A systematic review of inequalities in the use of maternal health care in developing countries: examining the scale of the problem and the importance of context
Lale Say a, Rosalind Raine b
Two decades after the launch of the Safe Motherhood campaign in Kenya in 1987, half a million women, most of whom live in developing countries, continue to die from maternal causes each year.1 Key health-care interventions can largely prevent women from dying of pregnancy-related causes. Attendance of antenatal care, delivery in a medical setting and having a skilled health worker at delivery improve maternal health.2–5 However, use of these interventions is limited in developing countries.6
Maternal health-care use is also reported to vary within developing countries, with most findings showing differences between affluent and poor women, and between women living in urban and rural areas.7–9 However, since the methodological quality of these studies has not been assessed systematically, it is difficult to draw conclusions on which to base policy recommendations.
Additionally, factors related to place of residence and socioeconomic status may account for variations in use of maternal health care. These factors include women’s age,10,11 ethnicity,12 education,7,10,13 religion,13,14 culture,15 clinical need for care15 and decision-making power.16 The costs,17 location7,18 and quality of health services are also important.19 These factors interact in different ways to determine use of health care. For example, rural women in northern India and those in KwaZulu Natal, South Africa, do not use antenatal care adequately, but for different reasons. In India, affluent rural women are unwilling to invite health workers into their homes;13 in KwaZulu Natal, women have little time left after attending to essential household tasks.18 When methodologically robust research shows variations in maternal health-care use according to women’s place of residence or socioeconomic status, an understanding of context is essential to design delivery mechanisms to redress such inequalities.
We therefore undertook a systematic review of the use of key maternal health-care interventions in developing countries by women’s place of residence and socioeconomic status. We sought to assess the extent and strength of evidence for variations in use, and to investigate the contextual circumstances of any variations shown in studies of moderate or high quality.
Identification of studies
Four electronic databases (MEDLINE, EMBASE, CAB Direct and POPLINE) were searched by developing search strategies specific to their medical subject headings and text words with the help of an expert librarian. Subject headings or text words related to reproductive and maternal health-care delivery (reproductive health services, reproductive health-care delivery, reproductive health-care utilization, maternal health services, obstetric care, women’s health services, pregnancy complications) were combined with those related to access variables or population characteristics (access, utilization, equity, health service access, equity, inequality, determinant, socioeconomics, income; detailed search strategies are available from the authors on request). Web pages of organizations known to be active in the field and reference lists of retrieved articles were screened for further relevant papers.
We identified population-based cohort studies, case-control studies or cross-sectional studies that were reported in the English language, that were undertaken in developing countries and that reported the effect of the selected non-clinical factors on the use of maternal health care by women aged 15–49 years. A country’s development status was determined by the United Nations classification. The published data had to report, or enable us to estimate, the association between non-clinical factors and use of health care by calculation of the ratio of the frequency of use for one group compared with that for another. If these data were not reported, we had to be able to cite the result of a statistical test for differences between the groups. Therefore, we did not consider qualitative studies of health-care use or health-care-seeking behaviour.
Research published before 1990 or including data from before 1985 was excluded, because the rapid development of maternal health care in the 1990s reduced the relevance of data from the 1980s. If publications of secondary analyses from large national surveys were identified for consecutive years from the same country, only the most recent version was considered. Data for only the most recent year were extracted if information existed for consecutive years within one publication.
Outcome measures, non-clinical factors and need
Outcome measures of interest were use of a skilled health worker at delivery (a medically trained person attending delivery, with or without specification of obstetric training), attendance of antenatal care in the first trimester of pregnancy (studies that reported frequency of antenatal visits without timing were excluded) and delivery in a medical setting (regardless of level of care). Non-clinical factors of concern were place of residence (urban versus rural) and economic status (assessed by possession of assets, household characteristics, income levels or expenditures). The population in need of health care was defined as pregnant women aged 15–49 years, although delivery in medical settings is not necessary for all women.20,21
Study quality and data extraction
The first author identified articles by examining titles, then abstracts, for relevance, and retrieved the full text of relevant abstracts for further assessment. Uncertainties were resolved through discussions with the second author. Both authors extracted data independently from each included study on a customized form. Information on contextual factors was extracted from the discussion sections of the articles. Methodological quality, in terms of internal validity and generalizability, was assessed with a checklist of quality criteria (Box 1) developed by both authors on the basis of existing instruments for observational studies.22–25 Criteria judged to be especially important were related to the ability of the studies to identify the entire population in need of care, provision of explicit definitions for outcomes and independent variables, and controlling for potential confounders. Each criterion was answered on a scale from “not reported” to “well-covered”. Disagreement between the authors about study quality was resolved by reviewing the article again together.
Box 1. Criteria used to assess quality of included studies
• Study addresses a clearly focused question
• Characteristics of study population are clearly described
• Clear eligibility criteria for selection of participants
• Participants are representative of target population
• Indicates how many of those asked to participate did so in each group
• Outcomes clearly defined
• Justification provided for independent variables and definitions
• Study-specific data collection
• Data collection tools defined
• Valid and reliable measurement of assessment of outcome and other independent variables
• Valid and reliable measure of assessment of exposures
• Indicates number of participants for whom data analysed in each group
• Main potential confounders identified and controlled for in the analysis [sociodemographic, health beliefs, attitudes and relevant health status measures (i.e. previous pregnancy outcomes, problems during ongoing pregnancy)]
• Factors influencing need for care are explored and taken into account in the analysis (i.e. health beliefs, attitudes, obstetric risk measures)
• Confidence intervals provided
• If individual and group level variables were analysed in the same model, multilevel modelling used
We extracted, or calculated from proportions, numerator data (use of health services) and denominator data (population in need of health care) for each comparison group where possible. We extracted crude odds ratios (OR), adjusted odds ratios, or both, as reported. If these data were not reported, we calculated crude odds ratios and 95% confidence intervals (CI) if numerator and denominator data were available. We converted associations presented as correlation coefficients (β) into odds ratios by taking the inverse natural log of the coefficients. We presented results of statistical significance tests if odds ratios could not be obtained.
Because of the wide variation in study populations, definitions of non-clinical factors, outcome measures and confounders investigated, it was not appropriate to obtain a statistical summary for the size of the effect of specific non-clinical factors on each outcome. We tabulated and summarized data as brief narrative for each outcome measure. Low-quality studies are included in the tables, but their findings were given less weight in the narrative summary.
We used an extended version of an analytical framework that we have previously used to explain variations in health-care use.23 The extended framework outlined the steps that must be followed to explain results. First, methodological quality was assessed to judge whether findings were likely to be artefactual and caused by chance or bias. Second, we attempted to account for results in terms of factors related to health-care users, factors related to the supply of health care, or an interaction between the two; we focused on whether findings could be explained by variations in women’s socioeconomic status or place of residence. Third, variations in use of health care by socioeconomic status or place of residence were placed in context to seek a complete explanation.
Description of studies
We identified 5575 citations for maternal health care and family planning outcomes, of which 418 were judged to be potentially relevant. Citations were excluded at this stage mostly because they reported no quantitative data or data on other outcome measures. When abstracts had been reviewed, 155 studies were retrieved for full evaluation, of which 93 were related to use of maternal health care. Fifty-six of these studies had to be excluded (Fig. 1); a further 10 could not be located (three were unpublished conference proceedings). Therefore, 27 studies were included. Identification of an additional three studies from the reference lists of included papers and websites brought the total number of included studies to 30, from 23 countries in three continents (Annex 1, available at: http://www.who.int/bulletin/volumes/85/10/06-035659/en/index.html). Three studies were of high quality, nine of moderate quality, and 18 of low quality. Twenty-one studies examined delivery in medical settings; the outcome of interest was the presence of a skilled health worker at delivery in 11 studies and antenatal care attendance in the first trimester of pregnancy in six. Thirteen studies were secondary analyses of large retrospective cross-sectional surveys, four were analyses of other existing data, and 13 were done prospectively. Two studies provided separate analyses from different country settings26,27 and one from different states of India.28
Skilled health worker at delivery
Seven studies (three of moderate quality, four of low quality) assessed urban-rural differences in having a skilled health worker at delivery in nine countries (Table 1, available at: http://www.who.int/bulletin/volumes/85/10/06-035659/en/index.html). The three moderate-quality studies reported significant differences in favour of urban women in Tajikistan (adjusted OR, 3.48; 95% CI: 1.50–8.06),29 Nepal (P < 0.01),30 and three of four Indian states [adjusted ORs between 1.42 (P < 0.10) and 2.30 (P < 0.01)].28 These findings seemed to be confirmed by those from low-quality studies in Ethiopia, Kenya, Turkey, and the United Republic of Tanzania,27,31–33 although two of these studies did not make adjustments for potential confounding factors. Other low-quality studies reported no differences (Paraguay, Uganda) or lower use by urban women than by rural women (India; P < 0.10).27
Nine studies on the effect of economic status on having a skilled attendant at delivery provided 15 analyses (Table 1). One study was of high34 quality and five were of moderate quality.16,28–30,35 The high-quality study showed no effect of economic status, defined by landholding size across three groups in Bangladesh.34 Instead, factors that affected this practice included the education of the woman and of her husband, complications of delivery and receipt of antenatal care. By contrast, all studies of moderate quality showed consistently greater use of a skilled health worker at delivery by affluent groups compared with poorer women. In four Indian states, economic status was defined by a standard-of-living index that included household assets and environmental characteristics. This study showed differences by standard of living that remained significant after controlling for potential confounders, including age, education of the woman and of her husband, obstetric history, caste and occupation.28 These findings were consistent with those of other moderate-quality studies, which defined economic status in terms of house-building materials in India (adjusted OR, 2.87; 95% CI: 1.70–4.84),35 income levels in China (adjusted OR, 1.38; 95% CI: 1.06–1.79),16 and total expenditure per capita in Tajikistan [adjusted OR between 2.09 (95% CI: 0.98–4.46) and 3.41 (95% CI: 1.58–7.06) across quintiles]29 and Nepal (P < 0.01).30 Results of low-quality studies were generally consistent with these findings27,32 or showed no difference.27,36
Medical setting for delivery
Ten studies on urban-rural differences in delivery in medical settings were identified, including one of high quality and four of moderate quality. These studies reported 16 analyses from eleven countries (Table 2, available at: http://www.who.int/bulletin/volumes/85/10/06-035659/en/index.html). The high-quality study showed that urban women in Jamaica were significantly more likely than rural women to deliver in medical settings (P < 0.01) after controlling for probable confounders, including pregnancy complications and attendance of antenatal care.3
These findings were confirmed by the studies of moderate quality in three states and a sub-district in India, Morocco and Tajikistan [adjusted ORs from 2.13 (95% CI: 1.21–3.76) to 3.64 (P < 0.01)]28,29,37,38 and in four low-quality studies in Belize, Ghana, Kenya, Uganda and the United Republic of Tanzania.27,32,39,40 Three of these studies controlled for factors that could pose a risk during delivery.28,37,38 No differences were found in India’s Kerala state28 or in Nepal.41
Seventeen studies looked at the association between economic status and use of medical settings for delivery (Table 2). The only high-quality study used per-capita household expenditure to measure economic status.3 Higher economic status was associated with an increased probability of using medical settings for delivery in Jamaica (P < 0.01) after controlling for probable confounders including education, living in the metropolitan area, pregnancy complications and receipt of antenatal care. Three of the five moderate-quality studies reported similar findings; in India, where economic status was assessed either by the financial value of household assets [adjusted OR, 1.23 (P < 0.05) for middle economic groups and 1.55 (P < 0.01) for wealthier groups, compared with poorer groups]37 or by different standards of living;28 and in Morocco (adjusted OR, 2.94; 95% CI: 1.45–6.04).38 By contrast, moderate-quality studies undertaken in Guatemala and Tajikistan showed no effect of economic status, measured by food consumption and per-capita household expenditure, respectively.12,29 Instead, determinants of delivering in a medical setting included education, age, obstetric risk measures, medical insurance, time to travel to site of care and openness to the outside world in Guatemala;12 and education, region, availability of polyclinic in the community and timing of the birth in Tajikistan.29 The low-quality studies tended to report either an association between higher economic status and increased use,27,32,42–45 or no difference with increasing wealth.19,46–49
Antenatal care in the first trimester
Five studies, one of high quality and three of moderate quality, examined the effect of urban-rural residence on antenatal care during the first trimester of pregnancy (Table 3, available at: http://www.who.int/bulletin/volumes/85/10/06-035659/en/index.html). The high-quality study showed that urban women in Jamaica were significantly less likely (P < 0.05) than rural women to attend antenatal care during this period.3 By contrast, the moderate-quality studies reported no significant difference between urban and rural women in India28,37 or reported that urban women were significantly more likely to attend antenatal care during the first trimester in Brazil (adjusted OR, 1.37; P < 0.01) after controlling for probable confounders, including sociodemographic and economic factors.26 Results of the low-quality study from Ecuador were consistent with those from Brazil.50
Six studies (two studies of high quality, three of moderate quality) assessed the effect of economic status on attendance of antenatal care during the first trimester (Table 3). The two high-quality studies from Jamaica found that an increased probability of early antenatal care attendance was associated with increased household expenditure (P < 0.001)3 after controlling for mainly sociodemographic factors and differences across income quartiles in favour of wealthier women.5 Economic status, measured by household assets score or living standards, was also significantly associated with antenatal care in the first trimester in Brazil (adjusted OR, 3.51; P < 0.001), South Africa (adjusted OR, 2.88; P < 0.001)26 and India’s Kerala state (adjusted OR, 1.51; P < 0.05).28 However, two other studies of moderate quality in other parts of India showed no effect of economic status after controlling for potential confounders.28,37 In these studies, factors associated with receipt of early antenatal care were education, pregnancy order, women’s autonomy and obstetric risk measures.
The importance of context
Contextual factors that could account for variations found in studies of moderate and high quality were identified and summarized (Tables 1–3). The diversity of findings between and within populations reflected a wide range of different ethnic, cultural and religious groups among the women surveyed, who reported important differences in beliefs and preferences relating to formal and informal maternal health care. These variations were explained by differences in women’s autonomy, gender relationships and social networks, which are influenced by embedded social structures, religion and cultural beliefs. For example, the extent of an Indian woman’s autonomy, which is often determined by continued links with her parental family after marriage, affected use of skilled delivery care, because living with or near a member of her birth family increased her ability to leave the house and go where she wanted.35 However, among rural Chinese women, freedom of movement did not affect rates of delivery with the help of a skilled health worker.16 Jamaican women did not want to be treated as ill during uncomplicated pregnancies, and so tended to delay initiation of antenatal care.3 Non-white South African women did not see the value of antenatal care, aside from it being necessary to allow access to care during delivery; therefore they began antenatal visits later in pregnancy.26
The wide range of health policies, financing arrangements and organizational structures also needed to be taken into account. Women who did not comply with policies to restrict numbers of children in China tended to avoid formal maternal health care to prevent being discriminated against by health-care providers.16 Following the end of apartheid, health policies in South Africa have been especially concerned with increasing access to primary health care, including maternal health care.26 Health care is free for pregnant women in Jamaica and South Africa.3,5,26 By contrast, the private and informal sectors are increasingly active in provision of maternal health care in India.37 Introduction of fees for maternal health care in Morocco during reorganization of the health sector was thought to limit access for poor women.38 Differences in use of skilled delivery care across economic groups were explained by informal charges in Tajikistan, where maternal health care was officially free of charge.29 Interventions aimed at poor areas did not benefit poor and underserved women in Brazil.26 By contrast, multipurpose health workers facilitated provision of antenatal care in remote parts of Karnataka in India, where rural residence and economic status did not affect this aspect of care.28
Interactions between factors at the level of the individual and those associated with supply or organization of health care were often crucial. Women in Tajikistan preferred to deliver at home because although medical settings were accessible and free of charge, women perceived these settings to be of very low quality and unsafe.29 Midwives’ advice on where to deliver was ignored by Jamaican women, who perceived the midwives to be too authoritarian.3 Women in rural Guatemala were less likely to deliver in medical settings because of the lack of social support provided by health-care professionals compared with traditional midwives.12
Overall, the evidence we reviewed showed that use of maternal health care varied greatly both within and between countries. Within countries, urban or wealthier women were usually more likely to deliver with the help of a skilled health worker than were rural or poor women. Urban women were more likely to use medical settings for delivery than were rural women. In some countries, wealthier women tended to deliver in medical settings, but in others (e.g. Guatemala and Tajikistan) economic status did not affect such practices. The association between place of residence and receipt of early antenatal care was not consistent. Some evidence suggested that wealthier women were more likely than poorer ones to receive early antenatal care, although no such difference was found in India.
We located few studies of high or moderate quality, and eight studies19,35,36,41,43,45,48,49 had sample sizes of less than 500 women, restricting the ability to draw conclusions. In some instances, the population in need of health care was not identified accurately because data were collected for other purposes, such as the evaluation of child health; for example, in one study, the population was restricted to married women younger than 35 years with at least one child younger than 5 years.37 The definition of need for maternal health care was also inconsistent: eligible groups included women who had given birth ever,29,46 in the past 45 days43 and in the past 5 years.12,32,33,38–40,47
The choice of potential confounders was another important reason for the difference in findings. Some studies did not adjust for any probable confounders.44,47 Those that did used a wide range of confounders in different combinations. Variables included user-related items such as education, religion, ethnicity, marital status, age, women’s or partners’ education and number of children,28,38–40,43 and supply-related items including distance to care and characteristics of health care-facility.12,27,29,38 However, other variables of probable importance, such as obstetric risk and users’ beliefs and preferences, were often not mentioned.
We also noted wide variation in definitions of non-clinical factors, confounders and outcomes, all of which probably contributed to the diversity of the findings. Measures of economic status varied greatly, ranging from landholding size34 or the type of walls35 or roof47 of housing to more conventional measures such as household income,36,45 expenditure3,29 or asset scores.26
Some investigators did not provide definitions for outcomes, probable confounders and non-clinical factors.33,41,44 Reliance on women’s self-reported histories of pregnancy and delivery up to 5 years previously made findings susceptible to recall bias.
First, the search strategy might not have identified all relevant papers. To check for completeness, we also searched reference lists of included papers, and identified only two additional articles. However, practical constraints restricted the search strategy to papers in the English language, meaning that relevant studies published in other languages were missed. Second, maternal health-care indicators were defined in broad terms; for example, it was not feasible to consider the relative effectiveness of different types of skilled health workers. The definitions chosen were, however, relevant from a policy perspective in the context of the millennium development goals. Third, the quality assessment instrument did not undergo formal psychometric evaluation, but was based on an existing widely used instrument, was deemed to have content validity, and was used by two authors independently. We felt that this instrument was adequate for our objective: to provide an explicit indication of study quality, rather than precise measurement.
This review demonstrates variations in the use of maternal health care across populations both within and between 23 developing countries. Variations were partly explained by methodological differences in study designs. However, important and diverse contextual factors were also identified, many relating to the funding and organization of health care. In addition, more subtle, but equally influential, context-specific individual level factors emerged, as did interactions between individual level and health service-related factors. Two reasons for the limited success of the safe motherhood campaign during the past two decades have been the lack of rigorous analysis of the data available on variations in use, together with an inadequate grasp of the contextual issues that must be addressed if inequalities in maternal health care use are to be reduced. Our results highlight the need to thoroughly explore and address context-specific causes of variable use of maternal health care if safe motherhood is to become a reality in developing countries. ■
- AM Starrs. Safe motherhood initiative: 20 years and counting. Lancet 2006; 368: 1130-2.
- T Adam, S Lim, S Mehta, ZA Bhutta, H Fogstad, M Mathai, et al., et al. Cost effectiveness analysis of strategies for maternal and neonatal health in developing countries. BMJ 2005; 331: 1107-.
- P Gertler, O Rahman, C Feifer, D Ashley. Determinants of pregnancy outcomes and targeting of maternal health services in Jamaica. Soc Sci Med 1993; 37: 199-211.
- Health Evidence Network. What is the effectiveness of antenatal care? Copenhagen: WHO Regional Office for Europe; 2005.
- A McCaw-Binns, J La Grenade, D Ashley. Under-users of antenatal care: a comparison of non-attenders and late attenders for antenatal care with early attenders. Soc Sci Med 2007; 40: 1003-12.
- WHO database on skilled attendant at delivery. Geneva: WHO: 2007. http://www.who.int/reproductive-health/global_monitoring/data.htm
- MA Magadi, EM Zulu, M Brockerhoff. The inequality of maternal health care in urban sub-Saharan Africa in the 1990s. Population Studies. 1993; 57: 349-68.
- C Ronsmans, JF Etard, G Walraven, L Hoj, A Dumont, L Bernis, et al., et al. Maternal mortality and access to obstetric services in West Africa. Trop Med Int Health 2003; 8: 940-8.
- M Wirth, D Balk, E Delamonica, A Storeygard, M Sacks, A Minujn. Setting the stage for equity-sensitive monitoring of the maternal and child health Millennium Development Goals. Bull World Health Organ 2006; 84: 519-27.
- NNA Al-Nadhedh. Factors affecting the choice of maternal and child health services in a rural area of Saudi Arabia. East Mediterr Health J 1995; 1: 261-9.
- MA Magadi, AO Agwanda, FO Obare. A comparative analysis of the use of maternal health services between teenagers and older mothers in sub-Saharan Africa: Evidence from Demographic and Health Surveys (DHS). Soc Sci Med 2007; 64: 1311-25.
- DA Glei, N Goldman. Understanding ethnic variation in pregnancy-related care in rural Guatemala. Ethn Health 2000; 5: 5-22.
- S Pallikadavath, M Foss, RW Stones. Antenatal care:provision and inequality in rural north India. Sos Sci Med. 2004; 59: 1147-58.
- SO Gyimah, BK Takyi, I Addai. Challenges to the reproductive-health needs of African women: On religion and maternal health utilization in Ghana. Sos Sci Med. 2006; 62: 2930-44.
- DA Glei, N Goldman, G Rodriguez. Utilization of care during pregnancy in rural Guatemala: does obstetrical need matter? Soc Sci Med 2003; 57: 2447-63.
- J Li. Gender inequality, family planning, and maternal and child health care in a rural Chinese county. Soc Sci Med 2004; 59: 695-708.
- BN Nwakoby. Use of obstetric services in rural Nigeria. J R Soc Health 1994; 114: 132-6.
- TM McCray. An issue of culture: the effects of daily activities on prenatal care utilization patterns in rural South Africa. Soc Sci Med 2004; 59: 1843-55.
- DV Duong, CW Binns, AH Lee. Utilization of delivery services at the primary health care level in rural Vietnam. Soc Sci Med 2004; 59: 2585-95.
- Care in normal birth: a practical guide. Geneva: WHO; 1996 (WHO/FRH/MSM/96.24).
- BK Paul. Maternal mortality in Africa:1980-87. Soc Sci Med 1993; 37: 745-52.
- S Zaza, LK Wright-De Aguero, PA Briss, BI Truman, DP Hopkins, MH Hennessy, et al., et al. Data collection instrument and procedure for systematic reviews in the “Guide to Community Preventive Services”. Am J Prev Med 2000; 18: 44-74.
- R Raine. Does gender bias exist in the use of specialist health care? J Health Serv Res Policy 2000; 5: 237-49.
- SH Downs, N Black. The feasibility of creating a checklist for the assessment of the methodological quality of both randomised and non-randomised studies of health care interventions. J Epidemiol Commun Health. 1998; 52: 377-84.
- SIGN. 50: a guideline developer’s handbook. Edinburgh: Scottish Intercollegiate Guidelines Network; 2001.
- S Burgard. Race and pregnancy-related care in Brazil and South Africa. Soc Sci Med 2004; 59: 1127-46.
- Tsui AO, Ukwuani F, Guilkey D, Angeles G. Health program effects on individual use of services. Chapel Hill: Measure Evaluation; 2002.
- K Navaneetham, A Dharmalingam. Utilization of maternal health care services in Southern India. Soc Sci Med 2002; 55: 1848-69.
- J Falkingham. Inequality and changes in women’s use of maternal health care services in Tajikistan. Stud Fam Plann 2003; 34: 32-43.
- DR Hotchkiss. Expansion of rural health care and the use of maternal services in Nepal. Health Place 2001; 7: 39-45.
- Y Celik, DR Hotchkiss. The socioeconomic determinants of maternal health care utilization in Turkey. Soc Sci Med 2000; 50: 1797-860.
- M Magadi, I Diamond, R Nascimento Rodrigues. The determinants of delivery care in Kenya. Soc Biol 2000; 47: 164-88.
- Y Mekonnen, A Mekonnen. Factors influencing the use of maternal healthcare services in Ethiopia. J Health Popul Nutr 2003; 21: 374-82.
- BK Paul, DJ Rumsey. Utilization of health facilities and trained birth attendants for childbirth in rural Bangladesh: an empirical study. Soc Sci Med 2002; 54: 1755-65.
- SS Bloom, D Wypij. das Gupta M. Dimensions of women’s autonomy and the influence on maternal health care utilization in a north Indian city. Demography 2001; 38: 67-78.
- C Phoxay, J Okumura, Y Nakamura, S Wakai. Influence of women’s knowledge on maternal health care utilization in Southern Laos. Asia Pac J Public Health 2001; 13: 13-9.
- JC Bhatia, J Cleland. Determinants of maternal care in a region of south India. Health Transit Rev 1995; 5: 127-42.
- Hotchkiss DR, Krasovec K, El-Idrissi MDZ, Eckert E, Karim AM. The role of user charges and structural attributes of quality on the use of maternal health services in Morocco. Chapel Hill: Measure Evaluation; 2003.
- I Addai. Demographic and sociocultural factors influencing use of maternal health services in Ghana. Afr J Reprod Health 1998; 2: 73-80.
- PW Stupp, BA Macke, R Monteith, S Paredez. Ethnicity and the use of health services in Belize. J Biosoc Sci 1994; 26: 165-77.
- A Bolam, DS Manandhar, P Shrestha, M Ellis, K Kalla, AM Costello. Factors affecting home delivery in the Kathmandu Valley, Nepal. Health Policy Plan 1998; 13: 152-8.
- G Letamo, SD Rakgoasi. Factors associated with non-use of maternal health services in Botswana. J Health Popul Nutr 2003; 21: 40-7.
- RR Wagle, S Sabroe, BB Nielsen. Socioeconomic and physical distance to the maternity hospital as predictors for place of delivery: an observation study from Nepal. BMC Pregnancy Childbirth 2004; 4: 8-.
- MAK Barbhuiya, S Hossain, MM Hakim, SM Rahman. Prevalence of home deliveries and antenatal care coverage in some selected villages. Bangladesh Med Res Counc Bull 2001; 27: 19-22.
- D Hodgkin. Household characteristics affecting where mothers deliver in rural Kenya. Health Econ 1996; 5: 333-40.
- O Anson. Utilization of maternal care in rural HeBei Province, the People’s Republic of China: individual and structural characteristics. Health Policy 2004; 70: 197-206.
- NV Toan, HT Hoa, NT Thach, B Hojer, LA Persson. Utilization of reproductive health services in a mountainous area in Vietnam. Southeast Asian J Trop Med Public Health 1996; 27: 325-32.
- OA Van der Heuvel, WG De Mey, H Buddingh, ML Bots. Use of maternal care in a rural area of Zimbabwe: a population-based study. Acta Obstet Gynecol Scand 1999; 78: 838-46.
- N Kavitha, N Audinarayana. Utilisation and determinants of selected MCH care services in rural areas of Tamil Nadu. Health Population. 1997; 20: 112-25.
- E Eggleston. Unintended pregnancy and women’s use of prenatal care in Ecuador. Soc Sci Med 2000; 51: 1011-8.
- UNDP/UNFPA/WHO/World Bank Special Programme of Research, Development and Research Training in Human Reproduction, Department of Reproductive Health and Research, WHO, 20 avenue Appia, 1211 Geneva 27, Switzerland.
- Department of Epidemiology and Public Health, University College London, London, England.