In search of a primary outcome for community-based newborn infection trials in Eastern Uganda: a nested cohort study within the BabyGel pilot trial

Abstract

Background: Due to their immature immune system, neonates are at high risk of infection. This vulnerability when combined with limited resources and health education in developing countries can lead to sepsis, resulting in high global neonatal mortality rates. Many of these deaths are preventable. The BabyGel pilot trial tested the feasibility of conducting the main randomised trial, with the provision of alcohol handgel to postpartum mothers for prevention of neonatal infective morbidity in the rural community. This secondary analysis sought to evaluate the methods of detecting infections in babies up to 3 months of age. Methods: The pilot two-arm cluster randomised controlled trial took place in 10 villages around Mbale, Eastern Uganda. Women were eligible and recruited antenatally if their gestation was ≥ 34 weeks. All infants of mothers participating in the BabyGel pilot trial were followed up for the first 3 months of life. Evidence for infant infection was collected using five different methods: clinician diagnosed infection, microbiologically confirmed infection, maternally reported infection, a positive infection screen using the World Health Organization (WHO) Integrated Management of Childhood Illness (IMCI) screening criteria, and reported antibiotic use identified during home and clinic visits. These methods were assessed quantitatively regarding the detection rates of suspected infections and qualitatively by exploring the challenges collecting data in the rural community setting. Results: A total of 103 eligible women participated in the BabyGel pilot trial, with 1 woman delivering twins. Of the 99 mother-infant pairs who consented to participate in the study, 55 infants were identified with infection in total. Maternal report of illness provided the highest estimate, with mothers reporting suspected illness for 45 infants (81.8% of the total suspected infections identified). The WHO IMCI screening criteria identified 30 infants with suspected infection (54.5%), and evidence for antibiotic use was established in 22 infants (40%). Finally, clinician-diagnosed infection identified 19 cases (34.5%), which were also microbiologically confirmed in 5 cases (9.1%). Data collection in the rural setting was hindered by poor communication between mothers and the research team, limited staff awareness of the study in health centres resulting in reduced safeguarding of clinical notes, and widespread use of antibiotics prior to notification and clinical review. Furthermore, identification of suspected infection may not have been limited to severe infections, with ambiguity and no official clinical diagnosis being given to those identified solely by maternal report of infection. Conclusions: A high rate of suspected infection was identified spanning the five sources of data collection, but no ideal method was found for detection of community neonatal infection. Although maternal self-reports of infant infection provided the highest detection rate, data collection via each source was limited and may have identified minor rather than major infections. Future studies could utilise the IMCI screening tool to detect severe community infection leading to referral for clinical confirmation. This should be combined with weekly contact with mothers to detect maternally suspected illness. Obtaining more details of the symptoms and timescale will improve the accuracy when detecting the total burden of suspected disease, and advising participants to retain medication packaging and prescriptions will improve identification of antibiotic use.