Skip to main content


Using machine learning to develop a risk stratification tool for neonatal sepsis in low-resource settings


CREATE (Africa Health Research Training Programme); Wellcome Trust

Type of Study



St George’s, University of London


Uganda, Zimbabwe

Chief Investigator

Dr Sarah Sturrock


Neonates born at ≥34 weeks’ gestational age


Current risk stratification systems are not able to predict neonatal sepsis risk in a low-resource setting with sufficient accuracy. Reliable prediction of neonatal sepsis risk would enable clinicians to prioritise early broad-spectrum antibiotics and close monitoring for the highest-risk neonates, reducing associated morbidity and mortality, and allow earlier discharge of low-risk neonates without antibiotic treatment, reducing the economic and psychological burden to families of remaining in hospital and antimicrobial overuse. In a wider context, the ability to predict sepsis risk more accurately would also aid the development of clinical pathways and service planning within maternity and neonatal services. Finally, the resulting model and code from this study could be used by other researchers in the future to model risk of other infective conditions, or neonatal sepsis in other settings.

To build a model for risk stratification, data are required that encompass not only a broad range of potential risk factors for sepsis, but also accurate outcome data confirming or refuting the diagnosis of sepsis. Previous attempts to create risk stratification systems have lacked blood cultures as the gold standard diagnostic method for sepsis, so this study will include a blood culture sample from every participant. This study will also include a range of patient-reportable potential risk factors to maximise the clinical utility of the model in the future. This study aims to:

1. Determine antenatal and perinatal factors associated with early-onset neonatal sepsis in low-resource settings, as diagnosed by a blood culture or a senior clinician.

2. Measure the strength of association between the above factors and early-onset neonatal sepsis.

3. Combine the above risk factors into a risk stratification model for early-onset neonatal sepsis using machine learning techniques.


September 2023 – May 2024