Development and evaluation of a clinical decision support tool for serious bacterial infection in neonates in Southern Africa

Neonatal bacterial infections are an important cause of morbidity and mortality in low-resource settings (LRS).

Accurate identification of those needing treatment while avoiding unnecessary antibiotics in those who do not is crucial.

Features used to identify neonates with risk factors for, or clinical features of possible serious bacterial infections (PSBI) in international guidelines merit re-evaluation, due to a massive shift to facility-based deliveries in LRS and a changing spectrum of pathogens.

I plan to reassess methods for identifying neonates with, and at risk of PSBI in LRS, to derive a patient-centred, contextually appropriate decision-support tool.

I will assess contextual influences on clinician decision-making e.g. resource availability, and seek families’ opinions of risk trade-offs in clinical decision-making about PSBI in Zimbabwe.

I will use national neonatal electronic healthcare records to prospectively evaluate features and risk factors for PSBI across a Zimbabwean province, with senior clinician diagnosis of PSBI and/or positive microbiology as the primary outcome. I will use machine learning to develop and externally validate a clinical prediction model.

Qualitative data be incorporated into the model to develop a clinical decision-support tool to triage neonates into high/moderate/low risk. I will pilot the tool in preparation for large-scale evaluation after this Fellowship.