HHS has a considerable investment in surveys and other data systems to support broad analytic and program objectives. These range from monitoring of the population and health system, to supporting biomedical, epidemiological, and health services research, to supporting the management and evaluation of HHS programs. Data from these surveys serve as the analytic base for formulating policy, modeling the impact of proposed changes in programs, and evaluating the impact of policies over time.
The operation of surveys is decentralized throughout HHS, with limited central strategic planning and direction. As a result, there are overlaps with respect to populations of interest, analytic capabilities, sample and questionnaire designs, and collection efforts. These overlaps sometimes place undue burdens on survey respondents, and there is a large "opportunity cost" in terms of lost analytic benefit. Despite HHS' large investment, there are still major gaps in information to assess the health status of the population, determinants to health, risks to health, and the functioning of the public health system, and lack the ability to analytically relate data across these areas. And it is clear that the data on health care costs and expenditures, which are collected in detail only every 10 years, are inadequate to understand and effectively model health care access, utilization, insurance coverage and health care costs on an ongoing basis.
HHS has targeted the improvement of the analytic capacity of HHS programs, the filling of major data gaps, and establishment of a survey consolidation framework in which HHS data activities are streamlined and rationalized. A detailed plan was developed as part of Reinventing Government Part II, and is currently being implemented. The National Center for Health Statistics (CDC) and the Agency for Health Care Policy and Research have key leadership roles in this initiative.
Major Goals of the HHS Plan for Survey Consolidation
• A more rational, systematic strategy for collection of data on key health issues.
• The filling of a critical gap by the production of annual estimates of health care expenditures, insurance coverage, and employer-related insurance costs.
• The continued ability to meet public health data needs now met through ongoing population and provider- based surveys, within a framework that also allows for expansion to meet unmet public health data needs.
• An enhancement of the analytic capabilities of HHS surveys, allowing multiple HHS data collection efforts to be linked analytically through the use of common core questionnaire, common sampling frames, and common definitions and terms.
• An overall reduction the burden imposed on survey respondents by HHS, below what would have been required in independent surveys to meet the same data needs.
• Efficiencies in sampling, data collection, questionnaire design, and survey operations, allowing more of HHS' resources to be focused on meeting high priority data needs.
This consolidation creates a framework in which nearly all of the major surveys of HHS will be significantly restructured and redesigned.
Major Elements of the HHS Survey Consolidation Plan
• Redesigning and automating the National Health Interview Survey (NHIS) to serve as the sampling "nucleus" for many HHS population surveys. The NHIS provides annual data on health status, access to care, health behaviors, and other topics with sufficient size and scope to cover many sub-population groups.
• Implementing an ongoing, longitudinal panel survey on insurance and expenditures - the Medical Expenditures Panel (MEP). The 1996 MEP will use the 1995 NHIS as a sampling frame. The MEPS will add to the NHIS data by providing in-depth data on expenditure-related topics, such as insurance and utilization.
• Consolidating HHS' surveys of employers so that the National Employer Health Insurance Survey and the insurance followup component to the expenditure survey (MEP) are jointly fielded.
• Implementing joint field operations and common core questionnaires for HHS' major surveys of health care providers.
• Redesigning the National Health and Nutrition Examination Survey so that it would be conducted using the NHIS as its sampling frame for its next cycle, beginning in 1998.
• Developing a new approach to characterizing the supply side of health care - provider organization, capacity, and affiliation - and rationalizing and coordinating HHS' provider inventories, health workforce analyses, and infrastructure surveys.
• Evaluating the issues involved in coordinating the field operations and sampling of the National Household Survey on Drug Abuse with similar approaches in the NHIS.
• Developing a modular design that will facilitate State-level estimates and provide a mechanism for States to "buy into" national survey efforts to meet their own needs.
HHS Plan for Consolidation of Surveys
[as submitted to OMB on April 11, 1995]
Working Group Charge
HHS identified "Consolidation of Surveys and Development of Data Standards" as a major issue for REGO II. The problem can be summarized as follows:
Inefficient and Overlapping Survey Efforts, Burden on Respondents. The Department of Health and Human Services (HHS) spends considerable resources each year on multiple, decentralized general purpose and program-relevant health surveys. These funds flow primarily from the budgets of multiple components within the Public Health Service, as well as from HCFA and ASPE. Many of these surveys overlap with respect to populations of interest, analytic capabilities, sample and questionnaire designs, and collection efforts. These overlaps sometimes place undue burdens on survey respondents.
Inadequate Survey Data. Despite the money and effort devoted to these activities, we lack the information necessary to assess the health status of the population, determinants to health, risks to health, and the functioning of the public health system, and lack the ability to analytically relate data across these areas. Similarly, data necessary to understand and effectively model health care access, utilization, insurance coverage and health care costs are inadequate. In addition, the HHS survey activities often fail to take into account the needs of -- and resources provided by -- States and private sector activities. This has been, in large measure, a product of inadequate coordination, cooperation, and consolidation.
Most HHS data systems have traditionally been developed independent of each other, with relatively focused analytic objectives to meet. While this might optimize the usefulness of the resulting data in the particular focus area, it results in a lack of data on the multiple, interconnected aspects of health and health care. The analytic potential of all data systems is thus greatly reduced.
Through Reinventing Government Part II (REGO II), HHS has targeted the improvement of the analytic capacity of HHS programs, the filling of major data gaps, and the establishment of a survey consolidation framework in which HHS data activities are streamlined and rationalized. A Survey Consolidation Working Group was charged with developing a consensus plan for meeting these objectives.
The working group was given one month to complete this task. This document is the report of that group.
Analytic Objectives, Customers, Data Users
HHS data systems cover a broad range of analytic objectives, paralleling the range of research, programmatic, and analytic issues addressed by HHS programs. Analytic objectives include:
• broad monitoring of the population and health system to identify trends, health problems, risk behaviors and conditions, measure the burden of illness, prevalence of conditions, and develop health accounts
• more specific data to support HHS program needs, such as management of the Medicare program, oversight of grants, and coordination of substance abuse/mental health programs
• analytic base for formulating policy, modeling the impact of proposed changes in programs, and evaluating the impact of policies over time
• public data resource for biomedical, epidemiological, and health services researchers to use in conducting independent research or HHS-supported analysis
• measuring, monitoring and evaluating performance of HHS-supported programs, including those proposed for consolidation in performance partnership grants
Health data systems must be broad, balanced, and comprehensive, obtaining information on 1) health status; 2) health risks (e.g., environmental, behavioral, genetic, exposure to infectious agents); 3) the capacity and functioning of the medical care and public health systems; and 4) health care costs, utilization, and associated areas. It is essential to be able to interrelate data from these areas in order to assess interactions and relationships.
In order to obtain information on the full range of issues that are encompassed by concepts of health and health care it is necessary to collect information on many, interrelated levels. It is necessary to obtain information on the characteristics of the population (including diverse subjective and objective measures of health status, functioning, insurance coverage, expenditures, utilization patterns, satisfaction with care), health behaviors, unhealthful exposures, behavioral and medical risk factors, and attitudes about (and knowledge of) health and health care. This information must be linked to information on the medical care and public health system's structure, employer provided health insurance, provider behavior, the distribution of public and private resources, and costs.
The interaction of information on populations and the health care system allows for the evaluation of effectiveness of care and interventions, the interplay of supply and demand leading to over or under supply of key providers, the investigation of how changes in the system will affect supply, total costs, and the shifting of costs to different payors, and adequacy of care. The lack of information in any one area makes it impossible to interpret the data that is collected, reducing its analytic usefulness. For example, it will not be possible to monitor the health of the Nation and to explain significant changes in these trends, to develop accurate projections and forecasts for planning purposes, or to evaluate the health care system and intervene to improve access and quality. As a result, HHS could not meet its primary objective to improve the health of the population.
Information is needed for the population as a whole, as well as for subgroups defined by demographic characteristics (age, race, gender, income) and geographic location, since in many cases HHS programs have particular responsibility for these populations.
To obtain such diverse information, it is imperative that a diverse array of survey mechanisms and approaches be employed. These include:
• personal interviews (e.g., to obtain information on the health and expenditure experience of families, the knowledge, attitudes, and risk behaviors of individuals, and how health conditions are translated into functioning in society);
• interviews or records from employers (e.g., to gauge the extent to which individuals are provided with health insurance, or the extent of occupational health problems and services);
• collection of information from providers of health care and essential public health services, including (e.g., to obtain medical or cost detail unknown to individuals, or to obtain a sufficient sample of rare events such as a surgical procedure or exposure to an infectious agent), or vital registration officials (e.g., to obtain complete coverage and detail on sentinel health events such as birth and death), as well as other data collection approaches to providers in order to determine their organization, distribution, behavior, and preferences; and
• direct physical examination of individuals (e.g., to identify undiagnosed or untreated illnesses, objectively measure and validate health status and disease prevalence, and analyze the relationship between risk factors and disease).
Key customers and constituents of HHS surveys include:
• agencies within HHS that use data for program management, actuarial projections, policy development, and evaluation;
• researchers within HHS that use data in epidemiologic studies, identification of hypotheses for biomedical research, demographic studies, etc.;
• analysts in other Executive and Legislative branch organizations, including OMB, CBO, OTA, and GAO;
• State and local governments involved in public health and health care financing;
• academic researchers working independently and with support from HHS;
• advocacy groups, including advocates for public health, disease-specific medical research, children, the aged, Americans with disabilities, and those advocating particular strategies for health reform; and
• the private sector, including providers, managed care organizations, and associations.
Major HHS Data Collection Efforts
HHS sponsors or operates a diverse number of surveys and other data collection efforts. Many of these have been conducted as independent efforts for technical reasons (e.g., specific design features, such as anonymity, that are not typical to other surveys), practical reasons (keeping burden on individual respondents to a minimum), funding reasons (e.g., dedicated or set-aside funding streams), and to maximize the flexibility and management control inherent in smaller-scale surveys.
Tab A summarizes the major parameters in designing surveys. Tab B is a matrix that summarizes key attributes and design parameters of selected major HHS surveys. The largest of these surveys, the primary focus of this report, are summarized below:
• The National Health Interview Survey (NHIS), conducted by the National Center for Health Statistics (NCHS), is a continuous household survey of approximately 40,000 households. The NHIS covers a broad range of health status, health insurance, utilization, and risk behavior topics on an annual basis, and numerous specialized supplements are added to the NHIS annually to address topics of current priority. A current example is the Disability Survey, which in 1994 and 1995 is using the NHIS to identify persons with disabilities. These persons are being recontacted for detailed study.
• The National Medical Expenditure Survey (NMES), conducted by the Agency for Health Care Policy and Research (AHCPR), is a study of 14,500 households conducted approximately every 10 years. The NMES collects detailed data on utilization, expenditures, and insurance in addition to related information on health status and disability. Multiple household interviews are augmented by collection of data from insurance plans and medical care providers, and an institutional component obtains data on nursing home expenditures.
• The Medicare Current Beneficiary Survey (MCBS), conducted by the Health Care Financing Administration (HCFA), is an ongoing longitudinal panel survey of approximately 12,000 individuals selected from Medicare administrative files. Household respondents provide data on utilization, expenditures, and insurance coverage, which is supplemented by linkage to Medicare claims information. The MCBS sample includes institutionalized beneficiaries.
• The National Employer Health Insurance Survey (NEHIS), a collaborative effort of NCHS, AHCPR, and HCFA, is a telephone survey of approximately 40,000 establishments to obtain information on the nature and cost of health benefits offered by employers. This survey was first conducted in 1994.
• The National Household Survey on Drug Abuse (NHSDA), conducted by the Substance Abuse and Mental Health Services Administration (SAMHSA), focuses on estimating the incidence, prevalence, consequences, and patterns of substance use and abuse. NHSDA surveys 18,000 respondents using special procedures to protect privacy and anonymity.
• The National Health Care Survey (NHCS) is a family of NCHS surveys that inventories providers and measures of the utilization of health services through a series of surveys of providers. Surveys under the NHCS include those of hospitals, physicians, emergency and outpatient departments, nursing homes, home and hospice care providers, and ambulatory surgery centers.
• Surveys of the Public Health Infrastructure. In FY 1996, in collaboration with the public health community, PHS will initiate a strategy for obtaining comprehensive baseline and trend data on the public health infrastructure at national, state, and local levels, to fill gaps in currently available data on the impact of proposed changes in the health care system on the capacity of communities, States, and the nation to carry out essential population-based public health functions or to provide safety-net services, and to evaluate and monitor our investment in public health or the relationship between the capacity and functioning of the public health system to health care costs and the health f the population. This effort will address the provision of "essential services" of public health, persons providing these services, revenue sources and expenditures, and the extent to which population-based public health services cross-subsidize personal health care services and vice versa. The need for this effort, and the general approach, is summarized in Tab C.
• The National Health and Nutrition Examination Survey (NHANES) collects data through direct physical examinations, laboratory analysis, and interviews. In the most recent NHANES, completed in 1994, approximately 30,000 persons were surveyed in standardized mobile examination centers to obtain a wide range of measurements. NHANES already represents a consolidation of HHS' direct physical measurement activities, in that the data needs of more than a dozen individual agencies are met through the NHANES mechanism. Tab D includes examples of the types of information obtained through direct physical examinations.
• The National Survey of Family Growth (NSFG) is an interview survey of women in their reproductive years. The subject matter focus includes family planning, contraceptive use and efficacy, adoption and abortion, and other aspects of reproductive health, family formation, and dissolution. The NSFG is conducted every five years, with intermediate longitudinal followup interviews. The NSFG is already consolidated with the NHIS in that the sample for the NSFG is selected from NHIS respondents.
Assumptions and Guiding Principles
The working group identified the following assumptions and principles as important to meeting the charge outlined in the REGO II process:
• The analytic capabilities of HHS surveys should be improved, particularly in terms of the ability to relate different aspects of the health system.
• Unnecessary overlap and duplication should be reduced.
• Burden on respondents to HHS surveys should be reduced in the aggregate, without increasing burden on any individual respondents to an unacceptable level.
• A framework should be established in which existing HHS surveys are streamlined and rationalized.
• Priority data needs that are currently met through existing survey efforts should continue to be met.
• Critical unmet data needs identified by HHS Data Planning Work Group should be met, including data on insurance and expenditure, capacity and performance of the personal and public health systems, and health status.
• HHS must maintain a balance in its survey programs so that the full range of needs of data users and customers within HHS and in the broader community will be met.
• In the aggregate, efficiencies from survey consolidation should mean that the cost of collecting these data are less than the cost of obtaining these data through separate, uncoordinated surveys.
HHS Plan for Survey Consolidation
Consensus was reached among the various HHS agencies on key features of a plan for consolidating HHS' survey. Major elements of this plan are summarized below, and detailed in Tab E.
The HHS Plan for Survey Consolidation will result in :
• A more rational, systematic strategy for collection of data on key health issues facing HHS.
• The filling of a critical gap by the production of annual estimates of health care expenditures, insurance coverage, and employer-related insurance costs.
• The continued ability to meet public health data needs now met through ongoing and planned population, provider-based, and infrastructure surveys, within a framework that also allows for expansion to meet unmet public health data needs.
• An enhancement of the analytic capabilities of HHS surveys, allowing multiple HHS data collection efforts to be linked analytically through the use of common core questionnaire, common sampling frames, and common definitions and terms.
• An overall reduction in the burden imposed on survey respondents by HHS, below what would have been required in independent surveys to meet the same data needs.
• Efficiencies in sampling, data collection, questionnaire design, and survey operations, allowing more of HHS' limited data collection resources to be focused on meeting high priority data needs.
This consolidation creates a framework in which nearly all of the major surveys of HHS will be significantly restructured and redesigned.
Key features of this plan include the following:
• Redesigning the core questionnaire and automating the ongoing NHIS, to continue broad-based monitoring of health issues and to establish the capability to use the NHIS as a sampling frame for multiple HHS population surveys.
• Implementation of an ongoing, longitudinal household panel survey on insurance and expenditures - the Medical Expenditures Panel (MEP). Growing out of the NMES design, the MEP will use the NHIS as a sampling frame, with coordinated questionnaires to enhance analytic linkages and reduce respondent burden. Immediate benefits of consolidation will be realized by using the 1995 NHIS as the sampling frame for the 1996 panel. The 1996 MEP would be introduced at a lower sample size than was planned for the 1996 NMES, but would be increased in size to meet precision requirements as the study continues in 1997.
• The MEP would be conducted on a periodic cycle so that insurance and expenditure data will be available annually, but with less detail and at lower cost than would be incurred if a full NMES-type plan were employed. The continuing annual sample size for the MEP would be 8,000 households, increasing to 12,600 households at five year intervals. An extensive provider followup would be conducted at these five year intervals, with more limited information being obtained from providers on the continuing annual sample.
• The MCBS will be closely coordinated with the MEP in terms of 1) greater questionnaire coordination; 2) analytic linkages of the MCBS and MEP samples to increase the power of each survey; and 3) possible offsetting reductions in sample size if analytic goals can be achieved through such linkages. Further steps to consolidate the MEP and MCBS will be evaluated as the MEP reaches a steady annual state in 1998.
• Consolidation of the NEHIS and NMES-HIPS, through a merger of field operations and the adoption of a common questionnaire for all employers. The consolidated NEHIS/HIPS would parallel the sample size cycles of the MEP. A sample of 42,000 employers every five years would coincide with the more extensive MEP, and provide both national and 50-State estimates in that year. The annual sample of 15,000 employers would provide for continuing national estimates.
• Joint field operations and common core questionnaires and forms for major surveys of health care providers, beginning with NCHS' National Health Care Survey components (hospitals, physicians, ER and OPD's, home and hospice care, ambulatory surgery) and the provider followup to the Medical Expenditure Panel. Additional evaluation efforts will address the potential for closer coordination or consolidation with the SAMHSA Drug Abuse Warning Network and the AHCPR Healthcare Cost and Utilization Project.
• Merger of the National Nursing Home Survey and the NMES Nursing Home Expenditure Survey into an integrated, periodic survey of nursing home capacity, services, utilization, and expenditures. Further evaluation will be made of the potential for coordinating the institutional component of the MCBS with this newly consolidated strategy for nursing home data.
• Redesigning the National Health and Nutrition Examination Survey so that it would be conducted using the NHIS as the sampling frame for its next cycle, beginning in 1998. The 1998 NHANES would begin with a more limited 2-year household-examination based survey, followed by continuous three-year cycles involving complete examinations, laboratory tests, and dietary interviews in mobile, field examination centers. In addition to being directly linked to the NHIS, the redesigned NHANES would serve as HHS' primary vehicle for obtaining detailed, objectively measured data on health status, prevalence of conditions, and distribution of medical risk factors in the population.
• Continuing the design of the National Survey of Family Growth as a survey directly linked to the NHIS sampling frame, with greater questionnaire coordination with other components of the consolidated survey plan.
• Development of a conceptual framework for characterizing health care capacity and public health infrastructure, and rationalizing and coordinating HHS' provider inventories, health workforce analyses, and data collection efforts to characterize the public health infrastructure. Building on this framework, HHS would develop the analytic capability to link capacity and infrastructure data more closely with health status, economic, and outcomes measures, and to look at interaction between functioning and costs in (and between) the medical care and public health systems.
• A design framework in which the National Household Survey on Drug Abuse could be consolidated with the NHIS in several respects, ranging from 1) closer coordination between the questionnaires of the two surveys; 2) using the NHIS as the sampling frame for the NHSDA; to 3) conducting the NHSDA as a
supplement to the NHIS. Some of these possibilities require that difficult (and possibly insurmountable) methodological issues be overcome. Tab F outlines these issues.
• Further efforts to integrate disease surveillance and grant reporting systems, which typically are conducted in partnership with State and local public health officials. These efforts will include 1) removing, to the extent possible, Federal requirements for States to maintain distinct, non-integrated data collection mechanisms; 2) support of streamlined, integrated data systems at the State level that can satisfy disease surveillance objectives; and 3) consolidation of categorical grants into performance partnership grants, with streamlined reporting requirements. Efforts to streamline and integrate these multiple, distributed data systems will depend on strong Federal leadership in the development of consensus standards, definitions, and approaches to recording and sharing of data.
• A periodic State-level telephone survey would be designed to obtain basic health status, access to care, insurance, and expenditure data of importance for national policy analysis, performance evaluation, and modeling. By using the available NHIS sample, already distributed widely in States, and using questionnaires from the consolidated national surveys, State level data can be obtained efficiently and be directly comparable to national data. If funding is available, this survey would be conducted at periods when the MEP was off-peak.
• A modular design that will allow surveys to be expanded to meet other critical data needs, such as for subpopulations (e.g., minorities, poor) or for State and local areas. This design will allow States to "buy into" the consolidated HHS survey to obtain data on their States. States may find this to be an efficient option for monitoring performance partnership grants. HHS would also devote efforts to developing modeling techniques that would enhance the value of national surveys for State purposes.
• Improved and more uniform policies regarding privacy, data access, public data release, and sharing among HHS agencies and with outside users, and development of new HHS policies that would assure that any new survey activities considered by HHS agencies would be developed within the consolidated survey framework, to the extent possible.
Consultation with Customers and Data Users. Before finalizing consolidation plans or moving to implementation, HHS will engage in an open, balanced process to gain input from the customers and data users described earlier in this document. HHS data systems are highly complex, and it has not been possible to address all potential implications of changes in survey design within the REGO II time frame.
Research Required to Develop Consolidated Designs. HHS will develop a more detailed research agenda to answer critical methodological questions that need to be resolved before new designs can be implemented. The agenda will include research addressing:
• effects of using the NHIS as a sampling frame for other surveys on response rates and other data quality measures;
• evaluation of alternative procedures and designs for obtaining data now obtained through the Drug Abuse Warning Network;
• refinement of an overall framework for a balanced health survey program that addresses health status, health risks, the health and public health system, and health care costs and utilization;
• development of a framework, or typology, of the health system that can be used in rationalizing inventories and other measures of the health care and public health infrastructure;
• optimum sampling/screening strategies to accommodate multiple population surveys, particularly when multiple surveys have demands for the same subpopulations;
• development of modeling techniques that would enhance the value of national surveys for State purposes;
• issues in determining whether it is possible to move from providing assurances of anonymity to guarantees of confidentiality in surveys of sensitive topics, such as the National Household Survey on Drug Abuse;
• operational aspects of consolidating field operations for provider surveys;
• approaches to improving data on managed care;
• approaches to capitalize on standardization of Federal programmatic data;
• continuing evaluation of the effect in changes in mode and frequency of data collection on costs and data quality; and
• research on approaches to capitalizing on emerging State health information systems, including those that focus on enrollment and encounter data.
Short term priorities also include assuring that the 1996 NHIS is implemented using Computer Assisted Personal Interview technology, and conducting a systematic review of existing and planned questionnaires in HHS surveys to achieve maximum analytic coordination and burden reduction.
Data Policy Development. In order to maximize the benefits from consolidation, HHS will develop more detailed policy statements in several areas including:
• more uniform policies regarding protection of privacy, data access, release, and sharing among agencies;
• improved coordination and decision making mechanisms within HHS, since the inter-dependency of various elements of HHS data systems will be greatly increased; and
• development of HHS policy for encouraging agencies to use the newly consolidated data mechanisms to the extent possible, rather than initiating new, independent data systems; a part of this policy will be a review and reissuance of HHS guidelines on the appropriateness of using extramural grants to support large scale survey activities.
HHS is addressing the process for development of consensus data standards in a related REGO II activity.
Maintaining Momentum for Survey Coordination and Consolidation. This report addresses the preponderance of HHS' large scale surveys, and creates a consolidated framework in which much of HHS' data needs can be met through a more rational and coordinated system of surveys. However, even if fully implemented, it is not the final step in the process. Additional steps include:
• identifying other data systems within HHS that might be potential candidates for consolidation, based on commonality of data sources, content, or field operations;
• continuing to identify data gaps that can be addressed efficiently within the consolidated framework, and developing proposals for enhancement of the designs proposed in this report;
• moving beyond HHS data systems to consider the relationship of HHS systems to systems at Census, BLS, and elsewhere in the Federal statistical system, including the potential that a continuous measurement program at Census has for efficiencies in HHS survey designs; and
• establishing approaches to partnerships with State and local agencies to maximize the utility of HHS surveys for State policy purposes, and to maximize HHS' use of data collected in State health information systems.
Risks of HHS Survey Consolidation Plan
In addition to the significant benefits outlined above, an effort to consolidate HHS surveys involves significant risks, including:
• The large overall budget for the Consolidated Survey serves as a more noticeable target for budget reductions.
• In a consolidated environment, if any of the components fall behind for any reason (e.g., funding or clearance is delayed, operational problems, etc.), all of the other components are affected. The budgets for each component will be more dependent on each other, and competing demands for the NHIS sample will make it important to keep to a rigorous planning schedule.
• There is the potential to over-consolidate, which may create a structure which yields diminishing returns as a consequence of its magnitude. For some existing data collection efforts, consolidation may prove to be counterproductive. For example, issues of respondent burden and the effects of changes in survey modes need to be researched.
• Although flexibility is a goal of consolidation, the inherent complexity of the consolidated survey will make it difficult to balance competing demands for supplemental questionnaires, special analyses, and unexpected uses of the sample.
The HHS survey consolidation plan offers value - first, a bigger "bang for the buck" for currently budgeted dollars, and second, a framework in which critical, longstanding data needs can be met with minimum additional investment.
The FY 1996 HHS Data Investment included funds to meet critical short-term data needs. This investment was only partly funded, and there was criticism that the investment strategy did not go far enough in rethinking the fragmentation of HHS data activities.
HHS continues to believe that the type of investment outlined in the FY 1996 Data Investment is still warranted, given the importance of good information in leveraging efficiencies in the multi-billion dollar health system. HHS has gone "back to the drawing board" to develop a more comprehensive, longer-term strategy. This consolidation plan goes considerably farther than the Investment in addressing the underlying structure and rationale for HHS data systems, proposing significant changes in the role and operations of nearly every major HHS survey. With the efficiencies brought about with these changes, HHS feels that at the level of investment sought in the FY 1996 Investment, a long-term strategy can be put in place for meeting ongoing data needs and responsibilities.
Tab G includes tables describing the current services funding baseline (reflecting the current funding and periodic cycling of existing HHS surveys), and funding for the proposed consolidated survey plan.
Tab A: Parameters in Survey Design
In order to clarify the terminology and outline the major parameters in survey design, this section provides a brief discussion of the major design features of surveys:
• Frequency options include data collection on a continuous basis, fixed periodic basis, irregular periodic basis, one-time or ad-hoc studies. Frequency is determined by need to detect changes over time and the likelihood of change.
• Geographic detail: national, regional, State, local. Most HHS surveys are designed to provide national and regional estimates.
• Population detail: subgroups of the population are defined by their race/ethnicity, poverty status, health characteristics, disability/functional limitation, age, and other characteristics. Most surveys require some detail on a variety of population subgroups in order to provide for adequate analytic capability. In order to provide the numbers to support such analysis, most HHS surveys make concerted efforts to screen and oversample population subgroups of interest. The groups oversampled vary by the objectives of the survey; for example, high expenditure groups are targeted by NMES; the young, race/ethnic minorities, and cigarette smokers are targeted by the National Household Survey on Drug Abuse; and race/ethnic minorities are oversampled in the National Health Interview Survey.
• Data sources: respondents to interview surveys include individuals, parents, family members or other proxy respondents, employers, and in some cases health care providers. Other surveys rely on administrative records maintained by health providers or employers. In some cases, providers are enlisted to record information about their encounters with individual patients.
• Mode of data collection: methods include household (in-person) interviews, telephone interviews (either RDD - random digit dialed, list-assisted, or calls made to known persons); self-administered questionnaires; direct physical and laboratory measurement; or abstracting of information from records. Each of these modes can be assisted by automated techniques, including Computer Assisted Personal Interview (CAPI), which uses laptop interviews in household interviews; Computer Assisted Telephone Interviews (CATI); Computer Assisted Self Interviewing, in which respondents interact directly with the CAPI system; and electronic access to provider records.
• Survey content. Depending on analytic needs and operational constraints, surveys range from highly detailed (such as NMES expenditure questionnaire and NHANES laboratory measurements) to more abridged or summary questionnaires.
• Data Quality, as measured by validity and reliability.
• Timeliness of data, both in terms of the length of time needed to obtain data (e.g., multiple visits over a year to the same individual, or surveying different individuals over a period of years to obtain sufficient sample size) as well as time needed to process data.
• Analytic uses, which include descriptive statistics, multivariate analysis, predictive modeling, and microsimulation.
Tab B: Summary of Selected HHS Surveys: Health Status, Access to Care, Health Care Use, Expenditures and Insurance Coverage
Health Care Use, Expenditures and Insurance Coverage
|National Medical Expenditure Survey (NMES): Household Survey
|Households; personal interviews
Person-level and family-level data (for calendar year):
|14,500 households, 38,000 individuals approximately every 10 years
Scheduled for 1996
|NMES: Health Insurance Provider Survey (HIPS)
|Employers, unions, and other sources of insurance
|Linked to household sample, to provide person and family level data on:
|Identified by household respondents; approximately 13,000 employers and 2,100 other providers of health insurance
|Medicare Current Beneficiary Survey (MCBS)
|Medicare beneficiaries (institutionalized and noninstitutionalized).
|Person-level, across calendar years:
|Approximately 12,000 individuals; rotating panel design with persons interviewed every 4 months for 3 years.
|National Employer Health Insurance Survey (NEHIS)
|Establishments; telephone interviews
|National and state-level estimates of:
|Planned as 52,000; 38,000 sampled in 1994. Annual (planned).
|National Health Interview Survey (NHIS)
|Households; personal interviews
|Annual data on:
Special topics include:
|National Health and Nutrition Examination Survey (NHANES)
|Personal interview; physical examination; laboratory tests; nutritional assessment
|30,000 persons aged 2 months and older; oversampling of Blacks and Mexican Americans; most recent cycle conducted over a 6 year period ending 1994
|National Health Interview Survey on Disability
|Households; personal interviews
|Phase 1 screens 50,000 households
Phase 2 is a followup of those persons with moderate to severe disabilities; approximately 20,000 per year
Special 2-year study
|National Survey on Family Growth (NSFG)
|personal interviews; telephone reinterviews
|10,000 women 15-44 years of age, drawn from the NHIS sample; oversampling of Blacks and Hispanics; conducted every 5 years, including 1995
|National Nursing Home Survey (NNHS)
|Institutions and next of kin;
|1,500 Nursing Homes
6,500 Nursing Home Residents
7,800 Nursing Home Discharges
Planned for 1995
|NMES - National Nursing Home Expenditure Survey
|Institutions and next of kin;
Personal and telephone interviews
|800 Nursing Homes
5,600 Nursing Home Residents
(3,200 1/1 residents)
Scheduled for 1996.
|NHSDA - National Household Survey on Drug Abuse
|personal interviews in households and group quarters
|18,000 household interviews of population aged 12 and older; oversampling of blacks and Hispanics, 12-34 year olds and cigarette smokers; conducted annually
Tab C: Data on the Public Health Infrastructure at Federal, State, and Local Levels
The FY 1996 Data Investment includes funding for a strategy to obtain comprehensive baseline and trend data on the public health infrastructure at national, state, and local levels. The strategy will build, to the extent possible, on data created by public health agencies in the normal course of business. It will be developed and implemented through a Federal/State/local partnership to assure that periodic information can be obtained which is useful at all governmental levels and to the general public health community . A steering group, including representatives of federal, State, and local health agencies, community public health providers, and the research community will advise and guide the overall progress and direction of the study.
The term "public health infrastructure" is used here to mean the federal, State, and local government agencies responsible for the health of the population. These agencies share the vision, mission, and responsibility for providing the essential public health services. At a minimum, these agencies include the U.S. Public Health Service; State and local public health, mental health, and substance abuse agencies; and that portion of the U.S. Environmental Protection Agency and State environmental health agencies directly committed to the protection of public health.
Many agencies and organizations have or are collecting data on aspects of the U.S. public health infrastructure. Nonetheless, during the course of the recent policy debate over health reform, the inadequacy of these data became apparent. For example, with data currently available, policymakers at all levels are limited in their ability to assess the impact of proposed changes in the health care system on the capacity of communities, States, and the nation to carry out essential population-based public health functions or to provide safety-net services. Equally important, available data does not adequately support the evaluation and monitoring of our investment in public health or the relationship between the capacity and functioning of the public health system to health care costs and the health of the population.
A more comprehensive data collection strategy is needed, encompassing information from substance abuse, mental health, environmental and other health-related agencies, which can characterize at national, State, and local levels:
• the provision of "essential services" of public health
• the persons providing these services
• expenditures related to their provision
• revenue sources and uses
• the extent to which population-based public health services cross-subsidize personal health care services and vice versa.
These data need to be obtained periodically and be sufficiently resolved to support the analysis of trends in public health resources, staffing, capabilities, and performance. This requires linking of data across different jurisdictional boundaries (for example, between city, county, multicounty or district health departments and agencies providing mental health and substance abuse services). Linkage to other datasets is also necessary, to relate capacity in population-based public health to sociodemographic factors, risks to health, the utilization of medical care, and health outcomes.
Tab D: Surveys to Objectively Measure Health
A system for the collection of data on objective measures of health, such as the NHANES operated by the National Center for Health Statistics, is an essential element of any national health data collection effort. Such surveys have been conducted as part of the overall data strategy of the Public Health Service since the 1960's, with examination surveys conducted at regular intervals from 1960 until the completion of the most recent survey in 1994.
Without objectives measures, collection on a representative population using standardized measures, any analysis of health status, health care utilization, expenditures, preventive strategies or performance monitoring will be severely compromised. Because of the costs and complexity of conducting nationally based examination surveys, it has been necessary for various agencies to consolidate their analytic needs. The NHANES survey is designed to obtain general health information needed by all agencies of the Department as well as the unique needs of each agency. Nearly all institutes of the NIH, FDA, EPA, CDC, and others have collaborated on the third NHANES which was recently completed.
Although the specific analytic objectives of data collected by examination surveys are extensive, they can be thought of as falling into two main classes.
• Information obtained from interviews or from medical records reflect how individuals view their own health and what they do about it in terms of seeking health care, the kind of care sought and their ability to perform normal functions. These mechanisms are not only a function of objective health status, but of social, environmental and behavioral circumstances. Examination studies provide information on objective indicators of health and disease, and therefore, provide the only source of information on undiagnosed conditions. This is particularly important when access and utilization of health care services, as well as knowledge of disease process, are not evenly distributed in the population. Without such information it is impossible to set research and intervention priorities, evaluate interventions, monitor performance and evaluate general health status. Objective measures of health also can be used to create predictive models of health status and the need for health care. Such measures are better predictors than are measures derived from other sources.
• Information from examination surveys provide standard distributions against which other data can be benchmarked. Most studies which rely on objective health measurement, such as clinical trails, are limited to select populations or geographic areas. It is essential that researchers be able to compare their populations to national norms in order to more fully interpret their findings. Nationally based examination surveys of sufficient size are the only means to develop these normative distributions. Information from national surveys is used to more efficiently plan intensive clinical or epidemiologic investigations, and also has clinical implications as evidenced by the use of growth charts to evaluate the development of children.
A few selected specific examples of the use of examination data collected by the NHANES is provided below.
Hepatitis viruses: NHANES is the only source of information on the number of people infected with the hepatitis viruses in the United States. Infection with these viruses is primarily asymptomatic (only 30 percent of infections have classic symptoms that would be diagnosed by a physician) but there are long term consequences with regards to chronic liver disease. Data from NHANES II (1976-1980) were the first evidence that hepatitis B was disproportionally infecting the minority population in the U.S. Until serologic tests were included in this survey, the epidemiology of this viral infection was estimated from blood donors, a very unrepresentative sample of the population. Based on NHANES II results, a universal vaccination program was implemented by CDC to eradicate hepatitis B transmission in the U.S. Future NHANES will help monitor the effectiveness of this program and further evaluate the epidemiology of all the hepatitis viruses on a national basis.
HIV: Measurement of prevalence of antibody to human immunodeficiency virus (HIV) on a representative sample of the U.S. population has provided CDC with a serologic basis for establishing an estimate of the number of people infected with this virus.
Tetanus: Tetanus antitoxin levels in the NHANES III population provided data for the first time to document that a substantial proportion of 10-16 year olds are not immune to tetanus. Although information obtained from interviews resulted in an estimate that 96 percent of children were vaccinated by the time they entered school, data from direct physical examinations revealed that these children had not received the required number of shots (5), and therefore were not fully protected. Only serologic data on a representative sample can provide this type of analysis of the immunization status of the population, to fully examine the medical practice and access to care in the U.S.
Herpes simplex type 2: Testing for antibody to Herpes simplex type 2 has been conducted in the previous two NHANES cycles as an index of sexually transmitted diseases (STD) in the population. These data linked to the STD risk behaviors provide evidence for an upward trend in STD that has helped redirect prevention efforts.
Helicobacter pylori: H. pylori has been confirmed as the causative agent of gastritic ulcers in adults and children. Recent evidence has suggested that infection with this organism is a critical precursor to the development of gastric cancer. This infection is primarily asymptomatic, therefore information from physician office visits or clinical trials do not provide the "big picture" on who is getting infected needed to establish effective screening and prevention programs. Testing of antibody to this organism had been performed on NHANES III sera and will be included in future surveys to provide a mechanism to estimate who is infected with this organism and help determine risk for infection.
Oral Health: Data derived from the NHANES have been critical for monitoring oral health status, risk factors for disease, access to preventive and treatment services and other variables among the general population and special subpopulations. Dental diseases have a major impact on the economic productivity of the Nation. In 1989, more than 164 million hours were missed from work and more than 51.6 million hours of school were lost because of dental treatment and problems.
Osteoporosis: Each year there are nearly 170,000 wrist fractures, 500,000 vertebral fractures, and 250,000 hip fractures in the U.S. population. It has been estimated that the annual cost of osteoporosis is about $10 billion. The magnitude of this problem is likely to increase dramatically over the next few decades as the population ages. Important pieces of data are not currently available about the changes in bone mass in the population. During NHANES III, dual energy x-ray absorptiometry of the hip was included in the examination of a subset of the examinees.
Osteoarthritis (x-rays): Arthritis and back problems (one of the musculoskeletal pain syndromes) are among the most frequently occurring chronic conditions affecting the U.S. population. They have a substantial impact on the quality of life of the individual, on the use of health care resources, and on the nation's economy. The most frequent form of arthritis, osteoarthritis (OA), has a formidable impact on the burden of disability and dependence among older Americans. Despite the public health impact, OA remains an enigmatic condition. There is no consensus as to the etiology, clinical features and natural history. Data from NHANES I have been useful in identifying obesity as a risk factor for symptomatic knee osteoarthritis. NHANES data were also valuable in identifying repetitive knee use in the work place as a risk factor.
Lung function (Spirometry): NIOSH must include in its lung function medical screening recommendations to OSHA and MSHA appropriate reference values. Until the completion of NHANES III, the most appropriate reference values recommended by NIOSH were based on less than 300 subjects. Although NHANES III provided critical information and a larger population base, reference values need to be continually updated as the population ages. The information generated is also used in medical care practice by physicians to directly assess the status of their patients. EPA is a major user in community studies.
Eye examination (Fundus photograph): The National Eye Institute has estimated that the cost of eye disease is in excess of $22 billion per year. However, no data exist on the current prevalence of visual acuity impairment and major causes of visual impairment in the population. In order to plan services and institute preventive programs, they have proposed the inclusion of an eye component in the next NHANES.
Cardiovascular disease (ECG): Cardiovascular disease is the leading cause of death and disability in the United States, and a primary cause of acute hospital bed days and physician visits. The assessment of cardiovascular disease-related risk factors has been a central component of the NHANES. Data describing rates of morbidity and disability related to cardiovascular disease have provided important information to researchers, health providers and policymakers from the public and private sectors of the health field. The population distribution of serum lipids and lipoproteins will provide reference data for evaluating progress in reducing serum cholesterol levels in the population through the National Cholesterol Education Program. Electrocardiograms will provide reference data for NHLBI studies of cardiovascular disease in older adults. High blood pressure control and reduction of serum cholesterol levels are among the priority areas specified in the 1990 Health Objectives for the Nation. NHANES III data will provide essential data describing the prevalence, determinants, and levels of public awareness of blood pressure and blood cholesterol and their relationship to cardiovascular disease morbidity and mortality.
Diabetes--Fundus photography : Diabetes mellitus is well documented to be a major public health problem in the United States. Fundus photography is included for adults 40 years of age and older for the purpose of diagnosis of retinal disease. Because retinopathy is one of the earliest complications of diabetes to appear, baseline fundus photographs will be used to document the natural history of diabetes and to quantify risk factors for the complications of diabetes. Of equal importance to NHANES III is the fact that photographs can be used to study the prevalence and natural history of macular degeneration.
Gallbladder disease: The results of the study of gallbladder disease in the NHANES III will effectively complement the clinical research also sponsored by the NIDDK. A better understanding of the prevalence of gallbladder disease and of the risk factors for this disease will result in greater opportunity for the primary prevention of gallbladder disease.
Allergy: The allergy data collected will be used by several agencies. The National Institute of Allergy and Infectious Diseases (NIAID) will use the information to further its knowledge about the distribution and determinants of allergies in the United States. The Health Resources and Services Administration (HRSA) will use the data to determine the demands allergies make on the health care delivery system. The FDA will use the data on the standardized allergens to assess reactivity in the general U.S. population. The NHLBI will use the information to increase its understanding of COPD. The history of asthma and skin test reactivity will be required to study COPD in NHANES III. A complete history of allergies and their effect on daily living is also required to describe child health accurately.
Food Fortification Policy: FDA, CDC, NIH and OASH are using the NHANES III nutrition biochemistries and dietary data to explore the safe levels of fortifying the food supply with folate. There are safety issues for older Americans with potential B12 deficiency and the detailed physiological data are necessary for changing the diet of the population with some confidence that no harm will be done. The public health issue is that folate levels may be related to neural tube defects and heart disease. The PHS group is currently active in examining this issue. Additional biochemistries are in progress that will enable the NHANES database to address the issue in sufficient detail. Historically, FDA has used the information to formulate or review current fortification and safety policies for iron and vitamin A in food products.
Exposure to toxic substances: NHANES information based on blood and urine specimens has produced comparative data for the study of toxic waste dumps, soldiers from Desert Storm, and exposure of the population to 15 to 20 pesticides and volatile toxic substances such as benzene, toluene, pentachlorophenol and others. Cadmium a highly toxic element has also been measured and exposure information will be produced. This type of information can be vital to administrators monitoring the potential exposures and risks as they are identified.
Tab E: Proposal for the Consolidated Health and Health Economic Survey System
Design Features, Efficiencies, and Analytic Capabilities
The system outlined below describes an integrated and consolidated data collection effort across four distinct areas of HHS surveys: 1) population-based surveys; 2) employer-based surveys; 3) efforts to characterize and measure the health and public health infrastructure; and 4) health care provider utilization surveys.
The National Health Interview Survey (NHIS) and the National Medical Expenditure Survey (NMES) serve as the foundations for the consolidated household-based health and health economic surveys. The NHIS would continue as an annual survey and serve as the nucleus of an integrated approach to the Department's personal household interview surveys. The NHIS sample of approximately 40,000 households (105,000 individuals) would be administered a core household interview based on the questionnaire redesigned for 1996 implementation. The core includes questions on demographics, SES, health status, utilization, insurance coverage and access. In addition to obtaining information on all household members, additional core questionnaires are administered to a sample adult and a sample child. Periodic modules in the area of utilization, behaviors, and health status collect more detailed information on a rotating basis.
The NHIS national core sample would serve as a sampling frame for the Medical Expenditure Panel (MEP), which would replace the periodic NMES. This panel survey would obtain national annual estimates on health care utilization, expenditures, insurance coverage and sources of payment for the noninstitutionalized population, and for policy relevant subgroups that include the poor and near poor, the elderly, individuals with functional limitations and individuals predicted to incur high levels of medical expenditures. The MEP would be implemented in 1996, based on the sample from the 1995 NHIS.
The MCBS will be closely coordinated with the MEP in terms of 1) greater questionnaire coordination; 2) analytic linkages of the MCBS and MEP samples to increase the power of each survey; and 3) possible offsetting reductions in sample size if analytic goals can be achieved through such linkages. Further steps to consolidate the MEP and MCBS will be evaluated as the MEP reaches a steady annual state in 1998.
This sample design integration would yield a significant reduction in data collection costs associated with the separate screening interview currently planned in NMES III to facilitate oversampling of policy relevant population subgroups.
The consolidated NHIS and MEP would serve as an integrated, flexible sampling frame for more specialized studies addressing a broad range of health issues, furthering HHS's goals for integration and consolidation of survey efforts. The NHIS would serve as the sampling frame for the MEP, and could serve as the basis for other surveys that would require large population size, geographic distribution, and a common core of information from which to select sample persons. Households that are included in the MEP could serve as the sampling frame where a study required a baseline of detailed expenditure data, a longitudinal design, or other features of the MEP design.
These specialized studies would take a number of forms, ranging from special topic supplements to the NHIS or MEP (such as the current NHIS immunization and Healthy People 2000 supplements); targeted followup studies that involve a re-interview with the household (such as the current disability survey); or more involved followup or panel studies.
Two that will be included in the consolidated survey plan include:
• A Health Examination Component - designed to obtain objectively measured data on health conditions for prevention, nutrition monitoring, monitoring environmental exposures and immunization protection levels. In 1998, a modification to the current NHANES will be implemented for two years, involving a household examination on a sample of persons identified in the NHIS. A limited physical examination (height, weight, blood pressure, etc.) will be administered, and blood specimens will be collected for extensive laboratory tests. In the year 2000, a more extensive NHANES, involving a complete examination, dietary interview, biochemistries, and more extensive and specialized tests, will be implemented. This NHANES would be instituted on a continuous basis, with 3 years of data collection required to reach a sample for national estimates. Samples each year would be drawn from the NHIS.
• A Component Addressing Family Related Issues - designed to obtain estimates of reproductive health, family formation, dissolution and other dynamics addressed by the current NSFG, already linked to the NHIS sample. This component would be part of the consolidated survey plan at the time the next NSFG would have been fielded, in the year 2000.
In addition, possible components include:
• A Consumer Survey - designed to obtain more in-depth consumer data on satisfaction, preference and access to the health care system. Such a panel, for example, might include persons that changed health plans, experienced high utilization, or incurred significant out-of-pocket costs, as identified through the MEP questionnaire.
• A Substance Abuse Survey - designed to address substance abuse and mental health issues similar to surveys such as the National Household Survey of Drug Abuse (see Tab F for discussion).
Design of the Medical Expenditure Panel (MEP)
The MEP is specified as a continuous survey with sample peaks at five year intervals. A 1996 MEP sample of 10,000 households, reduced from the intial plan for the 1996 NMES, will be drawn from the 1995 NHIS, with additional sample added from NHIS in 1997 to meet original precision requirements for groups such as the functionally disabled, children with functional limitations, individuals below 200 percent of the poverty level, and individuals likely to incur high medical expenditures.
The 1987 NMES II and the designed 1996 NMES III can be considered the conventional versions of data collection related to health care utilization, expenditures, employment, and health insurance coverage. There is no need for such an expansive data collection, with respect to both sample size and breadth of information, at intervals less than 9 or 10 years. What is needed is the ability to continuously monitor health care utilization, expenditures, and health insurance coverage and to examine changes over time.
To meet these goals and to permit limited resources to be allocated to other data collection efforts within a ten year cycle, a continuous MEP survey is specified, which has sample size peaks at five year intervals starting in 1997, that satisfy national precision requirements for policy relevant population subgroups (12,600 MEP households in addition to the MCBS). In the off-years of the survey (e.g. 1998-2001, 2003-2006), the sample would be reduced in scale (8,000 MEP households in addition to the MCBS), but with sufficient sample for national estimation and for major policy relevant population subgroups of interest (e.g, the poor/near poor). The survey would also include a longitudinal component, with a nationally representative subsample of the MEP households followed for a period of two years.
The following summarizes the phase-in and ongoing cycles of the MEP, by calendar year:
1995: The sample persons (approximately 10,000) from a sub-set of PSU's from the second and third quarter of the 1995 NHIS will be selected to participate in the 1996 MEP. This will require a change in the current plan for Phase II of the Disability Survey tied to the 1995 NHIS, and also require that the NHIS sample be released by November 1995. The NHIS includes an oversampling of blacks and Hispanics, but not other groups that would have been screened for the original 1996 NMES.
1996: Conduct MEP on a sample of 10,000 households (reduced from the original NMES III design of 14,500 households). Most of this sample (8,000 households) would be followed for two years; the remaining 2,000 households would be followed for three years, providing longitudinal data to facilitate analysis of change over time. Households sampled in 1996 and 1997 would be interviewed using an efficient conventional design--that is, the full breadth of data of interest (utilization, expenditures, health insurance, health status, access to care, long term care supplement), but to the extent possible, using more efficient data collection approaches such as telephone contacts. All households in 1996 and 1997 would be included in follow-back studies of medical providers and health insurance providers (MPS and HIPS).
1997: A new panel of 3,000 households selected from the 1996 NHIS would begin, in addition to the continuation of the 1996 panel of 10,000. This 1997 panel would be followed for two years. Coupled with data from the MCBS, this would provide the department with the analytic capabilities first proposed for the 1996 NMES III with respect to sample size. The same instruments used during 1996 would be administered during 1997.
1998: A new panel of 3,000 households selected from the 1997 NHIS would begin; coupled with the 2,000 households from 1996 and the 3,000 household panel from 1997, the overall sample size would be 8,000 cases. This 8,000 case sample of households is seen as the minimum size to permit monitoring of utilization, expenditures, and insurance among the population subgroups of policy interest (e.g., poor/near-poor).
The questionnaire used in 1998 would be a reduced questionnaire, with a central focus on utilization, expenditure, and insurance information. Limited MPS and HIPS information, with a 25 percent reduction in the HIPS sample.
The panel selected in 1998, and each subsequent year, would be followed for two years.
1999: A new panel of 5,000 households and a continuation of the 1998 panel of 3,000 households interviewed with the reduced questionnaire. Limited MPS and HIPS information, with a 25 percent reduction in the HIPS sample.
2000: A new panel of 3,000 households and a continuation of the 1999 5,000 household panel interviewed with the reduced questionnaire. Limited MPS and HIPS information, with a 25 percent reduction in the HIPS sample.
2001: A new panel of 5,000 households and a continuation of the Year 2000 panel of 3,000 households interviewed with the reduced questionnaire. Limited MPS and HIPS information, with a 25 percent reduction in the HIPS sample.
2002: To permit mid-cycle estimation (from a sample size perspective) paralleling the capabilities of the 10-year conventional version, in 2002, we would recommend a new panel of 8,000 households (added to the 5,000 panel from 2001). Half of this new panel would be interviewed only for one calendar year; the remaining would be followed into 2003. Content of the household questionnaire could be limited to the content from the previous three years or slightly expanded (e.g., administer the access to care supplement). Similarly, the questionnaire content of the MPS or HIPS follow-back could be assessed and if necessary, conducted for a subset of cases.
2005: Drop back to an overall sample of 8,000 households, with 4,000 continuing from the previous year for each of the years. Content similar to 1999-2001.
2006: Begin 10-year cycle again, with increase to 10,000 cases overall (new panel of 6,000 selected in 2006). Expanded household questionnaire and full sample of MPS and HIPS.
Expanded and Enhanced Analytic Capabilities
By moving to this consolidated, annual household data collection effort, the Department expands and enhances its analytic capabilities as described below:
• Retains the design of the core NHIS household interview. This core will provide cross-sectional population statistics on health status and health care utilization with sufficient sample size to allow for analyses based on breakdown of detailed age, race, sex, income and other socio-demographic characteristics, and also will allow for data on a broad range of topics currently provided by the NHIS.
• Retains the analytical capacity to obtain both annual and quarterly population estimates of health care utilization and the prevalence of health conditions for the nation and for policy relevant population subgroups.
• Provides the ability to model individual (and family-level) health status, access to care and use, expenditure, and insurance behavior over the year and examine the distribution of these measures across individuals. The longitudinal feature of the survey to collect data over multiple years further enhances the capacity to model behavior over time.
• Provides the ability to relate data from a detailed survey sample (e.g., MEP) to a larger population sample (e.g., NHIS) to enhance the utility of the MEP for national health account estimation and microsimulation modeling, including disaggregation by age group or geographic area.
• Provides the potential to expand to State-level estimates for marginal costs using the enhanced 358 PSU sample design of the NHIS.
• The longitudinal (over several years) aspect of the consolidated data collection effort provides the following:
- An increase in statistical power to examine change or make comparisons over time.
- The capacity to examine changes over time as well as changes in the relationship among measures of health status, access to care, health care use, expenditures, health insurance coverage, employment, functional limitations and disabilities, and demographic characteristics.
The National Employer Health Insurance Survey (NEHIS) was developed to obtain national and State level estimates of the number of employers offering health insurance, their costs, the coverage and characteristics of their respective health plans. In the NMES III Health Insurance Providers Survey, detailed information related to employer provided health insurance plans is also obtained, including details of plans held by household respondents. As currently designed, there is noticeable overlap in the focus of the two surveys.
The proposed design integrates the analytical capabilities of these distinct surveys. The overall survey design of the NEHIS survey would be modified to improve upon the limitations of the 1994 survey. Revisions would be made to the survey instruments based on results from an evaluation of the quality of the instrument used in the 1994 survey. The revised NEHIS Establishment Questionnaire and the revised NEHIS Plan Questionnaire would serve as the core questionnaires to be administered to all establishments associated with MEP survey respondents that comprise the Consolidated System of Health and Expenditure Surveys. The employers associated with the MEP survey respondents would receive a supplemental streamlined HIPS-type questionnaire to obtain person-based information on employer sponsored health insurance coverage (e.g., household members' specific coverage and premium).
The resulting design would reflect a consolidation of the questionnaire designs, imputation techniques, estimation tasks and data base designs across the surveys. Efforts will be made to develop estimation strategies to allow the two distinct employer samples to be combined to produce national estimates of the number of employers offering health insurance, their costs, the coverage and characteristics of their health plans.
The consolidated NEHIS/HIPS would parallel the sample size cycles of the MEP. A sample of 42,000 employers every five years would coincide with the more extensive MEP, and provide both national and 50-State estimates in that year. In the off-years of the survey (e.g. 1998-2001, 2003-2006), the sample would be reduced in scale to 15,000 employers, but with sufficient sample for national estimation and for major policy relevant employer population subgroups of interest (e.g, classification by size, industry classification).
Consolidated Design Features
The establishment-based consolidated health survey would incorporate the data collection efforts for NMES-HIPS and NEHIS into one effort with the following features:
• a cross-sectional design, in which establishments are interviewed once in any given year ;
• a continuous data collection effort, providing annual data for every year thereby maximizing on efficiency while minimizing costs associated with the start of a data collection effort;
• a sample size of 42,000 establishments every five years (at peak cycles), and 15,000 establishments annually; and
• a single questionnaire and data collection effort which satisfies the analytic capabilities of the individual data collection efforts.
Consolidated Design Efficiencies
The consolidated design reflects the following reductions in burden and associated costs:
• a reduction in the number of employer interviews;
• a reduction in the annual costs of data collection;
• efficiencies with respect to questionnaire design and implementation similar to those outlined for the consolidated household survey; and
• efficiencies with respect to post-data processing (editing, imputation, weighting, production of analytic data files), similar to those outlined in the consolidated household survey.
Expanded and Enhanced Analytic Capabilities
With respect to analytic capabilities, the consolidated effort provides:
• the ability to monitor changes in the coverage and costs of employer-sponsored health insurance at both the national and State level;
• provide measures of employer-sponsored health insurance coverage and costs for the national and State health accounts;
• assess changes in the provision of health insurance by employers;
• analyze person-level and family-level changes over time in the scope of health insurance coverage and options, for the proportion of the sample linked to the consolidated household survey;
• analyze behavior in the context of family choices and employer responses; and
• provide data on employer characteristics at both the establishment level and the firm level.
Measuring the Health Care and Public Health Infrastructure
The Department now obtains data on the level, characteristics and distribution of the health workforce and the physical capital in the health system through a number of separate inventories and surveys, with several more in the early planning stages. Data are obtained from a variety sources, including direct data collection, collaborative networks with State agencies, purchase of data from professional associations, and gathering of data available from public sources. In addition to data collection activities, the Department supports analytical programs in health resources and health workforce.
HHS programs include:
• National Health Provider Inventory, which focuses on nursing and related care homes;
• National Reporting Program for Mental Health Statistics - (Family of inventories and sample surveys of mental health organizations, and mental health services in prisons and general hospitals);
• Drug and Alcohol Services Information System (Annual inventory of alcohol and substance abuse treatment units and associated client data);
• HRSA's health professions analysis program (Analytical activities involving modeling and projections based largely on secondary data from other agencies and professional associations such as AHA and AMA);
• Area Resource File, a composite data base of county level information on health personnel, health facilities, health expenditures, socioeconomic and demographic data;
• The planned Public Health Infrastructure Data System, which will obtain data on the capacity, functioning, and funding of essential public health services in State and local public health agencies;
• HCFA National Provider Files; and
• HCFA health facility survey and certification activities.
The primary objectives of these activities include needs assessment; workforce planning; program administration; designation of medically underserved areas; descriptive analyses of the health care system; and sampling of providers for utilization surveys. Further, these data are essential to the evaluation and monitoring of our investment in public health and the relationship between the capacity and functioning of the public health system to health care costs and the health of the population.
The health facility and provider surveys have focused on selected providers of care rather than on the broad range of players in the health arena. As a result, there is a lack of systematic, comprehensive information that characterizes the capacity of the personal health care system and the infrastructure of the public health system over time. These efforts do not fully document and reflect the complexity of the underlying system, which is already heterogeneous and rapidly changing.
In the FY 1996 HHS Data Investment Initiative, funds have been earmarked for improvements in this area. The Investment provides for an expansion of the National Health Provider Inventory to additional health providers; development of a new data strategy for the public health infrastructure; new information on managed care settings; expansion of specialized inventories of substance abuse and mental health providers; and expansion of the Area Resource File.
What is needed is a conceptual framework--in effect, a broader systems view--that allows for characterization of the health system as a whole, the analysis of interactions between supply and demand, and the analysis of the relationship between capacity, functioning of the system, and cost. Such information would allow modeling of the impact of change in one aspect of the system on others (e.g., the interaction of the personal and public health systems under various health reform scenarios). It will allow HHS to monitor the capacity and performance of the providers, State and local agencies, and other players that are relied upon to accomplish national objectives.
In addition, regional, State and community level perspectives are essential in resource analysis and planning. Accordingly, a stronger focus on systems-wide or community perspectives would allow for analysis of the overall structure of the system in terms of regionalization, organization, redundancy, etc.
During the next year, HHS will take steps to develop such a conceptual framework, focusing on the characterization of communities and other geographic areas, relating systems capacity to measures of demand obtained through population surveys, and relating capacity measures to utilization surveys.
These steps will include:
• development of a common set of terms and definitions that can be used to characterize the health workforce and health establishments in all data collection systems;
• an HHS-wide approach to identifying commonalities that will allow linkages - such as standards for geocoding information, linkages between data systems obtaining information in the same areas, etc. These efforts can build on HCFA experience in developing national provider identification systems;
• development of approaches to defining and measuring trends in the structure, organization and operations of managed care entities;
• development of conceptual and analytic linkages that can tie separate data collection efforts into a systems-wide "virtual" inventory;
• coordinated approaches to data partnerships with States, as well as coordinated approaches to purchasing of secondary data from associations; and
• development of a strategy to direct HHS health workforce and resources data into accessible analytical data bases.
With such a conceptual framework, HHS can rationalize, consolidate, and coordinate existing and planned activities in this area.
Surveys of Health Care Institutions and Providers
HHS currently conducts multiple provider-based surveys, including components of NCHS' National Health Care Survey (hospitals, physicians, nursing homes, ambulatory surgery, and home and hospice care), the provider followup components of NMES, and the Healthcare Cost and Utilization Project. Moving beyond measures of capacity and infrastructure, these surveys measure the utilization of services by quantifying the types of diagnoses and services provided.
HHS' basic approach:
• In instances where multiple HHS surveys approach the same class of providers, efforts will be consolidated operationally so that there are common field staff, procedures, computer-assisted survey software, and post-processing capabilities.
• A common core questionnaire will be identified for use in surveys that would approach each type of provider.
• Common classification systems, standards, procedure coding, etc., will be adopted that would maximize efficiency as well as enhance data comparability and analytic utility.
Consensus has been reached on consolidation of the components of the National Health Care Survey and the provider followup components of the MEP. Additional steps will be taken to evaluate the extent to which other efforts, such as the Healthcare Cost and Utilization Project, the Drug Abuse Warning Network (DAWN), and the institutional component of the MCBS can be integrated into this framework.
A pending RFP for the continuation of DAWN provides for theredesign of DAWN. The redesign will be done in stages, with initial steps including extensive evaluation of the current design,determination of current data needs, and consultation with survey design experts and DAWN data users. The redesign issues that will be addressed are completely open, but it is anticipated that the scope of the survey, definitions of cases, sampling methods, and possible consolidation with other ongoing Federal government surveys will be studied. Later stages of the redesign will include conducting a pilot test and implementing the new design, assuming agreement is reached during the initial phase regarding the appropriate design.
Surveys of Nursing Homes and Related Institutions
Shortcomings with current surveys of nursing homes and related institutions include:
• The MCBS-institutional survey, while providing use and expenditure data for both nursing home residents and residents in personal care homes, is limited to a relatively small sample size of 1,000 individuals who are Medicare beneficiaries. The small sample size does not permit analysis for some subgroups of particular policy interest (e.g. individuals under the age of 65, minority populations), nor does the sample capture the non-Medicare population.
• The NNHS includes all residents of nursing homes (Medicare and non-Medicare eligible) as well as residents of personal care homes. Unlike MCBS and NMES-NNHES, the NNHS does not collect expenditure data; unlike NMES-NNHES the NNHS does not include a survey of the next-of-kin to collect information on family assets and community-based network of potential care givers.
• NMES-NNHES also is limited to nursing homes and does not include a sample of residents from personal care homes. While it does collect both use and expenditure data as well as data from the next of kin, the current design is relatively expensive in comparison to either MCBS or NNHS (approximately $14 million as compared to $1 or $1.5 million).
The proposed nursing home survey component of the Consolidated Survey would be collected every five years and jointly satisfy the analytical objectives of the respective AHCPR NMES III National Nursing Home Expenditure Survey and the NCHS National Nursing Home Survey. The survey design would be modelled to facilitate calendar-year use and expenditure estimates. The sample of facilities would be increased from 800 to 1,200 to satisfy precision requirements for facility level characteristics, while the sample of residents would remain at the level specified for the NMES III survey (3,200 residents in facilities at the start of the survey year and 2,400 first admissions over the course of the survey year). Data on the capacity, staffing, and services provided by the institutions would also be collected.
This sample of institutionalized residents would be coordinated with the institutional sample selected from the MCBS sample. Approximately 1,000 residents would be followed longitudinally across multiple years according to the MCBS data collection plan.
The design would include a more frequent survey of institutions, conducted in years between the full survey, to obtain necessary data on the capacity, staffing, and services provided by the institutions.
Survey Design Enhancements
To obtain complete annual profiles of health care expenditures at the person level, individuals sampled from the household component of the Consolidated Health and Health Economic Survey who entered long-term care facilities would be followed (approximately 250 individuals annually), and their institutional use and expenditure data collected. This is consistent with the current MCBS approach. This annual sample of institutional users selected from the MEP would be combined with the MCBS institutional sample to increase the precision of survey estimates that characterize the institutional population over levels currently attained through the MCBS. The current NMES III survey restricts coverage of the institutional population to individuals in nursing homes. The scope of the survey will be enhanced to attempt to represent individuals who reside in board and care homes.
Consolidated Design Features
The institution-based consolidated health survey would incorporate the data collection efforts for the NNHS, and NMES-NNHES into one effort, closely coordinated with the MCBS (institutionalized population) with the following features:
• a longitudinal design, in which nursing home residents would be followed for a period of 16 months (similar to NMES-NNHES), but with one less core round of data collection (from 4 to 3 core rounds) for the Consolidated Nursing Home Survey;
• a sample of 1,200 facilities; this sample would be merged analytically with the facilities associated with the MCBS institutional sample;
• a sample of 3,200 residents at the start of the calendar year and 2,400 new admissions over the course of the calendar year; this sample would be merged analytically with the MCBS sample of approximately 1,000 institutionalized residents that are surveyed each year;
• a single questionnaire and data collection effort which satisfies the analytic capabilities of the individual data collection efforts; and
• data collection repeated every five years.
Consolidated Design Efficiencies
The consolidated design reflects the following reductions in burden and associated costs:
• a significant reduction in the facility and resident sample sizes from the targets considered in the individual NCHS and AHCPR planned surveys;
• associated reduction in annual data collection costs;
• efficiencies with respect to questionnaire design and implementation, similar to those noted for the household consolidated effort; and
• efficiencies with respect to post-data processing (editing, imputation, weighting, production of analytic data files), similar to those discussed for the household consolidated effort.
Expanded and Enhanced Analytic Capabilities
The consolidated design provides the analytic capability to:
• examine the health status, medical care use and associated expenditures for nursing home residents over the course of a year, paralleling the data available for the noninstitutionalized population;
• assess the size of the Medicare-population institutionalized in personal care homes and explore the feasibility of using the Medicare beneficiary sample to identify personal care homes for estimating personal care home use by non-Medicare beneficiaries;
• examine acute care use (e.g., hospitalizations) for institutionalized individuals; and
• examine nursing home use for the non-Medicare population and changes in utilization by this population over time (a growing sector of the nursing home population).
Surveys of Hospitals
NCHS currently conducts the National Hospital Discharge (NHDS) as part of the National Health Care Survey. The NHDS is based on a list sample of hospitals, and discharge records for a sample of patients are abstracted to obtain data including diagnosis, surgical and other procedures, etc. The NMES medical provider study follows individuals in the household sample to hospitals where they obtained treatment, obtaining data similar to the NHDS but also including financial information.
Under the HHS plan, these surveys would be consolidated so that:
• Providers are identified from list frames or inventories and by respondents in the MEP;
• A single field agent will be used to collect data from hospitals;
• A common core of questions would be obtained from patient records regardless of which sample was used; and
• Gains would be achieved with respect to questionnaire design, training and fielding of interviewers, editing, imputation, and data based development.
Additional study will be given to the relationship of the Healthcare Cost and Utilization Project, which obtains data from State-wide discharge systems, to this consolidated activity.
Consolidated Design Efficiencies
Similar efficiencies to the nursing home component will be realized, including reduction in cost of field operations, savings in questionnaire design and implementation, and efficiencies in post-data processing.
Non-hospital Medical Care Providers
For NAMCS, NHAMCS, NHHCS, NSAS, and for portions of the NMES-MPS that approach hospital-based and office-based physicians and ambulatory surgery:
• Providers are identified from list frames or inventories and by respondents in the MEP.
• A single field agent will be used to collect data from all providers in these categories.
• A common core of questions will be asked of each provider. In addition, more detailed charge information will be obtained for the sample derived from the MEP and more detailed information concerning the content of care delivered will be asked on the broader sample.
• If feasible, multiplicity estimators can be used to combine the data from both frames.
• Gains will be achieved with respect to questionnaire design, training and fielding of interviewers, editing, imputation, and data based development.
The same approach will be explored for other provider surveys conducted by HHS, such as the Drug Abuse Warning Network (DAWN). A contract is currently being prepared by SAMHSA for the next cycle of DAWN, with the first phase of the contract being an evaluation of alternate means through which the data might be collected.
Public Health Surveillance and Grant Reporting Systems
Many of the information systems supported or used by HHS agencies are not national surveys, but surveillance and reporting systems designed to identify individual cases for public health intervention, track occurrence of infectious diseases in order to identify patterns, or provide accountability for Federal grant programs. Although there are many of these systems, the cost to the Federal government is not significant, since surveillance activities are typically based on voluntary reporting by State health departments.
There is wide consensus in the public health community that fundamental change is needed in the current patchwork of independent, uncoordinated, categorical disease surveillance and grant reporting systems. The report "Integrating Public Health Surveillance and Information Systems," drafted by the Centers for Disease Control and Prevention (CDC), outlines how this fragmentation has hindered public health efforts.
Efforts by CDC and PHS to foster the development of integrated information systems will accelerate, including:
• removal of unnecessary Federal requirements for States to maintain distinct, non- integrated data collection mechanisms, in order to allow States to develop integrated systems that satisfy both State and national needs;
• support of streamlined, integrated data systems at the State level that can satisfy disease surveillance objectives; and
• consolidation of categorical grants into performance partnership grants, with streamlined reporting requirements.
Efforts to streamline and integrate these multiple, distributed data systems will depend on strong Federal leadership in the development of consensus standards, definitions, and approaches to recording and sharing of data.
Modules for Obtaining State-Level Data
A State-level telephone survey would be designed by HHS to be conducted periodically to obtain basic health status, access to care, insurance, and expenditure data of importance for national policy analysis, performance evaluation, and modeling. By using the available NHIS sample, already distributed widely in States, and using questionnaires from the consolidated national surveys, State level data can be obtained efficiently and be directly comparable to national data. Linkage to the NHIS sample, and use of common questionnaires will greatly enhance HHS' and State's ability to create analytic approaches in which more detailed data available on national surveys can be modeled to the State level. If funding can be identified, this survey would be conducted at intervals when the MEP is off-peak.
The design will allow States to "buy into" the consolidated HHS survey to obtain data on their States on a more frequent basis than would be available in the HHS survey of all States, or possibly to target content relevant to State needs. States may find this to be an efficient option for monitoring performance partnership grants, and for meeting State policy, planning, and public health purposes. HHS would develop questionnaire modules, sampling strategies, and contract mechanisms through which States could participate in the consolidated HHS survey. Options might include 1) augmenting the existing sample from national surveys that already exists in a given State, either with additional household or telephone interviews; 2) using consensus questionnaire modules to replicate national surveys within a State; or 3) using national surveys as benchmarks for improved modeling and estimation at the State level.
The NHIS sample redesign implemented in 1995 increased the number of sampling units, eliminated sampling units that overlapped State boundaries, and made other changes to make State estimates more attainable. NCHS has already used the NHIS to provide State-specific data to New York and California, and the sample allows for direct estimates for additional States. States would be able to pay for the additional sample sizes necessary to meet their analytic objectives. This plan will take advantage of the combined features of the NHIS sample and the range of content and methods that will be included in the NHIS and the MEP to provide States with options that do not currently exist.
Tab F: Issues Related to Consolidating the National Household Survey on Drug Abuse
Description of NHSDA
The NHSDA is an ongoing survey of the civilian noninstitutionalized population age 12 and older in the U.S. It covers civilians living in households (including on military bases) and persons living in some group quarters such as homeless shelters and college dormitories. Active military personnel are excluded.
Conducted under contract, the sample is an area probability sample of about 120 PSUs, 2,000 segments, and 55,000 households per year. Household screenings at sample households are done to select sample persons within households (either one or none, and occasionally two). About 18,000 interviews are conducted each year, and the sample is allocated equally to the four calendar quarters, resulting in essentially continuous data collection in every PSU. Sampling rates of locations and of sample persons are controlled to accomplish oversampling of blacks and Hispanics, and also to oversample populations with elevated drug use rates (e.g., 12-34 year olds and cigarette smokers) to improve precision for rare drug use behaviors.
The survey is voluntary and no monetary incentives are offered. Response rates have typically been around 95 percent for screening and 80 percent for interviewing. Interviews are done anonymously and confidentially. No names are collected, and responses to sensitive questions are obtained using self-administered answer sheets that are not seen by interviewers. Other parts of the interview, such as demographic and health status questions, are interviewer administered.
The survey has undergone design changes regularly in the past, primarily in the sample design. For example, prior to 1990 the survey was conducted every 2-3 years with a sample of less than 10,000 per year. During 1991-93, six large metropolitan areas were oversampled, increasing the sample to 30,000 per year. In 1994-6, the oversample was eliminated resulting in a sample size of 18,000 per year.
The questionnaire has also changed. In 1994, SAMHSA implemented a newly designed questionnaire that incorporated a core-module structure to enhance trend measurement while encouraging outside agencies to sponsor modules of interest. The questionnaire was developed as a result of a series of methodological studies and field tests conducted by NIDA and SAMHSA during 1989-93. Other modifications that have been considered for future NHSDAs include (1) a modification to the sample design to improve estimation of mental health conditions, (2) incorporation of computerized interviewing methods such as audio-CASI, and (3) expansion of PSU coverage to enhance State-level estimation of substance abuse.
Issues Related to Consolidating NHSDA and NHIS
Anonymity, confidentiality and privacy are 3 different but interrelated issues that need to be addressed in surveys. Anonymity requires that no personal identifiers are collected or retained; confidentiality is the assurance that there will be no disclosure of individually identifiable survey data to unauthorized persons; and privacy (as narrowly used here) addresses issues related to keeping other persons present at the time of the interview and/or the interviewer from knowing the respondent's answers. Privacy is addressed through different interviewing methods and is essentially a technical issue, whereas anonymity and confidentiality involve many aspects of survey design including sampling, data collection, data processing, data release and perception, both of respondents and the broader public. Because the NHSDA focuses largely on sensitive behaviors, it utilizes special procedures to enhance anonymity, confidentiality and privacy for respondents to maximize response rates and accuracy. Research shows that these procedures enhance accurate reporting. Modifications in these procedures could effect the quality of the sensitive drug use data. Confidentiality and privacy for sensitive data could be built into the NHIS procedures, but the level of anonymity would depend on how the NHIS and NHSDA were linked, i.e., anonymity is lost once the NHIS core sample person is used, because identifying information such as the social security number are collected in NHIS. A drug use supplement (using self-administered forms to enhance privacy) on the 1991 NHIS found lower rates of adult cocaine and marijuana use than the NHSDA and also achieved a lower response rate. This issue cannot be resolved based on available research. It requires new methodological research that would address the following, for example:
• The importance of anonymity, confidentiality and privacy are likely to be different and possibly reduced with computerized interviews. For example, a technique that uses tape recorders and/or computers to administer the survey [e.g., audio-CASI] could enhance privacy, probably beyond paper and pencil self-administration. NCHS is currently analyzing drug use data from a 1992 NHIS supplement that used an audio-cassette interview to assure privacy to adolescents and youth.
• The perception may be more important than the reality in the actual effect of anonymity and confidentiality on reporting of sensitive behavior. Do respondents know the difference between anonymity and confidentiality and do they really believe the assurances they are given? We may learn something about this by conducting a small study of NHSDA respondents, using a short post-interview form asking about respondents' perceptions of anonymity and confidentiality and how these perceptions impacted on their reporting of behaviors and their decision to participate in the survey. NCHS data are protected from disclosure by the strongest confidentiality laws of any federal statistical agency, but what does this mean to respondents. Current research at NCHS is investigating the perceptions of respondents about the confidentiality of data collected about other sensitive topics, abortion and sexual behaviors.
Many of the most important behaviors measured by the NHSDA have very low prevalence. To maximize precision, the NHSDA sample design incorporates stratification and varying sampling rates that target the sample toward high risk populations. The need for data on minorities also dictates oversampling. If the NHSDA is to continue to provide data with the current level of precision for rare drug using behaviors and for race/ethnic and other subpopulations, the NHSDA sampling plan will need to be maintained within the NHIS design. This raises the following:
• Coverage of hard-to-reach populations will need to be more of a concern in the NHIS than it is currently, because substance abuse estimates are probably more subject to coverage bias than are many of the health measures estimated by NHIS. This could influence how the NHIS field work is managed, both in terms of conducting interviews and controlling response rates, and the need for detailed data on the field experience. In addition, certain types of group quarters, such as shelters, are not currently included the NHIS sample frame.
• It may not be possible to use the screened NHIS households as the frame for the NHSDA, because the oversampling of Hispanics in the NHIS will likely "use up" all the screened Hispanic households that would be needed for the NHSDA. The only way to alleviate this would be to screen more households to accommodate NHSDA needs, resulting in additional costs. However, the current NHIS sample list was designed to support a sample 4 times the current planned sample size.
The ability to link the NHIS behavior, health status (especially mental health), and utilization data with the NHSDA would be extremely valuable. It would reduce overall respondent burden because the health and demographic items currently collected by the NHSDA could be dropped. Even without complete linkage of the samples at the household level, the use of the same PSUs or segments in both surveys would allow contextual analyses to be done using survey data. This linkage could also benefit small area estimation of substance abuse, such as is currently being done by SAMHSA. Nonresponse and coverage adjustments for the NHSDA could potentially be improved with the linkage of the NHIS with the NHSDA samples.
Several issues concerning the operation of the survey will need to be addressed.
• The timeliness of the data is a key issue. NHSDA estimates are regularly released within 6-7 months of completion of data collection. This schedule is important to White House Office of National Drug Control Policy. The ability to meet this schedule is based on an intensive field management and data management effort involving data editing, weighting, table production, and report writing by contractor and SAMHSA staff. If the NHIS and NHSDA are integrated, we will either have to maintain this capability or convince data users to adjust their data needs.
• The timeliness of responses to ad hoc requests for data from the NHSDA involves a similar situation. SAMHSA currently is able to address key policy issues related to substance abuse on a quick response basis using the NHSDA data files.
• The NHSDA is conducted under contract. Issues concerning access to confidential data by SAMHSA and the NHSDA contractor will need to be addressed.
The current NHSDA contract runs through 1996, and a new RFP to conduct the 1997-2000 surveys is in preparation and will soon be released. It is anticipated that the earliest any merging could be accomplished would be for the 1997 survey. However, given the timing of the RFP, there will be substantial issues concerning modification to this contract if major changes in the NHSDA are made.
Summary of Options for Consolidating the NHSDA and NHIS
The options for incorporating the NHSDA into the NHIS can be summarized by the two extremes (a number of other options could be somewhere in between these two):
1. Select the NHSDA sample from the NHIS sample list frame. NHSDA and NHIS screening and interviewing would be done in separate samples of addresses.
Option 1 would accomplish the smallest cost savings and the least analytic advantage, but would allow the specialized NHSDA sample design and data collection methodologies to be virtually unaffected and most importantly would continue the assurance of anonymity.
2. Conduct the NHSDA as a sample of individuals in NHIS sample households. At the extreme end of this spectrum the NHSDA respondent would be NHIS sample person.
Option 2 would maximize cost savings and provide enhanced analytic capability with the NHSDA and NHIS data being collected on the same respondents, especially if the NHSDA sample person is the same as the NHIS sample person. However, methodological issues such as the importance of respondent anonymity and confidentiality would need to be addressed for this to be viable.
These two options at either end of a range of possibilities incorporate 3 different axes: no personal identifiers (anonymity) vs. personal identifiers (with confidentiality assured); independent screening for respondents vs. use of the NHIS respondents; and the coordination of questionnaire content, independent of survey methods. Coordination of content, especially the non-drug related items, now occurs but could be enhanced in the future, even with the current NHSDA design.
An anonymous survey design could have the following options:
• status quo, current NHSDA plan;
• current NHSDA, but incorporate additional health related data from the NHIS core;
• current NHSDA, but use of NHIS list frame to draw a non-overlapping sample for screening; or
• use same NHIS list for screening for both surveys.
A survey design with personal identifiers (confidentiality assured) could have any of the above options as well as the following options that would involve the direct use of NHIS sampled households:
• NHSDA as a followback to a respondent selected from the sampled NHIS household, and
• NHSDA as an integral part of the NHIS interview process, i.e., NHSDA sample person interviewed at time of selection or return visit.
Tab G: Baseline and HHS Consolidation Plan Budget Estimates
(This tab is not included in documents released to the public, since it includes information that has not yet been formally adopted as part of the President's budget for FY 1997, currently under development).
Tab H: Sample Size Specifications for the Design of the Medical Expenditure Panel (MEP)
NOTE: This tab includes information on the Medical Expenditure Panel that updates information included in the body of the report.
The 1996 MEP sample would consist of a sample of approximately 10,500 households selected as a nationally representative subsample of households that responded to the 1995 National Health Interview Survey (second and third quarters of the NHIS sample). The targeted overall response rate for obtaining calendar year 1996 data on health care use and expenditures is 80.75 percent (based on a 95 percent response rate for the NHIS and a 85 percent conditional response rate for three rounds of MEP data collection). Consequently, the MEP sample of 10,500 households responding to the NHIS interview would consist of approximately 8,925 households completing three core rounds of data collection to obtain calendar year 1996 data.
A new MEP sample of approximately 5,150 households would be selected as a nationally representative subsample of households that responded to the 1996 National Health Interview Survey. This sample would consist of an oversample of the policy relevant population subgroups identified for the 1996 NMES-3 design. In addition, the entire 1996 panel of 8,925 households would be continued to obtain calendar year 1997 data on health care use and expenditures (with a targeted round specific response rate of 96 percent). Consequently, the MEP sample for 1997 would consist of approximately 12,600 households completing three core rounds of data collection to obtain calendar year 1997 data (4,375 households from the new sample, 8,225 from the 1996 MEP sample). Coupled with data from the MCBS, this would provide the department with the analytic capabilities first proposed for the 1996 NMES-3 with respect to sample size.
A new MEP sample of approximately 4,700 households would be selected as a nationally representative subsample of households that responded to the 1997 National Health Interview Survey. In addition, the entire 1997 panel of 4,375 households would be continued to obtain calendar year 1998 data on health care use and expenditures (with a targeted round specific response rate of 96 percent). Consequently, the MEP sample for 1998 would consist of approximately 8,000 households completing three core rounds of data collection to obtain calendar year data (4,000 households from the new sample, 4,000 from the 1997 MEP sample). The 1996 MEP would be retired.
A new MEP sample of approximately 5,100 households would be selected as a nationally representative subsample of households that responded to the 1998 National Health Interview Survey. In addition, the entire 1998 panel of 4,000 households would be continued to obtain calendar year 1999 data on health care use and expenditures (with a targeted round specific response rate of 96 percent). Consequently, the MEP sample for 1999 would consist of approximately 8,000 households completing three core rounds of data collection to obtain calendar year data (4,300 households from the new sample, 3,700 from the 1998 MEP sample). The 1997 MEP would be retired.
A new MEP sample of approximately 4,700 households would be selected as a nationally representative subsample of households that responded to the 1999 National Health Interview Survey. In addition, the entire 1999 panel of 4,300 households would be continued to obtain calendar year 2000 data on health care use and expenditures (with a targeted round specific response rate of 96 percent). Consequently, the MEP sample for 2000 would consist of approximately 8,000 households completing three core rounds of data collection to obtain calendar year data (4,000 households from the new sample, 4,000 from the 1999 MEP sample). The 1998 MEP would be retired.
A new MEP sample of approximately 5,100 households would be selected as a nationally representative subsample of households that responded to the 2000 National Health Interview Survey. In addition, the entire 2000 panel of 4,000 households would be continued to obtain calendar year 1999 data on health care use and expenditures (with a targeted round specific response rate of 96 percent). Consequently, the MEP sample for 2001 would consist of approximately 8,000 households completing three core rounds of data collection to obtain calendar year data (4,300 households from the new sample, 3,700 from the 2000 MEP sample). The 1999 MEP would be retired.
A new MEP sample of approximately 10,120 households would be selected as a nationally representative subsample of households that responded to the 2001 National Health Interview Survey. This sample would consist of an oversample of the policy relevant population subgroups identified for the 1996 NMES-3 design. In addition, the entire 2001 panel of 4,300 households would be continued to obtain calendar year 2002 data on health care use and expenditures (with a targeted round specific response rate of 96 percent). Consequently, the MEP sample for 2002 would consist of approximately 12,600 households completing three core rounds of data collection to obtain calendar year data (8,600 households from the new sample, 4,000 from the 2001 MEP sample). Coupled with data from the MCBS, this would provide the department with the analytic capabilities first proposed for the 1996 NMES-3 with respect to sample size.
The 2000 MEP would be retired.
Drop back to an overall sample of approximately 8,000 completing three core rounds of data collection to obtain calendar year data on health care utilization and expenditures, with approximately 4,000-4,300 continuing from the previous year for each of the years.
Begin ten year cycle again with increase to 12,600 households completing three core rounds of data collection to obtain calendar year data on health care utilization and expenditures.