Abstract
INTRODUCTION: Heterogeneity in definitions of severe infection, sepsis and serious bacterial infection (SBI) in infants limits the comparability of randomised controlled trials (RCTs) of infection prevention interventions. To inform the design of infection prevention RCTs for infants in low-resource settings, we estimated the incidence of severe infection and death among Bangladeshi infants aged 0-60 days using variations in case definitions. METHODS: Among 1939 infants born generally healthy in Dhaka, Bangladesh, severe infection was identified through up to 12 scheduled community health worker home visits from 0 to 60 days of age or through caregiver self-referral. The primary severe infection case definition combined physician documentation of standardised clinical signs and/or diagnosis of sepsis/SBI, plus either a positive blood culture or parenteral antibiotic treatment for ≥5 days. Incidence rates were estimated for the primary severe infection definition, the WHO definition of possible SBI, blood culture-confirmed infection and five alternative definitions including non-injury death. RESULTS: Severe infection incidence per 1000 infant-days was 1.2 (95% CI 0.97 to 1.4) using the primary definition, 0.84 (0.69 to 1.0) using the WHO definition of possible SBI, 0.026 (0.0085 to 0.081) using blood culture-confirmed infection and 0.061 (0.029 to 0.13) for death. One-third of cases met criteria for the primary severe infection definition through physician diagnosis of sepsis/SBI rather than the standardised clinical signs, and 85% of cases were identified following caregiver self-referral despite frequent scheduled visits. CONCLUSIONS: Severe infection incidence in infants varied considerably by case definition. Using a clinical sign-based definition may miss a substantial proportion of cases identified by physician diagnosis of sepsis/SBI. A consensus definition of severe infection in infants that balances permissiveness and stringency and can be operationalised in low-resource countries would improve the comparability of RCTs. If health facilities are accessible and caregivers readily seek care for infant illness, frequently scheduled home assessments may not be necessary.