• We are sorry, but NCBI web applications do not support your browser and may not function properly. More information
Logo of amjphAmerican Journal of Public Health Web SiteAmerican Public Health Association Web SiteSubmissionsSubscriptionsAbout Us
Am J Public Health. 2006 January; 96(1): 84–93.
PMCID: PMC1470452

Contextual Influences on the Use of Health Facilities for Childbirth in Africa


Objectives. Previous studies of maternal health-seeking behavior focused on individual- and household-level factors. We examined community-level influences on the decision to deliver a child in a health facility across 6 African countries.

Methods. Demographic and Health Survey data were linked with contextual data, and multilevel models were fitted to identify the determinants of childbirth in a health facility in the 6 countries.

Results. We found strong community-level influences on a woman’s decision to deliver her child in a health facility. Several pathways of influence between the community and individual were identified.

Conclusions. Community economic development, the climate of female autonomy, service provision, and fertility preferences all exert an influence on a woman’s decision to seek care during labor, but significant community variation remains unexplained.

According to the World Health Organization, approximately half a million women die each year from complications of pregnancy or childbirth. The level of maternal mortality is disproportionately high in Africa, with a regional maternal mortality ratio of 1000 per 100 000 live births.1 Most maternal deaths occur during childbirth, and the presence of trained medical staff could substantially reduce the number.2 Therefore the need exists to understand the factors that encourage childbirth in a health facility attended by a trained medical professional.

Previous studies have approached this issue by examining individual- and household-level influences on the decision to deliver a child in a medical institution,35 but the role of community factors has been largely ignored. In recent years, there has been growing recognition of the importance of contextual influences on health outcomes; in particular, several studies have found significant effects of community-level factors on reproductive health outcomes.69 Furthermore, the application of multilevel modeling techniques has shown that spatial variations in reproductive health outcomes remain after control for individual- and household-level factors.10,11

We examined the influence of individual-, household-, and community-level factors on the decision to deliver a child in a health facility in 6 African countries. We build upon previous studies by including the role of the community in our analysis.


Community Influences on Health

Recognizing that the determinants of an individual’s health extend beyond individual and household risk factors, recent studies have examined community influences on health outcomes.1223 Such studies relate individual health outcomes to the socioeconomic characteristics of the community (e.g., levels of economic development) and the community health infrastructure (e.g., presence and quality of health services). The development of multilevel modeling techniques has created a mechanism for measuring the influence of community factors and unobserved community effects on health outcomes, while providing a robust method for analyzing hierarchically clustered data.2427 The incorporation of the role of the community in the analysis of health outcomes provides an opportunity to highlight health risks associated with particular social structures and community ecologies, providing a policy tool for the development of public health interventions.24,28

The growing interest in community influences on health arises from the recognition of a disjuncture between theory and research practice.12 For example, in the context of contraceptive behavior, a number of theories have hypothesized the influence of the community on a couple’s fertility decisions, yet until recently, studies of contraceptive use dynamics focused on individual determinants. The incorporation of community-level factors into multilevel models of fertility behavior has shown how community development, attitudes, norms, and availability of health service influence fertility.21,29 This new dimension in the study of health outcomes provides a means for substantiating theories of community influences on health, while observing interactions between the individual and the community.

In the context of reproductive health, studies of community effects have focused predominantly on fertility behavior,1214,21,29 although some studies have examined maternal health outcomes.6,18 Results indicate that the presence and quality of reproductive health services,13,14,21 levels of economic development,14,21 levels of school participation,21 economic roles of children,6,29 and community fertility norms30,31 all influence individual reproductive behavior. These studies have examined community influences on reproductive health in single country settings, but there is a dearth of studies that have quantified and compared community influences on health outcomes across several countries. Additionally, such studies have focused on one aspect of the community as an influence on health outcomes.

Childbirth in a Health Facility

The choice of place of delivery has consistently been found to be associated with maternal and neonatal outcomes.3234 Childbirth in a medical institution attended to by trained medical staff has been shown to be associated with lower rates of maternal and neonatal mortality and morbidity than home births.2,35,36 Given the demonstrated health benefits of institutional deliveries, it is necessary to understand the range of factors associated with the decision to seek care during delivery, and to understand the role that the community has in influencing this decision.

Previous studies have focused largely on the barriers and facilitators in the decision to seek health care. Studies of health care use have highlighted a range of potential influences on a woman’s propensity to seek care. Demographic factors that have been shown to increase the likelihood of health service use are low parities,6,37,38 younger maternal age,37 women’s employment in skilled work outside the home,39,40 and high levels of husbands’ education.41

Socioeconomic factors, however, have been shown to be of greater importance in determining health service use than demographic factors.42 The most consistently found determinant of use of reproductive health services has been a woman’s level of education.6,37,39,41 Obermeyer believes that increased education influences service use by increasing female decisionmaking power, increasing awareness of health services, changing marriage patterns, and creating shifts in household dynamics.43 Cost has often been shown to be a barrier to service use44 and also influences the source from which care is sought. Socioeconomic indicators such as urban residence,39 household living conditions,44 household income,38 and occupational status41 have also proven to be strong predictors of a woman’s likelihood of using reproductive health services.

Both demographic and socioeconomic determinants of use of reproductive health care are mediated by cultural influences on health-seeking behavior that shape the way individuals perceive their own health and the health services available.45,46 These community beliefs and norms are reflected in an individual’s health decisions; behavior is influenced by how a person thinks the community views his or her actions.46 For example, traditional beliefs about childbirth, coupled with misconceptions and fears of medical institutions, have led many women to maintain reliance on home births in India.45 Results from a study in Benin found that women giving birth unassisted were silently admired,47 and in West Africa childbirth is considered a woman’s battle.48 Thus, although demographic and socioeconomic factors are key determinants of health service use, the individual’s cultural environment provides a strong influence on the extent to which these factors can lead to the use of health services.

Previous studies have highlighted a range of factors associated with a woman’s decision to seek care during labor; however, the role of the community has been largely ignored. There are several pathways through which a community could influence an individual’s health. Community beliefs and norms relating to health behaviors are a strong influence on the health care decisions made by individuals.46 The level of community economic development may influence health directly, through an association between deprivation and poor health,49 and indirectly through access to health services and social support systems.50 Economic development is also positively related to health outcomes through its relationship with increased female decisionmaking power, an increased likelihood of female labor force participation, and positive attitudes toward health service use.12



The 6 African countries selected for analysis were divided into 2 regions: West (Ivory Coast, Burkina Faso, and Ghana) and East (Kenya, Malawi, and Tanzania). The selection of neighboring countries allowed identification of spatial variations in childbirth at a health facility that may transcend political boundaries. The selection of the countries was also informed by the availability of Global Positioning System data, which allowed the sampled communities to be mapped.

The data used in this analysis are from the Demographic and Health Surveys (DHSs) conducted in the 6 study countries (Kenya, 1998; Malawi, 2000; Tanzania, 1999 [the data set for Tanzania is the Reproductive and Child Health Survey 1999]; Burkina Faso, 1998; Ivory Coast, 1998; and Ghana, 1998). These surveys used a stratified multistage cluster sample design to collect nationally representative samples of women of reproductive age (15–45 years). Questionnaires were conducted with all eligible women in each sampled household; data on fertility, family planning, and health care seeking during pregnancy were collected, in addition to demographic and socioeconomic data. (Full descriptions of the study designs used in each country can be found at the Measure DHS Web site at http://www.measuredhs.com.) In addition, each of these data sets collected Global Positioning System locators for each of the primary sampling units (PSUs) included in the samples.51

The sample for each country represents all women of reproductive age covered in the national survey who had given birth in the past 3 years. Sample sizes are as follows: Burkina Faso, 3167; Ghana, 1785; Ivory Coast, 1131; Kenya, 3058; Malawi, 6318; and Tanzania, 1710. The Demographic and Health Survey data provide the individual- and household-level data for the analysis. Two approaches were used to obtain community data. Some community factors were taken from the Demographic and Health Survey data; this entailed averaging individual data to the PSU level (the PSU denotes the community in this analysis), thus producing derived community-level factors. Secondly, data were obtained from the following Geographic Information System sources: rainfall (UN Environment Program), habitat (ArcView World Map [ESRI International, Redlands, Calif]), and road and rail network data (Digital Chart of the World, ArcView 8.2 [ESRI International]). International and subnational boundary data were obtained from the African Population Database (National Center for Geographic Information and Analysis, University of California, Santa Barbara).


The dependent variable for analysis is a binary variable coded 1 if the woman delivered her last child in a health facility (including public and private facilities) and coded 0 if she delivered at home. Each of the Demographic and Health Survey data sets has a hierarchical structure, with women “nested” within households and households within PSUs, thus violating the assumption of independence of ordinary logistic regression models. A multilevel modeling technique was employed to account for the hierarchical structure of the data and to facilitate the estimation of community-level (PSU and district) influences on the outcome. The multilevel modeling strategy accommodates the hierarchical nature of the data and corrects the estimated standard errors to allow for clustering of observations within units.26

Multilevel models allow the identification of clustering in the outcome (also known as the random effect), which represents the extent to which the outcome of interest varies between each level of interest (PSU or district). A significant random effect may represent factors influencing the outcome variable that cannot be quantified in a large-scale social survey. A random effects model thus provides a mechanism for estimating the degree of correlation in the outcome that exists at the community level (PSU or district), while also controlling for a range of individual- and household-level factors that may potentially influence the outcome. Separate multilevel logistic models were fitted for each of the 6 countries through the MLwiN software package.52 Two levels of variance were considered, the PSU and the district. The models are written:

equation M1

where logeijk/(1 – πijk)) =α+βXTIJK + UJK + VK, Yijk , is a binary outcome (use of health facility at childbirth) for individual i in PSU j in district k, Yijk are assumed to be independent Bernoulli random variables with the probability of use of health facility for childbirth πijk =Pr(Yijk = 1). Consequently, to correctly specify the binomial variation, Zijk denotes the square root of the expected binomial variance of πijk and the variance of the individual residual term epsilonijk is constrained to be one. The outcome variable logeijk/(1 – πijk)) fitted in the model is the loge odds of use vs nonuse. This constrained the predicted values from the model to be between 0 and 1. α is a constant, whereas β is the vector of parameters corresponding to the vector of potential explanatory factors defined as χijk. The PSU (level-2) residual term is defined as Ujk~N(0,σ2 u) and the district (level-3) residual term is defined as Vk~N(0, σv2).

The variables to be entered into the models were grouped into individual/household and community variables; the same independent variables were entered into the models for all 6 countries. The choice of individual and household independent variables was informed by previous studies on the factors influencing the decision to deliver a child in a health facility: maternal age, parity, marital status, place of residence, education, religion, previous exposure to health services and media messages, and an index of household amenities. The index included the household drinking water source, toilet facility, and flooring material. Table 1 [triangle] shows the variables used in the final models.

Independent Variables Used in Modeling of a Woman’s Decision to Give Birth in a Health Facility in 6 African Countries

Several contextual influences on delivery in a health facility were considered in the original analysis design. It was intended to measure the health service environment, physical infrastructure, and prevailing cultural beliefs surrounding health care seeking. However, it proved difficult to obtain contextual data from many of the study settings, and indicators were often measured differently across the 6 countries, limiting the availability of standardized contextual data. Some contextual variables that were available for all countries proved not to be significantly related to the outcome; these were the transport infrastructure in the district (kilometers per 1000 km2), habitat type (forest/grassland), the predominant religion in the community, and the percentage of women in the community who desired to be tested for HIV. We thus present only the contextual variables that were significant in at least 1 of the models.

The contextual variables used in the final models were as follows. The mean number of children per woman in each PSU was used as a measure of traditional “pronatalist” community attitudes (i.e., in favor of large numbers of children). The percentage of husbands in the PSU who approve of family planning and the percentage of women in the PSU with secondary or higher education were entered to measure community attitudes toward women’s roles. The mean number of women in each PSU who have delivered at least 1 previous birth in a health facility was used as a measure of both the presence of services and the community attitudes toward the use of health services. Finally, the rainfall category of each PSU is used as an approximation of accessibility to services, with the assumption that services in places with poor infrastructures are harder to access during times of high rainfall.


Tables 2 [triangle] and 3 [triangle] show the results of the multilevel logistic models of the decision to deliver a child in a health facility in the 6 African countries. In terms of individual variables, few variables proved to be significantly associated with the decision in all 6 countries. The age of the respondent showed a significant association with the outcome in Malawi, Tanzania, and Kenya, but not in Burkina Faso, Ivory Coast, or Ghana. Relative to women aged 20 to 29 at the time of the survey, in Malawi and Tanzania, women of all age groups were more likely to have delivered their last child in a health facility. In Kenya, women aged 30 to 39 and women aged 40 to 49 were more likely than women aged 20 to 29 to have done so. Relative to women who had given birth once or twice, women at all higher parities were less likely to have delivered their last child in a health facility in all countries except Burkina Faso and Ivory Coast.

Multilevel Modeling of a Woman’s Decision to Give Birth in a Health Facility in 3 East African Countries
Multilevel Modeling of a Woman’s Decision to Give Birth in a Health Facility in 3 West African Countries

Mixed results were found in the association between a mother’s education and her decision to deliver her last child in a health facility. In Malawi, Kenya, Burkina Faso, and Ghana, women with a secondary education or higher were more likely than women with no education to have delivered in a health facility. However, in Tanzania, only women with primary education were more likely than women with no education to have done so, and there was no association with maternal education in Ivory Coast. With the exception of Burkina Faso, there was a linear relationship between the household amenities index and delivery in a health facility in all countries.

In Ivory Coast, women in polygamous marriages and women who were separated were less likely to report having their most recent child in a health facility than women in monogamous marriages; in Kenya, the same was true of women who were never married and women who were separated. In Ghana, Muslim women or women reporting another religion were less likely than Catholic women to report delivering their latest child in a health facility, whereas in Kenya and Tanzania, Protestant women were more likely than Catholic women to report having done so. Women who reported having been exposed to family planning messages in the media were more likely to report delivering in a health facility in Malawi, Kenya, and Tanzania.

Urban residence increased the likelihood of a woman reporting delivery of her latest child in a health facility in Malawi, Tanzania, and Ghana. The only individual variables that were significantly associated with delivery in a health facility in all countries were receipt of prenatal care during the last pregnancy and previous delivery in a health facility. Relative to women who had received 1 to 3 pre-natal care visits during their last pregnancy, women who had received 4 or more visits showed an increased likelihood of also reporting that their last child was born in a health facility in all 6 countries. Women who had no prenatal care were significantly less likely to report delivery in a health facility in Malawi, Kenya, Tanzania, Ghana, and Ivory Coast. In all countries, women who had delivered a previous child in a health facility were significantly more likely to report delivering their most recent child in a health facility.

The results for the contextual variables varied across the 6 countries. A significant positive association between the percentage of women in the PSU with secondary education or higher was found in Malawi, Kenya, and Ghana. In Malawi, Tanzania, Burkina Faso, Ghana, and Ivory Coast, a significant positive association was also found regarding the percentage of women in the PSU who had delivered at least 1 previous birth in a health facility. In Tanzania and Kenya, the percentage of husbands in the PSU who approved of family planning showed a significant positive association with a report that the last child was delivered in a health facility. The mean number of children per woman in the PSU showed a significant negative association with the outcome in Tanzania. In Kenya, a significant association was found between the mean rainfall of the PSU and the woman’s odds of reporting that her last child was delivered in a health facility.

Despite the inclusion of the individual and contextual variables, there was significant PSU-level variation in all 6 countries (Malawi, β = .27, SE = 0.04; Tanzania, β = .37, SE = 0.13; Kenya, β = .49, SE = 0.10; Burkina Faso, β = .65, SE = 0.14; Ghana, β = .54, SE = 0.15; Ivory Coast, β = .71, SE = 0.22). The district-level variation, however, remained significant only in Kenya (β = .40, SE = 0.13) and Tanzania (β = .28, SE = 0.13). Thus, the variables included in the models successfully explain the district-level variation in reporting delivery of the most recent child in a health facility in 4 of the 6 countries, but they are not successful in explaining variation at the lower-level PSU in any of the 6 countries. When individual-level variables are included in the models, there is significant variation at both the PSU and district levels, whereas the inclusion of the contextual variables acts to significantly decrease the degree of variation at both levels.

The maps in Figure 1 [triangle] show the variations in health facility use for childbirth in East and West Africa. The maps on the left show the weighted raw data, representing the percentages of women in each district who reported delivering their latest child in a health facility. In all 6 countries, there was a substantial variation in the outcome, although in West Africa there was less variation than in East Africa.

Actual and predicted levels (%) of use of health facilities for childbirth in 6 countries in West (Ivory Coast, Burkina Faso, and Ghana) and East (Kenya, Malawi, and Tanzania) Africa.

The maps on the right plot a comparison between actual use and the level of use predicted by the models. The level-3 (district-level) residuals are used to calculate the predicted level of use in each district given the individual variables included in the models. The residual variation is then calculated by comparing actual levels of use from the Demographic and Health Survey data with the level of predicted use; lower than predicted use in a district means that the level of use of health facilities for childbirth is more than 1.96 standard deviations lower than that predicted by the modeling process. The maps show the remaining variation unaccounted for by individual-level factors that are mostly accounted for by the community factors. In East Africa, areas of central Kenya, the Shihyanga and Dar es Salaam provinces of Tanzania, and Mzimba and Bilantyre provinces of Malawi show higher use of health facilities for childbirth than predicted by the individual factors. Conversely, parts of central Malawi, southern Tanzania, and Kenya display lower than predicted use. In West Africa, Burkina Faso and Ghana display generally higher levels of health service use than predicted by the individual factors.


The results demonstrate the impact of individual-, household-, and community-level influences on the decision to deliver a child in a health facility. The significant individual-level factors reflect relationships identified in previous studies. Maternal age, parity, educational status, religion, and marital status were all influential in a woman’s decision to deliver her last child in a health facility, although the pattern and magnitude of these relationships varied across the 6 countries, indicating geographic and cultural variations in the pathways through which these variables influence health behavior. The significance of urban residence in 4 countries highlights the benefits of greater service availability afforded to urban residents. Two variables were consistently related to the decision to deliver a child in a health facility in all 6 countries: receiving prenatal care and delivering a previous child in a health facility. The latter demonstrates a simple relationship: women who have delivered a child in a health facility are the most likely to continue to deliver future children in health facilities, irrespective of maternal age and parity.

The significant effect of prenatal care points to the role that care during pregnancy has in informing women of the benefits of institutional deliveries and in connecting women to appropriate services. The result also highlights a selectivity effect: the characteristics that predispose women to seek pregnancy care also make them more likely to seek care during labor. There is obviously an influence of previous exposure to maternal health care services on a woman’s decision to seek care during pregnancy, suggesting that other reproductive health services can be used as an opportunity to inform women of the benefits of institutional deliveries. The variables measuring previous exposure to maternal health services are also likely to reflect the availability of such services in the community.

The main focus of this research is on the role of community-level factors on the decision to deliver a child in a health facility, and the results point to several pathways through which the community can influence individual behavior. The significant effect of the percentage of women in the community with secondary education and higher in Malawi, Kenya, and Ghana points to 2 potential pathways of influence: the role of community economic development and the influence of community attitudes on female roles.

In less developed societies such as those analyzed here, levels of female education are often low, and the attainment of secondary education or higher often reflects higher socioeconomic status. Communities in which a higher percentage of women are achieving these levels of education are therefore likely to be communities with higher percentages of socioeconomically advantaged households. Greater household wealth may enable women to seek care during pregnancy, with the costs of seeking care acting as a significant barrier to women from poorer households. Higher levels of female education in the community may also point to greater awareness of the need for care during childbirth. Although the content of formal education may not include health information, higher levels of education may create a greater awareness of health services and the need for care.

In more traditional societies, higher levels of female education may also indicate greater female autonomy, as education is often restricted to male children, and earlier ages at marriage and childbearing may restrict female access to higher levels of education. The positive association between the percentage of husbands in the community who approve of family planning and a woman’s decision to deliver her child in a health facility also highlights the influence of female autonomy on health behavior. High levels of approval of family planning are associated with less conservative communities, which may also be less conservative in their attitudes toward women’s roles. Hence, women living in communities with higher levels of female education and approval of family planning may also be living in climates of greater autonomy, allowing them greater decision-making power and the opportunity to seek care during pregnancy and labor. The significance of education at the individual and community levels suggests the importance of both individual autonomy and the climate of autonomy that exists in the community. It also suggests that the influences on individual health behavior extend to beliefs and practices of others in the community.

In Malawi, Tanzania, Ghana, Burkina Faso, and Ivory Coast, the percentage of women in the community who had delivered a child in a health facility had a strong positive influence on a woman’s decision to seek care. There are several possible pathways of influence. The high percentage of women in the community who had delivered their child in a hospital may simply reflect the presence of maternal health services in the community. Data were not available to measure the actual presence of health services, so this variable may be acting as a proxy for service availability. Previous studies have shown that women’s decisions regarding health seeking are strongly influenced by the practices of others in the community53; in a community in which a high percentage of women are using health services for childbirth, the practice is therefore likely to be seen as a norm, influencing individual behavior.

A high mean number of children per woman in a community had a negative influence on a woman’s decision to deliver her child in a health facility. Communities with higher fertility may be more conservative in their attitudes toward service use and the expected roles of women, and may have a lower level of economic development, all of which influence a woman’s ability to seek care during labor. High fertility may also reflect a lack of reproductive health services and a lack of awareness of such services, both of which have obvious implications for maternal health service use.

After control for individual, household, and community factors in the models, significant variation in the outcome still exists at the PSU level in all 6 countries, indicating that the models do not fully explain the community-level variation in the decision to deliver a child in a health facility. This residual variation, sometimes known as unobserved heterogeneity, may represent 1 of 2 sets of factors, or a combination of the 2: factors omitted from the models or factors that cannot be measured in a large-scale social survey. Factors influencing delivery in a health facility that are difficult to capture in a survey include cultural influences on service use, which may vary across the 6 study settings. Community levels of education, approval of family planning, and fertility measures have captured some of these cultural influences, but others, such as traditional views on childbearing, are harder to record in a survey. The residual variation may also reflect omitted factors, the most obvious being the presence of maternal health services in the community.

Other factors that were not measured, but that may have helped to reduce this variation, include the type and quality of health services and the financial accessibility of services. The presence of persistent residual variation at the PSU level should act as a motivation for agencies responsible for data collection to improve the quality and breadth of community-level data collection. Specific areas of improvement are increased data on health services (including their presence and quality) and measures of community attitudes and practices regarding childbirth.

The models were, however, more successful in explaining district-level residual variation, with significant variation remaining only in Kenya and Tanzania. It seems that the contextual variables chosen are more appropriate for explaining larger area variations, and that more research is needed to understand the factors influencing health behavior at the local community level. The maps shown in Figure 1 [triangle] show that the actual levels of service use do not always match those predicted by the individual factors, indicating that although they act to highlight some of the main determinants of service use, they do not necessarily capture the range of community factors acting to influence the decision to deliver a child in a health facility.

The results highlight several important implications for policymakers and program managers. The identification of several community-level characteristics that are significantly associated with a woman’s decision to deliver her child in a health facility highlights the potential for such factors to be harnessed for the development of public health interventions that aim to increase service use. The range of community factors identified and their variation across the study settings demonstrate that any such interventions must be context specific, and should reflect the characteristics and dominant influences present in the community. The methodology used in this research can be applied to other health outcomes, and it provides policymakers and researchers with an opportunity to incorporate existing data sources


Our analysis, which highlighted the range of influences on a woman’s decision to deliver her child in a health facility, found many similarities in these influences across 6 African countries. A range of community-level influences have been identified. These influences have been surprisingly similar across the 6 settings, illustrating the influence that indicators of community-level socioeconomic development, female autonomy, and fertility norms have on individual health-seeking behavior. There is, however, sufficient variation in the significant community-level variables between the 6 countries to suggest the need to examine the local cultural context when identifying community-level interventions. The persistence of significant community-level variation in the outcome illustrates 2 points: current social surveys are insufficient in measuring the range of cultural influences on health-seeking behavior, and more research is needed to understand the dynamics of community influences on individual health.

From a methodological perspective, this research has incorporated existing social survey data with contextual data to provide a fuller understanding of the determinants of the decision to use a health facility for childbirth, and it has illustrated the role that Global Positioning System–linked data has in explaining health behavior. However, the difficulties experienced in obtaining standardized contextual data from each of the countries point to the need to improve community-level data collection techniques. From a public health perspective, this research has shown that the community plays an important role in shaping an individual’s health behavior; although we can identify some of the pathways of influence, there remains much to be learned of community-level influences on health seeking in the contexts studied here.

The latter illustrates the need for existing data sources to include more comprehensive community-level data and for measures of community norms and practices to be included in the analysis of individual health-seeking behavior. Community-level influences on health-seeking behavior can be harnessed to develop community-level health interventions. This research has extended previous studies by incorporating Global Positioning System data into this process, but more research is needed to understand the existing residual variation in a woman’s decision to deliver her child in a child facility.


This study was funded by the Economic and Social Research Council, United Kingdom (award no. R000239664).


Peer Reviewed

R. Stephenson and S. Clements designed and conceptualized the study. A. Baschieri and S. Clements conducted the data analysis, with input from R. Stephenson, M. Hennink, and N. Madise. Rob Stephenson wrote the final article, with input from A. Baschieri, S. Clements, M. Hennink, and N. Madise. All authors assisted in interpreting the results.


1. Maternal Mortality in 1995: Estimates Developed by WHO, UNICEF, UNFPA. Geneva, Switzerland: World Health Organization; 2001:56.
2. Howlader AA, Kabir M, Bhuiyan MM. Health-seeking behaviour of mothers and factors affecting infant and child mortality. Demography India. 1999; 28(2):225–238.
3. Amooti-Kaguna B, Nuwara F. Factors influencing choice of delivery sites in Rakai district of Uganda. Soc Sci Med. 2000;50:203–213. [PubMed]
4. Griffiths P, Stephenson R. Understanding users’ perspectives of barriers to maternal health care use in Maharashtra. J Biosoc Sci. 2001;33:339–359. [PubMed]
5. Kyomuhendo GB. Low use of rural maternity services in Uganda: impact of women’s status, traditional belief and limited resources. Reprod Health Matters. 2003;11(21):16–26. [PubMed]
6. Magadi MA, Madise NJ, Rodrigues RN. Frequency and timing of antenatal care in Kenya: explaining the variations between women of different communities. Soc Sci Med. 2000;51:551–561. [PubMed]
7. Pebley AR, Goldman N, Rodriguez G. Prenatal and delivery care and childhood immunization: do family and community matter? Demography. 1996; 33(2):213–247. [PubMed]
8. Stephenson R, Tsui AO. Contextual influences on reproductive wellness in northern India. Am J Public Health. 2003;93(11):1820–1829. [PMC free article] [PubMed]
9. Stephenson R, Tsui AO. Contextual influences on reproductive health service use in Uttar Pradesh, India. Stud Fam Plann. 2002;33(4):309–320. [PubMed]
10. Amin S, Basu AM, Stephenson R. Spatial variation in contraceptive use in Bangladesh: looking beyond the borders. Demography. 2002;39(2):251–267. [PubMed]
11. Madise NJ, Diamond I. 1995. Determinant of infant mortality in Malawi: an analysis to control for death clustering within families. J Biosoc Sci. 1995; 27(1):95–106. [PubMed]
12. Grady WR, Klepinger DH, Billy JOG. The influence of community characteristics on the practice of effective contraception. Fam Plann Perspect. 1993;25(1): 4–11. [PubMed]
13. Chacko E. Women’s use of contraception in rural India: a village-level study. Health Place. 2001;7: 197–208. [PubMed]
14. DeGraff D, Bilsborrow RE, Guilkey DK. Community-level determinants of contraceptive use in the Philippines: a structural analysis. Demography. 1997;34(3): 385–398. [PubMed]
15. Manda SOM. Unobserved family and community effects on infant mortality in Malawi. Genus. 1998; 54(1–2):143–164. [PubMed]
16. Matteson D, Burr J, Marshall J. Infant mortality: a multilevel analysis of individual and community risk factors. Soc Sci Med. 1998;47:1841–1854. [PubMed]
17. Anderson R. Mortality effects of community socioeconomic status. Epidemiology. 1997;8:42–47. [PubMed]
18. Pebley AR, Goldman N, Rodriguez G. Prenatal and delivery care and childhood immunization: do family and community matter? Demography. 1996; 33(2):213–247. [PubMed]
19. Congdon P. The impact of area context on long term illness and premature mortality: an illustration of multilevel analysis. Re.g., Stud. 1995;29:327–344.
20. Von Korff M, Koepsell T, Curry S, Dieh P. Multilevel research in epidemiologic research on health behaviors and outcomes. Am J Epidemiol. 1992;135: 1077–1082. [PubMed]
21. Entwisle B, Casterline JB, Sayed AA. Villages as contexts for contraceptive behavior in rural Egypt. Am Sociol Rev. 1989;54(6):1019–1034.
22. Entwisle B, Mason WM. Multilevel effects of socioeconomic development and family planning programs on children ever born. Am J Sociol. 1985;91(3): 616–649.
23. Tsui AO. Community effects on contraceptive use. In: Casterline JB, ed. The Collection and Analysis of Community Data. Voorburg, Netherlands: International Statistical Survey; 1985:77–99.
24. Duncan C, Jones K, Moon G. Context, composition and heterogeneity: using multilevel models in health research. Soc Sci Med. 1998;46(1):97–117. [PubMed]
25. Diez-Roux AV. Investigating neighborhood and area effects on health. Am J Public Health. 2001;91(11): 1783–1789. [PMC free article] [PubMed]
26. Goldstein H. Multilevel Statistical Models. 2nd ed. London, England: Edward Arnold; 1995.
27. DiPrete TA, Forrostal JD. Multilevel models: methods and substance. Annu Rev Sociol. 1994;20: 331–357.
28. Pickett KE, Pearl M. Multilevel analysis of neighborhood socioeconomic context and health outcomes: a critical review. J Epidemiol Community Health. 2001; 55:111–122. [PMC free article] [PubMed]
29. Hirschman C, Guest P. Multilevel models of fertility determination in four south-east Asian countries: 1970–1980. Demography. 1990;27(3):369–396. [PubMed]
30. Dixon-Meuller R. Gender Inequalities and Reproductive Health: Changing Priorities in an Era of Social Transformation and Globalization. Liege, Belgium: International Union for the Scientific Study of Population; 1999. Policy and Research Paper 16.
31. Greenwell KF. Contraceptive method mix menu: providing healthy choices for women. World Health Stat Q. 1996;49(2):88–93. [PubMed]
32. Giri K. Discussion. Int J Gynecol Obstetr. 1995; 50(supp1 2):S43.
33. Tsu VD. Antenatal screening: its use in assessing obstetric risk factors in Zimbabwe. J Epidemiol Community Health. 1994;48:297–305. [PMC free article] [PubMed]
34. Thaddeus S, Maine D. Too far to walk: maternal mortality in context. Soc Sci Med. 1994;38(8): 1091–1110. [PubMed]
35. Luther NY. Mother’s tetanus immunisation is associated not only with lower neonatal mortality but also with lower early childhood mortality. Natl Fam Health Surv Bull. April 1998;10:1–4. [PubMed]
36. Jejeebhoy SJ, Rao SR. Unsafe Motherhood: A Review of Reproductive Health. In: Das Gupta M, Chen LC, Krishnan TN, eds. Women’s Health in India: Risk and Vulnerability. Bombay, India: Oxford University Press; 1995.
37. Bhatia J, Cleland J. Determinants of maternal care in a region of South India. Health Trans Rev. 1995;5: 127–142.
38. Kavitha N, Audinarayana N. Utilization and determinants of selected MCH care services in rural areas of Tamil Nadu. Health Popul Perspect Issues. 1997;20(3): 112–125.
39. Addai I. Demographic and sociocultural factors influencing use of maternal health service in Ghana. Afr J Reprod Health. 1998;2(1):73–80. [PubMed]
40. Estrin DJ. Reproductive behavior is linked to work autonomy, not to employment itself. Int Fam Plann Perspect. 1999;25(1):50–51.
41. Nuwaha F, Amootikaguna B. Predictors of home deliveries in Rakai District, Uganda. Afr J Reprod Health. 1999;3(2):79–86.
42. Obermeyer CM, Potter JE. Maternal health care utilization in Jordan: a study of patterns and determinants. Stud Fam Plann. 1991;22(3):177–187. [PubMed]
43. Obermeyer CM. Culture, maternal health care, and women’s status: a comparison of Morocco and Tunisia. Stud Fam Plann. 1993;24(6):354–365. [PubMed]
44. Bloom SS, Lippeveld T, Wypij D. Does antenatal care make a difference to safe delivery? A study in urban Uttar Pradesh, India. Health Policy Plann. 1999; 14(1):38–48. [PubMed]
45. Basu AM. Cultural influences on health care use: two regional groups in India. Stud Fam Plann. 1990; 21(5):275–286. [PubMed]
46. Rutenberg N, Watkins SC. The buzz outside the clinics: conversations and contraception in Nyanza Province Kenya. Stud Fam Plann. 1997;28(4): 290–307. [PubMed]
47. Sargent C. The politics of birth: cultural dimensions of pain, virtue, and control among the Bariba of Benin. In: Handwerker WP, ed. Births and Power: Social Change and the Politics of Reproduction. Boulder, Colo: Westview Press; 1990:69–79.
48. Diallo AB. “A Tora Mousso Kele La”: A Call Beyond Duty Often Omitted Root Causes of Maternal Mortality in West Africa. New York, NY: UN Development Programme, HIV and Development Programme; 1991.
49. Kriegar N, Rowley D, Herman A Avery B, Phillips MT. Racism, sexism and social class: implications for studies of health, diseases and well-being. Am J Prev Med. 1993;9(6 suppl):82–122. [PubMed]
50. Diez-Roux AV. Bringing context back into epidemiology: variables and fallacies in multilevel analysis. Am J Public Health. 1998;88:216–222. [PMC free article] [PubMed]
51. Montana L, Spencer J. Incorporating Geographic Information Into MEASURE Surveys: A Field Guide to GPS Data Collection. Calverton, Md: Macro International; 2001.
52. Multilevel Modeling Project. Institute of Education, University of London. Available at: http://www.mlwin.com. Accessed November 2004.
53. Stephenson RB, Matthews Z. Uptake of maternal health care services among migrant and nonmigrant groups in Maharashtra, India. Asia Pacific Popul J. 2004; 19(1):39–60.

Articles from American Journal of Public Health are provided here courtesy of American Public Health Association
PubReader format: click here to try


Related citations in PubMed

See reviews...See all...

Cited by other articles in PMC

See all...


  • PubMed
    PubMed citations for these articles

Recent Activity

Your browsing activity is empty.

Activity recording is turned off.

Turn recording back on

See more...