Factors associated with health facility childbirth in districts of Kenya, Tanzania and Zambia: A population based survey

Background: Maternal mortality continues to be a heavy burden in low and middle income countries where half of all deliveries take place in homes without skilled attendance. The study aimed to investigate the underlying and proximate determinants of health facility childbirth in rural and urban areas of three districts in Kenya, Tanzania and Zambia.Methods: A population-based survey was conducted in 2007 as part of the ‘REsponse to ACcountable priority setting for Trust in health systems’ (REACT) project. Stratified random cluster sampling was used and the data included information on place of delivery and factors that might influence health care seeking behaviour. A total of 1800 women who had childbirth in the previous five years were analysed. The distal and proximate conceptual framework for analysing determinants of maternal mortality was modified for studying factors associated with place of delivery. Socioeconomic position was measured by employing a construct of educational attainment and wealth index. All analyses were stratified by district and urban-rural residence.Results: There were substantial inter-district differences in proportion of health facility childbirth. Facility childbirth was 15, 70 and 37% in the rural areas of Malindi, Mbarali and Kapiri Mposhi respectively, and 57, 75 and 77% in the urban areas of the districts respectively. However, striking socio-economic inequities were revealed regardless of district. Furthermore, there were indications that repeated exposure to ANC services and HIV related counselling and testing were positively associated with health facility deliveries. Perceived distance was negatively associated with facility childbirth in rural areas of Malindi and urban areas of Kapiri Mposhi.Conclusion: Strong socio-economic inequities in the likelihood of facility childbirths were revealed in all the districts added to geographic inequities in two of the three districts. This strongly suggests an urgent need to strengthen services targeting disadvantaged and remote populations. The finding of a positive association between HIV counselling/testing and odds in favor of giving birth at a health facility suggests potential positive effects can be achieved by strengthening integrated approaches in maternal health service delivery. © 2014 Ng’anjo Phiri et al.; licensee BioMed Central Ltd.

The survey was conducted in 2007. Selected sites were Malindi in Kenya, Mbarali in Tanzania, and Kapiri Mposhi in Zambia. The sites were selected based on the assumption of their similarities in disease burden and health systems [29], although, the organization of the healthcare system in the three districts appears to differ substantially [30]. The estimated population in the study areas was 350,000 in Malindi, 235,000 in Mbarali and 200,000 in Kapiri Mposhi [31-33]. Malindi is located in the Coast Province along the Kenyan border with the Indian Ocean. It is a major tourist destination, and its main economic activities include fishing, forestry, tourism, and agriculture. Malindi has a crude birth rate of 48 per 1,000, growth rate of 3.9%, and total fertility rate (TFR) of 6.1, with contraceptive acceptance rate of 29% [31]. It has three hospitals (one government and two private) and 24 dispensaries (17 government and seven non-governmental organizations) [31]. Mbarali is situated in the Mbeya region of Tanzania. It is along the major road connecting Mbeya city with Dar es Salaam, and a railway line that joins Kapiri Mposhi in Zambia to Dar es Salaam. The main economic activity is agriculture. The district’s growth rate is 2.8%, and the TFR for the southern region is 4.4 [27]. Mbarali has two hospitals in Chimala and Rujewa, two health centres and 39 dispensaries. Most of the health facilities are supported by health insurance services from government, National Health Insurance Fund [34]. In contrast and similarities, Kapiri Mposhi is in the Central Province and a gateway to the north of the country via road and railway connections. The main economic activity is agriculture. The TFR for Central Province is 6.4 [28]. The crude birth rate is 43 per 1,000 and a growth rate of 2.7% [33]. Kapiri Mposhi has 25 public health centres, two health posts, one private and two mission health centres [35]. A second level referral hospital located about 50 kilometres south of the district in Kabwe, served referrals from Kapiri Mposhi until 2011 when a new hospital was opened within the district. The data stem from a population-based survey employing multi-stage stratified random cluster design. Stratification was by district and within district by rural–urban residence. Rural and urban were defined according to population censuses of 1999, 2002 and 2000 in Kenya, Tanzania and Zambia, respectively [36]. A combination of politico-administrative perspective and human settlements perspective were used with Kenya defining urban as municipalities with 2,000 inhabitants or more; whereas Tanzania defined this by size and density with majority of their inhabitants in non-agricultural occupations; and Zambia defining urban as localities of 5,000 inhabitants or more and a majority of the labour force in non-agricultural activities [36]. Proportions of urban population were 21, 23 and 35% for Kenya, Tanzania and Zambia respectively [26-28]. Standard enumeration areas (SEAs) were used as basic sampling units. First, clusters that corresponded with the SEAs at the district were selected using probability proportional to size. The listing of SEAs provided information on households based on the population census. A total of 49 clusters (Malindi 10, Mbarali 19, Kapiri Mposhi 20) were selected from the urban stratum, and 70 clusters (Malindi 19, Mbarali 26, Kapiri Mposhi 25) from the rural stratum. The second stage involved randomly selecting a fixed number of households from the list that was compiled consisting of all households in the selected SEAs in each district. The aim was to select 2,000 individuals in each district. One male and one female aged between 15 and 49 years of age were randomly selected as participants in each household. This study analysed women respondents who had delivered within five years prior to the study and information obtained about the most recent childbirth. A conceptual framework by McCarthy and Maine [37] was employed to guide analysis. The framework guides analysis of maternal mortality determinants and could be applied to research and programmes [37]. The concept grouped determinants as distant factors that are underlying socio-economic and cultural factors; intermediate or proximate factors, such as healthcare seeking behaviour and use of health services, health status, and access to services, which directly influence pregnancy outcomes of morbidity and mortality. The distal socio-economic and cultural factors are mediated through the health seeking behaviour and access to health care service to result in pregnancy outcome [37]. Considering evidence from studies that relate EmONC services to reduced maternal mortality [38], we found this framework applicable to our study on use of health facilities for childbirth, with the presumption that facilities provide skilled birth attendance and EmONC services. Selection of variables to include in our model was guided by previous studies and the above framework. It was hypothesized that underlying socioeconomic position, age and marital status were associated with health facility childbirth. Women’s educational attainment and wealth status have been associated with health facility childbirth in sub-Saharan Africa and other regions [12-14,39,40]. Single women seem to have more autonomy than married women, and conversely young women may not have the financial capability to access health services [41]. In sub-Saharan Africa, older age has been associated with reduced health facility childbirth [16,42]. It was hypothesized that proximate factors of access to health services measured by perceived distance and perceived cost were associated with reduced health facility childbirth, as observed in previous studies [16-18,43]. We also hypothesized that trust and perceived quality of care, use of health care services during ANC visits and exposure to HIV counselling and testing were associated with increased health facility childbirth. Perceived quality of care has been associated with by-pass of the nearest health facility, and ANC visits with increased use of health facility at delivery [16,18,44-46]. Data was collected in 2007 by trained enumerators using a structured questionnaire. The data collection tools were developed and standardised for application in the three countries within a standard operating procedure for training of staff, and pilot testing of the tools. EpiData version 3.1 was used for data entry. The dependent variable was place of childbirth dichotomised as: home delivery = 0, all health facility deliveries = 1. Only women having given birth the previous five years or less were included in the analysis. Independent underlying variables included socioeconomic position (SEP), age and marital status categorised as single (never married, widowed, separated and divorced), and married (married, cohabitating). SEP was created by summation of wealth index and the woman’s educational attainment in school years. Wealth index was a summation of information on electricity; asset ownership of radio, television, refrigerator, bicycle, plough, donkey, cattle; and type of housing construction material. SEP is seen as a multidimensional and multilevel construct which is partly determined by structural relations. We used educational attainment and wealth index as indicators of SEP with the intention to capture the association between facility childbirth and SEP as part of the assessment for equity. SEP was categorised as low, middle and high to indicate proportions of facility childbirths but retained as a continuous variable in the model. Educational attainment and wealth index were also analysed as separate variables in the model to compare with the model using SEP variable. However, only educational attainment remained as a significant factor. Since the two indicators were highly correlated (r = 0.2), we used the composite variable in analysis. Proximate variables included a composite “trust-quality”, which was created using self-rating of local health services, perceived drug availability and perceived attitudes of the health care staff at the nearest clinic. These three variables were correlated and principal component analysis was used for data reduction. The trust-quality variable was categorized into four groups ranging from ‘very bad’ to ‘very good’ and later condensed the adjacent groups to create two categories, ‘bad’ and ‘good’. The non-categorized variable was retained and used as a continuous variable in the model. Other proximate variables were perceived cost and perceived distance which were categorized using a Likert scale ranging from “not at all” to “very much” and later condensed as “not at all/little”, “fairly” and “much/very much” while maintaining the uncondensed variable as a continuous quantity in the model. The number of ANC attendance and ever tested for HIV (yes, no) were used as proxies to use of health services during pregnancy. ANC attendance was grouped as 0–3 visits, 4 or more visits, and was used as continuous in the model. Data analysis was done using SPSS for Windows Version 19, SPSS Inc. Chicago, Illinois. Descriptive statistics and multivariate analysis were done and complex sample design used to take into consideration the design effect. Stepwise multivariate logistic regression was used to estimate adjusted associations. In step1 only the underlying (or distal) factors were included in the model, whereas in step 2 the proximate factors were also included. The variables selected in the model were those that were found significant in bivariate analysis in any of the three districts. Analyses were stratified by district and by rural–urban residence. Ethical clearance was obtained in Kenya from Kenya Medical Research Institute (KEMRI) and from the National Ethical Review Committee (NERC); in Tanzania from the Medical Research Coordinating Committee (MRCC) of the National Institute of Medical Research (NIMR); and in Zambia the University of Zambia Research Ethics Committee. Written informed consent was obtained from all participants of the population based surveys prior to being interviewed. Confidentiality and anonymity of the study participants was maintained. This study was specifically approved by the Steering Committee of REACT, which is also the project review board including representatives from all three study countries.

N/A

• Dima Health