Mortality risk associated with clinical signs of possible serious bacterial infection (PSBI) in young infants in Africa and Asia: protocol for a secondary pooled analysis

TL;DR

This study aims to analyze global data on young infants with possible serious bacterial infection signs to better understand their link to mortality risk.

Contribution

The study introduces a pooled analysis combining global data and novel machine learning methods to assess PSBI signs and mortality in young infants.

Findings

The study will analyze individual and combined PSBI signs for their association with mortality.

Machine learning will be applied to PSBI data for the first time.

Findings will inform global recommendations for managing young infants with PSBI.

Abstract

The WHO’s Integrated Management of Childhood Illness (IMCI) in young infants <2 months of age includes the identification and management of signs of possible serious bacterial infection (PSBI). However, equal importance is given to all the PSBI signs, which signal the need for referral and hospital management, except for fast breathing in infants aged 7–59 days, for which outpatient treatment by clinical staff working at a health facility is recommended. Moreover, studies to validate the importance of clinical signs of PSBI have mostly used the need for hospitalisation as the outcome. There is a need to further examine the association of signs of PSBI individually and in combination with risk of mortality and to analyse global data to inform global recommendations.

We will create a dataset that integrates data from population-based studies globally with similar designs that have examined the presence of signs of PSBI identified by frontline health workers throughout the young infant period (days 0 to <60) and that have also recorded infant vital status. We will conduct pooled, individual-level analyses of the frequency of identification of signs individually and in combinations and will conduct three types of analyses of association of signs of PSBI with mortality: (1) case fatality, which has been used in a multisite study of mortality risk associated with signs of PSBI in young infants in Africa; (2) Cox regression, which will enable time-varying analysis of exposure in relation to mortality, as has been done in a multisite study in Asia and (3) machine learning analysis, which has not previously been applied to any of the available data.

All prior studies incorporated into our pooled analysis were approved by the independent local ethics committee/institutional review board (IRB) at each study site in each country, and all study participants provided informed consent. This project was approved by the Stanford University School of Medicine IRB protocol 74456. Study findings will be disseminated through publications in peer-reviewed journals, WHO documents, and presentations at maternal and child health meetings.