Background: As perinatal and neonatal mortality move to the foreground of the debate on
how best to improve child survival in poor countries, there appears a pressing need to test
potential interventions. Implicit in testing is the ability to document birth and neonatal
outcomes in rural communities. The thesis sets perinatal events in the context o f the child
survival literature and critically examines current knowledge about practices, outcomes and
interventions. This knowledge is found to be limited, particularly in terms o f practices and
outcomes in rural settings in developing countries.
Methods: Two methods are described to document perinatal events in villages in
Makwanpur district, Nepal. (1) a census of women o f reproductive age, collecting
information about previous pregnancies and most recent live births; (2) a prospective system
of registration that tracked women through childbirth and infants through the neonatal
period. Both methods yielded information on pregnancies, birth outcomes, care practices and
health care seeking patterns.
Results: The census collected information from 12,170 women, o f whom 4867 had given
birth in the preceding two years. The prospective surveillance collected information on 3522
pregnancies over two years. The thesis presents results under two broad themes: (1) a
description o f the birth experience o f women in rural Makwanpur, its outcomes, practices
and care seeking patterns; (2) a comparison of the two methods of data collection.
Poverty was the norm in rural Makwanpur and only a quarter o f participants were literate.
30% of pregnant women had any antenatal care, which tended to be both late and limited,
and 95% gave birth at home. Three percent of women were helped by trained health workers
and there were compromises in hygiene and warmth at the time o f delivery. Breastfeeding
rates were high. The prospective registration process suggested a neonatal mortality rate of
37 per thousand live births, a figure higher than that from the census (27 per thousand), and
thus higher than would be estimated by existing methodologies.