Oversampling in the NSCH is an option to enable reporting for specific population groups, conditions, or localities. Oversampling increases the households sampled in order to increase the number of completed surveys.
When there are more data, patterns within specific populations can be examined [e.g., looking at specific racial and ethnic groups within the more general population of children and youth with special health care needs (CYSHCN)]. By allowing more focused assessment, oversampling extends the utility of the survey to meet state and local needs, helping them to plan and direct investments for greatest impact.
We offer two types of oversampling:
The cost of oversampling depends on the type, size, and number of years to achieve sample size objectives. Since 2020, actual annual costs have ranged from $20,000 to $500,000 per oversample.
Federal Title V MCH Block Grant allocations can fund oversamples prior to state disbursement. The state Title V Director must make a request through their Project Officer to use this mechanism. Other state and private funds can also be used by establishing a contractual agreement directly with the U.S. Census Bureau.
A state must finalize their plans by July before the data collection year (e.g., July of 2023 for a 2024 oversample). This allows time for routing and signatures.
The U.S. Census Bureau releases NSCH data in the October following the oversample year (e.g., 2021 data were released in October 2022).
Our grantee, The Center for Adolescent Health and Measurement Initiative, provides an interactive web query system to view pre-calculated data tables, charts, and maps for a variety of indicators and characteristics. Public use files are also available for download for those wishing to conduct their own analysis; however, the U.S. Census Bureau requires an approved project proposal from a Research Data Center to access sub-state geographic identifiers. Research Data Center access may take up to a year to establish and may involve additional fees. To protect confidentiality, any sub-state estimates produced from restricted access files must be approved by the Census Bureau’s Disclosure Review Board prior to public release.
Since 2020, 15 states and one metro area have sponsored oversamples that have included a range of both single and multi-year efforts. In alignment with the our Maternal and Child Health Bureau strategic plan (PDF - 590 KB) focus on equity, the most common objective included in nine oversample plans is to improve estimates for specific racial and ethnic groups.
In 2020, Colorado, Nebraska, Oregon and, Wisconsin started sponsoring an oversample.
In 2021, Ohio, Louisiana, and Atlanta, Georgia started sponsoring an oversample.
In 2022, California, New York, Pennsylvania, Tennessee, and Wyoming started sponsoring an oversample.
In 2023, Illinois, Kansas, Minnesota, and New Mexico started sponsoring an oversample.
| State | Type | Objective | 2020 | 2021 | 2022 | 2023 |
|---|---|---|---|---|---|---|
| California | Local Area + Race/Ethnicity | To enable county estimates (150+ interviews in each of 41 counties) and regional reporting by race/ethnicity and CYSHCN using 3-year data. Learn more about California’s current effort to use oversampling | X | X | ||
| Colorado | Local Area | To enable regional estimates (150+ interviews per year in each of eight regions) | X | X | X | X |
| Georgia (Atlanta) | Local Area | To achieve 1,000 total interviews in the Atlanta metro area for reliable anxiety and depression estimates by age | X | X | X | |
| Illinois | Race/Ethnicity | To achieve 300 Black interviews for subgroup analyses using 2-year data | X | |||
| Kansas | General | To achieve 4,000 total interviews and ~1,000 CYSHCN for subgroup analyses using 2-year data | X | |||
| Louisiana | General + Race/Ethnicity | To achieve ~1,500 total interviews for fully reliable Title V performance measures and 300 Hispanic interviews for reportable Title V performance measures using 2-year data | X | X | ||
| Minnesota | General + Race/Ethnicity | To achieve 500 CYSHCN interviews for subgroup analyses and boost Black and American Indian/Alaska Native interviews for reportable estimates | X | |||
| Nebraska | General | To achieve ~1,500 total interviews for fully reliable Title V performance measures using 2-year data | X | X | X | X |
| New Mexico | Race/Ethnicity | To increase American Indian/Alaska Native interviews (150+) for reliable overall estimates using 2-year data | X | |||
| New York | Local Area + Race/Ethnicity | To enable regional estimates for 12 counties or county groups (~300 interviews each) and subgroup analyses for Black and Hispanic populations (~2,000 extra interviews) | X | |||
| Ohio | Race/Ethnicity | To achieve 200+ Black and Hispanic interviews for reliable estimates of CYSHCN and other key indicators using 2-year data | X | X | X | |
| Oregon | Race/Ethnicity | To enable reporting for all racial/ethnic groups (30+ interviews for each group) using 2-year data | X | X | X | |
| Pennsylvania | General | To improve overall precision through a general oversample achieving 1,500+ interviews per year | X | X | ||
| Tennessee | Race/Ethnicity | To achieve 300+ Black and 250+ Hispanic interviews for subgroup analyses using 2-year data | X | X | ||
| Wisconsin | General | To enable CYSHCN subgroup analyses (500+ interviews) using 2-year data | X | X | X | |
| Wyoming | General | To improve overall precision, particularly for CYSHCN, through an oversample achieving 600 additional interviews | X | X |
States who oversampled significantly increased their sample size and achieved their objectives.
Wisconsin sponsored a general oversample to enable CYSHCN subgroup analyses and more than doubled their 2020 and 2021 sample size with over 800 CYSHCN respondents. Oversampling provided more precise estimates of Title V National Performance Measures (NPMs), particularly for CYSHCN, and helped to reveal a statistically significant improvement in the percent of CYSHCN who have received services to prepare for the transition to adult health care (NPM-12) compared to the 2016 baseline (34.6% in 2020-2021 versus 21.1% in 2016). Without oversampling, this estimate would have been considered unreliable with wide confidence intervals and would not have been significantly different. Wisconsin is developing a CYSHCN surveillance report, the first of its kind to describe their CYSHCN population in terms of health status, access to care, and disparities.
Oregon sponsored a targeted oversample to enable reporting for all racial/ethnic groups. Prior to oversampling, child health estimates were only reportable (sample size of 30+) for non-Hispanic White, Hispanic, and non-Hispanic Asian populations. After oversampling, estimates are now reportable for all racial/ethnic groups (a first for non-Hispanic Black, American Indian/Alaska Native, and Native Hawaiian/Other Pacific Islander populations) and sufficient to support subgroup analyses for Hispanic and non-Hispanic Asian populations. Their large oversample also quintupled the size of their CYSHCN sample (~1,400 in 2020-2021) and their CYSHCN program is now working on an analysis of CYSHCN national priorities and social determinants of health by race/ethnicity.
Children’s Healthcare of Atlanta began sponsoring an annual oversample to produce reliable estimates of anxiety and depression by child age for the Atlanta metro area in 2021. As designed, over 1,000 interviews in the Atlanta metro were completed in 2021. These data will be used to support surveillance, healthcare planning, and treatment efforts.
State Oversampling in the National Survey of Children’s Health: Feasibility, Cost, and FAQs (PDF - 275 KB)
Ashley Hirai, PhD, Senior Scientist
Office of Epidemiology and Research