Hospital-acquired neonatal infection in very low birth weight infants: Optimising diagnostic strategies for low-resource settings

Neonatal healthcare-associated infection (HAI) is a significant contributor to neonatal morbidity and mortality globally, especially in low- and middle-income countries (LMICs). Rapid diagnosis and treatment are critical, especially in very low birth weight (VLBW) and preterm neonates, where a delay in initiating treatment may have serious consequences. However, diagnosis is challenging owing to the non-specific presenting clinical signs and symptoms, delay in the availability of the diagnostic gold standard (blood culture), and absence of an ideal biomarker to inform antibiotic decision-making. Additionally, the epidemiology of neonatal HAI in LMICs needs to be better described, owing to challenges with harmonised sepsis definitions, weak health information systems, and limited access to microbiology laboratory tests. High rates of empiric antibiotic use for suspected neonatal HAI may contribute to the rising incidence of antimicrobial resistance in LMIC neonatal units. In Part I, Chapter 2, we described neonatal HAI's epidemiology and disease burden among VLBW at a tertiary neonatal unit in South Africa. We demonstrated that proven (culture positive) and presumed (culture negative) HAI are equally prevalent in our neonatal population. Part II explores infection prediction scores as an adjunctive diagnostic tool for neonatal HAI. In Chapter 3, we conducted a performance comparison of existing infection prediction scores and retrospectively evaluated five scores. The NOSEP1 score ≥ 8 achieved the highest diagnostic accuracy but did not perform well enough to recommend routine clinical use. Chapter 4 presented the development and internal validation of a novel infection prediction score (NeoHoP score – Neonatal Healthcare-associated infection Prediction score) using local VLBW neonatal patient characteristics. We showed that a NeoHoP score of ≥ 2 had high specificity and positive prediction values, positioning it as a ‘rule-in’ test for neonatal HAIs in VLBW infants. In Chapter 5, we externally validated the NeoHoP score using an international research dataset (NeoOBS), which included a more diverse neonatal population. We demonstrated that the score achieved comparable results to the internal validation, with high specificity and positive predictive value, confirming that it is well positioned as a ‘rule-in’ test for neonatal HAI. We conducted a prospective study in Part III using point-of-care C-reactive protein (CRP). In Chapter 6, we demonstrated that combining the NeoHoP score and serial POC CRP tests in a stepwise algorithm could support antibiotic decision-making and potentially significantly reduce the length of antibiotic therapy for neonatal HAI.