The setting and research design
The study was carried out in León municipality, a city located in the Pacific part of Nicaragua 93 km from the capital. It has a population of 174 051 inhabitants (2005), where 80% live in urban areas, and 46% of the population is under 18 years of age [
18]. León municipality is considered the second most important city in Nicaragua after the capital with a university hospital, which it is a referent hospital for the closest cities. According to the Human Development Report, in 2002 the human development index in Nicaragua ranked from 0.37 to 0.82 [
19]. León municipality had a human development index of 0.74 a value shared for most cities located close to the capital. From 2002 and onwards a Health and Demographic Surveillance System (HDSS) has been running, updating demographic information every six to twelve months and providing prospective information on pregnancies, births and child survival until the end of the current study period December 31, 2005 [
20]. This HDSS was based on a linkage on household and individual level to a cross sectional study data base. The data in this data base was collected from a 22% sample of the total population in León municipality selected in 1993, using a cluster-sampling technique described elsewhere [
14]. The follow-up survey was performed in 2002, providing updated and linked information on demographic characteristics and selected health indicators. This 2002 HDSS data base included a total of 54 647 people residing in 10 994 households. Only eleven of the households (0.1%) declined to participate in the survey.
Female field workers performed the data collection and interviewed women of reproductive age in their homes. Quality control routines were realized during fieldwork and data entry.
Women of reproductive age (15-49) were interviewed with respect to their reproductive and birth histories including dates of births, outcome of each pregnancy and, when relevant, the date of the child's death. Data collection was supported by the use of a local events calendar. Data were also collected on educational level of the mother, age at delivery, parity, and domicile. However, for place and attendance at delivery data was collected only from the start of the surveillance (2003) and only for women resident in the study area.
Under-five mortality rate was defined as the number of deaths before five years of age per 1000 live births. The U5MR was disaggregated into neonatal (< 28 days), post neonatal (28 days-11.9 months) and child mortality (1-4.9 years). Mother's age and parity were calculated for the time of each childbirth. Parity was categorized into three levels; one, two to four, and more than four live births, respectively. Place of delivery was categorized as hospital, health centre, private clinic, home or other place. Attendant at delivery was classified into physician, nurse, midwife or other person. Women's educational levels were classified into no formal education (either illiterate or literate but had not completed primary school) or formal education (completed primary school or higher level of education). Domicile of the mother was defined as urban or rural.
Analyses
Place of and attendance at delivery during the last 3-year period were analysed and differences between rural and urban areas were tested (Chi2). Data on births, deaths and neonatal, post-neonatal and child mortality rates were analysed for 3-year periods from 1970 to 2005. Mortality curves were prepared with 3-year moving averages in order to smooth random fluctuations in the material. The U5MRs for rural and urban areas were graphically represented. Comparisons of rates or proportions were supported by Chi2 tests. Relative risk of neonatal death (Hazards ratios with 95% confidence intervals) and relative risk of under-five death were analysed by Cox proportional hazard analyses, considering educational level of the mother during three different time periods from 1970 to 2005. Adjustments were done for parity and domicile. Data entry and managing were done using Microsoft Access 2000 (Microsoft Corporation, Washington), while analyses were performed in Statistical Package for Social Sciences (Version 14.0; SPSS Inc, Chicago, USA).