In this thesis, I explored this problem, and potential solutions in critically ill children.

Firstly, I evaluated the use of microbiology and molecular diagnostic tests in a cohort of mechanically ventilated children and assessed the baseline use of antimicrobial therapy in the paediatric intensive care unit (PICU). I identified that lower respiratory tract infection (LRTI) was the leading indication for antimicrobial treatment; however, a bacterial cause was not found in most patients. Therefore, I developed a single-centre study to evaluate the implementation of a culture-independent, 52-pathogen custom TaqMan array card (TAC) in critically ill children who had suspected LRTI. This semi-quantitative PCR array had been previously validated in a study of adults with suspected ventilator associated pneumonia (VAP). I adapted this protocol to be suitable for children and broadened the investigation to include community acquired LRTI – the leading cause of PICU admissions. I found that TAC was reliable and faster than routine microbiological culture. Doctors frequently used TAC results to make antimicrobial prescribing decisions. Despite the test being more sensitive than microbiological culture, it did not result in an overall change in the use of antimicrobial therapy.

After completing the diagnostic study, I interviewed PICU staff regarding their experiences of using the TAC in clinical practice. The staff described the utility of TAC within the study's boundaries for two primary reasons. Firstly, TAC was helpful as a screening test in situations where bacterial infection was unlikely, and the test supported clinicians in the decision to cease antimicrobial therapy. Secondly, TAC was requested as a diagnostic test, where there was a high pre-test probability of bacterial infection, to identify the pathogen causing infection and optimise ongoing antimicrobial therapy. I also undertook an international survey to understand the information PICU prescribers use to diagnose and treat LRTI. Clinicians frequently reported that they relied on their clinical assessment rather than investigation results to make antimicrobial decisions in the PICU. Although considered to be highly relevant when making prescribing decisions, microbiological cultures are slow and have poor sensitivity. Therefore, molecular diagnostic tests could have a greater role in such scenarios.

Finally, I validated and assessed the feasibility of using an AMR gene TAC (AMR-TAC) as an additional tool to guide antimicrobial decision-making in the PICU. Although I found a high prevalence of AMR gene carriage in critically ill children, this was infrequently associated with phenotypic AMR. Whilst AMR-TAC should not have widespread routine use, further evaluation is required in a select group of children at high risk of AMR or with treatment-resistant infection. AMR-TAC could also have a role as an AMR screening tool in healthcare environments.

In summary, in this thesis, I describe through mixed methods research, the potentially transformative impact that syndromic molecular arrays may have for diagnosing infection in the PICU.