Title: Using the Sampling Frame for Major National Health Surveys for Other HealthRelated Surveys
1Using the Sampling Frame for Major National
Health Surveys for Other Health-Related Surveys
Trena M. Ezzati-Rice National Center for
Health Statistics Centers for Disease Control
and Prevention
2(No Transcript)
3Overview of data collection methods
- Vital statistics data collected in partnership
with states - Household interview surveys
- Examination surveys
- Health care provider surveys
4Four major data systems
- National Vital Statistics System and surveys
- National Health Interview Survey
- National Health and Nutrition Examination Survey
- National Health Care Survey
5NCHS data are used for
- Comparisons across populations, providers, and
geographic areas - Planning/assessing public health programs
- Identifying health problems, risk factors, and
disease patterns
6Outline
- Overview of U.S. Dept. of Health and Human
Services (DHHS) Survey Integration Plan - Overview of the National Health Interview Survey
(NHIS) sample design - Overview of surveys targeted for linkage to the
NHIS
7Outline
- Goals and objectives of the surveys targeted for
linkage - Summary of sample design and how linked to the
NHIS (current or in past) - Summary of advantages and limitations of linkage
- Other strategies for linkage
- Methodological research issues for integration of
health surveys - Summary
8Impetus for Survey Integration
- DHHS - number of population-based surveys
- Operation of surveys decentralized
- Perceived overlap and duplication
- Operational inefficiencies
- Major gaps in data or health care costs
expenditures - on an annual basis
9Impetus for Survey Integration
- Inability to analytically link data from
different sources - Reduce costs
- Emerging health data needs at the State level
10Key Elements of DHHS Survey Integration Plan
- Use of common sampling frames - NHIS as sampling
nucleus - Use common core questionnaires and common
definitions and terms - Reduce overlap
- Increase analytic capabilities
- Ease of use for research analysts
- Consolidate field operations for surveys of
employers and health care providers
11Key Elements of DHHS Survey Integration Plan
- Sharing knowledge and resources to reduce costs
- instrument testing
- sample design techniques
- estimation techniques
- Implement an ongoing longitudinal panel survey
- Redesigning NHANES using NHIS as its sampling
frame - Develop capacity for State-level estimates
12Household Surveys Targeted for Linkage to NHIS
- National Survey of Family Growth (NSFG)
- Medical Expenditure Panel Survey (MEPS)
- National Health and Nutrition Examination Survey
(NHANES)
13Alternative Levels of Linkage
- Respondent level Same respondents in both
surveys - HH (address) level Same HH - but possibly
different respondents - Segment level Same second-stage units (segments)
- but different HH within segments - PSU level Same first-stage units (PSUs) - but
different segments within PSUs
14Key Features of the NHIS
- Survey of the U.S. civilian noninstitutionalized
population - Continuously conducted since 1957
- Face-to-face interviews
- Broad range of health issues
- Stratified multistage area probability sample
design - Sample redesign every 10 years following
decennial census
15NHIS Sample Design 1995-2004
- Oversampling of black and Hispanic persons
- State and Census region stratification
- 358 sample PSUs
- Sample partitioned into subdesigns
- Assignment of sample to weekly and quarterly
subsamples - 4 panels or subsamples
16National Survey of Family Growth (NSFG)
- Survey of women 15-44 years of age
- Reproductive history and health of women and
children - First national household survey linked to the
NHIS
17National Survey of Family Growth (NSFG)
- Sample of women for NSFG Cycles IV (1988) and V
(1995) selected from women previously sampled in
the NHIS - NSFG IV sample selected from 156 NHIS PSUs based
on the 1985-1994 design - NSFG V sample selected from all 198 PSUs based on
the 1985-1994 design
18Advantages of NHIS NSFG Linked Design
- Avoids the costs of listing and screening with
identification in the NHIS of HH containing women
15-44 years of age - Particularly useful for the oversampling of
Hispanic and non-Hispanic black women
19Advantages of NHIS NSFG Linked Design
- Linked analyses, combining NHIS and NSFG data,
can be done - Use of NHIS data to adjust for nonresponse to the
NSFG to compute the sample weights - Use of NHIS data on income to help impute NSFG
missing income data
20Limitations of NHIS NSFG Linked Design
- Size of the NSFG sample is limited by the size of
the NHIS sample - Number of women aged 15-44 cannot exceed the
number of women in the NHIS sample (without
supplemental area sample) - Accumulation of more than one year of NHIS data
may be needed to meet sample size requirements
21Limitations of NHIS NSFG Linked Design
- Need to trace persons who moved since their NHIS
interview (moving rates vary by age) - Potential for decreased response rates
- Tracing costs and lack of clustering of NHIS PSUs
may increase survey costs - Need for high completion rate in the NHIS of
locator data (name, SSN, contact persons) - Impact on field data collection
22National Health Nutrition Examination Survey
(NHANES)
- Collection of health and nutrition data via
personal household interviews and physical
examinations - Periodic survey up until 1994
- Continuous annual survey starting in 1999
- Annual estimates for broad subdomains
- Linkage with other surveys (NHIS and CSFII)
23National Health Nutrition Examination Survey
(NHANES)
- Possible linkage at
- PSU level
- HH (address) level
- Person level
- Survey content level
24Advantages of NHIS NHANESLinked Design
- Common sampling frame
- Reduced cost of screening
- Enhanced analytic capabilities and modeling
- Correlation of self-reported disease incidence
and examination results - Enhanced nonresponse adjustment
25Limitations of NHIS NHANES Linked Design
- Survey operation constraints
- Standardized environment, equipment, and data
collection through Mobile Exam Centers - Logistics of moving/setting up exam center
26Limitations of NHIS NHANES Linked Design
- Fixed number of PSUs per year (about 15)
- Larger PSUs (single counties) needed in NHANES
than NHIS - Fixed sample size per PSU (about 450 examined
persons)
27Limitations of NHIS NHANES Linked Design
- Fixed sample size requirements by age, sex, and
race/ethnicity - Tracking and tracing (approximately 17 of
population moves every year) - Linkage at sample or address level not possible
without supplemental independent area sample
28Medical Expenditure Panel Survey (MEPS)
- Type, frequency of use, and cost of health
services - How health services paid for
- Cost, scope, and breadth of health insurance
coverage
291996 MEPS Sample Design
- 1996 MEPS sample linked to 1995 NHIS subsample
(2nd and 3rd quarter from 2 panels) - 1995 NHIS subsample nationally representative
- Minority oversample
- Fixed precision requirements
301996 MEPS Sample Design
- 195 PSUs
- 1,675 area segments
- 10,597 NHIS dwelling units selected for the MEPS
311996 MEPS Round 1 Response Rates
- 1995 NHIS response rate (93.9)
- 99.6 of NHIS eligible units fielded
- MEPS Round 1 response rate (83.1)
- Overall response rate of 77.7
32MEPS Design Improvements
- Savings achieved through elimination of
independent sample and screening - Analytical capacity enhanced
- Data linkage to NHIS
- Longitudinal 2-year panel
- Annual expenditure survey
- Efficiencies through continuous operation
- Improved nonresponse adjustments
33Limiting Features of the MEPS Linkage to NHIS
- Lower response rates relative to NMES (1987)
- Greater restrictions in access to data related to
confidentiality of data - Lower cost savings in non-peak years with limited
oversampling of specific population subgroups
34Limiting Features of the MEPS Linkage to NHIS
- Need to track movers and re-enumerate NHIS
households to capture changes in HH composition - Dispersion of NHIS PSUs constrains flexibility of
survey operations
35Other Survey Design Integration Strategies
- Use of NHIS questions in other surveys
- Integration of NHIS for state-level data
- Use of NHIS data to adjust for noncoverage of
nontelephone HHs in a large RDD survey - Dual frame NHIS/RDD design
36National Immunization Survey
- Large ongoing random-digit-dialing (RDD) survey
(initiated in 1994) - Provide vaccination coverage data at state and
local level (50 states and 28 urban areas) - Target population children 19-35 mo. (lt 5 of
HHs in U.S.)
37NIS Data Collection
- Over 2 million telephone numbers sampled per year
- 1.6 million numbers dialed per year after
pre-screening for business and non-working
numbers - About 929,000 with working numbers
- 900,000 screened for children 19-35 months
- 36,000 eligible HHs interviewed
38NIS and NHIS Integration
- Same questions on immunizations
- Immunization information on both phone and
non-phone HHs in the NHIS - Use of ratio of immunization status in phone and
non-phone HHs from the NHIS used to adjust for
non-telephone coverage in the NIS
39Expansion of the NIS
- Use of large random sample of telephone numbers
- Collect more broad based health and welfare data
for adults and children - New survey State and Local Area Integrated
Telephone Survey (SLAITS)
40Key Features of SLAITS
- Consistent with DHHS survey integration plans
-
- Decreased survey costs
- Use of questions from NHIS and other national
surveys - Use of NHIS data to adjust for nontelephone
coverage - Comparisons across states and with national data
- Complements other state and local surveys
41Summary
- Survey integration - goal for HHS surveys
- Coordination of survey efforts
- Fill gaps in data
- Enhanced analyses
- Efficiencies in sampling, data collection,
questionnaire design, and survey operations - Progress on implementing MEPS-NHIS linkage
42Summary
- Progress on implementing NHANES on annual basis
and linked to NHIS - Progress on capacity for State level data
- Ongoing evaluations and research on
methodological issues for the integration of
health surveys
43Methodological Research Integrated Health Surveys
- Redesign NHIS with linkage as major goal
(optimize NHIS design for family of surveys
linked to it) - Reduce percent of cases with missing recontact
information - Reduce time lag between surveys to reduce tracing
and improve response rates
44Methodological Research Integrated Health Surveys
- Continued research on capacity for State-level
estimates - Assess risk of new sources of survey error
- Combined effect of area frame coverage and
nonresponse in NHIS - Recontact of NHIS respondents
45(No Transcript)