Menu
Menu
Menu
Menu
Background: Maternal health service coverage in Kenya remains low, especially in rural areas where 63% of women deliver at home, mainly because health facilities are too far away and/or they lack transport. The objectives of the present study were to (1) determine the association between the place of delivery and the distance of a household from the nearest health facility and (2) study the demographic characteristics of households with a delivery within a demographic surveillance system (DSS). Methods. Census sampling was conducted for 13,333 households in the Webuye health and demographic surveillance system area in 2008-2009. Information was collected on deliveries that had occurred during the previous 12 months. Digital coordinates of households and sentinel locations such as health facilities were collected. Data were analyzed using STATA version 11. The Euclidean distance from households to health facilities was calculated using WinGRASS version 6.4. Hotspot analysis was conducted in ArcGIS to detect clustering of delivery facilities. Unadjusted and adjusted odds ratios were estimated using logistic regression models. P-values less than 0.05 were considered significant. Results: Of the 13,333 households in the study area, 3255 (24%) reported a birth, with 77% of deliveries being at home. The percentage of home deliveries increased from 30% to 80% of women living within 2km from a health facility. Beyond 2km, distance had no effect on place of delivery (OR 1.29, CI 1.06-1.57, p = 0.011). Heads of households where women delivered at home were less likely to be employed (OR 0.598, CI 0.43-0.82, p = 0.002), and were less likely to have secondary education (OR 0.50, CI 0.41-0.61, p < 0.0001). Hotspot analysis showed households having facility deliveries were clustered around facilities offering comprehensive emergency obstetric care services. Conclusion: Households where the nearest facility was offering emergency obstetric care were more likely to have a facility delivery, but only if the facility was within 2km of the home. Beyond the 2-km threshold, households were equally as likely to have home and facility deliveries. There is need for further research on other factors that affect the choice of place of delivery, and their relationships with maternal mortality. © 2014Mwaliko et al.; licensee BioMed Central Ltd.
This study was conducted using data from the Webuye Health and Demographic Surveillance System. The DSS is located in Bungoma County of the former Western Province, and is approximately 400 km west of Nairobi. The study site is an area approximately 24km from north to south, and 2–6km east to west. The total area is 130km2 with a population of about 77,000 people living in 13,333 households. About 61% of the population lives below the poverty line, and social amenities like water and electricity are not readily available to the majority. There is one 100-bed mission hospital within the study area and one 200-bed district hospital adjacent to the study area, both offering comprehensive emergency obstetric care. There are also several dispensaries, staffed by nurses and offering outpatient care, and one health center offering 24-hour delivery services but without the capacity to perform cesarean sections. This was a cross-sectional community-based study using data obtained from the Webuye health and demographic surveillance system (HDSS) database between 2008 and 2009. Each household was geo-referenced using the Global Positioning System (GPS). The study included all households within the Webuye HDSS that were registered during the baseline and subsequent censuses, and had reported at least one birth within one year preceding the census. Data were collected via structured interviews with the assistance of trained field assistants. The contents of the interview schedules were adapted from the standard INDEPTH [9] questionnaires developed by various HDSS sites. Various stakeholders in the surveillance activities met to discuss key contents of the questionnaires, modified some of the existing questions and designed new questions to reflect the local situation. The questionnaires were further refined after a pilot study prior to the distribution of the final versions to the field staff. All household data were collected via interviews with the head of the household and from GPS coordinates of each household; therefore, we present data of the women’s immediate environment (household) rather than her individual characteristics (Table 1). Descriptive statistics The household questionnaire gathered basic information from the head of the household on usual members of and visitors to the household, including age, sex, education level, and relationship to the head of the household. Information was also collected on deliveries that had occurred during the previous 12 months and socio- economic characteristics of the household’s dwelling unit, such as the source of water, property ownership and possession of mosquito nets. Digital coordinates were also collected for the households and sentinel locations such as health facilities using GPS units. Completed questionnaires were first checked in the field by the field supervisors for completeness. The questionnaires were then sent to the field office where data- quality checkers reviewed the forms for completeness, logic and consistency. The incorrectly filled questionnaires were returned to the respective field interviewers for correction. The correctly filled questionnaires were passed over to the data entry clerks for data entry. After data entry, questionnaires with questionable records identified through automated internal consistency checks were sent back to the field interviewers for verification and correction. The data were stored in a Mysql database (Mysqlab Inc., Uppsala, Sweden). All data were organized and analyzed using STATA version 11 (StataCorp, 2011). Distance from households to health facilities was calculated as Euclidean distance using WinGRASS version 6.4. Hotspot analysis was done in ArcGIS using Hotspot Analysis within the Spatial Statistics toolset to detect clustering of facility deliveries. The demographic and baseline outcomes were recapitulated using descriptive summary measures expressed as the sum, mean, median and standard deviation for continuous variables and percentages for categorical variables. Unadjusted and adjusted odds ratios were estimated using logistic regression models. P-values less than 0.05 were considered significant. Three multivariate models using different covariates to describe access to facilities were explored. Model 1 included distance to any facility as a continuous measure and the type of nearest facility, Model 2 categorized distance to the nearest facility using a threshold, and Model 3 categorized distance to the hospital using a threshold. The best model was selected using Akaike Information Criterion (AIC, Additional file 1: Table S1). The study received ethical clearance from the joint Institutional Research and Ethics Committee of Moi University and Moi Teaching and Referral Hospital. Clearance certificate number IREC/2008/05 (for the period 24th April 2008 to March 2009) was obtained before commencement of the data collection.
