Poverty, prenatal care, and infant health in Puerto Rico
Oropesa, R S
ABSTRACT: Using data from a survey administered to a representative sample of mothers who gave birth in Puerto Rico in 1994-95, we investigate whether prenatal care and infant health outcomes are associated with family poverty and neighborhood poverty. The results show that infant health outcomes are unrelated to both family poverty and neighborhood poverty, despite the association of family poverty with the adequacy of prenatal care and the content of prenatal care. However, the poverty paradigm does receive some support using measures of participation in government programs that serve the low-income population. Women who rely on the government to fund their medical care are more likely than women who rely on private health insurance to have an infant death. They are also less likely to receive the highest levels of prenatal care. Nonetheless, targeted government programs can have an ameliorative impact. The analysis shows that participants in the Women, Infants, and Children (WIC) program are more likely than non-participants to receive superior levels of prenatal care and are less likely to have negative infant health outcomes.
During the last half of the twentieth century, both scholars and policy makers focused attention on the linkage between economic development and poverty in Latin America. The macro-level promise of economic development for reducing poverty is linked to the enhancement of economic, educational, transportation, communications, and health care infrastructures. At the microlevel, the promise of economic development lies in increased individual access to resources that reduce poverty. Poverty, in turn, is of interest for reasons other than material deprivation per se. For example, poverty compromises the health and well– being of the most vulnerable members of any society-infants and children.
Low birth weight and infant mortality are commonly accepted indicators of infant health. Indeed, Gortmaker and Wise (1997) describe infant mortality metaphorically as a “barometer” of population health that is sensitive to poverty. According to the poverty paradigm, poverty exposes women and children to a range of factors that increase the risks of inadequate medical care and negative infant health outcomes. Exposure to risks occurs through the family and neighborhood environments in which individuals live. Indeed, those who experience both family and neighborhood poverty may be in “double jeopardy” of experiencing negative health outcomes.
Although the poverty paradigm is intuitively appealing, efforts to document the relationships between poverty, medical care, and infant health outcomes in Latin America are fraught with difficulties. These difficulties stem from the inadequacy of available vital records data for the measurement of poverty and the limited generalizability of small-scale surveys administered to local populations. Using data from a survey of mothers in Puerto Rico, we are able to overcome these limitations. Our analysis answers the following set of interrelated questions: Are impoverished women more likely than women who are not impoverished to experience negative infant health outcomes such as low birth weight and infant mortality? Are impoverished women more likely than women who are not impoverished to have inadequate prenatal care utilization and prenatal care content? Are differences in the risks of negative infant health outcomes by poverty status due to differences in prenatal care? Do the combined effects of family poverty and neighborhood poverty put women in “double jeopardy” of experiencing negative outcomes and receiving inadequate prenatal care?
Because efforts to reduce the potentially deleterious consequences of poverty for birth outcomes are likely to focus on access to the health care delivery system and the delivery of services through publicly funded programs, we pay particular attention to the role of government programs in providing resources to help indigent populations. Thus, we ask whether public medical insurance and public programs that target at-risk mothers play a role in prenatal care and infant health. After a brief description of the Puerto Rican context, we provide a fuller discussion of how poverty and government programs may affect both infant health and prenatal care.
FAMILY POVERTY AND INFANT HEALTH1
The infant mortality rate for Puerto Rico was 8.9 (per 1,000 live births) in 1996 (U.S. Bureau of the Census, 1998).2 Although this is only slightly higher than the rate of 7.5 for the United States (Population Reference Bureau, 1999), the current level of infant mortality is the product of major declines in mortality over a short period of time in Puerto Rico. The infant mortality rate stood at 43.3 in 1960, 28.5 in 1970, and 18.5 in 1980 (U.S. Bureau of the Census, 1998).
The level and trend in infant mortality in Puerto Rico are of interest, especially against the backdrop of the level and trend in poverty. In the 1990 Census, approximately 66 percent of children (Oropesa and Landale, 2000) and 57 percent of families (Rivera-Batiz and Santiago, 1996) lived below the official poverty line. Moreover, the percent of families in poverty has exhibited modest declines over time, from 62.8 percent in 1970 and 59.8 percent in 1980 to 57.3 percent in 1989 (Rivera-Batiz and Santiago, 1996). These numbers suggest that a reduction in poverty is not a driving force in the long-term improvement in infant health in Puerto Rico.
Although the magnitude of the change in infant mortality is greater than might be expected from the magnitude of the change in poverty over time, some cross– sectional evidence suggests that mortality is generally related to the economic resources of both individuals and the communities in which they live (Anderson et al., 1996; LeClere, Rogers, and Peters, 1997; Starfield et al., 1991 ). Indeed, there are several reasons why poverty should be an important determinant of infant health outcomes such as infant mortality and low birth weight. These reasons pertain to the ability to afford adequate medical care, nutrition, demographic circumstances, exposure to stressful life circumstances, and social ties.3
Medical Care and Nutrition: The impoverished lack monetary resources to purchase goods and services that reduce the likelihood of negative outcomes for children. Specifically, the receipt of medical care should be related to poverty status because the practice of medicine is market driven. Access to medical care requires income to pay for the direct costs of care and the indirect costs of accessing care (e.g., transportation and child care). Income also potentially affects the adequacy of nutrition, which in turn affects resistance to disease and pregnancy outcomes.
Stressful Life Circumstances: Poverty and residence in distressed neighborhoods also affect exposure to stressors that have negative implications for physical and mental health (Krieger et al., 1993). Economic disadvantage increases exposure to the stress of family instability, as well as the strains of “making ends meet” and inadequate living environments. McLean et al. (1993) suggest that such stressors increase the risk of undesirable birth outcomes such as low birth weight and pre-term delivery. The mechanisms for these relationships involve both the central nervous system and the immune system (Kelly, Hertzman, and Daniels, 1997). Stress may change neuroendocrine and immune system functioning, as well as promote negative health-related behaviors such as smoking (Chomitz et al., 1995).
NEIGHBORHOOD POVERTY AND INFANT HEALTH
Up to this point, we have focused on micro-level links between poverty and health. Such an emphasis, however, decontextualizes the individual (Diez-Roux, 1998). Individual choices are made within social and economic contexts that extend beyond the individual. Potentially relevant contexts include the larger living environments of cities and their constituent communities and neighborhoods. These environments are important as the geographic loci of various services and facilities that comprise the public health infrastructure, as well as the various institutions and people with whom individuals come into contact as they carry out their daily round of activities. The level of poverty in the local community may influence health by affecting the socioeconomic status of individual residents, but also the “social, service, and physical environments . . . shared by residents” (Robert, 1999).
While the aforementioned linkages between neighborhood poverty and infant health are potentially important, skeptics have argued that there is a lack of correspondence between the physical location of patients and the physical location of health care services. Expectant mothers do not necessarily receive health care in the areas in which they live because hospitals and clinics are not heavily concentrated in residential areas. Moreover, transportation between locations is not necessarily problematic for most pregnant women (Oropesa et al., 2000). Thus, residence in a high poverty area might not have an impact on infant health through physical access to the health care infrastructure.
At the same time, recent empirical research connects neighborhood poverty to both general health and infant health (Anderson et al., 1996; Fiscella and Franks, 1997; Haan, Kapland and Camacho, 1987; O’Campo et al., 1997; Waitzman and Smith, 1998). Previous studies suggest that the neighborhood environment may affect outcomes independently of individual socioeconomic characteristics for reasons that extend beyond the neighborhood social service infrastructure (Diez-Roux, 1998). Neighborhood poverty may affect health through linkages with crime, stress, social isolation, fear of public travel, social capital, trust, pollution, and exposure to persons who engage in unhealthy lifestyles (Robert, 1999: 494). The neighborhood environment may also work in conjunction with the family environment to exacerbate negative outcomes. It should be noted that much research is limited by the inability to include measures of the poverty status of individuals and measures of areal poverty in the same analysis (see below).
RESEARCH ISSUES
The foregoing has established theoretical and empirical grounds for hypothesizing positive relationships between poverty and both low birth weight and infant mortality. However, these hypotheses may not be supported because Puerto Rico, as a Commonwealth of the United States, participates in federal programs that provide income and medical assistance to the low– income population. The poor participate widely in programs such as Medicaid and the Special Supplemental Food Program for Women, Infants, and Children (WIC) that are designed to ensure access to medical care and to provide medical referrals, nutritional information, and nutritional supplements to the poor.
The empirical evidence is also less convincing than it appears at first glance. In reviewing the general literature on mortality, it was just ten years ago that Starfield et al. (1991: 1168) noted that “there have been no population-based studies that used individual data rather than ecologic data to characterize family income.” Despite notable exceptions (Frisbie et al., 1997; Moss and Carver, 1998), similar observations could be applied to studies of infant health. For example, most studies of infant mortality per se rely on vital records, but vital records do not include measures of income. Because vital records lack information that can be used to measure individual or family income, researchers tend to “aggregate up” and analyze relationships across census tracts or link individual vital records to information on the areas in which individuals live (Clarke, Farmer, and Miller, 1994; Guest, Almgren, and Hussey, 1998; O’Campo et al., 1997). The former strategy runs the risk of committing the “ecological fallacy” and the latter strategy overestimates the effect of community income because individual or family income is excluded from models.
The importance of taking family income into account is demonstrated by Moss and Carver’s (1998) analysis of infant mortality with the 1988 National Maternal and Infant Health Survey (NMIHS). Their analysis shows that the risk of infant mortality from both endogenous and exogenous causes of death is inversely related to family income in models that control for government program participation, health behaviors, and bio-medical risks. Moreover, the resilience of income in models of infant mortality that include bio-medical risk factors is consistent with Frisbie et al.’s (1997) demonstration with the NMIHS that financial resources do not have a direct effect on bio-medical indicators of compromised pregnancy outcomes– intrauterine growth retardation and prematurity. Neither of these studies investigates the association between outcomes and the income levels of the neighborhoods where people live.
The lessons to be drawn from previous studies that focus on income can be extended without difficulty to our effort to focus on family poverty and neighborhood poverty. The effect of neighborhood poverty cannot be properly understood from statistical models that exclude family poverty. To demonstrate the role of neighborhood poverty, it is necessary to show that neighborhood poverty has a direct effect on an outcome of interest in models that include family poverty or to show that neighborhood poverty interacts with family poverty. The latter is necessary to document support for a “double jeopardy” hypothesis-“that living in communities with lower socioeconomic status levels may be particularly detrimental to the health of individuals who have lower socioeconomic position themselves” (Robert, 1999: 497). In other words, the double jeopardy hypothesis suggests that the risks of negative outcomes among those who are impoverished should increase with the poverty of the areas in which they live.
We address these issues using data collected from the 1994-95 Puerto Rican Maternal and Infant Health Survey. The primary research issue concerns the relationships between family and neighborhood poverty and two infant health outcomes: low birth weight and infant mortality. Another research issue focuses on the relationships between poverty and prenatal care utilization and content. Women who live below the poverty threshold should be more likely than those who live above the poverty threshold to have inadequate levels of prenatal care utilization; they should also receive fewer prenatal care services and less advice. This is an important issue because public health efforts to reduce the consequences of poverty for infant health focus largely on prenatal care utilization, despite the fact that many scholars are skeptical of the role of prenatal care in the reduction of infant mortality. Consequently, prenatal care is a potential mechanism for the association between poverty and infant health outcomes.4
As noted above, there are many other possible mechanisms through which poverty is linked to infant health outcomes. An additional factor that needs to be considered is participation in government programs such as Medicaid that are designed to mitigate the negative consequences of inadequate financial resources for infant health. Puerto Rico is well integrated into the federal Medicaid program. Around the time of the survey, approximately 39 percent of the island population had been evaluated, and 79 percent of those evaluated had been certified as medically indigent and eligible for Medicaid. Medicaid recipients were typically treated in health centers run by the Puerto Rico Department of Health, but a series of health care reforms were initiated in 1994. The purpose of these reforms was to “guarantee to all Puerto Ricans, regardless of their economic status, quality health care and related services through a managed care model . . . for the medically-indigent and underinsured population to have access to private medical providers through a health insurance plan financed by the Government mainly with state funds” (Puerto Rico Department of Health, 1995: 6). Thus, a greater prevalence of negative outcomes among those who finance their medical care through government programs in this study should reinforce efforts to adopt models for health care delivery based on the private sector.
Another government-funded program that may have a positive impact on infant health is WIC. This is a means-tested program that is designed to foster healthy pregnancies and healthy children during the first five years of life. The program provides expectant mothers whose family incomes are at or less than 185 percent of the poverty threshold with medical referrals, information about health and nutrition, vouchers for food, and a health screening. Although some scholars suggest that WIC may not have an impact on various outcomes (Besharov and Germanis, 1999), others suggest that WIC reduces the incidence of undesirable health outcomes such as infant mortality (Moss and Carver, 1998; Stockbauer, 1987). Needless to say, most of the literature is based on the mainland United States; to our knowledge, no study examines the impact of WIC participation on infant health outcomes in Puerto Rico.
DATA AND METHODS
This study utilizes data from the Puerto Rican Maternal and Infant Health Survey (PRMIHS). The PRMIHS is a survey of 2,763 mothers of infants sampled from birth and death certificates in 1994-95 from the island of Puerto Rico and six administrative areas on the U.S. mainland (New York City and the states of Florida, Connecticut, Massachusetts, Pennsylvania, and New Jersey). The birth sample, which was drawn from birth certificates, is representative of infants born in the study area. It was stratified by infant health outcome (low versus normal birth weight) and the month in which the birth occurred. A separate death sample was drawn from death certificates to provide sufficient death cases for the study of infant mortality.
After the samples were drawn from vital records, address information was used to locate mothers in order to secure their participation in the survey. In Puerto Rico, approximately 83 percent of those who were selected into the study were located and agreed to be interviewed in person. Mothers of infants who were not of “Puerto Rican descent” were screened from the survey.
The present study is restricted to births and deaths that occurred in Puerto Rico. The Puerto Rico birth sample is weighted to represent all infants born in Puerto Rico in 1994-1995. The weighted birth sample is used for all analyses except the analysis of infant mortality, which is based on both the birth and death sampies. The weights used in our analyses were re-scaled to preserve the unweighted N’s for the birth sample (N = 576) and the combined birth and death samples (N = 974) for tests of significance. These N’s represent the responses of all mothers with singleton births whose census tract of residence could be identified from information on the birth certificate (discussed below).
INFANT HEALTH AND PRENATAL CARE
This analysis focuses on infant health outcomes and prenatal care. The infant health outcomes examined are low birth weight and infant mortality. Low birth weight is a dichotomous measure that contrasts infants who weighed less than 2500 grams at birth (1) with those who weighed 2500 or more grams at birth (0). Our measure of infant mortality contrasts infants who died during the first year of life (1) with those who survived the first year (0).
Two dimensions of prenatal care are measured-utilization and content. Utilization is tapped by the Adequacy of Prenatal Care Utilization Index (APCUI) (Kotelchuck, 1994). The APCUI is based on the month prenatal care began and the ratio of the observed number of prenatal care visits to the expected number of visits given gestational age, according to standards developed by the American College of Obstetricians and Gynecologists (ACOG). “Inadequate” care is defined as care initiated later than the first trimester or care with less than half of the expected number of visits to a provider. “Intermediate” care is received by those initiating care in months 1-4 who have between 50 percent and 79 percent of the recommended visits. “Adequate” care is initiated in months 1-4 with the observed number of visits between 80-109 percent of the number recommended. “Intensive” (or “adequate plus”) care is defined as care that exceeds ACOG recommendations. This is care that begins in months 1-4 of the pregnancy with the observed number visits exceeding the expected number of visits by at least 10 percent. Women who did not receive care are included with those who received inadequate levels of care. Three dummy variables represent these four categories of care, with adequate care serving as the reference.
The APCUI is widely used, but the causal inferences that can be made about the consequences of intensive prenatal care for infant health are tenuous due to potential endogeneity. Those who receive intensive care may be more likely to have deleterious outcomes because intensive care is required to monitor high-risk conditions that are diagnosed during the course of prenatal care itself. Thus, intensive prenatal care may result from conditions that ultimately eventuate in negative outcomes. Intensive care might also contribute to negative outcomes if it reflects overly aggressive treatment.
Prenatal care content is measured in terms of services and advice received. All respondents were asked whether they received the following services from their medical care provider: a physical exam, pelvic exam, pap smear, and ultrasound/ sonogram. The respondents also indicated whether their medical provider took a health history of the mother and a family health history that included queries about mental retardation or physical deformities. Given that an additive index for these items is heavily skewed, we employ a dichotomy with those who received all services coded as I and those who received less than the full array of services coded as 0.5
Responses to questions about whether or not the medical care provider advised the respondent about using vitamin/ mineral supplements, nutrition, weight gain during pregnancy, and using over– the-counter drugs are also utilized. Those who received advice in all areas (coded as 1) are contrasted with those who did not receive all advice (coded as 0).6
POVERTY
Our measure of poverty is based on the income-to-needs ratio. The income– to-needs ratio expresses the financial circumstances of a family as the ratio of its total income to the poverty threshold established for families of the same type and composition by the U.S. Federal Government.7 Indeed, 48 income thresholds take into account the size of household, the number of children, and the age of the householder (over or under 65). This measure is treated here as a categorical variable because of a special interest in those who are most impoverished, even though inferences are not sensitive to whether it is treated as a continuous or categorical variable. Individuals who live in families whose pre-tax income (excluding the value of noncash benefits) is less than 50 percent of the poverty threshold are in “deep poverty.” This group serves as the reference category for those whose income is 50 to 99 percent, 100 to 149 percent, and 150 or greater percent of the appropriate threshold.
To measure neighborhood poverty, we coded the addresses provided on the birth certificates to the census tract level.’ Census tracts are small units within counties that have permanent visible boundaries, such as streets and rivers. When first established, census tracts in a county can range from several blocks to several square miles in size, but the overall average for the county must be about 4,000 people (with a range from 2,500 to 8,000). The population of each tract is also required to be relatively homogeneous on various socioeconomic characteristics and living conditions. Using data from the 1990 U.S. Census, neighborhood poverty is measured as the logged percentage of the tract population that falls below the poverty threshold.9
A caveat is in order regarding the measurement of family poverty and neighborhood poverty-federal poverty thresholds do not take differences in the cost of living across areas into account. This issue is pertinent to nearly all studies that examine the causes and consequences of poverty, including studies of Puerto Rico. However, there is no straightforward resolution because no study to our knowledge proposes and critically evaluates alternative thresholds. This is undoubtedly why the federal thresholds are used in the benchmark studies of poverty in Puerto Rico (e.g., Rivera-Batiz and Santiago, 1996; Oropesa and Landale, 2000).10 In addition, the federal poverty thresholds created for the contiguous 48 states are relevant here because they are used to define eligibility for many federal assistance programs in Puerto Rico, including WIC (U.S. Department of Agriculture, 2001).11
UTILIZATION OF GOVERNMENT PROGRAMS
Puerto Rico benefits from many federal assistance programs that may have a bearing on prenatal care utilization and health outcomes. The respondents indicated whether they participated in WIC and how they paid for their medical care during pregnancy. WIC participants are coded as 1 and non-participants are coded as 0 for the measure of WIC participation. Method of payment is measured with two dummy variables. Women who relied on private insurance and other sources are contrasted separately with those who relied on the government to pay for medical expenses.12
COVARIATES
Several sets of covariates are included in multivariate models as controls:
Human Capital: Completed years of education is the indicator of human capital.
Social Capital: Social capital refers to resources that emanate from social relationships that are based on trust and felt obligations. One indicator of social capital is whether there was anyone that the respondent could rely on for emotional support or advice during the pregnancy. Other measures indicate the number of relatives who lived nearby and whether the respondent lived with any members of her extended family. Marriage creates social capital as well. Those who are single and cohabiting are contrasted separately with those who are married.
Demographic Characteristics: Demographic characteristics include age and the number of previous births.
Lifestyle Behaviors: Lifestyle behaviors are denoted by indicators of whether the respondent smoked or whether she drank alcohol. Another measure records the number of different types of stressful events that occurred during the pregnancy: 1) someone very close had drinking or drug problem; 2) husband or partner went to jail; 3) homelessness; 4) job loss; 5) bills that could not be paid; 6) involvement in physical fights; 7) hit or physically hurt by partner.
Biomedical Risk Factors: Biomedical risk factors include whether the mother had at least one previous miscarriage, whether the mother had at least one previous low-birth-weight birth, whether the mother had low weight gain (
ANALYSIS PLAN
Given the complex sampling design, the analysis of these data is not straight– forward. Indeed, there are several interrelated issues that are related to the sample design and the treatment of missing data. Specifically, we have complete data for our measures of birth weight and infant mortality, but the data are incomplete for covariates. In general, observations with missing data are not rejected from the analysis (an exception is discussed below). Instead, we impute missing values with a Markov chain Monte Carlo method that employs a Bayesian noninformative prior distribution (see Schafer, 1997). This procedure generates N datasets with imputed values substituted for missing data that can be subsequently analyzed with complete-data methods. We conducted analyses with five different sets of imputations using SUDAAN (release 8.2) to generate the correct parameter estimates and standard errors for data collected using a stratified sampling design. The five sets of parameter estimates and standard errors were subsequently combined, following Rubin’s rules (Rubin, 1987), to arrive at parameter estimates, standard errors, and p-values that reflect the underlying uncertainties about the missing data (Schafer, 1997).
In addition to taking the complex sampling design into account, SUDAAN was used to deal with another issue-the clustering of observations within the census tracts that are used to calculate poverty rates. If uncorrected, this may deflate the standard errors of parameter estimates and produce misleading test statistics. The statistical tests that we report are based on standard errors that have been corrected for clustering within census tracts.13
The significance criterion has also been adjusted due to the relationship between statistical power and sample size. Specifically, the power of a statistical test is a positive function of sample size, as well as the magnitude of the true effect and the significance criterion itself. Because the power of statistical tests with our sample may be too low using the conventional .05 significance criterion to reject the null if the true effect is weak, we also identify coefficients that are significant at the .10 level.
The analysis focuses primarily on logistic regression models of prenatal care utilization, prenatal care services, prenatal care advice, and infant health outcomes. Each dependent variable is regressed on the central variables of interest-family poverty, neighborhood poverty, WIC participation, and method of payment-along with the other covariates that are of secondary interest. The analyses build sequentially upon one another. After examining the adequacy of prenatal care utilization, we include adequacy of prenatal care utilization with all of the covariates to predict the services and advice received from prenatal care providers. Adequacy of prenatal care utilization and both dimensions of the content of care are then included in analyses of low birth weight, which in turn is included in the analysis of infant mortality. It should be noted that these analyses are presented for the total sample and a subsample restricted to those with family incomes below poverty to gain special insights into the circumstances of those who are impoverished. To maintain parsimony in the presentation of results, results will be presented for only the main variables of interest. The results for the other covariates are available on request.
RESULTS
Table 1 provides preliminary insights into the relationships between poverty, prenatal care, and infant health outcomes. The pattern of results for low birth weight and the infant mortality rate (infant deaths per 1,000 live births) initially appear to be weakly consistent with the expectation that negative outcomes are associated with poverty. Approximately 10 percent of mothers in deep poverty gave birth to an infant weighing less than 2,500 grams. The percent low birth weight for infants of more affluent mothers ranged from 7.9 to 8.8. Similarly, the infant mortality rate is highest for those in deep poverty at 12.2, while the infant mortality rates for the other groups hover between 9 and 11. Obviously, these differences are small, and the tables that follow will show that they are not statistically significant.
Adequacy of prenatal care is clearly related to family poverty status. Approximately 46 percent of pregnant mothers in deep poverty received adequate (31 percent) or intensive care (15 percent). This stands in sharp contrast to the percentages for those in other categories of the income-to-needs ratio. Approximately 65 percent of mothers who were just below poverty, 70 percent of mothers who were slightly above poverty, and 80 percent of mothers whose family incomes were at least 1.5 times the poverty threshold received the highest levels of care. Moreover, the difference between those in deep poverty and those in other categories is reflected in the percentages receiving intermediate and inadequate care. About 24 percent of those in deep poverty had inadequate levels of prenatal care utilization, compared to about 5 percent of those with incomes at least 1.5 times the poverty threshold and 8-12 percent of those in the two middle categories.
Differences in the content of prenatal care services are less pronounced than differences in prenatal care utilization. About 58 percent of the respondents in deep poverty reported receiving all prenatal care services. This is only slightly lower than the 61-66 percent of respondents in the other three categories. As for prenatal care advice, about 70 percent of those in deep poverty and 54 percent of those just below poverty reported receiving all types of prenatal care advice. These figures are substantially lower than those for women who are above poverty. Nearly 87 percent of those whose income-to-poverty ratio exceeds 1.5 received all types of prenatal care advice.
The remaining rows demonstrate the variation in neighborhood poverty and government program participation by family poverty status. As might be expected, the typical mother in deep poverty lives in a census tract in which two-thirds of the population is also in poverty. This declines to 61 percent for the two middle categories and 54 percent for those in the most affluent group. As for government program participation, nearly all of those with incomes below one-half of the poverty threshold received WIC (94 percent) and relied on government programs to pay for prenatal care (85 percent). In contrast, WIC participation is 58 percent and reliance on government programs to pay for prenatal care stood at 15 percent for those in the most affluent group.
Additional insights into the relationships between poverty, prenatal care, and health outcomes are available from the logistic regressions presented in Tables 2-4. Each of these tables presents at least two models. Model 1 provides bivariate odds ratios that describe the association between poverty and a given outcome or behavior. Model 2 adds the full set of covariates. Model 2 is presented for the full sample and for those with incomes that are below the poverty threshold. This latter restriction provides an opportunity to test the central claim of the double– jeopardy hypothesis: the likelihood of negative outcomes among those who are impoverished will increase with neighborhood poverty. In keeping with a framework that suggests prenatal care utilization and content are possible antecedents of infant health outcomes, we first discuss the results for prenatal care.
Table 2 presents odds ratios from multinomial logistic regressions for the adequacy of prenatal care utilization. As noted in the previous table, Model 1 is consistent with the poverty paradigm. Using adequate care as the reference category, we see that the odds of receiving inadequate care among those whose incomes are at least 1.5 times the poverty threshold are one-sixth the odds generated by those in deep poverty (odds ratio = .16). The odds of inadequate care for those who are near poverty (1-1.49 the poverty threshold) and those who are .5-.99 of the poverty threshold are about one-fourth and one-half the odds of those in deep poverty, respectively.
The pattern of coefficients for intermediate care is similar to that for inadequate care, despite the fact that the only contrast that achieves statistical significance for intermediate care is for the most affluent category. With an odds ratio of 2.3, the most affluent are also significantly more likely than those in deep poverty to receive intensive care. The odds ratios for those who fall nearer the poverty threshold exceed 1.5, but they are non– significant.
Although the odds ratio for intensive care among the most affluent group is not significant and less than one in Model 2, the overall pattern of results for inadequate and intermediate care is consistent with the bivariate results after the covariates are controlled. Most notably, the most affluent women are less likely than those who are the least affluent to receive inadequate care (.27, p = .063) and intermediate care (.39, p = .099). The remaining coefficients for the inadequate– adequate contrast are consistent as well, but they are not significant. Additional analyses suggest that these associations become weaker because poverty is associated with prior fertility. Poverty and fertility are positively associated and women with higher fertility are less likely to have adequate prenatal care utilization.
Model 2 also demonstrates the role of areal poverty and program participation. Adequacy of prenatal care utilization is not associated with areal poverty, but women who participate in WIC are less likely than women who do not participate in WIC to have inadequate prenatal care utilization. This is important because this odds ratio (.22) is net of family poverty and method of payment, as well as all of the other covariates. At the same time, participation in government programs is not necessarily advantageous. On the contrary, the odds of intensive prenatal care utilization among women with private health insurance are about three times the odds for women who rely on government programs.
Parameter estimates for those who live below the poverty line reinforce conclusions about the benefits of WIC for reducing the likelihood of inadequate utilization and the benefits of private insurance for the receipt of intensive care. However, the analysis of the below– poverty population is primarily of interest for the light it sheds on the double jeopardy hypothesis. The odds ratio generated by the log of the poverty rate for inadequate care exceeds one and is consistent with expectations, but does not approach statistical significance. The risk of receiving sub-adequate care among those who are impoverished does not increase with neighborhood poverty. The only odds ratio to approach significance is that for intensive care, but this contrast is not of central interest and the direction of the association is counterintuitive.14
One question that remains unanswered by this analysis is: Why is the higher likelihood of intensive care among the most affluent group shown in Model 1 reduced to insignificance in Model 2? A supplementary analysis (not shown) indicates that this is due in part to the association between poverty and participation in government programs for medical care. In particular, the relatively affluent are more likely to receive intensive care because they rely on private health insurance for medical care. Private insurance, in turn, increases the likelihood of intensive prenatal care utilization. The controls for method of payment and WIC are also responsible for reducing the other coefficients for family poverty (i.e., 1-1.49, .5-.99 for the inadequate-adequate contrast) to insignificance as well. Thus, government programs play a role in reducing socioeconomic differentials in prenatal care utilization in Puerto Rico, but they do not eradicate these differentials.
Table 3 shows the impact of poverty and adequacy of prenatal care on the services and advice that women received from prenatal care providers. The overall pattern of results in Model 1 for the receipt of routine services is consistent with expectations. However, the odds ratios for family poverty do not achieve statistical significance in either Model 1 or Model 2 for the full sample. The receipt of routine services also is not associated with neighborhood poverty for the full sample or the sample restricted to those who live below the poverty threshold. Consistent with the previous results, impoverished families in high poverty areas are not necessarily in double jeopardy of being shortchanged on the routine services examined in this study.
The pattern of odds ratios for the adequacy of prenatal care utilization index is consistent with expectations; that is, those with intensive care should be more likely than those who receive adequate care to receive a full range of services and those who receive intermediate and inadequate levels of care should be less likely to receive a full range of services. Similarly, the pattern for the method of payment is consistent with expectations. However, none of these odds ratios is significant.15 WIC participation is the only variable to generate a significant parameter estimate. Women who participated in WIC were more likely than non– participants in WIC to receive the full range of services (1.89, p = .085). Interestingly, WIC is not significant for those who are below poverty.
Columns 4-6 shift attention to advice received. Model 1 indicates that the affluent are more likely than the impoverished to receive advice from prenatal care providers. The odds of receiving the full range of advice for those whose family incomes exceed the poverty threshold by 50 percent or more are twice the odds of those whose family incomes are less than 50 percent of the poverty threshold (2.19). This odds ratio is reduced to insignificance in Model 2, after the various covariates are controlled. Interestingly, the introduction of controls does not affect an idiosyncratic finding (see also Table 1) women with family incomes that are .5 to .99 of the poverty threshold are less likely than women in deep poverty to receive all types of advice.16 Advice is also unrelated to areal poverty, although odds ratios less than one are consistent with expectations.
These results raise the question of what accounts for the reduction in the parameter estimate from 2.19 to 1.12 for those whose family income is at least 1.5 times the poverty threshold. This was explored with additional models (not shown) that included poverty along with each variable and set of variables separately. These analyses indicate that it is easier to eliminate suspected variables than to identify the variable that is the proverbial “smoking gun.” Various lifestyle behaviors (e.g., stress, smoking, alcohol) and medical risk factors are not responsible for this reduction. The odds ratio for the most affluent group is not significant in models that include prenatal care utilization, education, fertility, and method of payment. Of these variables, however, only prior fertility is associated with advice received. The impoverished tend to have more children than the affluent and those with more children are less likely to receive all the forms of advice examined here.
Model 2 for the poverty population also reveals that the receipt of advice from prenatal care providers is unrelated to the other covariates except WIC participation. The odds for WIC participants are over three times the odds for non– participants (3.16, p = .076). This finding is consistent with the focus of the dependent variable and WIC; the dependent variable focuses primarily on nutrition– related advice and WIC provides food supplements and nutrition counseling.
Table 4 focuses on low birth weight (columns 1-3) and infant mortality (columns 4-8). In contrast to the results for prenatal care utilization and advice, both family poverty and areal poverty are not significant in any model for low birth weight. The only parameter estimates that are significant are those for intensive care and WIC. For reasons noted above, women who receive intensive care are more likely than women who receive adequate care to have a low-birth-weight infant. At the same time, the risk of having a low-birth-weight infant is unrelated to the receipt of routine services and advice. The second variable that achieves significance in the analysis of low birth weight is WIC participation. Among women in poverty, the odds of having a low-birth– weight infant for WIC participants are less than one-third (.28) the odds for non– participants.
The modeling strategy for infant mortality is the same as that for low birth weight, with one modification. Because low birth weight is a major risk factor for infant death, we include low birth weight along with the other covariates in a separate model (Model 3). The results for these models are similar to those for low birth weight in one important respect– infant mortality is not significantly associated with family poverty despite a bivariate pattern in Model 1 which would otherwise suggest that those in deep poverty are at the greatest risk. The risk of infant mortality also does not increase as a function of the percentage of the neighborhood population that is below poverty. This is shown for both the full sample and the subsample that lives below the poverty threshold. Again, these findings do not support the double jeopardy hypothesis.
The pattern of results for the indicators of prenatal care utilization and content in Model 2 for infant mortality are more consistent with expectations. For prenatal care utilization, we see that the likelihood of infant mortality is greater for those who receive intensive care than for those who receive adequate care. Those who receive intermediate and inadequate care also generate odds ratios that are consistent in sign, but fail to achieve significance. As for the content of care among the full sample, women who receive the full range of routine services and advice are less likely than their counterparts to have an infant die in the first year of life. In Model 2 for both the full sample and the below-poverty sample, the odds of mortality among infants whose mothers received all services are half the odds of mortality among infants whose mothers received less than the full complement of services.
Infant mortality is also associated with both WIC and the method of payment. In general, WIC participants are less likely than non-participants to experience the death of an infant. This is especially noteworthy for women who are below poverty. Among those who are impoverished, Model 2 shows that the odds of mortality among the infants of WIC participants are one-fourth (.27) the odds of mortality among infants of non-participants. Thus, the results suggest that programs that specifically target at-risk groups and the quality of services provided to expectant mothers are important for infant health outcomes.
The last set of coefficients in Model 2 shows that government programs that pay for medical services are not able to eradicate inequalities in health outcomes. Women who carried private insurance were less likely than women who relied on government insurance to have an infant who died. The odds for carriers of private insurance were half (.54) the odds of those who used government aid to pay for care. This relationship does not vary by poverty status.
The addition of birth weight in model Model 3 is revealing as well, if for no other reason than the substantial magnitude of the association between infant mortality and low birth weight. Regardless of one’s poverty status, the odds of dying in the first year of life are over 20 times greater for low-birth-weight infants than normal-birth-weight infants. Moreover, low birth weight also serves as an important explanatory factor for WIC in both the full sample and the subsample that is below poverty. WIC participation reduces the likelihood of infant mortality through its association with low birth weight. Low birth weight also appears to play a similar role for prenatal care advice in the full sample, but there is some uncertainty due to the failure of prenatal care advice to approach significance in the previous results for low birth weight.
The association between private insurance and infant mortality for those who are impoverished is not reduced to insignificance when low birth weight is controlled. It should also be noted that the results of Model 3 for the poverty population also indicates that those who receive adequate care have the lowest risk of having an infant die, despite the fact that the odds ratio of 1.6 for those with inadequate utilization does not approach significance.
SUMMARY AND CONCLUSION
We began this paper by noting that the poverty paradigm underlies much current thinking about prenatal care and infant health outcomes in both developed and developing countries. However, rigorous investigations of this paradigm are relatively rare because of data limitations. Information on income typically is not available in vital records, and there are few alternative data sets that are both representative and include an adequate number of cases with poor birth outcomes (especially infant deaths) to allow for analysis. Using data from a representative sample of mothers who gave birth in Puerto Rico during 1994-95, this research provides mixed support for the poverty paradigm. On the one hand, little support for the paradigm is evident in the results for infant health outcomes. Low birth weight and infant mortality are not associated with family poverty or neighborhood poverty. The latter finding is particularly noteworthy because it is inconsistent with the expectation that the likelihood of deleterious infant health outcomes among those living in poverty increases with the prevalence of neighborhood poverty. There is no evidence of a double jeopardy effect.
The poverty paradigm does receive some support from analyses of prenatal care. Both prenatal care utilization and the advice received from health care providers (not services received) are associated with family poverty in bivariate models. Those who live in deep poverty are more likely than those who live above the poverty threshold to have inadequate prenatal care utilization and to receive inadequate advice. Yet, no dimension of prenatal care examined here is associated with neighborhood poverty for the total sample or the subsample of respondents that live below the poverty threshold (notwithstanding the one counterintuitive exception for utilization noted above). Consistent with the results for infant health outcomes, the results for prenatal care do not support the double jeopardy hypothesis.
The role of prenatal care in infant health also was examined. This is an important issue because public health initiatives frequently emphasize the prenatal care component of health care delivery. Disregarding the relatively high risks of negative outcomes among those who receive intensive care, the results suggest that initiatives are more likely to be successful in the improvement of infant survival than birth weight. Except for the elevated risks of delivering low birth weight infants among mothers with the most intensive levels of utilization, low birth weight is generally unrelated to the adequacy of utilization and the content of care. At the same time, the adequacy of utilization and the content of care play roles in the reduction of infant mortality. Women who receive the full range of routine services and advice about nutrition from their prenatal care providers are less likely to have an infant die in the first year of life. However, the latter association weakens considerably in models that include low birth weight. Models in which birth weight is controlled also suggest that sub-adequate prenatal care utilization increases the risk of infant mortality, especially for those who are below poverty.
Attention to the potential deleterious consequences of poverty directs attention to the role of the government and the private sector in financing medical care. Government programs such as Medicaid and WIC are designed to compensate for the failure of the private marketplace to meet the needs of the indigent population. Our results demonstrate the benefits of WIC for those who are below poverty. WIC participants are more likely than non-participants to receive nutrition advice, to give birth to a normal-birthweight infant, and to have an infant who survives the first year of life. The association between WIC and infant mortality is reduced to insignificance after low birth weight is controlled. Thus, WIC reduces the risk of infant mortality by reducing the risk of having a low-birth-weight infant. This demonstrates the positive role that government programs that are targeted to at-risk populations can have in reducing the risks of negative outcomes.17,18
While such findings are encouraging, government programs are not able to equalize access to health care. Women who rely on private health insurance are more likely than women who rely on government insurance to receive intensive prenatal care and less likely to experience the death of their infant. These patterns are consistent with the health care reforms that were being initiated in Puerto Rico at the time of the survey to reduce inefficiencies in the delivery of health care to the Medicaid population. Over the past several years, the Government of Puerto Rico has modeled the public health care delivery system after managed care plans offered in the private sector.
The poverty paradigm is clearly entrenched in current thinking about the social and economic foundations of infant health outcomes. Although some of our findings are inconsistent with this paradigm, various methodological challenges must be met before a final evaluation can be made. For example, the measurement of poverty hinges on the use of income to measure the resources at the disposal of a family and the assumption that resources are roughly equally distributed to all family members. Income may indicate the inflow of resources into a family, but it does not necessarily indicate the amount of resources that are available to families or the resource constraints of families. Families must make decisions on how to allocate assets to meet the needs of expectant mothers and their infants within the context of the debts that they have as well. Wealth-based measures that take into account both the assets and liabilities of families may provide results that are more consistent with expectations about how economic deprivation influences infant health. Further, inequities within families in control over resources may place some family members at a disadvantage in the expenditure of resources to meet their health needs.
Multi-level studies that focus on neighborhood conditions in developing settings using census data also face special challenges. One challenge is posed by the ubiquity of poverty across spatial units. The typical respondent in the birth sample for Puerto Rico lives in a census tract in which nearly two-thirds of the population is below the poverty threshold (i.e., the median is 64 percent and the interquartile range is from 51 percent to 71 percent). And over three-fourths of the respondents live in census tracts in which the majority of the population lives below poverty. The pervasiveness of deprivation that is evident from the concentration of cases in areas in which at least 50 percent of the population lives below poverty may be responsible for the null findings for our measure of tract-level poverty. Too few persons may live in areas in which the majority of persons have abundant resources to generate significant results for neighborhood poverty.
A related challenge is posed by the use of census tracts as a proxy for neighborhoods. Despite efforts to create tracts that are homogeneous, census tracts are less than ideal as spatial units. Census tracts are often large in scale and include multiple neighborhoods (or parts of neighborhoods) that circumscribe the routines of daily life and social interaction. This measurement difficulty may be compounded by the inadequacy of aggregate measures of economic circumstances that can be derived from the census as proxies for the conditions in neighborhoods that matter for the health of local populations-criminal activity, exposure to environmental toxins, and sanitation. More direct measures of these conditions may provide insight into how neighborhood contexts affect infant health.
In closing, these results suggest that poverty does not play a direct role in infant health outcomes in Puerto Rico. However, additional research questions remain to be explored. In addition to addressing the issues described above, future research should employ a comparative framework to determine whether similar findings hold for Puerto Ricans on the mainland. Such comparisons may shed light on how the larger contexts of the island and the mainland may shape the relationship between economic deprivation and infant health. Comparative analyses may also provide additional information on the limitations of conventional measures of poverty at both the neighborhood and individual levels. Thus, the last word on the health consequences of economic deprivation among Puerto Ricans remains to be written.
ACKNOWLEDGMENTS
A preliminary version of this paper was presented at the CROP/CLASCO Conference on “The Demography of Poverty in Latin America,” Buenos Aires, Argentina (November 9-11, 2000). This research was funded by the National Institute of Child Health and Human Development, the Maternal and Child Health Bureau, and the Centers for Disease Control. Support services were provided by the Population Research Institute, The Pennsylvania State University. The authors appreciate the programming expertise of Jeanne Spicer and Cynthia Mitchell.
Please address all communications regarding this manuscript to either R. S. Oropesa or Nancy S. Landale, both of whom may be reached at: Department of Sociology, The Pennsylvania State University, 601 Oswald Tower, University Park, PA 16802, e-mails: oropesa@pop.psu.edu, landale@pop.psu.edu.
1The U.S. Federal Government defines poverty in terms of the combined income of all family members and the size/composition of the family. Thus, an individual’s status with respect to poverty is determined by the circumstances of his/her family, and all family members have the same poverty status.
2The infant mortality rate for Puerto Rico is substantially lower than the rates for most countries in Latin America (which range from 11.7 for Chile to 67 for Bolivia) and the Spanish Caribbean (which range from 7.2 for Cuba to 47 for the Dominican Republic) (Population Reference Bureau, 1999).
‘Substandard living conditions also place the most vulnerable members of a family at risk of a variety of infectious diseases. In less developed settings, crowding can contribute to the spread of respiratory diseases. Inadequate sanitation, poor water quality, and a lack of running water can contribute to the spread of gastrointestinal diseases. Substandard living conditions may also make it harder for the poor to follow hygienic practices that inhibit the spread of disease, including hand washing. Classical models that link infant mortality to the spread of infectious diseases through poor sanitation and water quality in developing countries may not be relevant to Puerto Rico, a “country” whose per capita income places it amongst “middle income” countries in international comparisons.
‘Utilization and content are not necessarily independent. Most studies of prenatal care focus on the timing of the initiation of prenatal care (Ist trimester) or utilization measures that combine both the timing and frequency of care. These practices are predicated on the assumption that the content of care is a function of utilization. This issue is discussed in the context of the association of each of these factors with poverty.
The survey also asked whether the respondents had amniocentisis or chorionic villas sampling. This information is not included here because these are not routine procedures for all women.
‘his variable largely reflects the receipt of advice about diet and nutrition. Different questions deal with smoking, alcohol, and street drugs. One set of questions asked women if they were advised to cut down or stop using such substances, after another set of questions about whether they smoked, drank alcohol, or used street drugs. These items are not included in the summary index because of differences in question wording and the lack of relevance of the former questions to those who did not use such substances. The creation of a separate index from these items in the preliminary analysis did not yield significant results.
‘The current criteria for the poverty threshold originated in the Social Security Administration in 1964. The key component of the definition of poverty is the amount of income that is necessary to support a nutritionally adequate diet. This was determined by the cost of the least costly food plan-the “economy” food plan–established by the U.S. Department of Agriculture. Since various consumer surveys indicated that families typically spent about one-third of their income on food, the poverty threshold was set at three times cost of the economy food plan for family units of various types.
‘Birth certificates were obtained for cases drawn from death certificates.
`We also analyzed a categorical measure of tract– level poverty in the preliminary analysis. The results are not sensitive to how the measure is coded.
“Although the assumption that the cost of living is lower in Puerto Rico than on the mainland may be generally true, various “salary calculators” suggest that this assumption bears further scrutiny (see www.homefair. corn). For example, costs are I I percent higher for homeowners and 4 percent higher for renters in San Juan than in Philadelphia. Compared to the Bronx (New York), costs are about 15 percent higher for homeowners and 3 percent lower for renters in San Juan.
“The federal offices administering particular programs in “outlying areas” such as Puerto Rico are responsible for determining whether thresholds for determining poverty in the contiguous 48 states are used (Department of Health and Human Services, 2000).
“Medicaid is a federal program that funds medical care for the indigent population. However, special allocation formulas direct proportionately less Medicaid funding to Puerto Rico and other outlying areas than to the states. Health care reform began in Puerto Rico in 1994. The purpose of the reform was to initiate a managed care model for the indigent population to provide access to private medical providers through government insurance (Puerto Rico Department of Health, 1995).
13 This required us to modify our treatment of missing data by excluding those who could not be assigned a tract ID because of incomplete address information on the birth certificate. Although the poverty rate of the tract of residence can be imputed for these individuals, the tract of residence that is used to define the territorial unit that observations are clustered into cannot be imputed.
“All conclusions in this paper regarding the effect of areal poverty are supported by conventional tests for interaction that add multiplicative terms (areal poverty*family poverty) to a baseline model.
“This could be due to the fact that these services are routine services that may be provided at any time during the course of a pregnancy. In theory, all of these services could be provided to those who start their prenatal care beyond the first trimester or have just a few visits to a prenatal care provider. Indeed, there is some sentiment for reducing the number of prenatal care visits for expectant mothers but increasing the range of services provided during each visit. Still, the pattern of results is suggestive and should be investigated further with studies that are able to measure a wider range of services and the frequency with which various services are provided.
“Additional analyses explored the possibility that grouping error (i.e., cut-off points) is responsible for the counterintuitive odds ratio for those who are .5 to .99 of the poverty threshold. The results are not sensitive to the cutoffs used. However, the substantive importance of this result is unclear, especially in light of the fact that the income-to-needs ratio is positive and significant when it is treated as a continuous variable.
“The findings for WIC are replicated when the sample is restricted to the eligible population; that is, those with a family income of 1.85 times the poverty threshold or less.
“The extent to which selectivity bias is responsible for these findings cannot be determined. As noted by Currie (1993), the effects of WIC are probably overestimated if women who select themselves into the program are more motivated. The effects of WIC may be underestimated, however, because the program selects “at risk” women. Currie (1993: 27) argues that findings which demonstrate the beneficial effects of WIC are probably not driven by positive selection.
REFERENCES
ANDERSON, R. T., P SORLIE, E. BACKLUND, N. JOHNSON, and G. A. KAPLAN. 1996. Mortality effects of community socioeconomic status. Epidemiology 8:42-47.
BESHAROV, D. J., and P. GERMANIS. 1999. Is WIC as good as they say? The Public Interest 34:21-36.
CHOMITZ, V. R., L.W.Y. CHEUNG, and E. LIEBERMAN. 1995. The role of lifestyle in preventing low birth weight. The Future of Children 5:121-138. CLARKE, L. L., F. L. FARMER, and M. K. MILLER. 1994. Structural determinants of mortality in metropolitan and nonmetropolitan America. Rural Sociology 59:84-99.
CURRIE, J. 1993. Welfare and the well-being of children: the relative effectiveness of cash and inkind transfers. National Bureau of Economic Research, Cambridge, MA.
DEPARTMENT OF HEALTH AND HUMAN SERVICES. 2000 (February 15). Federal Register 65 (31): 7555-7556.
DiETz, J. L. 1986. Economic history of Puerto Rico: institutional change and capitalist development. Princeton, NT: Princeton University Press.
DiEz-Roux, A. V. 1998. Bringing context back into epidemiology: variables and fallacies in multivariate analysis. American Journal of Public Health 88:216-222.
FISCELLA, K., and P. FRANKS. 1997. Poverty or income inequality as predictor of mortality: longitudinal cohort study. British Medical Journal 314:1724-1728.
FRISBiE, W. P., M. BIGLER, P. DE TURK, D. FORBES, and S. PULLUM. 1997. Racial and ethnic differences in determinants of intrauterine growth retardation and other compromised birth outcomes. American Journal of Public Health 87:1977-1983.
GoRTMAKER, S. L., and P. H. WISE. 1997. The first injustice: socioeconomic disparities, health services technology, and infant mortality. Annual Review of Sociology 23:147-170.
GUEST, A, M., G. ALMGREN, and J. M. HUSSEY. 1998. The ecology of race and socioeconomic infant and working-age mortality in Chicago. Demography 35:23-34.
HAAN, M., G. A. KAPLAN, and T. CAMACHO. 1987. Poverty and health: prospective evidence from the Alameda County study. American Journal of Epidemiology 125:989-998.
KELLY, S., C. HERTZMAN, and M. DANIELS. 1997. Searching for the biological pathways between stress and health. Annual Review of Public Health 18:437-62.
KoTELcHUCK, M. 1994. An evaluation of the Kessner Adequacy of Prenatal Care Index and a proposed Adequacy of Prenatal Care Utilization Index. American Journal of Public Health 84:1486-1489.
KRIEGER, N., D. L. ROWLEY, A. A. HERMAN, B. AvERY, and M. T. PHILLIPS. 1993. Racism, sexism, and social class: implications for studies of health, disease, and well-being. American Journal of Preventive Medicine 9 (supplement): 82-122.
LECLERE, F. B., R. G. ROGERS, and K. D. PETERS. 1997. Ethnicity and mortality in the United States: individual and community correlates. Social Forces 76:169-198.
MCLEAN, D. E., K. HATFIELD-TIMAJCHY, P. A. WiNGO, and R. L. FLOYD. 1993. Psychosocial measurement: implications for the study of preterm delivery in black women. American Journal of Preventive Medicine 9:39-81.
Moss, N. E., and K. CARVER. 1998. The effect of WIC and Medicaid on infant mortality in the United States. American Journal of Public Health 88:1354-1361.
O’CAMPO, P., X. XUE, M-C. WANG, and M. 0. B. COUGHY. 1997. Neighborhood risk factors for low birthweight in Baltimore: a multilevel analysis. American Journal of Public Health 87:1113-1118.
OROPESA, R. S., and N. S. LANDALE. 2000. From austerity to prosperity? Migration and child poverty among mainland and island Puerto Ricans. Demography, 37:323-338.
OROPESA, R. S., N. S. LANDALE, M. INKLEY, and B. K. GORMAN. 2000. Prenatal care among Puerto Ricans on the U.S. mainland. Social Science & Medicine 51:1723-1739.
POPULATION REFERENCE BUREAU. 1999. World population data sheet. Washington, DC.
PUERTO Rico DEPARTMENT OF HEALTH. 1995. Title V application for Maternal and Child Services Block Grant.
RIVERA-BATIZ, F L., and C. E. SANTIAGO. 1996. Island paradox: Puerto Rico in the 1990s. New York: Russell Sage.
ROBERT, S. A. 1999. Socioeconomic position and health: the independent contribution of community socioeconomic context. Annual Review of Sociology 25:489-516.
RUBIN, D. B. 1987. Multiple imputation for nonresponse in surveys. New York: J. Wiley & Sons. SCHAFER, J. L. 1997. Analysis of incomplete multivariate data. London: Chapman & Hall.
STARFIELD, B., S. SHAPIRO, J. WEISS, K-Y. LIANG, K. RA, D. PAIGE, and X. WANG. 1991. Race, family income, and low birth weight. American Journal of Epidemiology 134:1167-1174.
STOCKBAUER, J. W. 1987. WIC prenatal participation and its relation to pregnancy outcomes in Missouri: a second look. American Journal of Public Health 77:813-818.
U.S. BUREAU OF THE CENSUS. 1998. Statistical abstract of the United States. Washington, DC: U.S. Government Printing Office.
U.S. DEPARTMENT OF AGRICULTURE. 2001. WIC income eligibility guidelines. http://www.fns.usda. gov/wic/incomeeligguidelines01-02.htm.
WAITZMAN, N. J., and K. R. SMITH. 1998. Phantom of the area: poverty area residence and mortality in the United States. American Journal of Public Health 88:973-976.
R. S. Oropesa, Nancy S. Landale, and Ana Luisa Davila
Department of Sociology, The Pennsylvania State University, 611 Oswald Tower, University Park, PA 16802; Department of Sociology, The Pennsylvania State University, 611 Oswald Tower University Park, PA 16802; Demography Program, Graduate School of Public Health, Medical Sciences Campus, University of Puerto Rico, San Juan, Puerto Rico
Copyright Society for the Study of Social Biology Spring 2001
Provided by ProQuest Information and Learning Company. All rights Reserved
