Children's Health, the Nation's Wealth: Assessing and Improving Child Health (2004)

Existing state and local data constitute an underutilized resource formonitoring children’s health and its influences. Effective use of existing data could serve many diverse functions, for example:

• a data system could provide a state or local health department with data on a range of health indicators to monitor children’s health over time and inform policy decisions;

• a user-friendly web site could enable community groups to compare child health and environmental risk factors in their own communities with those in neighboring communities and with state and national averages;

• an integrated data system could enable public health professionals to identify geographic clusters of birth defects diagnosed months or years after birth, to link these birth defects with the prenatal histories of the mothers and to the newborn screens, and to determine whether environmental toxins are identified in the cluster areas;

• a data system could enable a pediatrician to determine whether a child being seen for the first time has received a full course of immunizations or enable the coordination of care among multiple providers serving the same child.

All of these information needs could be addressed by combining existing data from such sources as vital records, birth defects registries, Medicaid records, immunizations registries, the decennial census, and education records. But doing so requires attention to privacy concerns, standardization of data, political will, and financial resources. While there are challenges to creating an integrated data sys-

tem, the opportunities far outweigh the challenges. In addition to an ethical mandate to know what is happening to children (e.g., when they are at risk, whether their health is changing for the better or worse), a solid data foundation to monitor risks and changes in health is crucial to evidence-based policy making and accountability, particularly in an environment in which states are being given more decision-making responsibilities. There is also a clear need to better understand and act on disparities in health to ensure the long-term productivity of the nation and the health of its citizens.

The chapter begins with definitions and then provides descriptions of data systems, including examples of state and local efforts to build children’s health data systems and federal efforts to encourage such systems. It then outlines the needed steps that would facilitate all states and communities using available data to monitor their children’s health and the influences on it.

For purposes of this report, data are tools to measure health, influences, and their indicators. A data element is a specific component of data with a clear, standardized definition to ensure that the numbers collected accurately represent the component. Age, birthweight, income, source of insurance, educational level completed, and race are examples of data elements. A dataset refers to observations and measurements collected through a single mechanism, effort, or type of scientific investigation. For example, data from a range of questions asked by a survey, such as the census or the National Health Interview Survey, is a dataset. Other examples of datasets are law enforcement records (e.g., incidence of domestic violence, juvenile arrests, and homicides), vital records (e.g., birth, death, marriage, and divorce records), immunization registries, newborn metabolic screening results, program encounter data, results of school readiness assessments, reports of childhood abuse, and health data for children in foster care. A data system is an integrated system of multiple datasets, including the combination of software and hardware that makes possible the manipulation or interpretation of data and datasets.

Data integration refers to the combination of data from more than one dataset into a data system. Data integration facilitates maximal use of data and improves surveillance of children’s health and the factors that affect it. Single datasets allow an assessment of only a narrow set of aspects of health. Data integration can involve either aggregation of data from multiple datasets or linkage of multiple datasets at the individual level.

Good integrated data systems increase knowledge that can be used by states, communities, and policy makers who design policies and fund interventions that

affect children’s health. Data systems can also be used to communicate knowledge about children’s health to the public. Integrated data systems provide early information to identify areas of vulnerability, monitor health disparities, and detect manifestations of adverse effects on children’s health across time, across domains of health, or for a variety of subpopulations defined by geography, ethnicity, or other characteristics.

Integrated data systems with data linked at the individual level also allow individual children or groups of children to be followed longitudinally across datasets to better understand why a characteristic or problem develops in some children but not others, and they support epidemiological studies of the potential effects of certain health influences. Linked datasets from multiple service systems, such as public health, education, human services, child welfare, juvenile justice, and health care, enable particularly useful longitudinal analyses. Based on guidelines to protect privacy and confidentiality, there should be limited access to linked data systems for a range of purposes. For example, policy makers can forecast and monitor the effect of their policies on children; researchers can determine best practices for health, social, and education services; the particular impact of multiple health influences on children can be determined; and management of patient care can be improved.

A community can be empowered when it has information specific to its children. States and some communities collect a wealth of data about children and their health. If data at the state or community level are scattered among several agencies or departments, it is difficult for community groups to compile them. Community groups and local policy makers are much more likely to use data systems that combine administrative datasets from multiple state and community sources (without identifiers) and are easily accessible to the public. Making children’s data systems accessible on the Internet can enhance distribution and use by individuals outside government and the research community. The responsibility for providing much of these data is the public health department at the state and local levels. Since one of the three core functions of public health is assessment, data collection and dissemination is a major activity. The Institute of Medicine recommended in 1996 that the state health department be responsible for the provision of data. More specifically, within the public health department, the Title V maternal and child health program is mandated by federal law to be the lead on maternal and child health issues.

Advances in the systematic collection, analysis, and reporting of data have been made in recent years. Standardizing data definitions and collection methods is an important step in improving the value of data elements, datasets, and data systems and has been a priority of the National Center for Health Statistics (NCHS). Federal, state, and community programs have been encouraged to adhere to guidelines established by NCHS in order to improve the comparability of data. Without standardization, even sophisticated, integrated data systems cannot provide good information.

In addition, advances in technology and architecture for linking data systems allow for the integration or linkage of a large number of datasets on different platforms. Information technology systems in the future will increasingly separate the data elements from the business rules for combining the data elements; this will allow for more uses of the data elements and less recoding or reliance on one or more software packages.

Both aggregation and linkage can serve a range of state and local purposes. While both strategies can be utilized for the same datasets, aggregation alone is most common because it offers fewer challenges; the technology, expertise, confidentiality, and security issues are relatively simple. There is increasing interest in linked datasets given the expanded analyses they enable, but for some jurisdictions the expense, privacy obstacles, or other implementation issues make this an unreasonable short-term objective.

The oldest and easiest method is to aggregate data from multiple datasets of counted events, problems, or traits for a population group defined by residence in a geographic area or some other factor. Population subgroups may be further defined according to age, race, gender, or other data elements included in the datasets. For example, each of several datasets collected by the federal, state, and local government could be aggregated to the same geographic level to provide a more detailed picture of the health of the entire population of children or a subpopulation. Individual identifiers are typically removed before data are shared for aggregation. Many data-holding agencies (e.g., state or local departments of education) are prohibited by state privacy laws from sharing data unless personal identifiers are removed.

Policy and program decisions about children’s health and environments can be enhanced when aggregated data are provided on the web in a user-friendly manner. Massachusetts and Missouri are two examples of states with web-based systems that provide children’s health data from a variety of sources at several levels of geography. Aggregated data can be used to compare a multitude of health characteristics or influences on health for children in one community or across communities. Local and state programs use data such as these to better plan for services and improve programs and policies as well as evaluate programs over time; several states have systems that make aggregate data available on a regular basis. Consistent with the expectation that collection and provision of data are core functions of public health (Institute of Medicine, 1988, 1996, 2001a), public health departments are responsible for the provision of these aggregate-level data on a timely basis.

Aggregated data cannot be used to track individual children or groups of children over time, so they cannot be used to fully understand the significance of particular health factors or influences. Surveillance efforts that count events or conditions are useful for determining when there is an unusual incidence or cluster of a disease or condition that deserves closer scrutiny or identifying problems, such as high levels of substance abuse or injury, but they do not usually provide data that would help in determining why the condition has occurred.

Although states and communities may be organized differently, all have similar departments or divisions providing services to families and children: Medicaid and the State Children’s Health Insurance Program, public health departments that provide such services as maternal and child health (Title V), family planning, and communicable disease services, child welfare, public assistance, education, mental health, housing, developmental disabilities, and juvenile justice. States are required to aggregate data for some federal reports, such as the one associated with the Maternal and Child Health Block Grant (Title V), but many states and several large cities have gone beyond mandated requirements to improve their health assessment and planning capabilities for children.

There is wide variation among states and communities on the number of datasets they aggregate and the accessibility of this information to communities and the public. The final report on the Evaluation of State-Based Integrated Health Information Systems commissioned by the Centers for Disease Control and Prevention (CDC) attests to the variation among states in the distribution of health data on the Internet (ORC Macro, 2000). The report summarizes an inventory of 53 state and local public health department web sites. Data were found on 51 of these 53 web sites. Key findings were:

• The most common datasets contained population data on death, birth, and low birthweight.

• Few sites provided data on health care or hospital utilization, social programs, or behavioral risk factors, such as smoking, alcohol use, obesity, and seat belt use, despite the fact that half of the sites contained other behavioral risk factor survey data.

• Only 10 of the 51 web sites were designed to allow users to formulate their own queries of the data system in addition to the graphs, charts, and tables created and provided by the states.

• About half the sites disaggregated data for geographic units—typically by county.

• Some web sites were geared for public use, others for use by researchers.

• The assistance provided for use of the web sites varied according to the complexity of the site, including whether the site could be queried and users could create both tables and graphs.

The study was limited to an evaluation of web-based efforts since a key purpose was to evaluate the accessibility of health data. The study of web sites also discussed some of the difficulties facing a state that donates nonproprietary software to help another state develop its health data system. The donating state often lacked the training and technical resources to assist the recipient state in getting the system operational. At the same time, states often were reluctant to purchase proprietary systems due to the initial and ongoing costs involved.

Many more states and communities are aggregating health data that were not included in this study, but they have not yet fully utilized the web to make aggregate, queriable data publicly available. In the committee’s view there is great value in federal government efforts to partner with states to develop web-based data systems that, at a minimum, make aggregate data available in a form that can be queried for various geographic levels and for different measures over time. For many states, such a system would be a significant advance over current single dataset analyses.

Some states and communities have aggregated children’s data without individual identifiers from datasets in two or more governmental departments or agencies. Over the past decade, several larger cities and counties have developed children’s data systems to assist with policy development, service planning, and program accountability. While states may aggregate data by county or city, communities often need to analyze data for smaller, neighborhood-based units, typically a census tract. This is important for identifying areas with the greatest problems and the highest priority for resources. Some critical children’s data are often more accessible to cities than to states. For example, a police department can contribute data regarding arrests for juveniles by geographic location, age, race, and reason for arrest. Education data can include absentee rates, suspensions, expulsions, number of special education students, number taking medications for chronic conditions, and proficiency scores by school. In contrast, education data in many states are not standardized, computerized, or in a central database, so collecting this information may require contacting each school. If a community has concerns about a particular area or group of the population and wants additional information, it may develop or use existing tools to collect these data.

Many communities have developed efforts to monitor and track the health and well-being of children in their communities and develop strategic priorities. Los Angeles County has one of the more comprehensive children’s data systems and, through the Los Angeles Planning Council, produces a Children’s ScoreCard every 2 years with data by services planning area. This data system contains aggregated, geographically coded data on children from health, education, social services, juvenile justice, and law enforcement agencies at the county, city, and state levels. However, Los Angeles goes beyond integrating the existing administrative

datasets collected by government agencies and designs its own survey to capture data elements needed for planning children’s health services. For the 1999–2000 health survey done by the Department of Health, 6,000 interviews of parents with children under age 18 were completed in a county with 1.3 million families with children. The sampling for this survey was designed to provide estimates for the service planning areas of the county. While only large communities like Los Angeles may have the resources for such a data system, they also must deal with such challenges as collecting data from many agencies, schools, and municipalities, as well as with geocoding data according to service planning areas or the equivalent (see Box 6-1).

The MassCHIP system in Massachusetts provides data on the web that can be queried at various substate levels. The data come from multiple state agencies as well as some sources external to government. Data include expanded Behavioral Risk Factor Surveillance System data on children and families that can be aggregated at various substate levels. MassCHIP also produces reports on the Kids Count indicators and the Maternal and Child Health Healthy People 2010 objective for each city, town, and region of the state (see Box 6-2).

Many states or communities produce periodic reports, sometimes referred to as “report cards,” modeled on the types of indicators identified by the Federal Interagency Task Force on Child and Family Statistics or the KIDS COUNT initiative. In some cases, these reports are produced by government agencies and in other cases by community groups. For example, Philadelphia Safe and Sound is an organization dedicated to improving the lives of Philadelphia’s children through “a committed public-private collaborative” and publication of Report Card: The Well-Being of Children and Youth in Philadelphia, which monitors key indicators of childhood health, safety, and development (Philadelphia Safe and Sound, 2001).

There is a growing interest in building data systems that are more sophisticated and have greater potential for policy and research purposes by linking individual-level data from two or more datasets. Linking datasets using consistent individual identifiers significantly increases the benefits of a system, even if personal identifiers are removed after the linkage. For example, linking would allow determination of whether conditions or behaviors co-occur, in what proportion of children, and in which areas of the state or locality. The advantage of this model is that it offers an opportunity to identify clusters of children who have specific characteristics, traits, behaviors, and health conditions and to follow individual children in a population longitudinally. This type of data system, if it provided valid and reliable data on a large and diverse population of children, could allow all the functionalities (surveillance, monitoring, forecasting, indicators to be used for accountability or quality improvement, and research) envi-

sioned by the committee as desirable for a data system of the future. The disadvantage of linking datasets is that more concerns are raised about confidentiality and privacy, discussed later in this chapter.

A linked data system must adhere to state and federal rules and regulations regarding the collection and dissemination of confidential data; the Health Insur-

ance Portability and Accountability Act and similar state laws provide the context for how data systems can be developed and used in the future. All states have procedures for how to obtain access to confidential information; these rules are followed internally within health departments for linking data between key surveillance systems. Unique identifiers that allow data for the same individual to be linked without that person’s identity revealed are often used when data are shared.

Two decades ago, the University of Miami undertook one of the earliest efforts to link data at the state level with financing from the Florida departments of education and health (personal communication, Fredia Wadley with Keith Scoot and Rachel Spanjer, 2003). The Florida Department of Education initiated the linking of datasets because it would help elucidate the problems of future cohorts of students and support better policies and planning to meet student needs. However, Florida is still one of the few states in which the education department shares its data with individual identifiers with another state agency (see Box 6-3).

Tennessee has been developing a data system on children over the past 5 years. The first phase aggregated datasets across five state-level departments serving children. The second phase linked the datasets with individual identifiers to allow case managers in the various state departments to access information on the Internet for their clients.¹ Initially the information provided by each program was limited. Case managers could find out what other state services were being provided and whether other case managers were assigned to the same child, but they had to contact other relevant programs to gain more information. The final phase is being developed and is to contain a record of all state services and health care information for children in foster care.

Health providers who have been approved to access the system for their foster care patients will be able to view records for their clients and input specific diagnostic, treatment, and referral information. Although there are only approximately 10,000 children in state custody at any time in Tennessee and a total of 15,000 in an average year, these children present an enormous challenge to state and to health providers. Locating and making available social, education, and health records to those service and health providers attempting to make adequate assessments and plans for these vulnerable children has been a long-standing problem. A second challenge has been tracking these children as they receive multiple services and therapies and sometimes relocate to other regions of the state. The type of system being developed would not only help in resolving these two problems common among states but would also serve as a prototype for broader children’s health data systems (Urbano, 2001).

The Robert Wood Johnson Foundation All Kids Count initiative has a recommended approach to creating an integrated data system for the early years that grew out of efforts to develop and implement immunization registries. Rhode

Island is one of the states that have linked their immunization registry with other state datasets. Rhode Island’s children’s data system, KIDSNET, links nine datasets from programs that serve children in the state health department. While other state health departments have integrated many of their datasets, Rhode Island is unique in that it has concentrated on integrating the data available on children. Data from KIDSNET is made available in both aggregated and linked form, depending on the user (see Box 6-4).

Other examples include the systems in Missouri (MOHSAIC), Oregon (FamilyNet), and Utah (CHARM). The Maternal and Child Health Bureau and the CDC have over the years funded various states to develop integrated child health data systems. The technology available today can produce linked data systems. The challenges stem from a lack of awareness about the potential ben-

efits, a need for strong commitment and leadership at all levels of government among the various data owners, the reality of government budget crises, and increasing concerns about confidentiality and privacy. Even if these challenges are overcome, it will take time and resources to build linked systems.

Ethics pervades discussions of the collection and use of data. It is accepted that certain types of health data serve a public good and are in the interest of the community so can and should be collected. For example, certain vital statistics are required to be reported in all jurisdictions. Similarly, reporting of some conditions is mandatory. In the case of children, this includes birth defects (in some jurisdictions), newborn metabolic screening, newborn hearing screening (in some jurisdictions), certain infectious diseases, all cancers, and suspected abuse and neglect. The exact list of reportable conditions is further guided by state law. There is also general consensus that it is appropriate to track the rates of some potentially sensitive population characteristics using aggregated data (e.g., suicide rates, adolescent pregnancy). Much of the information that is collected in national and state surveys is also reported as aggregated data. However, reporting of data also raises issues of privacy and confidentiality.

Virtually all of the datasets used to obtain aggregated data include detailed information on individuals or in some cases on all the individuals in a family or household. This provides the mechanism for review and analysis of data pertaining to individuals and the cross-tabulation of data to assess how two or more characteristics are associated. For example, one could determine the proportion of white males in rural areas who are immunized. In order to examine such relationships, identifiers are routinely removed to preserve anonymity.

Although the agency that collected the data could theoretically identify the individual, and in some cases does go back to collect additional information (as for example in a follow-back or longitudinal study), the information that could be used for identification is stripped from the file before the data are analyzed. This protects the individual who provided the information, but at the same time allows analysts to query the data so that multiple pieces of information can be combined to assess the relationships between elements in the dataset.

Numerous data are available in public use datasets that can be analyzed both by public health officials and by investigators in research institutions. This mechanism has provided a great deal of current information about children’s health. The removal of individual identifiers protects the individual’s privacy while advancing the public interest to learn more about children’s health. Furthermore, there are requirements to ensure that individuals with rare conditions are not identifiable by prohibiting the reporting of units of analysis (such as geographic sites) that would allow the person to be traced.

Data anonymity is now possible with current technologies, even for linking

individual data across time within a survey or agency and also across surveys and agencies. In such cases, individual identifiers are removed from the data and are known only to the persons who have access to the linking process. The data stripped of identifiers can be used for a wide range of purposes, while still allowing for the protection of individual interests. Again, standards would be needed to ensure that individuals with particular profiles of rare occurrence could not be identified.

Private Lives and Public Policies, a National Research Council and Social Science Research Council report of the Panel on Confidentiality and Data Access, made recommendations designed to aid federal statistical agencies in their stewardship of data for policy decisions and research (National Research Council and Social Science Research Council, 1993). Recognizing the tension between individuals’ rights for privacy and the public’s need to know certain information for the good of society, the panel stated that “Sharing of identifiable data for statistical purposes can have many potential benefits, including the enrichment of cross-sectional and longitudinal datasets, evaluation and improvement of the quality of census and survey data, improvement of the timeliness and consistency of statistical reporting, development of more complete sampling frames, and improvement of comparability between data developed by different statistical agencies” (p. 40). The panel also posed questions to consider before linking records:

1. Will the linkage process conform to all statutory confidentiality standards of the agencies involved?

2. Will the linkage conform to pledges made to data providers concerning the use of their information?

3. Is linkage of data the only feasible way to develop the desired statistical products?

4. Are the uses of the data sufficiently important to justify introducing additional risks of disclosure?

5. Should resources and costs be considered in obtaining the desired data by other means?

6. Who will have access to the data? Who should decide?

These questions form a good basis for making decisions affecting the development and maintenance of data systems at all government levels. Since federal and state laws vary according to the data elements and datasets involved, general statements cannot be made to fit all federal and national efforts to develop systems of data. However, some overriding issues must be addressed at the federal level before states can move forward.

The first question of the Panel on Confidentiality and Data Access related to barriers to agencies sharing data. The Federal Education Right to Privacy Act presents some legal concerns for sharing education data on individual students. An exception provides that schools can share individual data on students if they

are used to improve instruction (National Education Goals Panel, 1997). Nevertheless, it is unlikely that state education departments will participate in linking their datasets with those in other departments as long as they are unsure of the interpretation of the law and this exception. Sharing of education data and health data collected in schools (e.g., school-based health centers, school nurses) could be facilitated with an amendment to clarify that student data can be used for research and planning when they (1) benefit the health and well-being of students and (2) confidentiality is protected.

Some state Medicaid programs also cite Section 1092 (a) (7) of the Social Security Act as limiting the sharing of Medicaid data with individual identifiers. This section limits the use of individually identifiable Medicaid data to purposes directly connected with the administration of Medicaid. Some Medicaid programs believe that linking data helps in meeting the needs of their participants and improving programs and do not see this as a barrier, while others take a more restrictive view.

The Health Insurance Portability and Accountability Act of 1996 (HIPAA) also raised concerns about sharing individual data. The final rule required health plans, health care clearinghouses, and health care providers who conduct certain financial and administrative transactions electronically to disclose how they will (1) use, store, and share health information; (2) ensure patient accessibility to their medical records; and (3) obtain patient consent before releasing patient information (U.S. Department of Health and Human Services, 2003). The act establishes mandatory procedures for the linkage of data sets. Although HIPPA is explicit about covered health entitities under the law, very little reference was made to uncovered entitites, especially those in public health departments. All state public health mandated data systems, such as registries and communicable disease reporting systems, must include all relevant cases in the dataset.

Linkages that do not remove identifiers can be desirable for purposes of care management. In these situations, the goal of linking the data is to allow relevant service providers to know what other service sectors are involved in an individual’s care. In these situations, the balance of individual interests with public interests (for example, to save money, avoid duplication of services) may become more complex. These issues are sometimes handled by suggesting that consent be obtained for such linkages.

In order for consent to be informed, parents must be told and must understand who will have access to the data, how they will be used, and any risks associated with their child’s data being in the system. Parents must also be assured that no negative action will occur if they refuse consent. Consent forms can be designed to have a list of data elements that can be shared and those that are not to be released. A critical question is whether a parent can consent for a child,

especially if, as is often the case, the ultimate uses or potential uses of the data are not known at the time that the permission is obtained. In addition, new knowledge might make parents want existing data elements to be kept confidential. State laws and regulations differ on this point, though generally uses of data are restricted to what was included in the original consent. Federal guidance on this point will be increasingly important as new technologies push the current boundaries of research and make possible previously unforeseen analyses.

With any collection of data containing personal identifiers or a potential for breach of confidentiality, there will be questions about who has access to the data and who makes access decisions. Advisory groups consisting of data experts, data providers, and data users, frequently used by state health departments to guide data system development, can be created to provide guidance consistent with federal, state, and local laws and policies for dealing with issues of access, privacy, and ethics. Advisory groups can decide which data elements should be excluded in the data system, or should be accessible only after the identifiers have been removed, or should be shared only after a unique identifier has been assigned. This same advisory committee could also develop protocols on what data could be available to researchers or other groups and under what conditions. Policies and procedures should be developed before proceeding with the development of the data system. Guided by state laws regarding confidential data, state health departments have such procedures and policies in place for health department data.

Even data that are a matter of public record have the potential to stigmatize counties. Many states distributing data for HIV/AIDS cases do not report a number for a county until the total number of cases reaches a specified level. The concern is that if only a couple of cases of HIV/AIDS are reported in a small community, certain individuals might be labeled, correctly or incorrectly, as having AIDS due to their sexual preference or life-style. Communities have also expressed concern about reporting certain youth characteristics (e.g., a high percentage of sexually active adolescents or a high incidence of sexually transmitted diseases). Schools have refused to participate in surveys designed to determine risk behaviors of students if the results were to be made known by school or community, fearing that characteristics perceived as negative can stigmatize the school and create problems for the school with community leaders. Many state public health departments have developed their own rules about suppression of small numbers of cases. HIPAA also includes data suppression and reporting rules. Additional federal standards on this issue could facilitate a consistent approach across states.

Some security concerns relate to the technological design and maintenance of a system. Over the past few years, several widely publicized breaches of network

security and global viruses have elevated concerns about the security of online data. Most of these incidents have not involved health-related data, but they have fostered the perception that any online data poses a security threat (Eng, 2001). Today’s technology allows databases to be designed to provide security against direct query of certain attributes. Any specific user can be given restricted access to specific parts of a database. Such a multilevel database stores data according to different security classifications and allows users access to data only if their security level is greater than or equal to the security classification of the data (National Research Council, 1993).

Another concern is what Lunt and collegues (1990) call an inference channel. An inference channel is said to exist in a multilevel database when a user can infer information classified at a high level for which he or she does not have access based on repeated queries of information classified at a lower level to which the user does have access. Techniques (e.g., query restriction, response modification) exist to limit the potential for such inferences, and consideration should be given to these approaches in designing data systems. Extensive security protocols in place for public health bioterrorism systems also provide possible models.

The Census Bureau has implemented an approach that makes sensitive individual-level data available to researchers in seven data centers across the country (see Box 6-5). This approach demonstrates that there are viable strategies for making individual data available without compromising confidentiality, representing an initial step toward improving access to data. However, accessibility under this model is limited, the relatively small number of centers makes the process somewhat cumbersome, and it can require significant expense by researchers who are not located near a center. Additional and expanded approaches should be developed to continue to make Census Bureau and other datasets, such as those available through the NCHS, more accessible. Although NCHS has a mechanism in place to make sensitive data available at their headquarters, the process is more cumbersome than the Census Bureau’s approach and currently does not meet the needs of data analysts outside the center.

Another set of complex ethical issues arises from the relatively new ability to collect and store biological specimens for long periods of time. For example, all states collect drops of blood from newborn infants on special absorbent cards. These are used for mandatory newborn screens for inborn errors of metabolism and other congenital conditions. There is inconsistency in the screening procedures used across states and in the procedures used to handle such specimens after collection. However, it is technically possible to take anonymous fragments of these specimens to look at population exposures to infectious diseases (such as to determine rates of HIV transmission) or noxious chemicals (e.g., pesticides or

persistent organic pollutants—Burse, DeGuzman, Korver, Najam, Williams, Hannon, and Therrell, 1997).

In addition, it would be possible for an investigator to approach families several years after a baby’s birth to obtain permission to use identified samples to conduct a retrospective study of a toxic agent to assess the body burden at the time of birth. In this case, presumably the family could make an informed decision about the merits of the study and the risks to the child, if any, of making this determination. It is likely that interest in conducting such retrospective analyses will grow with the further sophistication in understanding of biological mechanisms of gene-environmental interaction.

What is much less clear is whether it is ethical to ask parents for consent to use their baby’s biological samples for open-ended purposes at the time of the infant’s birth, since it is not clear what the implications are for future risk to the child of information that might be obtained at a later point. For example, if genes are identified that are associated with increased risks for antisocial behavior,

might a child’s genetic profile be used to exclude him or her from some activities, or if a child is found to have a genetic predisposition for an illness, might he or she be charged a higher premium for life insurance? Many would argue that the parents could not anticipate the potential consequences of the permission they were granting.

These examples illustrate some of the complexity of the ethical issues involving the use of different methods for monitoring children’s health and its influences. Because of this complexity, the committee recommends that special attention be given to these issues. In the committee’s view, linkage of currently available data, using existing and emerging technologies, can advance understanding of disparities in health and ways to address them. Not utilizing these capacities also raises ethical concerns, because it means that information within our grasp is not used to improve children’s health.

Standardizing data elements and the methodology for their collection is vital to having reliable analyses. Various agencies of the federal government have taken the lead in this area, and there is good rationale for strengthening the coordination and collaboration of all the federal agencies providing funds and guidelines for the many programs that require collection of administrative data on children’s health or health influences. States must also be partners in the process, since they not only collect data for the federal programs but also for state programs. States are most likely to adopt standardization guidelines if they are developed by the federal government in collaboration with the states. Standardization would produce comparable data across states and reduce the time and resources needed for individual states.

Medicaid and education programs represent a large part of government budgets at all levels. There are federal requirements for both programs to collect a significant amount of data, but it is not specified or mandated that states collect all data elements in a standardized way. For example, Medicaid requires a comprehensive physical (including behavioral) examination and developmental assessment on children receiving an Early, Periodic Screening, Diagnosis and Treatment (EPSDT) examination. The federal Medicaid guidelines do not require the use of standardized tools for these assessments, or that there be a reduction in the reimbursement if all sections of the EPSDT exam are not completed. As a result, documentation for these assessments may be absent. Even when they are present, there is no way to aggregate these data in any meaningful way, since the tools and documentation vary so greatly.

A similar problem exists for education data. States and school districts do not always require their schools to use the same tools and methodology for collecting data. While all schools are now required by the No Child Left Behind Act of 2001 to conduct annual assessments at specific grade levels, it is left to the

states, and often the school systems within a state, to determine which tools will be used for assessment. Similarly, Readiness for School: A Survey of State Policies and Definitions documented in fall 2000 that only 13 states conducted statewide screenings or assessments when children enter kindergarten; 5 states required the screenings or assessments but local school districts decided how to do them; and 26 states did not mandate any readiness assessments. The National Education Goals Panel (1997) has made recommendations for defining and assessing school readiness, but educators have not been able to reach consensus on if, what, or how this should be done (Saluja, Scott-Little, and Clifford, 2000). Standardized data on school readiness and performance in grades 4 and 8 could advance efforts to identify practices most effective in helping children perform at the same level of their peers even though they are behind when they start kindergarten (see Box 6-6 for an approach in Vancouver using school readiness data).

In developing a data system, it is important to follow a logical cycle from planning to development to implementation and to involve relevant program and community leaders in developing system specifications. Such a system should be developed with a clear set of requirements based on the various processes or

outcomes desired. When system needs are well defined, technical solutions can be purchased to meet specified requirements. Although building or purchasing a system is likely to be feasible, doing so will require resources. Federal programs such as the Title V program should consider a dedicated funding stream for the design and implementation of well-developed data systems to guide policy development and aid program monitoring.

However, states and communities that are developing information systems on children are struggling to develop or purchase software as well as obtain necessary hardware. Building the infrastructure for such a system may require steps as simple as having enough computers for field staff in the various children’s services agencies who will be putting data into the system. Another challenge is acquiring the hardware and processes necessary to address access and security concerns. Proprietary systems that come with support services may be too expensive for a state or community to purchase. States willing to donate their system to another state may not have the manpower resources to help in the training and implementation phase of the project, even in the absence of patent or copyright protection. Obtaining software can therefore be a major barrier for a state or community trying to develop a data system on children.

In addition, data system developers must grapple with datasets that sometimes contain incomplete and inaccurate data. Standardization of data elements will help to reduce inaccuracies. Involvement of the people and programs responsible for providing relevant data in the development of system specifications will also help address these issues, particularly if the system is designed to provide useful information to the people responsible for providing the data.

Personnel time and expertise are necessary to identify and integrate datasets, to create and maintain data at the substate level (e.g., county, city, or census tracts), to make the data available for others to disaggregate at these levels, and to create web sites with tutorials on how to use the data. A greater level of manpower and expertise is needed to create linked data systems and protect the data in those systems than for aggregated systems. Hiring adequate staff to develop and maintain such data systems is a barrier for many states and communities. However, every Maternal and Child Health (MCH) Title V program has one or more personnel trained in data analysis and applications, and there is a national network of state MCH data contacts. CDC and HRSA have helped to develop and train data expertise in each state; CDC has fostered the development of a network of maternal and child health epidemiologists.

The National Action Alliance started in 1997 as the national action agenda for building data capacity for maternal and child health. The initial leadership group, which consisted of the Association of Maternal and Child Health Programs, CityMatCH, HRSA, and CDC, has now expanded to include other state

affiliates (the Council for State and Territorial Epidemiologists, the National Association of Public Health Statistics and Information Systems), the National Association of County and City Health Officials, the Association of State and Territorial Health Officers, the Association of Schools of Public Health, and the Association of Teachers of Maternal and Child Health. The National Action Alliance has advocated for better maternal and child health data and information systems, increased opportunities for field-based capacity building, sufficient well-trained people in the field, as well as increased communication and opportunities for research and evidence-based practice.

Data linkage raises concerns as more information becomes available about specific individuals. Parents may be reluctant to have their children’s data in a linked system. Government agencies, health providers, health insurers, and even legislators may oppose data linkage because of the accountability issues it raises. Data are very important for ensuring accountability, but not everyone welcomes accountability with the same enthusiasm. Competing private entities may have concerns about how data will be collected, analyzed, utilized, and reported and may fear that negative data could be presented in a manner that threatens the stability and future of a facility, organization, company, or community.

Geographically coded data can help communities understand children’s health and its influences in their communities. They provide communities with an efficient method for determining what or where their problems are and allow them to compare their community with other similar communities. Most data now are geographically coded at the state and county level, and some public health data, such as cancer data, includes geocodes. However, few communities have data other than decennial census data available at smaller geographic levels, such as the census tract level. In some cases, administrative data sets that incorporate geocodes do not retain multiple geocodes, which negates geographic tracking over time.

Geocoded data can also assist states and communities in their efforts to target limited resources to meet the greatest need, forcing communities to recognize problems that they might not acknowledge without data to apply to them. For example, although a state may recognize that it has a higher incidence of youth alcohol and drug use or juvenile arrests than surrounding states or the nation, communities can deny that the problem relates to them if there are no data specific for their community. Rural communities may be surprised at their incidence of specific social and health problems because they have associated these problems with urban areas (e.g., drug abuse). Middle and upper income commu-

nities may also underestimate the risk behaviors of their children and youth without data that are specific to them.

Most national and state surveys are not designed to produce valid data at the community level, whether that is a county, city, or census tract. While the expense of producing tract-level survey data may be prohibitive, county, city, or town-specific data are often feasible and enable meaningful geographic analyses within a state. Some states have invested in expanding both the Behavioral Risk Factor Surveillance System and the Youth Behavioral Risk Factor Survey, funded by the Department of Health and Human Services and designed to produce state-level estimates, in order to have county-specific data. Expanding these surveys (and others) to obtain census-tract-specific data is also an option for communities. Many states are using this approach to develop substate estimates as well as stable estimates for ethnic and minority groups. States should ensure that data are captured at substate levels that make sense for the geographic characteristics of their state. Additional federal resources to states and communities targeted to data and analysis would advance efforts to make substate data available and ultimately improve efforts to target resources to needs.

As the technology is available to build systems of data providing benefits for children, how does the nation move forward? Some states and large cities have had success in building children’s health data systems; these systems need to be expanded to other areas.

The federal government has a key role to play in establishing data standards, providing resources to states to develop data systems, providing guidance on data system development, and facilitating the availability of state or local-level data. The vision for what these data systems can produce to improve children’s health should be shared by the multiple federal agencies serving children and communicated to the leadership of all states and to the public to generate commitment. The federal government should take the lead in convening and determining the elements and standards for a children’s data system for use at the national, state, and local levels. Several federal efforts have supported or could facilitate state and local-level data systems, but increased efforts are needed.

As mentioned in Chapter 4, the MCHB and CDC have led government efforts to collect and report state-level health indicators for children. The MCHB has played an important role designing a unique set of performance standards. Created in partnership with the Association of Maternal and Child Health Programs, all states receiving federal maternal and child health block grant dollars are required to collect and submit certain data elements that are used to create pro-

files for states and the nation. These data, however, are largely anonymous and cross-sectional. As such, they do not allow assessment of change over time except for individual elements at the population level.

The Division of Services for Children with Special Health Needs of the MCHB recently began a new effort to develop comprehensive, community-based systems of care for children with special health care needs. As states develop systems of care, they are being asked to measure and monitor their progress in meeting specific goals and to build state capacity for data-based decision making. The MCHB has developed a standard definition for children with special health care needs and supported research used to develop a screening tool to identify them. The screener has been used in a survey using the State and Local Area Integrated Telephone Survey (SLAITS) platform and by the Consumer Assessment of Health Plans Survey used by Medicaid and commercial managed care organizations. This is a good example of the federal leadership required to establish standardized data collection protocols for states.

CDC has been the leader in defining standard definitions for key data registries and systems used to monitor health outcomes in states; these include definitions for HIV/AIDS, sexually transmitted diseases, birth defects, and cancers. CDC has also developed the National Electronic Disease Surveillance System (NEDSS), originally piloted in Nebraska and Tennessee, which establishes standardized data elements and can link with state systems to use state data to get national totals for specific indicators. However, data are retained and analyzed at the state level. NEDSS is important because the system is using a set of data and technology standards based on the Public Health Information Network to collect basic data from a variety of settings rather than requiring reports using standard software or hardware across states. While this makes it difficult to pool data at the national level, it provides advantages for states and anyone reporting data to the state. There are great expectations for NEDSS as a prototype to facilitate better aggregated and linked data in the future at the state and national levels. Although its design is intended for surveillance of infectious diseases, the approach could be readily used for assessment of children’s health over time and across groups as defined by the committee. Some states currently building child health tracking systems (e.g., Utah, Massachusetts, Tennessee, Washington) are using the NEDSS approach and standards.

The National Committee on Vital and Health Statistics, which advises the secretary of health and human services on national health information policy, presented the report of its Workgroup on the National Health Information Infrastructure (NHII) in November 2001 (U.S. Department of Health and Human Services, 2001a). The report, Information for Health: A Strategy for Building the National Health Information Infrastructure, envisions a process for building a 21st-century health information system, emphasizing the importance of an effective, comprehensive health information infrastructure that links all health decision makers, including the public. The heart of the plan is sharing information

and knowledge appropriately so that they are available to people when they are needed to make the best possible health decisions. It seeks to serve all individuals and communities equitably. The plan does not include a large centralized data warehouse with information on individuals stored at one site. Instead, it is to be a system with a set of technologies, standards, applications, systems, values, and rules that allow access to specific data only on an approved and a “need to know” basis. Personal health, health care providers, and population health are the three key dimensions. These dimensions provide a means for conceptualizing the capture, storage, communication, processing, and presentation of information for each group of information users.

While the NHII would offer opportunities for standardization and linkage of datasets, it is not currently designed to be an information system to monitor children’s health. As proposed, it provides a framework for integrating data on health services and population health. To be used for children’s health, it would have to be expanded to reflect a board definition of children’s health and to include specific child health variables, and standard definitions for additional child health variables would need to be developed. This might be done through a partnership of the HHS agencies that collect data on children’s health and could provide a valuable tool for states designing data systems.

Although linking existing data systems will be an important first step in monitoring children’s health at the state and local levels, additional data may also be required. The SLAITS platform developed by the NCHS in CDC provides a valuable mechanism for states to collect additional data. When collecting additional data, states would be well served to use existing survey questions and mechanisms developed at the federal level to allow comparisons with national standards and avoid investing resources to develop instruments that are already in place. The federal government could play a lead role in guiding states in such efforts.

The Institute of Medicine’s report Fostering Rapid Advances in Health Care also provides recommendations on health data systems (Institute of Medicine, 2002a). It proposed federal funding of 8 to 10 demonstration projects to establish computerized clinical information systems and Internet-based communication in the health field. These proposals have concentrated on designing a national information infrastructure for health services and if implemented should explicitly include child health measures as a core component of the infrastructure.

The MCHB and CDC have funded some states to develop integrated child tracking systems that combine datasets, such as vital statistics data on births, newborn screening, and early childhood programs. Funding for continued and expanded demonstration projects to create and share children’s health data systems can greatly facilitate the process of building information systems for children. To facilitate analysis of the costs and benefits of these data systems, an area about which there are very few data, demonstration projects could be required to collect relevant data. Multiple variables will affect the cost of these data systems, including the method for integration, the software used, the number of datasets

included, and the existing technological NCHS infrastructure. The benefits will also vary depending on whether aggregated or linked data systems are developed, how the data are used, who has access to the data, and how many critical datasets are included. Expanded model demonstration projects would offer continued opportunities to develop comprehensive data systems and to estimate costs and benefits based on the types of data system and their intended uses.

The federal government could also advance the development of data systems by convening relevant federal, state, and community stakeholders to explore the range of privacy and confidentiality issues, including the sharing of individual data (with identifiers) among departments of government and to develop clear guidance related to privacy and confidentiality.

Several governors and mayors have recognized the value of appointing a Children’s Cabinet, a public health agency, or similar entity to coordinate and improve services influencing children’s health across multiple state departments and agencies. This same type of leadership and political will is needed to create data systems across departmental lines and to resolve funding and data sharing issues. States and localities will need the vision to see that these data systems have the potential to increase the effectiveness and efficiency of their services to children, provide data for accountability, and empower communities to act (see Box 6-7 for one state example). But this vision has to spread across the many departments and agencies that are involved in providing relevant services. A financial investment will be required for the development of any data system, and states and localities without an adequate infrastructure will require a greater investment. States might take a graduated approach. An initial first step is the enhancement of state public health web sites with aggregated health data. The next objective could be aggregated children’s data available on the Internet followed by the linkage of many of the state’s datasets on children for government and private research purposes. These three steps are well within the reach of all states within the next few years.

It is equally important for communities within states to evaluate child health. While it may not be feasible to collect every relevant data element at the census tract level or even the city or county level, states and communities should improve efforts to work together and share the financing of data collection efforts. Just as it can be more economical for states to enhance national surveys to get valid data for the county level, the same is true for communities that might want to expand either a national or state survey to meet their own needs. This could involve increasing the sample size to obtain data valid for a smaller geographic unit or adding new questions to an existing survey, as some communities do now using the Behavior Risk Factor Survey or the Youth Risk Behavior Survey.

State-level aggregated data that are available to the public on the Internet will

be most accessible to communities for planning and monitoring the health of their children. In addition to making the data available, states should provide the tutorials and explanations that can promote utilization of the data by the public, interested organizations and researchers. While accessibility via the Internet is very important, states will often need to provide technical assistance to communities in understanding and utilizing their data. This is especially true for small, rural communities with very limited resources. Similarly, one of the most important steps that a community can take is to ensure that there is a designated agency or entity responsible for reviewing available data on children and making a report to the community on an annual basis. Knowledge is empowering, and a community empowered with data is more likely to develop necessary interventions to improve their children’s health.

A logical point of accountability for data collection and provision and in some cases leadership is the state maternal and child health agency, which is directed by statute to ensure the health and well-being of all mothers, children, and youth. This includes making data readily available; in recent years this includes reporting consistent data to the federal government as part of the block grant. In collaboration with the federal MCHB, states are also working to improve state data capacity. The state maternal and child health office may be a logical locus for leadership on state data systems; at a minimum, state efforts should attempt to build on this existing capacity.

Many steps have already been taken to develop data systems that can better evaluate and monitor the health of children at the federal, state, and community levels. The technology and some models are available to direct the next steps leading to better data systems and improved children’s health that can result from use of these systems. Vision, leadership, political will, and investments will be needed to meet this challenging but attainable goal.