March 1-2, 2004
Holiday Inn Select
Bethesda, MD
This meeting was held in conjunction with the National Children’s Study, which is led by a consortium of federal agency partners: the U.S. Department of Health and Human Services (including the National Institute of Child Health and Human Development [NICHD] and the National Institute of Environmental Health Sciences [NIEHS], two parts of the National Institutes of Health, and the Centers for Disease Control and Prevention [CDC]) and the U.S. Environmental Protection Agency (EPA).
Background
Co-Chairs: Daniel Lichter, Ph.D., Ohio State University, and Ann Tickamyer, Ph.D., Ohio University
Rural populations comprise one fifth of the nation, yet are often neglected in research and the public consciousness. Rural children face unique challenges to healthy development associated with poverty, spatial and social isolation, inadequate health, educational, commercial, and social service infrastructure, and unique environmental hazards. The National Children’s Study (Study) provides an unprecedented opportunity to learn how these challenges influence the health and development of rural children, and what can be done to improve their lives. However, to realize this potential, careful attention must be given to rural populations in the design of the Study. This workshop explored how the Study can contribute to a better understanding of the health and development of rural children.
The Study has a clear responsibility to address the unique health and developmental challenges of rural children, but doing so will require careful thought and planning. Rural populations are typically poorly represented in national studies because their dispersion makes them more difficult to sample and interview. These sampling issues need to be identified, and creative solutions to them advanced. Rural environments are highly diverse, and sampling strategies must be theoretically guided to capture those environments of greatest interest to the Study. In addition, rural children may experience different threats to health and development, and different protective influences, than children in urban or suburban settings; measurement protocols for the Study must be designed to encompass these influences.
This workshop was designed to cast a spotlight on rural children and to identify what the Study needs to do in order to effectively address their health and development. The purposes of the workshop were to:
- Highlight the importance of rural children’s health issues and identify high-priority questions for the Study relevant to the health and development of children.
- Address methodological challenges in designing the Study so that it can address these issues effectively, including informing alternative sample design approaches currently under consideration.
Overview of the National Children’s Study
Marshalyn Yeargin-Allsopp, M.D., CDC, DHHS
The Rural Workshop began the evening of March 1 with an orientation to the Study given by Dr. Yeargin-Allsopp, a member of the Study’s Interagency Coordinating Committee. Dr. Yeargin-Allsopp discussed the Study’s origins in the President’s Task Force on Environmental Health and Safety Risks to Children. Planning for the Study was mandated in the Children’s Health Act of 2000 (PL 106-310), which called for a national longitudinal study of environmental influences on children’s health and development. She then described the involvement of multiple agencies in development of the Study.
Dr. Yeargin-Allsopp outlined the major concepts of the Study:
- It will be a national, high quality, longitudinal study of 100,000 children, their families, and their environment.
- Environment is broadly defined to include chemical, physical, behavioral, social, and cultural aspects.
- The design includes a common range of environmental exposures, less common outcomes, and gene-environment interactions.
- State-of-the-art technology will be used for tracking, measurements, and data management.
- Extensive public-private partnerships are planned.
- Data collected will be a national resource for future studies.
The Study will examine five different priority outcomes: pregnancy outcome, neurodevelopment and behavior, asthma, injury, and obesity and physical development. Exposures that the Study will measure include:
- Environmental samples of air, water, dust, and soil
- Exposure biomarkers and genetic factors in blood, breast milk, hair, tissue, and urine
- Information on occupation, diet/nutrition, medicines, supplements, and herbals
- Housing and living situations, family and social experiences, and neighborhood and community characteristics.
Demographic Overview of Children in Rural America
Leif Jensen, Ph.D., Pennsylvania State University, and William O’Hare, Ph.D., The Annie E. Casey Foundation
Dr. Lichter began the second day of the workshop by introducing Dr. Jensen and Dr. O’Hare, who provided participants with an overview of the demographics of rural children. Dr. Jensen’s presentation focused on three topics, including the definition of "rural," comparisons of rural and urban children, and migration between rural and urban settings. Dr. O’Hare’s presentation addressed poverty and diversity among rural children and public policy issues relating to rural children.
Dr. Jensen explained differences among various concepts of "rural." Ecological concepts define rural based on the distribution of people across space and identify as rural areas with low and sparsely settled populations. Economic/industrial concepts define rural areas as those in which farming or other extractive industries are more likely to occur. Sociocultural concepts define rural areas as locations where people adhere to a set of values that differ from residents of more urban areas. Dr. Jensen said that the most commonly used definitions of "rural" are the ecological ones.
Dr. Jensen next discussed operational methods for defining "rural." The most commonly used operational definition of "rural" is as a nonmetropolitan area. To explain the meaning of nonmetropolitan, Dr. Jensen began by describing metropolitan areas as counties that have one or more urbanized areas with a population of 50,000 or more, plus surrounding counties with economic ties as determined by commuting patterns. All other areas are considered to be nonmetropolitan. He said that many people use the terms "rural" and "nonmetropolitan" interchangeably. However, according to the definition used by the U.S. Census Bureau, "rural" means areas of open country or settlements with fewer than 2,500 people. Dr. Jensen pointed out that the definitions of "nonmetropolitan" and "rural" overlap.
Because nonmetropolitan areas encompass many types of places, researchers have refined that definition in various ways. Mr. Calvin Beale, a senior demographer at the Economic Research Service (ERS), U.S. Department of Agriculture, developed a nine-category, rural-urban continuum that blends metropolitan and rural definitions, taking into account metropolitan status, size of settlements, and adjacency to metro areas. Another method to operationalize "rural" uses urban influence codes. These incorporate micropolitan areas--nonmetropolitan counties that have an urbanized area of 10,000 or more--plus the surrounding areas tied through commuting patterns. The ERS uses codes that include county typologies (for example, by persistent poverty or the most common industrial or economic activity).
Dr. Jensen said that nonmetropolitan counties outnumber metropolitan counties by almost 2:1 and cover more land, but they contain less than 20 percent of the U.S. population. They are distributed throughout the country, but a band exists from the northern plains south into Texas. He then showed a map that indicated the number of counties that changed categories over time.
Dr. Jensen noted that the definition of "rural" has implications for the Study, because rural areas are heterogeneous and the full array of types of rural areas needs to be included in the Study sample. He stressed that it would be unacceptable to only define rural as nonmetropolitan, because that large category includes children from areas that are fairly urban to those that are extremely rural. It is important for the Study to include as much geographic specificity as possible (for example, by collecting Global Positioning System readings for every place the child lives), and to provide rural-urban continuum codes and urban influence codes.
Dr. Jensen discussed the reasons why the Study should include rural children. Fourteen million children live in nonmetropolitan areas in the United States, and 15.4 million children live in rural areas. Children in rural and nonmetropolitan areas are important to study because they have less access to quality services and institutions (for example, schools and health care).
Dr. Jensen then provided a brief demographic overview of the differences between children in nonmetropolitan areas and those in metropolitan areas. He pointed out that nonmetropolitan children are:
- Concentrated in the Midwest and the South
- More likely to live in a home that is owned rather than rented
- More likely to be White (20-25 percent of Whites are nonmetropolitan) or American Indian (close to 50 percent are nonmetropolitan)
- More likely to be covered by Medicare or Medicaid but less likely to be covered by other health insurance than metropolitan children.
Nonmetropolitan children are as likely to be living with both parents as metropolitan children, but central-city children are less likely to be living with both parents. Fewer children in nonmetropolitan areas are rated as having "excellent" health than children living in metropolitan areas.
Dr. Jensen said that 15 percent of children in the U.S. moved between 2002 and 2003. Approximately 50 percent of families that moved stayed within the same county. The most common movement was within metropolitan areas, and the next most common move was within nonmetropolitan areas. Of the children who moved to an area in a new category, nonmetropolitan to metropolitan moves were twice as common as those in the other direction. The trend during the twentieth century has been migration from rural to urban areas, except for the 1970s and the 1990s when that flow was reversed. Over the last 5 decades, many young adults have left rural areas for education, employment, and social opportunities in urban areas.
Dr. O’Hare continued the presentation with a discussion of child poverty, racial and ethnic diversity, and public policy. In 2000, child poverty rates were 15 percent in metropolitan areas, 18.5 percent in micropolitan areas, and 21.2 percent in rural areas. Child poverty is higher in nonmetropolitan areas for all racial groups except Asian Americans. Further, the gap between child poverty in metropolitan and nonmetropolitan areas is growing.
Dr. O’Hare noted that three institutions contribute to whether children are likely to be poor:
- Work (if parents work, children are less likely to be poor)
- Education (the higher the educational level of parents, the less likely children are to be poor)
- Marriage (children who live with two parents are less likely to be poor than those who live with single parents).
These three factors affect children in both metropolitan and nonmetropolitan areas, but they are more effective in preventing poverty for children living in metropolitan areas.
Dr. O’Hare said that rural child poverty is highest in Appalachia (mostly White), the Northern Plains (mostly American Indian), the Rio Grande Valley (mostly Hispanic), and the Southern Delta region (mostly Black). The overlay of race and region is very important to consider when studying rural children who live in poverty. A key finding is the major distinction between urban child poverty and rural child poverty based on the regionalization of child poverty by race.
There are 382 persistently impoverished counties in the U.S. (poor every census from 1959 to 1999) with poverty rates of 20 percent or higher. Almost all of these counties are nonmetropolitan. There is a large overlap between persistently poor counties and the racial distribution of poor children. Of the 50 counties with the highest child poverty rates (defined as having more than 43 percent of the children being poor), 48 are nonmetropolitan with high numbers of minority children. Of the 48 counties:
- Twenty-four are Black majority
- Nine are American Indian majority
- Eight are Hispanic majority
- Seven are White majority.
Dr. O’Hare discussed the relevance of federal policies for rural children. He explained that poor rural children are more likely to be in working families, thus policies that are aimed at the working poor disproportionately affect the rural poor. These include policies related to the Earned Income Tax Credit, the minimum wage, childcare subsidies, health insurance for low-income adults and their families, and low-income housing programs. He explained that devolution--the shifting of power to the states--increases the importance of measuring social policies at the state level. Compared to their urban counterparts, poor rural families usually live in states with lower per capita incomes, fewer services and institutional resources, and less supportive policies and benefits for low-income families. As a result, these families are less likely to get cash public assistance.
Dr. O’Hare stressed two implications of his comments for the design of the Study. First, the sample should capture the different subgroups of the rural population, defined in terms of race and region of the country. Second, it will be important to design the Study to provide representation at the state level as much as possible so that the effects of state-level policies on children’s health and development can be studied.
During discussion following the presentations, James J. Quackenboss, M.S., EPA, said he is interested in state representation in the Study but sample sizes required in states are something to consider. He emphasized that the Study is being designed to answer questions about relationships between social, demographic, behavioral, and environmental exposures and relatively rare health outcomes. He said that a large sample is necessary to assess rare outcomes, and there may be difficulties when dealing with small populations. Nevertheless, some environmental information may possibly be summarized at a state level.
Janet Bokemeier, Ph.D., Michigan State University, explained that a large number of rural children change schools within a year, and this can have an impact on interventions. In addition, recent trends towards the consolidation of rural schools have taken children out of smaller, local schools and moved them to larger schools at great distances from home. These developments may have profound impacts on their activities and development.
In response to a question about rural/urban differences in children’s health and well-being, Dr. O’Hare said that a book to be released in fall 2004 will compare state-level data on child well-being in city and rural populations on 10 census-based indicators. He added that a special report on measures of child well-being, which is published every year, will include a set of rural measures, and it will soon be available on the Population Reference Bureau Web site.
Key Aspects of Rural Environments Relevant to Children’s Health and Development
During the next portion of the workshop, three breakout groups--focused on chemical exposures, physical health and development, and mental health and development--considered the following questions:
- What are the most important questions for the Study to answer about rural children?
- What are the questions about the health and development of children that are unique to rural children?
- What is it about rural settings that produce different (negative or positive) health and developmental outcomes in children?
- What characteristics, exposures, and processes are especially characteristic of rural settings and how do they affect children?
The groups were asked to consider the five focal areas of the Study: pregnancy outcome, neurodevelopment and behavior, asthma, injury, and obesity and physical development.
Following their deliberations, members of the breakout sessions reported back to the larger group.
Haluk Ozkaynak, Ph.D., M.S., EPA, summarized the discussions of the Chemical Exposures Breakout Session. The group considered several types of rural/urban differences that could affect differences in chemical exposures and their impact on children. Rural areas may differ not only in the prevalence of various toxins, but also in behavioral and climatic patterns which affect the extent of exposure and its impact. The group noted many factors that produce differences in chemical exposures for rural children, including:
- Outdoor and indoor environments
- Commuting in cars
- Daily behaviors that depend on community and social context
- Exposures to environmental tobacco smoke and parental use of alcohol
- Methods of cooking and heating (increased emissions of particulates, PAHs, and nitrogen oxides)
- Environmental factors (houses with lead paint, pesticides in water and air)
- Mitigating factors (more fresh air, less lead in gas, and less exposure to motor vehicle exhaust)
- Compositional differences in outdoor air pollution (particulate matter, gases, sulfur dioxide, nitrogen dioxide, ozone, volatile organics, and metals)
- Compositional differences in indoor air
- Differences in climate (hot moist air in the South can increase mold and mildew and affect the atmospheric formation of acids, metals, and other toxic pollutants)
- Grasshopper effect from pesticides blown by the wind to neighboring areas
- Regional differences due to crops, grain dust, dry or moist air, pollen, and water sources
- Crop and livestock interactions
- Dietary differences
- Greater proportion of time spent outdoors (perhaps with a greater ingestion of outdoor soil but lower exposure to indoor dust)
- More physical activity (children allowed more freedom of movement)
- Caregiving provided more often by family and friends, rather than a daycare setting
- Presence of toxic dumps and Superfund sites
- Water sources (runoff from fields with pesticide application or biological waste runoff from animals on farms)
- Occupation of parents.
Dr. Ozkaynak said that rural preschool children might be exposed to urban environments when accompanying their parents who work in urban employment centers. He added that rural children sometimes are exposed to poor housing conditions such as mobile home dwellings, which have a higher potential for mildew, mold, and formaldehyde. These can increase the chances for infections, respiratory problems, and exacerbation of asthma. Older homes in rural communities may have lower indoor pesticide concentrations than urban homes, but higher outdoor concentrations may be present due to greater numbers of home gardens and nearby agricultural fields.
Dr. Ozkaynak suggested that the Study relate sources and concentrations of exposures to ensuing health effects while accounting for the relationships among exposures, genetic factors, and individual predispositions. The group recommended that exposures may be evaluated collectively for certain groups with similar exposure characteristics, rather than individually. However, the group also expressed the concern that some proposed sampling methods might not adequately incorporate rural communities if sampling rates are driven by large primary sampling units (PSUs) or areas with high population densities.
Christine Bachrach Ph.D., NICHD, NIH, DHHS, summarized the report of the Physical Health and Development Breakout Session. She said that rural children have worse overall health and higher mortality rates than do urban children, however, data that show the extent of disparities in specific health outcomes (for example, diabetes, obesity, and asthma) are very limited. The National Health and Nutrition Examination Survey cannot provide adequate measures because it does not represent the rural population well enough. She added that it is important for the Study to address health disparities.
The group felt that the Study is a unique opportunity to evaluate how rural children’s health and development are influenced by the distinctive characteristics of rural areas, including characteristics of the people who live in rural areas (population composition), cultural characteristics, and characteristics related to rural place (infrastructure and types of industry). The group developed a model to summarize the pathways hypothesized to be relevant to rural children’s outcomes (see Figure).
Cultural aspects of rural life relevant to children’s health include diet and attitudes towards health and health care. The group discussed diets of rural children and countered the myth that rural populations grow and consume their own food. Diets of rural children may be affected because it is expensive to transport fresh fruits and vegetables to rural areas. Parents may be suspicious of health care providers and reluctant to bring children to a doctor.
A major feature of rural environments is the lack of infrastructure, including health, social, and educational services. Hospitals are poorly funded, struggle to remain open, and in some cases are dilapidated. Health care is provided mostly by family practitioners, and few specialists are available. There may be poor continuity of care, and many physicians are on J-1 visas, have poor command of English, or are from a different culture. Many rural families must travel great distances to obtain health care or dental care, and families often lack health insurance.
Infrastructure for education is also often substandard in rural areas. Concerns include dilapidated school buildings, distance to schools, lower quality of teachers, and the need to change schools often because of school consolidation.
The group identified a number of factors and processes that are likely to mediate the effects of rural residence on health outcomes, but noted that very little research has studied these mediating processes. Mediating pathways include poverty, increased stress, and patterns of family and social interactions. Poverty is both a function of diminished opportunities in rural areas and, in addition to increasing family stress, reduces families’ abilities to invest in their children’s health and development as well as protect them from health threats.
The social worlds of rural children may be different in ways that can affect health both positively and negatively. In many rural communities, communities and families may be highly cohesive. On the other hand, in widely dispersed communities the family may be the entire social world for children, and children in dysfunctional families may have few other resources to turn to for support.
The group discussed several health outcomes that might emerge from these characteristics of rural environments, including:
- Stress and alcohol leading to obesity and diabetes
- Poor nutrition as a function of poverty
- Asthma (wood or coal used for heating, mold and mildew, or mushroom farms)
- Injury due to child labor on farms.
Lynne Vernon-Feagans, Ph.D., University of North Carolina, presented a summary of the Mental Health and Development Breakout Session. This group addressed the question, "What are unique characteristics of rural environments that affect children’s mental health and health outcomes?" After much discussion, participants concluded that there are real differences in opportunity structures and resources available in rural areas, and that these have significant effects on mental health and developmental outcomes.
Many rural areas are resource-poor. Compared to non-rural areas, they have:
- Higher rates of poverty
- Lower availability of mental health services
- Lack of specialized programs in schools
- Less access to childcare (especially licensed childcare)
- Less access to support services for disabled parents
- Less access to advanced media and technology.
This lack of resources is complemented by social and cultural differences that moderate rural children’s outcomes. For example, poverty is likely to compromise rural children’s development less in communities with strong and cohesive social structures. Such communities certainly exist in rural areas, but they are not universal. The physical isolation of rural settings can adversely affect social networks and support structures. Shared concepts of safety and what is defined as a threat to safety differ in rural areas, and these can influence how resources are deployed to protect children.
Thus, the processes contributing to positive or negative outcomes may be different in rural areas. Parenting processes are different in rural areas because the rural environment allows types of parental monitoring not feasible in cities. With respect to developmental outcomes, people in a rural setting might characterize a condition differently than people in an urban area. One example is the diagnosis of Attention Deficit/Hyperactivity Disorder, which may be overdiagnosed and overmedicated in children from urban areas. In a rural area, a child might be viewed as simply rambunctious, and that might be advantageous for the child.
In rural areas, the mentally ill may never be defined or diagnosed. Perception, treatment, and care of the mentally ill may differ dramatically from that in urban areas. Rural areas have fewer resources for early intervention and, often, stigma is attached to mental health services. One-stop shopping for health services (including mental health services) sometimes prevents stigma in rural settings. The role of faith-based institutions in dealing with mental health problems was discussed.
The group stressed the need for sufficient variation in rural communities’ circumstances to permit the rigorous testing of hypotheses related to size of place and children’s outcomes. It also stressed the need for the designers of the Study to be sensitive to differences in interpretation of health-related issues across rural and urban areas, differences that need to be considered in the design of Study materials.
Current Sampling Design Options Under Consideration
James J. Quackenboss, M.S., EPA
In the afternoon, discussion at the workshop shifted to methodological challenges in designing the Study to address outcomes for rural children. Mr. Quackenboss provided an initial presentation on sample design options currently under consideration for the Study. He explained that decisions about the Study design will be based on a set of "givens," including:
- The Study will be national in scope, but not necessarily nationally representative.
- A large sample is required to allow for evaluation of infrequent outcomes.
- Recruitment needs to occur as early in pregnancy as possible.
- Access to and collection of biological samples at birth is needed.
- Stratification is needed to obtain adequate ranges of exposures.
- Recruitment/retention, efficient collection of measures, access to infrastructure, community engagement, and flexibility to conduct special studies are additional considerations.
- Organizational structure(s) necessary to operationalize the design will also be considered.
He said that a decision has not yet been made on the sampling design for the Study, and a Sampling Design Workshop later in the month would examine a variety of potential options. These options are based on two basic sampling strategies:
- National household probability sampling
- Recruitment through medical centers of excellence, which could include a probability sample of center patients, or the area surrounding the center.
Probability-based sampling can use stratification and oversampling (sampling at higher rates than other populations), and can generalize to a broader population in a straightforward manner. Nonprobability-based sampling selects participants from the population in a nonrandom manner. It may involve the use of quotas and purposive selection of certain types of participants. With nonprobability sampling, generalization to a broader population depends on model-based analysis and additional assumptions. Most of the options under consideration are hybrids involving combinations of these strategies. The options are being evaluated for cost, statistical power, recruitment and retention, and other criteria.
The Sampling Design Workshop will have an expert panel to review the options and develop a report. In addition to this report, Study decision-making will be informed by input from the National Children’s Study Advisory Committee, the Study Design Working Group, the Interagency Coordinating Committee, and the Study Program Office. Duane Alexander, M.D., NICHD Director, NIH, DHHS, will be responsible for the ultimate decision.
Mr. Quackenboss next discussed the designs currently under consideration and showed a slide representing groupings of designs. Mr. Quackenboss was asked if center-based designs would have no representation of rural areas. He answered yes, but added that one of the issues to consider was how to incorporate rural children in a center-based design.
Mr. Quackenboss concluded by identifying issues for participants of the Rural Workshop to consider:
- Center-Metropolitan Statistical Area and patient frames may exclude rural areas and populations.
- Should centers be required to provide some rural coverage?
- Should centers be established in rural areas?
- National probability frame
- Use household or physician office frame?
- Oversample rural PSUs so that they are proportionally represented in the Study?
- Combine PSUs to screen an adequate number of households?
Considerations in Designing the Study to Address Rural Children
William Kalsbeek, Ph.D., Ohio University, and Ann Tickamyer, Ph.D., Ohio University
The final hours of the workshop were devoted to discussion of design and logistical issues relevant to ensuring that the Study will be able to answer important questions about rural children. Drs. Tickamyer and Kalsbeek began the discussion with some initial comments.
Dr. Tickamyer noted that the morning’s discussion indicated consensus that there are reasons to study rural children because processes, outcomes, and exposures differ between rural and urban children. She said that existing national studies do not adequately represent rural populations. Other research that has been conducted has been limited to case studies or studies of small communities in particular contexts. The research is often excellent, but it is not generalizable to a larger population. It is important to deal seriously with the issue of how to include rural children in the Study because it is a study of importance and magnitude. Dr. Yeargin-Allsopp agreed on the importance of including rural children the Study.
Dr. Kalsbeek noted two key issues in sampling children who live in rural areas. These were the need to:
- Ensure an adequate sample size of rural children
- Encompass adequate diversity (including race/ethnicity, poverty, and geographic region).
He said that planners must consider if the sample size will be adequate. Specifically, if 20 percent of the 100,000 children come from rural areas, is this an adequate number to answer relevant questions? If not, oversampling may be needed to obtain adequate numbers.
Dr. Kalsbeek said that, given a defined population, random selection in probability samples achieves diversity naturally and more effectively than non-probability samples. However, probability sampling cannot reflect diversity in the population as a whole if only a portion of the population is sampled. For example, a probability sample of urban communities only will not capture variation that exists in urban and rural areas together.
Dr. Ozkaynak reminded participants that the Study is not a surveillance survey but a testing of specific hypotheses. Thus, appropriate sample size and diversity in independent variables are the key factors needed in order to obtain answers.
Participants considered the potential of a Study design based on competitive recruitment of medical centers of excellence, and the steps that could be taken to include rural populations in such a design. Such designs could include probability samples of local residents or center patients, but the areas to be studied would be selected on the basis of proximity to medical centers rather than probability sampling methods.
Mr. Quackenboss responded to a question about the advantages of a nonprobability sample by suggesting that such a sample has many strengths, including:
- Retention (self-selected volunteers have more motivation to remain in the Study)
- Potential to collect more frequent measures
- Ability to collect more technically advanced measures in medical settings
- Potential lower cost of recruiting and keeping a sample.
Joseph Waksberg, M.S., Westat, agreed that medical centers are advantageous because they are able to do the measurements in a standardized way, and they have the intellectual capacity to analyze data. However, he suggested that there is confusion about two issues: how a sample is obtained versus who carries out the measurement. He said that one could select areas on a probability basis and then recruit centers to do the measurement work. He argued that there is no way to assure adequate diversity without using a probability sample. A nonprobability sample requires assumptions, brings up issues of credibility, and creates uncertainties.
Dr. Kalsbeek pointed out that if a center model were used, there would be a limited number of rural respondents because most of the centers are in urban areas. If hospitals are used as the centers, there are also the issues of reaching the people who do not go to hospitals and whether the rural people who go to the hospital represent the needed diversity. One way to address this would be to expand the definition of center to include larger community hospitals. Dr. Bokemeier agreed that relying on medical centers for data collection would introduce bias because people living in rural areas often do not have access to that care. Eileen M. Holloran, Office of Rural Health Policy, HRSA, DHHS, noted also that the Study would be missing rural if it only includes counties that have major medical centers. Rand Conger, Ph.D., University of California, Davis, said that medical centers and physician’s offices may improve efficiency, but may so bias the results that they aren’t useful.
Mr. Quackenboss reminded participants that the Study was trying to get away from the idea of probability sampling versus a center model. Instead, the current direction considers options that combine center-based samples with a national household probability sample or combine center-based samples with household probability samples drawn from the center’s metropolitan area. He also noted that location and population coverage could be made a requirement in the Study. Centers could be required to serve other areas (including rural areas). He pointed out that large medical schools are usually located in inner city areas, and that children in these areas are also at risk.
Dr. Bachrach suggested two models that would enable a center-based design to include rural populations:
- Major medical centers would be given responsibility for probability sampling and measurement in rural as well as urban areas in a geographic region.
- Coordinating bodies or partnerships involving medical centers and survey units would share responsibility for the probability sampling and measurement for a geographic region.
Participants considered the challenges associated with designs that use probability sampling of households. Mr. Waksberg explained how multi-stage probability sampling could be used to get the diversity needed in a rural sample as well as a sample of women who were not yet pregnant:
- Select the areas (usually counties)
- Use stratification based on area characteristics such as geography, mean income, and race distribution
- Stratification might be limited by Study resources.
- Select blocks, then households, within the selected areas
- From census information, stratify blocks on relevant characteristics
- Select a sample of blocks and then select a sample of houses within selected blocks
- Interview each household
- Conduct a 10-minute interview
- Recruit women of childbearing age and contact them on a quarterly basis for 3 years to build up a sample of pregnant women.
Appropriate stratification is important in order to capture adequate diversity within the rural population. Dr. Bokemeier said that there are significant regional differences that need to be considered. Stratification on either race/ethnicity or region would suffice to ensure including distinct areas of persistent rural poverty such as the Ozarks and Appalachia. Stratification should be based on measures of exposures, not outcomes. Measures of rurality used in stratification should be fine-grained, like those developed by the U.S. Department of Agriculture.
Several participants discussed the possibility of oversampling rural populations. Oversampling increases sample size for oversampled groups, but reduces the relative representation of--and ability to answer questions about--other groups, raising questions of equity. If oversampling is used, Dr. Kalsbeek suggested that planners should carefully consider how to define "rural," since various definitions of this concept exist. For example, rural has been defined at the county level or at finer levels of aggregation, such as the population density of block groups. Finer levels of aggregation in combination with disproportionate stratified sampling at those same levels leads to greater statistical and cost efficiency in the oversampling process.
Rural areas pose special challenges for the Study because of their population dispersion and lack of infrastructure. Mr. Quackenboss suggested that household sampling might not work well in rural settings, because of the distance between households. Dr. Bachrach replied that many ongoing studies that use household probability sampling are successfully conducted in rural areas. Others suggested that, rather than sending people to a central location for the Study to collect data, data collectors should be sent to the people in rural areas. It was pointed out that other agencies already visit households in rural areas, including the U.S. Department of Agriculture, which goes out to farms on a yearly basis. Rogelio Saenz, Ph.D., Texas A&M University, agreed and said that relying on physicians or centers excludes people who do not have access to health care, including the very poor or immigrants. Patterns may be missed unless an effort is made to access hard-to-reach populations. Others also stressed the importance of including hard-to-reach populations such as migrant workers and noncitizens, but noted this would be challenging.
Dr. Bachrach asked how the Study could overcome barriers to obtaining advanced medical measures and tests in rural settings. Ann Bullock, M.D., Eastern Band of Cherokee Indians, suggested that physician’s offices would be a good place to collect data because the staff already has excellent relations with their patients, which will help to increase retention. Using physician’s offices allows data to be collected at the local level, and in order to accomplish this, doctors and patients can be given incentives. Data could be collected at the local level and then sent elsewhere for analysis. Dr. Bachrach asked if collection of placentas at local hospitals would be a problem and she was told that many hospitals sell placentas for a profit and might be reluctant to provide them to the Study without appropriate incentives.
Arthur M. Bennett, M.E.A., B.E.E., NICHD, NIH, DHHS, noted that without collecting information at medical centers, some data would be missed. For instance, data collectors cannot obtain a three-dimensional (3-D) ultrasound in rural areas. Dr. Bullock responded that instead of using 3-D, they could do two-dimensional ultrasound, allowing the Study to include a large high-risk rural population.
To address the challenges in rural areas, the Study will need to use creativity in building on the existing structures, such as the Critical Access Hospitals Program (a federal government program that supports small rural hospitals), community health services, rural utility services, nutrition education programs through the Cooperative Extension Service, churches, and other faith communities. Gladys H. Reynolds, Ph.D., CDC, DHHS, recommended hiring people who are known and trusted in the community to help with recruitment and retention.
Concluding Thoughts
The group was asked to share any concluding thoughts. Remarks included general comments about the Study:
- The Study has a huge potential to identify mediators between risks and outcomes that are not currently known and that cannot be obtained other than from a study of this type and magnitude.
- The Study provides an opportunity to examine the childhood roots of adult health (for example, obesity and diabetes in American Indians).
- Plan the Study to have broad objectives and build in flexibility to respond to new issues and questions that will arise over the next 20 years.
Participants also stressed the need for obtaining a diverse sample:
- Rural areas are diverse, so sampling design is very important.
- In designing the sample, distinguish rural from suburban areas and be sure to represent both adequately.
- The Study should ensure that a sufficient sample is obtained in different populations, for example, by race and ethnicity.
- In 20 years the racial mix will be different than it is today, and families that are currently in rural areas may no longer be in rural America in the future. This argues for oversampling racial and ethnic minorities in rural areas.
- Keep sampling and measurements planning separate and coordinate them later.
Others made suggestions and observations about measurements:
- If measures are selected that cannot be obtained in rural areas, then an opportunity to learn will be lost. It is necessary to balance getting the best biological measures with the newest technology against what is practical in rural areas.
- An effort to include multiple methods and measurements (for example, measurement of cultural and social environment) would be helpful.
- Measure the level of difficulty for obtaining services in rural areas ("opportunity structure").
- Consider the family’s health as a unit, not just the child’s health, in order to determine how family health may affect outcomes.
- It is important to look at risks to populations from common exposures (for example, family dysfunction, abuse, and neglect), not just rare exposures.
- Family interactions can be risk factors as well as protective factors for rural children.
- The manner in which questions are asked in rural areas will have to be carefully crafted in order for the answers to be comparable to those from other areas.
Other suggestions included:
- It would help to develop a white paper or literature review identifying key issues unique to rural areas.
- In order to make the results of the Study publicly available, many confidentiality issues need to be addressed.
Dr. Tickamyer and others ended the workshop by saying that this has been a stimulating, thoughtful discussion that showed the complexities of the task ahead. These are things that should be done and can be done. She thanked participants for giving rural children the attention that they need and deserve.
Participants
Christine A. Bachrach, Ph.D., Center for Population Research, NICHD, NIH, DHHS
Adelaide Barnes, B.A., National Children’s Study Program Office, NICHD, NIH, DHHS
Arthur M. Bennett, M.E.A., B.E.E., Office of the CIO, NICHD, NIH, DHHS
Janet Bokemeier, Ph.D., Michigan State University
Ann Bullock, M.D., Health and Medical Division, Eastern Band of Cherokee Indians
Linda M. Burton, Ph.D., Pennsylvania State University
Rand Conger, Ph.D., University of California, Davis
Eileen M. Holloran, Office of Rural Health Policy, HRSA, DHHS
Leif Jensen, Ph.D., Pennsylvania State University
William D. Kalsbeek, Ph.D., University of North Carolina
Daniel T. Lichter, Ph.D., Ohio State University
Mark Mather, Ph.D., Rural Families Data Center, Population Reference Bureau
William O’Hare, Ph.D., KIDS COUNT, Annie E. Casey Foundation
Haluk Ozkaynak, Ph.D., M.S., Office of Research and Development, EPA
Alfred M. Pheley, Virginia College of Osteopathic Medicine
James J. Quackenboss, M.S., Office of Research and Development, EPA
Gladys H. Reynolds, Ph.D., Office of the Director, CDC, DHHS
Rogelio Saenz, Ph.D., Texas A&M University
Ann R. Tickamyer, Ph.D., Ohio University
Lynne Vernon-Feagans, Ph.D., University of North Carolina
Joseph Waksberg, M.S., Westat
Bruce Weber, Ph.D., Oregon State University
Marshalyn Yeargin-Allsopp, M.D., National Center on Birth Defects and Developmental
Disabilities, CDC, DHHS