Healthcare Utilization Patterns for Acute Febrile Illness in Bangladesh, Nepal, and Pakistan: Results from the Surveillance for Enteric Fever in Asia Project

Background: Characterizing healthcare-seeking patterns for acute febrile illness is critical for generating population-based enteric fever incidence estimates from facility-based surveillance data. Methods: We used a hybrid model in the Surveillance for Enteric Fever in Asia Project (SEAP) to assess incidence of enteric fever at 6 study hospitals in 3 countries. We recruited individuals presenting to the hospitals and obtained blood cultures to evaluate for enteric fever. For this analysis, we undertook cluster random household surveys in Dhaka, Bangladesh (2 sites); Karachi, Pakistan; Kathmandu, Nepal; and Kavrepalanchok, Nepal between January 2017 and February 2019, to ascertain care-seeking behavior for individuals with 1) fever for ≥3 consecutive days within the past 8 weeks; or 2) fever resulting in hospitalization within the past year. We also collected data about disease severity and household demographics and assets. We used mixed-effect multivariable logistic regression models to identify determinants of healthcare seeking at study hospitals and determinants of culture-confirmed enteric fever. Results: We enrolled 31 841 households (53926 children) in Bangladesh, 25510 households (84196 children and adults) in Nepal, and 21310 households (108031 children and adults) in Pakistan. Children <5 years were most likely to be taken to the study hospitals for febrile illness at all sites. Household wealth was positively correlated with healthcare seeking in 4 of 5 study sites, and at least one marker of disease severity was positively associated with healthcare seeking in 3 of 5 catchment areas. Wealth and disease severity were variably predictive of blood culture-confirmed enteric fever. Conclusions: Age, household wealth, and disease severity are important determinants of healthcare seeking for acute febrile illness and enteric fever risk in these communities, and should be incorporated into estimation models for enteric fever incidence.