• 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. 2003 November; 93(11): 1820–1829.
PMCID: PMC1448058

Contextual Influences on Reproductive Wellness in Northern India


Objectives. There has been a growing recognition of the importance of contextual influences on health outcomes. This article examines community-level influences on 5 reproductive wellness outcomes in Uttar Pradesh, India.

Methods. Multilevel modeling is used to estimate household and community-level effects on wellness, with hierarchically organized data from a statewide survey of villages, urban blocks, households, women, health providers, and staff.

Results. The household and community have a strong contextual influence on wellness, although the models explain more of the variation in outcomes between households than between communities.

Conclusions. Communities influence wellness outcomes through the socioeconomic environment and the characteristics of the health infrastructure. The specific dimensions of the community and health infrastructure varied between the outcomes.

A recent growth of interest in measuring community-level effects on health outcomes has given rise to the increased availability of community-level data and a recognition that the determinants of health extend beyond individual factors to the community in which the individual lives. The development of multilevel modeling techniques has provided the means for estimating community effects on health.1–3 This focus on the community places the health outcomes and behaviors of individuals in their socioeconomic context and provides a policy tool for measuring group-level effects on individual health.1

This has important implications for public health initiatives; information on the interaction between the individual and the community can be used to inform the development of community-based health programs. This report adds to the existing literature by examining 5 different, although interrelated, reproductive wellness outcomes in Uttar Pradesh, India, facilitating a comparison of community-level effects across a range of wellness outcomes. It examines influences on wellness outcomes at the household, facility, and community levels, extending the work of previous studies that have focused on effects operating at each of these levels in isolation. The study thus provides an opportunity to understand the hierarchy of influences operating to influence an individual’s health.


Community Influences on Health

Studies of the determinants of health outcomes have long focused on individual risk factors, neglecting the socioeconomic environment in which the outcomes occur.1,3 Recently there has been growing interest in examining community influences on health outcomes, so as to put health in its socioeconomic environment.3–15 Such studies relate individual health outcomes to socioeconomic characteristics of the community (e.g., levels of economic development) and the community health infrastructure. 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.2,16–18 An examination of community influences on health provides an opportunity to highlight health risks associated with particular social structures and community ecologies, and it can provide a useful policy tool for the development of community-level health interventions.1,5

The growing interest in community influences on health arises from the recognition of a disjuncture between theory and research practice.3 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,19,20 yet until recently studies of contraceptive-use dynamics have focused on individual determinants. The incorporation of community-level factors into multilevel models of fertility behavior has allowed the identification of the influence of community development, attitudes, norms, and health service availability on fertility.13,21,22 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,3–5,13,21 although some studies have examined pregnancy care.6,10 Findings have shown that the presence and quality of reproductive health services,4,5,13 levels of economic development,5,13 levels of school participation,13 economic roles of children,5,21 and community fertility norms23,24 all influence individual fertility behavior. These studies have focused on community effects on single reproductive health outcomes, and there is a dearth of studies that have examined community influences on a range of related health outcomes.

This report looks at community influences on 5 reproductive wellness outcomes in Uttar Pradesh, India, to highlight the variety of community effects that may exist within a field of health research and to contrast the size and significance of community influences between related health outcomes. In examining the contextual influences on individual well-being, we follow the logic put forth by Stokols25 to adopt a broader conception of health by focusing on wellness as a combined and positive state of well-being, rather than simply the absence of disease or illness.

Reproductive Health in Uttar Pradesh

With a population of 166 million according to the 2001 census, Uttar Pradesh is the most populous state in India.26 Relative to other Indian states, Uttar Pradesh fares more poorly in terms of demographic indicators. Fertility and mortality are higher there than in many other states, with a total fertility rate of 4.0 (2.9 for all of India) and an infant mortality rate of 87 per 1000 live births (68 per 1000 live births for all of India).27,28

Reproductive health indicators for Uttar Pradesh are equally poor. Maternal health is characterized by low uptake of prenatal care (35% of pregnant women vs 65% for all of India), with only 7% of pregnant women receiving the recommended 4 prenatal care visits (30% for all of India), and a reliance on home births (85% of all pregnant women vs 65% for all of India).27 Contraceptive use is low, with only 28% of married women in Uttar Pradesh using contraception (48% for all of India), while 25% of married women experience an unmet need for family planning (15% for all of India).28 A large number of married women report experiencing reproductive tract or sexually transmitted infection (RTI/STI) symptoms (38% for Uttar Pradesh vs 39% for all of India), while awareness of AIDS is low (20% of all married women in Uttar Pradesh vs 40% for all of India).27 These indicators exist in an environment of high illiteracy, low female autonomy, and inadequate quality of and access to health services.

The economic, social, and health care environments of Uttar Pradesh provide a number of possible contextual influences on a woman’s reproductive wellness. Uttar Pradesh is a poor, predominantly rural state with an inadequate health infrastructure. As such, there is wide variation in individuals’ access to health care, and the ability of an individual to access adequate care may be strongly influenced by the presence and quality of health services in their community. The economic development of the community in which the individual lives may influence individual health through opportunities for employment, increased income, and education. The level of economic development may also be linked to greater health awareness through the presence of health campaigns and higher-quality health services in more developed communities.

The continued reliance on traditional childbirth practices in Uttar Pradesh, including the use of untrained birth attendants, suggests the strong role that community norms and attitudes have in shaping an individual’s health outcomes, and the influence of a lack of quality health services on individual health. An understanding of how these potential contextual influences on health in Uttar Pradesh vary between reproductive wellness outcomes provides an opportunity to identify potential areas for health intervention, and the homogeneity of the population suggests the potential for implementing such interventions on a wider scale.


The data set for this analysis was the 1995–1996 PERFORM System of Indicators Survey (PSIS).29 The survey was designed to provide representative estimates of the levels and patterns of contraceptive practice and service delivery for the 28 districts, 14 divisions, and 5 regions of Uttar Pradesh. Data were collected via 7 questionnaires: individual (women aged 13–49 years), household, community (separate questionnaires for village and urban block), private and public health facility (separate questionnaires for fixed service delivery points and individual service delivery agents), and health facility staff.

The survey employed a stratified multistage cluster sample design for households and service delivery points. Two districts in each of the 14 administrative divisions were selected. The health facility questionnaires included a component on reproductive health services, collected in only 5 districts. The 5 districts in which the reproductive health component of the facility questionnaire was conducted were randomly selected from the 28 districts in the study; thus, these 5 districts should not differ from other districts in the state.

The analysis is restricted to 5128 women from these 5 districts, to allow an investigation of the influence of reproductive health services on reproductive wellness. When there was more than one health facility per community, the responses were averaged across health facilities. Hence, the health facility variables used in the analysis represent average values for the health facilities in the community.


Five reproductive wellness outcomes were modeled. Three of these were chosen to represent physiological outcomes, which allowed an examination of how individual, household, and community factors influence a woman’s risk of experiencing reproductive morbidity and the extent to which these risks vary by community. The 3 outcomes are as follows:

  • Whether the woman experienced pregnancy complications in her last pregnancy. Pregnancy complications were self-reported and included bleeding, swelling, fever, jaundice, high blood pressure, anemia, convulsions, and pain during urination. The analysis was restricted to women who had a birth in the 3 years prior to the survey (n = 3106).
  • Whether the woman experienced labor complications at the birth of her last child. Labor complications were self-reported and included prolonged labor, high fever, cesarean delivery, use of forceps, excessive bleeding, delayed delivery of placenta, and convulsions. The analysis was restricted to women who had a birth in the 3 years prior to the survey (n = 3106).
  • Whether the woman experienced symptoms of RTI/STI in the past 12 months. RTI/STI symptoms were self-reported and included abnormal vaginal discharge, itching/irritation in the vaginal area, and pain or burning during urination. Only women who had a birth in the 3 years prior to the survey were asked the questions on RTI/STI symptoms; hence the analysis is restricted to this group (n = 3106).

Two outcomes were chosen to represent reproductive wellness behaviors—that is, outcomes that required an element of decisionmaking or preference on behalf of the respondent. These outcomes reflect 2 different aspects of fertility decisions:

  • Whether the previous pregnancy was unwanted. This outcome reflects the respondent’s ability to avoid unwanted pregnancies. The analysis was conducted on women who had a child in the 3 years prior to the survey (n = 3139).
  • Whether the desired family size had been achieved. This outcome reflects individual control over fertility and the ability to achieve the desired number of children. Failure to attain the desired family size may reflect infertility. The desired family size was measured from a question asking how many more children the woman would like to have. When women reported that they wanted no more children, it assumed that the desired family size had been achieved. The analysis included all women from the 5 districts (n = 5128).

A multilevel modeling strategy was employed to account for the hierarchical structure of the data. Ordinary regression models assume that all observations are independent. The PSIS has a hierarchical structure, with women clustered within households, which are in turn clustered within communities. Hence, the odds of women experiencing the outcome of interest are not independent, as women share common exposure to household and community characteristics. A multilevel modeling strategy accommodates the hierarchical nature of the data and corrects the estimated standard errors to allow for the clustering of observations with units.17,30

The use of multilevel models also allows the identification of clustering of the outcome at different levels. This clustering, known as the random effect, represents the extent to which the outcome of interest varies between each higher order unit (household or community). The random effect can reflect factors influencing the outcome that have been omitted from the model, or factors that cannot be quantified in a large-scale social survey (e.g., variations in health beliefs). A random effects model thus provides a mechanism for estimating the degree of correlation in the outcome that exists at the household or community level, while also controlling for a range of individual, household, and community factors, and the random effects can thus be thought of as the residual variation in the outcomes.

Separate multilevel logistic models are fitted for each of the 5 outcomes. The models take the form of 3-level models, with women (level 1) nested within households (level 2), which are in turn nested within communities (level 3). The models are written

equation M1

where pijk is the probability of experiencing the outcome for the ith respondent in the jth household in the kth community, xijk is a vector of covariates corresponding to the ith respondent in the jth household in the kth community, β is a vector of unknown parameters, ujk is the random effect at the household level, and vk is the random effect at the community level. The distribution of the random effects is assumed to be normal, with mean zero and variance su. When su = 0, the model reduces to the ordinary logistic model, indicating that there is no significant correlation in the risk of the outcome between households or communities. The testing of the null hypothesis su = 0 against the alternative hypothesis su > 0 is used to assess the significance of random effects terms, using a modified likelihood ratio test.

The variables to be entered into the models are grouped into individual, household, community, and health facility factors. The individual variables are chosen to represent demographic and socioeconomic factors that previous research has shown to be associated with various reproductive wellness outcomes: parity, age, educational attainment, media exposure, and experience of child death. In the absence of data on household income, an asset index is used to represent the socioeconomic status of the household.31 The index includes the ownership of several consumer goods and vehicles.

Table 1 [triangle] shows the variables used in the analysis. Four indices were created to represent differing aspects of the community environment. The economic index reflects the level of community economic development. The health infrastructure index reflects the presence of formal health facilities in the community and the traditional medicine index reflects the community’s reliance on traditional forms of care. The community population size is used to represent the level of demand for health services in the community. The facility variables were chosen to represent both the presence of differing types of health services in the community and the characteristics of the services in terms of the services they offered and their staffing levels.

Variables Used in Modeling of Reproductive Wellness Outcomes: Uttar Pradesh, India, 1995

The analysis uses a cumulative approach to model building. Model 1 includes only individual and household factors. Model 2 includes individual and household factors together with village-level variables. Model 3 includes the individual/household, village, and facility-level factors. This approach allows the identification of the relative impact of each set of factors in explaining the household and community variation in the outcomes. All models include household- and community-level random effects terms.


Individual and Household Effects

Tables 2 [triangle] through 4 [triangle] show the results of the multilevel modeling of the 5 outcomes. Five individual and household factors were entered into the models. Parity was not used in the model of achieved desired family size, as it is part of the outcome variable. Instead, maternal age was entered into the model. Parity was significantly related to unwanted pregnancies (Table 3 [triangle]) and labor complications (Table 2 [triangle]). Not surprisingly, compared with women with a parity of 1 or 2, women with a parity of 3 or higher were more likely to report their last child as unwanted (parity 3–4 coefficient = 1.254, SE = 0.139; parity 5–6 coefficient = 1.702, SE = 0.163; parity >7 coefficient = 2.520, SE = 0.188), and they were less likely to experience labor complications at last childbirth (parity 3–4 coefficient = −0.301, SE = 0.093; parity 5–6 coefficient = −0.229, SE = 0.120; parity >7 coefficient = −0.434, SE = 0.154).

Multilevel Modeling of Pregnancy and Labor Complications: Uttar Pradesh, India, 1995
Multilevel Modeling of Wanted Status of Previous Child and Achieved Desired Family Size: Uttar Pradesh, India, 1995
Multilevel Modeling of Reporting of Symptoms of Reproductive Tract Infection or Sexually Transmitted Infection: Uttar Pradesh, India, 1995

Women with a parity of 3 or more were clearly at a higher risk of having an unwanted pregnancy since they had had more pregnancies in total. The respondent’s exposure to family planning messages in the media was associated with an increased likelihood of reporting RTI/STI symptoms (coefficient = 0.498, SE = 0.107) (Table 4 [triangle]) and pregnancy complications (coefficient = 0.271, SE = 0.094) and labor complications (coefficient = 0.654, SE = 0.101) (Table 2 [triangle]). Exposure to family planning messages was self-reported and thus reflects the respondent’s awareness of health campaigns in her community, an increased awareness thus leading to a greater likelihood of recognizing and reporting symptoms.

The experience of previous child loss was significantly associated with a woman’s reduced likelihood of achieving her desired family size (coefficient = −0.932, SE = 0.080) and of reporting the previous child as unwanted (coefficient = −0.393, SE = 0.123). Those women who had attained a high school education or more were less likely to report RTI/STI symptoms than illiterate women (coefficient = −0.481, SE = 0.195). Women with middle-level education were more likely to report that their previous child was unwanted than were illiterate women (coefficient = 0.358, SE = 0.174). Relative to women aged 20 to 24 years, those aged 13 to 19 years were, not surprisingly, less likely to have achieved their desired family size (coefficient = −1.493, SE = 0.212), while the likelihood of achieving desired family size rises after age 25.

Village-Level Effects

Higher levels of economic development were significantly associated with increased reporting of RTI/STI symptoms (coefficient = 0.113, SE = 0.055) and a decreased likelihood of experiencing pregnancy complications (coefficient = −0.120, SE = 0.050) and unwanted pregnancies (coefficient = −0.150, SE = 0.062). The presence of modern health facilities was associated with an increased likelihood of experiencing labor complications (coefficient = 0.161, SE = 0.070) and decreased likelihood of an unwanted pregnancy (coefficient = −0.189, SE = 0.080). The presence of traditional health facilities was significantly associated with a decreased likelihood of experiencing labor complications (coefficient = −0.196, SE = 0.067). The population size of the community was negatively associated with the likelihood of achieving the desired family size (coefficient = −0.135, SE = 0.060).

Facility-Level Effects

Higher numbers of doctors in a community were associated with an increased likelihood of achieving the desired family size (coefficient = 0.032, SE = 0.014) and experiencing an unwanted pregnancy (coefficient = 0.114, SE = 0.032). The presence of a health campaign in the community was significantly associated with decreased reporting of RTI/STI symptoms (coefficient = −0.207, SE = 0.082) and experiencing pregnancy complications (coefficient = −0.250, SE = 0.115). Unlike the individual’s exposure to family planning messages, the presence of a health campaign in the community is an objective measure of the availability of health messages in the local community. The decreased likelihood of reporting RTI/STI symptoms and pregnancy complications in communities with health campaigns in operation may reflect the nonrandom placement of health campaigns. Health campaigns are most likely to be targeted toward communities with the greatest need, which may also be the communities in which individuals are least likely to recognize and report symptoms.

The number of family planning methods available in a community was significantly associated with an increased likelihood of achieving the desired family size (coefficient = 0.061, SE = 0.017) and of experiencing labor complications (coefficient = 0.181, SE = 0.076). The presence of a secondary health facility in the community was associated with a decreased likelihood of experiencing pregnancy complications (coefficient = −0.462, SE = 0.171) and labor complications (coefficient = −0.545, SE = 0.190), while the presence of a tertiary facility was significantly associated only with the decreased likelihood of labor complications (coefficient = −0.324, SE = 0.160). The greater the number of pregnancy care facilities in the community, the lower the likelihood of a woman experiencing pregnancy complications (coefficient = −0.013, SE = 0.006) or labor complications (coefficient = −0.042, SE = 0.007).

Random Effects

The bottom of Tables 2 [triangle] through 4 [triangle] shows the changing size and significance of the household- and community-level random effects terms as each set of factors is entered into the models. Substantially large random effects terms were present at both the household and community levels for each of the outcomes. With the inclusion of only individual and household factors (Model 1), significant community-level random effects were present for all outcomes, while significant household-level random effects were present only for labor complications, pregnancy complications, and achieving desired family size.

This pattern remains with the introduction of community-level factors (Model 2) and health facility variables (Model 3). Thus, when all levels of variables (individual, household, community, and facility) were included in the models, the 5 outcomes still varied significantly between communities, and 3 of the outcomes (pregnancy and labor complications and achieved family size) varied significantly by household. Although the community-level random effect remains significant in Model 3, the magnitude of the random effect decreases with the introduction of each set of factors for all 5 outcomes. Thus, each set of factors adds to the explanation for variation in the outcomes between communities, justifying their inclusion.

The pattern of household-level random effects is more irregular, with the size of the random effect fluctuating with the addition of each set of variables. For pregnancy complications and wanted status of the previous child, the addition of community-level factors increased the size of the household random effect, while the random effect is reduced for experiencing labor complications, achieving desired family size, and RTI/STI symptom reporting. When facility factors are entered into the models, the household-level random effect increased for labor complications and decreased for the remaining outcomes.

The intrahousehold and intracommunity correlation coefficients measure the extent to which the outcomes under observation cluster at the household and community levels. The correlation coefficient is expressed as the ratio of the unit variance (household/community) to the total variance.32 The household-level correlation coefficient was larger than the community-level correlation for 3 of the outcomes: labor complications (0.34), pregnancy complications (0.33), and unwanted pregnancy (0.23). These 3 outcomes also had substantial community-level correlation coefficients: labor complications (0.27), pregnancy complications (0.21), and unwanted pregnancy (0.22). The community correlation coefficient was larger than the household correlation coefficient for the 2 remaining outcomes: RTI/STI symptoms (0.25, 0.21) and achieved desired family size (0.41, 0.10). Hence, each of the 5 outcomes clusters at both the household and community levels, although the magnitude of the clustering at each level varies between the outcomes.


The modeling of 5 reproductive wellness outcomes has highlighted the contextual influence of the community on health outcomes and behaviors and, in particular, has shown that the impact of community factors varies according to the outcome under observation. The analysis included indices to measure 4 aspects of the community: economic development, health infrastructure, reliance on traditional medicine, and sociopolitical development.

The significance and magnitude of impact of each of these indices varied across the 5 wellness outcomes, illustrating that there is no single community effect on reproductive wellness. This result highlights the need for health researchers to look further than the broad categorization of “community effects” and to recognize the variation in health impact of the differing aspects of communities. Aspects of the health service environment also proved to be significant determinants of the wellness outcomes, illustrating that it is not the mere presence of services in a community that influences health but also the characteristics of the services offered.

Several dimensions of the community were important in determining wellness outcomes. The level of socioeconomic development was associated with increased reporting of RTI/STI symptoms and decreased likelihood of experiencing pregnancy complications and unwanted pregnancies. These relationships were present after the presence of health services was controlled for. Quite likely, it reflects an association between economic development and greater health awareness, with those in more developed communities the most likely to recognize and thus report RTI/STI symptoms and pregnancy and labor complications.

Higher levels of economic development may offer environments conducive to enhancing female status through the increased availability of employment outside the home. This may increase access to and use of health services, particularly in north India, where a lack of autonomy and unequal social status have limited women’s access to health care.32 The relationship between economic development and wellness may also reflect differences between communities in the quality of health service, with the more developed communities having health services that function better, thus offering a greater prospect for improving reproductive health.

Characteristics of the health service environment were consistently related to reproductive wellness. The presence of secondary, tertiary, and pregnancy care health services was associated with a reduced likelihood of experiencing pregnancy and labor complications. A high health infrastructure index also reduced the likelihood of unwanted pregnancy. These results highlight a simple relationship: services can be used to improve reproductive wellness only if they are available in the community. A high value of the health infrastructure index was associated with a reduced likelihood of experiencing labor complications, while a negative association was found with the index of traditional health. Similarly, there is a positive relationship between the number of doctors in a community and the odds of experiencing an unwanted pregnancy.

Although one would expect a greater availability of services to reduce the likelihood of pregnancy and labor complications, this result may reflect the greater propensity for complications to be identified and treated in communities with more health facilities. The reduced likelihood of reported labor complications in communities with a greater presence of traditional health services might reflect the inability of these services to detect such complications. There may also be censoring—that is, women who rely on traditional services are more likely to suffer mortality due to complications and thus will not be captured in the survey. Additionally, communities with a reliance on traditional health services are likely to be poorer communities, and this relationship may reflect the decreased likelihood of women in poorer communities to recognize and report labor complications.

A greater number of family planning methods available in the community resulted in greater likelihood of achieving the desired family size. The importance of choice in contraceptive methods in influencing use of family planning is well documented.34,35 The ability to achieve the desired family size was also greater when there were more doctors in the community, suggesting the importance of choice in health service providers in influencing fertility decisions. Just as a greater number of methods increases the likelihood of a woman finding a suitable method, the greater number of service providers increases the likelihood of a woman finding an appropriate provider (in terms of cost and physical accessibility). The presence of health campaigns in the community decreased the likelihood of respondents reporting RTI/STI symptoms and pregnancy complications, pointing to the role such campaigns have in improving the health knowledge of communities.

Large and significant community-level random effects were found for each of the outcomes. While the finding of significant community-level variation confirms the presence of a community effect on each of the outcomes, its presence after a number of individual, household, facility, and community factors are controlled for points out that the models do not fully capture all the determinants of each outcome. Random effects typically signify the omission of explanatory factors from the model and are often attributed to the omission of less quantifiable or unobservable influences on health—for example, variations in health beliefs or genetic traits.36

The largest community-level random effects are found for 2 morbidity outcomes (pregnancy and labor complications), possibly reflecting the absence of direct measures of physiological well-being in the survey. Smaller community-level random effects are found for unwanted pregnancies, an outcome with an attitudinal dimension. Social surveys such as the PSIS and the Demographic and Health Surveys, commonly used for social demography research, collect data relating to the individual, household, and community structural factors. The magnitude of the random effects for the morbidity outcomes highlights the absence of biological markers and relevant epidemiological measures in such data sets and illustrates the need for these data collection systems to adopt a wider view of health determinants.37

It seems plausible that the presence of biological markers in the models (e.g., anthropometric measures or hemoglobin levels) would add to the explanation of both pregnancy and labor complications, reducing the random effects for each. The inclusion of biological markers would do little to reduce the random effects terms for behavior-related outcomes such as the risk of experiencing an unwanted pregnancy. The smaller random effects for the fertility-related outcomes suggest that conventional social structural variables are more relevant for explaining the determinants of these outcomes and less relevant for explaining the occurrence of morbidity outcomes.

Although large community-level random effects were found for all outcomes, large significant household-level random effects were also observed for pregnancy and labor complications. The presence of household-level random effects reflects heterogeneous excess risk between households, perhaps the product of genetic factors shared by people from the same family or the influence of the family-level traditions on health-seeking behavior. Additionally, the household-level random effects may reflect socioeconomic variations between households. In the absence of income data, the models use an asset index to estimate household socioeconomic status, which provides a less accurate measure than income data. The lack of data on household income may contribute to the household-level random effects terms observed in the models.

The intrahousehold correlation coefficients were larger than the intracommunity correlation coefficients for 3 of the 5 outcomes, indicating that greater clustering is observed at the household level than the community level. This illustrates the importance of the household context relative to the wider community context in shaping wellness outcomes. Although correlation coefficients are largest for the household, the random effects are smaller for the household. Thus, although the outcomes cluster most at the household level, the models are more successful in explaining this clustering than the community-level variation.

There are, however, a number of limitations to this study. First, the analysis relies largely on the self-reporting of RTI/STI symptoms and pregnancy and labor complications. The reporting of these symptoms and complications are likely to be highly correlated with educational attainment and measures of health awareness, thus providing a potential source of bias in the results. The presence of biological markers in the data would surmount this problem. Additionally, the cross-sectional nature of the PSIS data means that it is impossible to identify the temporal aspects of the relationships between the provision of health services and the experiencing of wellness outcomes.


This analysis illustrates the importance of understanding reproductive wellness outcomes in terms of their contextual influences. Significant community-level effects were found for each of the 5 outcomes, although their magnitude varied among outcomes. Communities influence wellness outcomes, through both the general socioeconomic environment and the characteristics of the health infrastructure, although the specific dimensions of the community and health infrastructure varied between the outcomes. The range in size and type of community effects observed suggests that there is no single community effect on reproductive wellness, and that the aspect of the community that influences wellness is specific to the outcome under observation. The results, however, indicate that for reproductive morbidity, the influence of the household on reproductive wellness is stronger than that of the community, pointing to the influence of shared biological factors and household and family dynamics on wellness outcomes.

The presence of both household and community clustering in each of the outcomes suggests that those women living in households and communities with differing characteristics will vary dramatically in their risk of experiencing reproductive wellness outcomes. This has 2 important implications. From an analytical perspective, it is therefore important when studying wellness outcomes to address them in the social context in which they occur. From a public health stance, the presence of community-level effects indicates that changes in community and service delivery characteristics (e.g., health campaigns, number and type of services) can influence wellness outcomes for whole communities. The growing interest in population health and health outcome disparities creates an imperative for such investigations.

The analysis also highlights the inadequacies of large-scale social surveys, in the absence of epidemiological measurements, to fully explain variation in physical wellness outcomes. The conventional socioeconomic variables are useful for assessing the determinants of behavioral outcomes, but there is a need for large population-based surveys to collect data that can measure the epidemiological influences on health as well, thereby permitting a more complete understanding of the array of factors that jointly influence reproductive wellness.


The Andrew W. Mellon Foundation made training support available to the Carolina Population Center, which enabled a substantial part of the research reported in this article.

We acknowledge the fieldwork efforts of the following organizations: the Centre for Population and Development Studies, Hyderabad; the Operations Research Group, New Delhi; the Marketing and Research Group, New Delhi; and the Indian Institute for Health Management Research, Jaipur.


R. Stephenson and A. O. Tsui jointly planned the study, performed the statistical analyses, and wrote the article.

Peer Reviewed


1. 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]
2. Duncan C, Jones K, Moon G. Context, composition and heterogeneity: using multilevel models in health research. Soc Sci Med. 1998;46:97–117. [PubMed]
3. Grady WR, Klepinger DH, Billy JOG. The influence of community characteristics on the practice of effective contraception. Fam Plann Perspect. 1993;25:4–11. [PubMed]
4. Chacko E. Women’s use of contraception in rural India: a village-level study. Health Place. 2001;7:197–208. [PubMed]
5. DeGraff D, Bilsborrow RE, Guilkey DK. Community-level determinants of contraceptive use in the Philippines: a structural analysis. Demography. 1997;34:385–398. [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. Manda SOM. Unobserved family and community effects on infant mortality in Malawi. Genus. 1998;54:143–164. [PubMed]
8. 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]
9. Anderson R. Mortality effects of community socioeconomic status. Epidemiology. 1997;8:42–47. [PubMed]
10. Pebley AR, Goldman N, Rodriguez G. Prenatal and delivery care and childhood immunization: do family and community matter? Demography. 1996;33:213–247. [PubMed]
11. Congdon P. The impact of area context on long term illness and premature mortality: an illustration of multilevel analysis. Reg Stud. 1995;29:327–344.
12. 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]
13. Entwisle B, Casterline JB, Sayed AA. Villages as contexts for contraceptive behavior in rural Egypt. Am Sociol Rev. 1989;54:1019–1034.
14. Entwisle B, Mason WM. Multilevel effects of socioeconomic development and family planning programs on children ever born. Am J Sociol. 1985;91:616–649.
15. 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.
16. Diez Roux AV. Investigating neighborhood and area effects on health. Am J Public Health. 2001;91:1783–1789. [PMC free article] [PubMed]
17. Goldstein H. Multilevel Statistical Models. London, England: Edward Arnold; 1995.
18. DiPrete TA, Forrostal JD. Multilevel models: methods and substance. Annu Rev Sociol. 1994;20:331–357.
19. Casterline JB. Community effects on fertility. In: Casterline JB, ed. The Collection and Analysis of Community Data. Voorburg, Netherlands: International Statistical Survey; 1985:65–67.
20. Potter JE. Effects of societal and community institutions on fertility. In: Bulatao R, Lee R, eds. Determinants of Fertility in Developing Countries. Vol 2. New York, NY: Academic Press; 1983:627–625.
21. Hirschman C, Guest P. Multilevel models of fertility determination on four south-east Asian countries: 1970–1980. Demography. 1990;27:369–396. [PubMed]
22. Alauddin M. Rural development and family planning behavior in Bangladesh villages. Bangladesh Dev Stud. 1979;7:4–58. Cited by: Tsui AO. Community effects on contraceptive use. In: Casterline JB, ed. The Collection and Analysis of Community Data. Voorburg, Netherlands: International Statistical Survey; 1985:77–90.
23. Dixon-Meuller R. Gender Inequalities and Reproductive Health: Changing Priorities in an Era of Social Transformation and Globalization. Belgium: International Union for the Scientific Study of Population; 1999. Policy and Research Paper No.16.
24. Greenwell KF. Contraceptive method mix menu: providing healthy choices for women. World Health Stat Q. 1996;49:88–93. [PubMed]
25. Stokols D. The social ecological paradigm of wellness promotion. In: Jamner MS, Stokols D, eds. Promoting Human Wellness. Berkeley: University of California Press; 2000:21–37.
26. Census of India, 2001. New Delhi: Government of India. 2001. Available at: http://www.censusindia.net/results. Accessed October 9, 2003.
27. International Institute for Population Sciences. National Family Health Survey (NFHS-2), India. 1998–99. Bombay, India: International Institute for Population Sciences; 2000.
28. Bloom S, Lippeveld T, Wypij D. Does antenatal care make a difference to safe delivery? A study in urban Uttar Pradesh, India. Health Policy Plann. 1999;14:38–48. [PubMed]
29. MEASURE Evaluation. PERFORM System Indicators Survey Uttar Pradesh 1995–1996. Chapel Hill, NC. Available at: http://www.cpc.unc.edu/projects/measure/data/india/description.html. Accessed August 14, 2003.
30. Skinner C, Holt D, Smith TF, eds. The Analysis of Complex Surveys. 1989. New York, NY: Wiley; 1989.
31. Filmer D, Pritchett L. Estimating Wealth Effects Without Expenditure Data or Tears: An Application to Educational Enrollments in States of India. Washington, DC: Development Economics Research Group, The World Bank; 1988. World Bank Policy Research Working Paper No. 4. [PubMed]
32. Im S, Ginola D. Mixed models for binomial data with an application to lamb mortality. Appl Stat. 1998;37:196–204.
33. Visaria L, Jejeebhoy S, Merrick T. From family planning to reproductive health: challenges facing India. Int Fam Plann Perspect. 1999;25:S44–S49.
34. Pariani S, Heer DM, Van Arsdol MD. Does choice make a difference to contraceptive use? Evidence from East Java. Stud Fam Plann. 1991;22:384–390. [PubMed]
35. Young MK, Kols A, Mucheke S. Informed choice and decision-making in family planning counseling in Kenya. Int Fam Plann Perspect. 1998;24:4–11. [PubMed]
36. Das Gupta M. Death clustering, mother’s education and the determinants of child mortality in rural Punjab, India. Popul Stud. 1990;44:489–505.
37. Finch CE, Vaupel JW, Kinsella K, eds. Cells and Surveys: Should Biological Measures Be Included in Social Science Research? Washington, DC: National Academy Press; 2001. [PubMed]

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...