Download:
pdf |
pdfAppendix F
2019 National Survey of Children’s Health
Methodology Report
September 15th, 2020
2019 National Survey of Children’s Health
Methodology Report
The U.S. Census Bureau reviewed this data product for unauthorized disclosure of confidential information and
approved the disclosure avoidance practices applied to this release. CBDRB‐FY20‐POP001‐0166
Contents
Abstract ......................................................................................................................................................... 5
Objectives.................................................................................................................................................. 5
Methods .................................................................................................................................................... 5
Results ....................................................................................................................................................... 5
Introduction .................................................................................................................................................. 6
Survey History ............................................................................................................................................... 7
Challenges faced by NSCH/NS-CSHCN and Subsequent Redesign ............................................................ 7
Frame, Sample, and Selected Child Subsample .......................................................................................... 10
Frame and Sample Selection ................................................................................................................... 10
Treatment Groups ................................................................................................................................... 11
Selected Child Subsample ....................................................................................................................... 12
Data Collection ............................................................................................................................................ 14
Survey Content ........................................................................................................................................ 14
2019 Content Changes ............................................................................................................................ 15
Data Collection Instruments ................................................................................................................... 16
Mailout Content and Schedule ............................................................................................................... 20
Response Analysis ....................................................................................................................................... 25
Response Rates ....................................................................................................................................... 25
Item-Level Response ............................................................................................................................... 27
Treatment Groups and Response ........................................................................................................... 29
Data Processing ........................................................................................................................................... 32
Unduplication .......................................................................................................................................... 32
Paper to Web Standardization ................................................................................................................ 33
Data Edits ................................................................................................................................................ 33
Recoded and Standardized Variables ..................................................................................................... 35
Missing Values and Imputation............................................................................................................... 39
Suppressed Variables .............................................................................................................................. 41
Geography Variables ............................................................................................................................... 42
Weighting Specifications ............................................................................................................................. 45
Overview ................................................................................................................................................. 45
Population Controls ................................................................................................................................ 48
Limitations............................................................................................................................................... 49
Estimation, Hypothesis Testing, and Data Use Guidelines ......................................................................... 50
Variance Estimation ................................................................................................................................ 50
Combining Data across Survey Years ...................................................................................................... 51
Confidentiality ......................................................................................................................................... 51
Guidelines for Data Use .......................................................................................................................... 51
Supporting Material .................................................................................................................................... 53
References .............................................................................................................................................. 53
2
2019 National Survey of Children’s Health
U.S. Census Bureau
Attachment A: Estimated State-Level Production Sample Sizes for the 2019 National Survey of Children’s
Health .......................................................................................................................................................... 54
Attachment B: Traditional and Redesigned Envelope Designs ................................................................... 56
Attachment C: Determined Household Type from the 2019 National Survey of Children’s Health .......... 59
Attachment D: Weighted Response Rates by State .................................................................................... 61
3
2019 National Survey of Children’s Health
U.S. Census Bureau
Table of Figures
Table 1. Web Submission Times (in minutes) ............................................................................................. 17
Table 2. TQA Purpose Codes used in ATAC System .................................................................................... 19
Table 3. Production Mailout Schedule ........................................................................................................ 21
Table 4. Screener Test Card Mailout Schedule and Content ...................................................................... 24
Table 5. 2019 Final Dispositions (Unweighted)........................................................................................... 25
Table 6. 2019 NSCH Weighted Response Rates .......................................................................................... 27
Table 7. 35 Lowest Item Response Rates.................................................................................................... 28
Table 8. Average Cost per Completed Questionnaires and Percent of Eligible Households that Completed
Questionnaires by Screener Cash Incentive Group .................................................................................... 30
Table 9. Average Cost per Completed Topical Questionnaires and Completion Rate by Topical Cash
Incentive Group .......................................................................................................................................... 30
Table 10. Screener and Topical Response Probability (Incentive versus No Incentive) by Education, Race,
and Poverty Status ...................................................................................................................................... 31
Table 11. Unduplication Criteria for both Web and Paper Returns............................................................ 32
Table 12. Unduplication Criteria for Two Paper Returns ............................................................................ 32
Table 13. List of Standardized Variables ..................................................................................................... 35
Table 14. List of Derived and Recoded Variables ........................................................................................ 35
Table 15. List of Imputed Variables and their Imputation Flags ................................................................. 40
Table 16. List of Suppressed Variables ........................................................................................................ 41
Table 17. List of Geography Variables......................................................................................................... 43
Table 18. Geographies Identified at the Intersections ............................................................................... 43
4
2019 National Survey of Children’s Health
U.S. Census Bureau
Abstract
Objectives
This report details the development, plan, and operation of the 2019 National Survey of Children’s
Health (NSCH). This survey is designed to provide national and state-level estimates on key indicators of
the health and well-being of children, their families and their communities, as well as information about
the prevalence and impact of special health care needs. Funding and direction for this survey was
provided by the Health Resources and Services Administration’s Maternal and Child Health Bureau
(HRSA MCHB) within the U.S. Department of Health and Human Services. The U.S. Census Bureau
conducted the survey on behalf of HRSA MCHB.
Methods
The 2019 NSCH used a national sample of 184,000 addresses. During data collection, a screener
questionnaire was used to identify households with children and roster children in the household. The
screener questionnaire also included a battery of questions to identify children with special health care
needs. One child was randomly selected from each eligible household, and that child was the subject of
a more detailed topical questionnaire. Responses to the screener and topical questionnaires were
collected, processed, and published in the Screener Public Use File and Topical Public Use File.
Results
The weighted Overall Response Rate for the 2019 NSCH was 42.4%. A total of 68,500 1 screener
questionnaires were completed, and of those 35,760 were eligible for topical questionnaire follow-up.
Of those topical-eligible households, 29,433 completed a topical interview. Weighted estimates from
the Topical file generalize to state and national resident child populations. Weighted estimates from the
Screener file generalize to state and national resident child populations (using the child weight) and
households with children by state and nationally (using the household weight).
1
Rounded to the nearest five hundred in accordance with the U.S. Census Bureau disclosure avoidance practices.
2019 National Survey of Children’s Health
U.S. Census Bureau
5
Introduction
The 2019 National Survey of Children’s Health (NSCH) was conducted by the U.S. Census Bureau for the
Health Resources and Services Administration, Maternal and Child Health Bureau (HRSA MCHB) within
the U.S. Department of Health and Human Services (HHS). As stated in the Office of Management and
Budget Clearance Package, the purpose of the NSCH is to “collect information on factors related to the
well-being of children, including access to and quality of health care, family interactions, parental health,
school and after-school experiences, and neighborhood characteristics.” 2 This document details the
objectives, methodologies, and results of the 2019 NSCH into seven sections.
•
Survey History. The 2019 NSCH is the fourth annual production following the redesign and
merging of the previous NSCH and National Survey of Children with Special Health Care Needs
(NS-CSHCN).
•
Frame, Sample and Selected Child Subsample. A screener questionnaire identified households
with children and rostered the children in those households. A topical questionnaire collected
detailed information about one child selected at random from the household.
•
Data Collection. Data were collected using a two-stage paper survey instrument and a singlestage web-based survey instrument. This section discusses the mail schedule and data capture
methods for web, paper, and telephone questionnaire assistance operations.
•
Response Analysis. This section discusses the calculation of response rates along with analyses
of survey breakoffs, item nonresponse, and treatment group comparisons.
•
Data Processing. Web and paper survey responses were cleaned for analysis, including
unduplication of responses, edits for data quality, creating standardized and derived variables,
and imputation of missing values.
•
Weighting Specifications. Weights allow for estimates to be generalized to state and national
child resident populations (Screener and Topical file) and households with children (Screener
file).
•
Estimation, Hypothesis Testing, and Data Use Guidelines. A discussion for the best practices for
data users and limitations of the 2019 NSCH.
The Office of Management and Budget Clearance Package is available at
https://www.reginfo.gov/public/do/PRAViewDocument?ref_nbr=201903-0607-004
2
2019 National Survey of Children’s Health
6
U.S. Census Bureau
Survey History
The Health Resources and Services Administration’s Maternal and Child Health Bureau (HRSA MCHB),
within the U.S. Department of Health and Human Services (HHS), has sponsored the National Survey of
Children’s Health (NSCH) 3 and its companion survey, the National Survey of Children with Special Health
Care Needs (NS-CSHCN), 4 since 2001. HRSA MCHB has provided funding and direction for the two
periodic surveys in order to provide both national and state estimates of key indicators of child health
and well-being for children ages 0-17 years.
Together, these surveys provided critical data on key measures of child health; the presence and impact
of special health care needs; health care access, utilization, and quality; and the family and community
factors that impact child and adolescent health and well-being. Both surveys were fielded three times
(NS-CSHCN 2001, 2005-06, and 2009-10; NSCH 2003, 2007, and 2011-12) as modules of the State and
Local Area Integrated Telephone Survey (SLAITS) system by the Centers for Disease Control and
Prevention’s National Center for Health Statistics. As part of the SLAITS system, the surveys utilized a
random-digit-dial sample of landline telephone numbers, with cell-phone supplementation in the last
year of administration for both surveys.
While the geographic representation, sample size, and content breadth remained significant strengths
of the surveys, over time HRSA MCHB and its stakeholders came to realize that a redesign of the two
surveys was warranted. Declining response rates, along with the declining proportion of households in
the U.S. with landline telephones, led to the decision to change the underlying sampling frame from
telephone numbers to household addresses. Efforts were made to moderate this trend through the
addition of a cell-phone frame to the last administrations of both the NSCH and the NS-CSHCN.
However, consistent with industry-wide challenges, the inclusion of cell-phone samples proved to be
both costly and inefficient.
In 2015, HRSA MCHB redesigned the NSCH and the NS-CSHCN into a single combined survey that utilized
an address-based sampling frame. This newly consolidated survey incorporated questions from both of
the former surveys and retained the NSCH name. The U.S. Census Bureau now conducts the NSCH
annually on behalf of HRSA MCHB and HHS under Title 13, United States Code, Section 8(b), which
allows the Census Bureau to conduct surveys on behalf of other agencies.
Challenges faced by NSCH/NS-CSHCN and Subsequent Redesign
The telephone interview methodology utilized for the former NSCH and NS-CSHCN allowed for a
complex questionnaire as it ensured that skip patterns were properly followed. Furthermore, it
Blumberg SJ, Foster EB, Frasier AM, et al. 2012. Design and Operation of the National Survey of Children’s Health,
2007. National Center for Health Statistics. Vital Health Stat, 1(55).
http://www.cdc.gov/nchs/data/series/sr_01/sr01_055.pdf
4
Bramlett MD, Blumberg SJ, Ormson AE, et al. 2014. Design and Operation of the National Survey of Children with
Special Health Care Needs, 2009–2010. National Center for Health Statistics. Vital Health Stat, 1(57).
http://www.cdc.gov/nchs/data/series/sr_01/sr01_057.pdf
3
7
2019 National Survey of Children’s Health
U.S. Census Bureau
protected against data entry error through preprogrammed range and logic checks on responses.
Interviewers were able to address respondent questions and concerns as they arose, helping reduce
response error. However, in recent years declining willingness of the public to participate in surveys and
changes in household telephone use resulted in declining response rates for Computer-Assisted
Telephone Interviewing surveys. 5 Of particular concern was the increasing prevalence of households
substituting wireless service for their landline telephone. Efforts to include these non-landline
households within the telephone sampling frames for the former NSCH and NS-CSHCN through the
addition of cell-phones to the frame were ultimately not cost efficient or effective. Furthermore,
because the former NSCH and NS-CSHCN were administered using the Centers for Disease Control and
Prevention’s National Immunization Surveys (NIS) sampling frame and followed behind the NIS
interview, they experienced additional impacts in response rates when cases failed to move through the
NIS itself.
The surveys were no longer sustainable in the face of declining response rates and rising costs.
Therefore, considerable work was done to determine how to address these issues, and the decision was
reached to utilize a two-phase multimode data collection design for a combined NSCH/NS-CSHCN
survey, henceforth known as the NSCH. The proposed approach to data collection and nonresponse
follow-up was based on previous project experience and recommendations made by Dillman and
colleagues (2009). 6
The redesigned NSCH consists of two questionnaires: (1) an initial household screener to assess the
presence of children in the home and facilitate the selection of a target child within the household (with
oversampling of children with special health care needs and young children ages 0-5 years), and (2) a
substantive topical questionnaire that combines selected content from the former NSCH and NS-CSHCN
questionnaires along with new content to address emerging public health topics.
Revisions to existing items were generally made for the following reasons: 1) a desire for consistency
with federal policies or programs and harmonization of content across U.S. Department of Health and
Human Services surveys (e.g., the item on physical activity was edited to reflect the new Dietary
Guidelines for Americans); 2) changes in the field or the understanding of a topic or issue (e.g., with
direction and support from co-sponsors, content on attention deficit/hyperactivity disorder treatment
was expanded to include separate items on behavioral and medication treatment); and 3) selfadministered surveys require wording and framing that differs from interviewer-assisted surveys (i.e.,
instructional text throughout the instrument was refined and simplified).
Concomitantly, the addition (or deletion) of content was driven by four factors: 1) the need to include
the most critical content from both former surveys; 2) the prioritization of topics highly relevant to HRSA
MCHB investments (e.g., items required to track National Performance and Outcome Measures for the
Blumberg SJ, Luke JV. 2010. Wireless Substitution: Early Release of Estimates from the National Health Interview
Survey, January–June 2010. National Center for Health Statistics. Available from:
http://www.cdc.gov/nchs/nhis.htm
6
Dillman DA, Smyth JD, Christian LM. 2009. Internet, Mail and Mixed-Mode Surveys: The Tailored Design Method,
3rd edition. Hoboken, NJ: John Wiley & Sons.
5
8
2019 National Survey of Children’s Health
U.S. Census Bureau
Title V Maternal and Child Health Services Block Grant program); 3) the commitment to improve
methods for assessing key topics; and 4) the desire to address emergent priorities as identified by states
and the broader maternal and child health field (e.g., the addition of items to assess readiness to learn
among children aged 3-5 years).
9
2019 National Survey of Children’s Health
U.S. Census Bureau
Frame, Sample, and Selected Child Subsample
This section covers frame development, sample selection, experimental treatment group assignments
within the sample, and the selected child subsample process.
Frame and Sample Selection
The 2019 NSCH uses an address-based sample selected from an extract of the Census Bureau’s Master
Address File (MAF) 7. It covers the 50 states and the District of Columbia 8. The sample frame uses
administrative records-based flags to identify three mutually exclusive strata:
•
•
•
Stratum 1: Addresses that are explicitly linked to children using administrative records.
Approximately 80% of these addresses are households with children.
Stratum 2a: Addresses that are probabilistically linked to children using administrative records
and block group characteristics. Approximately 15% of these addresses are households with
children.
Stratum 2b: The remaining addresses. Less than 3% of these addresses are households with
children.
Addresses assigned to Stratum 1 are explicitly linked to a child record either directly or through a parent
using administrative records. Administrative record sources include:
•
•
•
•
•
•
•
•
•
•
Master Address Auxiliary Reference File (MAF-ARF) (for household addresses)
2010 Census unedited file
Social Security Data
the IRS 1040, and 1099 files
the Medicare Enrollment Database (MEDB)
Indian Health Service database (IHS), Selective Service System (SSS)
Public Indian Housing (PIC)
American Community Survey and CPS-ASEC data (for parent-child links)
Tenant Rental Assistance Certification System (TRACS) data from the Department of Housing and
Urban development (HUD)
National Change of Address data from the US Postal Service
For 2019, the sample added Department of Housing and Urban Development (HUD) data to supplement
Stratum 1. Administrative HUD PPIC and TRACS data identified additional children in public housing and
voucher households.
In 2019, there were 40 million unique addresses linked to households with children.
The MAF is a Title 13 data source, and all data collected are confidential under 13 U.S.C. Section 9. All access to
Title 13 data from this survey is restricted to Census Bureau employees and those holding Census Bureau Special
Sworn Status pursuant to 13 U.S.C. Section 23(c).
8
Hereafter, ‘state’ will include the District of Columbia.
7
2019 National Survey of Children’s Health
U.S. Census Bureau
10
Among the remaining addresses, a linear probability model was developed against American Community
Survey (ACS) returns to predict child presence using block group7 characteristics and administrative
records associated with the address (e.g., presence of adults 20-50 years old and child-related tax
deductions). Addresses were sorted on the probability of child presence by state. The size of Stratum 2b
in any state was constrained so that it represents no more than 5% of households with children in that
state.
The sample for the 2019 NSCH includes approximately 184,000 unique addresses. The sample includes
addresses from Strata 1 and 2a; Stratum 2b addresses were excluded from sampling. For the sample
selection:
•
•
•
•
The addresses within the state were first sorted by strata, then organized into two groups by the
block group 9 poverty rate to ensure states had proportional representation of addresses in high
poverty areas selected for the sample.
The sampling rate by strata in each state is optimized to maximize the number of completed
interviews per address selected (selecting more addresses from Stratum 1) while constraining
the impact of the design effect on variance. Nationally, 60% of addresses came from Stratum 1
and 40% from Stratum 2a.
The sample was distributed across states to produce a roughly equal number of completed
interviews per state (see Attachment A).
To minimize respondent burden, addresses can be included in the sample only once in any five
year period.
Treatment Groups
The survey sample of 184,000 was divided into treatment and control groups for various experiments.
The 2019 NSCH groups are:
•
•
•
•
•
Screener cash incentives: $0 (control), $2, or $5 screener incentive
High Web/High Paper: Web push invitation suite, or web and paper invitation suite
Envelope design: Traditional (control) or redesigned envelope invitation suite
Screener card test: Traditional (control) or screener card invitation suite
Topical cash incentives: $0 (control), $2, or $5 topical incentive
For the screener cash incentive, 90% of the sample received a small denomination bill with the initial
invitation as an incentive to complete the survey, half receiving a $2 bill and the other half a $5 bill. The
other 10% of the sample did not receive an incentive and represented the control group for monitoring
the effectiveness of the incentive treatments.
A Census block group is a geographical unit with 600 to 3,000 population. Census blocks are grouped into block
groups; block groups, in turn, are grouped into Census tracts. The block group is the smallest scale geographical
unit for which the Census Bureau publishes sample statistics, i.e., estimates based on a sample of residents in the
block group. Consequently, it is the smallest scale geographical unit that could be used for this exercise.
9
11
2019 National Survey of Children’s Health
U.S. Census Bureau
The High Paper treatment group was composed of the 30% of the addresses identified as having the
highest probability of responding by paper and not by web. These addresses received both a paper
screener questionnaire and an invitation for the web screener questionnaire with the initial invitation.
The remaining 70% of addresses were assigned to the High Web group. These addresses received only a
web invitation in the first two mailings, but received a paper questionnaire in the final two mailings.
More information about the mailout schedule is included in the Data Collection section.
For the envelope design test, 50% of addresses were mailed the initial invitation in a redesigned
envelope and the other 50% received the invitation in the traditional design from 2018. The redesigned
envelope added colors and graphics to the traditional envelope design. See Attachment B for
information copies.
Finally, 4,000 addresses were assigned to test a new response mechanism, the screener card. Screener
card test case responses were not included in the production data files.
Selected Child Subsample
The screener questionnaire collects information on the presence of children within the household, child
demographic information, and basic questions about each child’s health. 10 One child is selected from
the completed screener, and one of the three age-based topical questionnaires is provided to the
household based on the sampled child’s age:
•
•
•
NSCH-T1: children aged 0 through 5,
NSCH-T2: children aged 6 through 11, or
NSCH-T3: children aged 12 through 17
The probability of selection for a child is based on the number of children in the household, the special
health care needs status, and the age of the child. When appropriate, an 80% oversample is applied to
children with special health care needs and a 60% oversample to young children (ages 0-5). 11 See
attachment C for more details. 12
Upon completion of the screener questionnaire, Web respondents are immediately brought to the
appropriate topical questionnaire web page. For mailed-in screener responses, the appropriate topical
questionnaire is mailed to the household, and mail materials indicate which child has been selected.
Bethell CD, Read D, Neff J, Blumberg SJ, Stein RE, Sharp V, Newacheck PW. 2002. “Comparison of the Children
with Special Health Care Needs Screener to the Questionnaire for Identifying Children with Chronic Conditions—
Revised.” Ambulatory Pediatrics, Jan-Feb 2(1): 49-57.
11
The 80% oversample is applicable only for those households with both CSHCN and Non-CSHCN present. The 60%
age-based oversample is applicable when the conditions of the CSHCN oversample are not met and there are both
young (ages 0-5) and older (ages 6-17) children present.
12
Eligible children in a household are sorted first by special health care needs status (CSHCN then Non-CSHCN) and
then by age (youngest to oldest). Additionally, children with the same special health care needs status and age are
sorted by name. In households with four or more eligible children, children are sorted first on special health care
needs status, then alphabetically by name, and then by age.
10
12
2019 National Survey of Children’s Health
U.S. Census Bureau
Additionally, incentives are used in the first mailing to each topical group. 90% of these households
receive a $5 incentive while 10% do not receive an incentive.
13
2019 National Survey of Children’s Health
U.S. Census Bureau
Data Collection
Data collection efforts for the 2019 National Survey of Children’s Health (NSCH) began on June 28, 2019
and continued until January 17, 2020. The 2019 NSCH retained a two-phase data collection approach:
(1) an initial household screener to assess the presence, basic demographic characteristics, and special
health care needs status of any children in the home; and (2) a substantive topical questionnaire to be
completed by a parent or caregiver of the selected child. The data collection methodology follows
strategies to increase response rates, including offering multiple ways to respond, treatments and nonresponse follow-up.
This section covers survey content and 2019 content changes, data collection instruments, and the data
collection process.
Survey Content
Questionnaires were designed to encourage cooperation by prospective respondents, enhance
respondent comprehension, and make instructions clear and simple. Questions were developed and
grouped by subject area to create logical, clear questionnaires with concrete question wording and
simple grammar.
The screener questionnaire consisted of two sections. The first section contained four questions about
the presence of children in the home, the primary language spoken, and home tenure (rent or own). The
next section contained detailed questions about the demographics and health of up to four children,
from youngest to oldest. If there were more than four children in a household, the first name (or initials
or nickname), age, and sex were asked for up to ten children.
There were three different topical questionnaires tailored to three age groups of the selected children:
NSCH-T1 for 0 to 5 year old children, NSCH-T2 for 6 to 11 year old children, and NSCH-T3 for 12 to 17
year old children. All three questionnaires contained 11 sections about the child, their family, and
neighborhood, but the specific questions were tailored to be relevant to children in that age specific
range. Copies of the screener and topical questionnaires can be found at
https://www.census.gov/programs-surveys/nsch/technical-documentation/questionnaires.html.
Section A. This Child’s Health
Questions about whether the child has current or lifelong physical, mental, behavioral, learning, or
developmental conditions. Additional questions on whether the child’s health conditions affect their
ability to do things.
Section B: This Child as an Infant
Birth-related questions including birth weight, breastfeeding, and use of formula. Infant feeding
questions are only included on NSCH-T1.
Section C: Health Care Services
Questions about source of a usual place for health care, need for and use of medical, dental, mental,
and specialized health services in the last 12 months.
14
2019 National Survey of Children’s Health
U.S. Census Bureau
Section D: Experience with This Child’s Health Care Providers
Questions about frequency of care and satisfaction with the child’s health care providers. Also,
questions about how the child’s doctor or health care provider worked with the child. NSCH-T3 includes
questions about the child’s preparation for transition into adult health care.
Section E: This Child’s Health Insurance Coverage
Questions about whether the child has adequate health care insurance coverage, and whether there
were any gaps in health care insurance coverage in the past 12 months, including at the time of the
survey.
Section F: Providing for this Child’s Health
Questions on cost of health care in the past 12 months and time spent providing and arranging for the
child’s health care.
Section G: This Child’s Learning/Schooling and Activities
Questions on early language development and learning for children ages 1 to 5 years. For children ages 6
to 17 years, questions about experiences at school, participation in organized activities, and physical
activities.
Section H: About You and This Child
Questions about daily life and household activities, including the child’s sleep habits, screen time, and
the demands of parenting/caregiving on the respondent.
Section I: About Your Family and Household
Questions about the frequency of family meals, the use of tobacco in the household, how the family
copes with problems, and if any assistance is needed to provide food for the family. Also questions
about the respondent’s perception of their neighborhood (e.g., amenities, safety), and questions about
whether the child has ever experienced any adverse childhood experiences.
Section J: Child’s Caregivers
Questions on demographic information about up to two adults in the household who are the child’s
primary caregivers.
Section K: Household Information
Questions on household count, family count, and family income.
2019 Content Changes
Three variables were added in 2019 NSCH questionnaires and reported on the public use data
files:
•
•
•
BIRTH_YR (“What year was this child born?”)
BIRTH_MO (“What month was this child born?”)
BIRTH_YR_F (a data quality flag for BIRTH_YR and BIRTH_MO)
The BIRTH_YR and BIRTH_MO variables are in the topical data file and captured in NSCH-T1,
NSCH-T2, and NSCH-T3. The BIRTH_YR_F variable is a flag in the topical data file to indicate if
15
2019 National Survey of Children’s Health
U.S. Census Bureau
the topical questionnaire birth month and year is consistent with the screener questionnaire
birth month and year.
Additionally, the response option “Some other race” was removed from SC_RACE_R and C_RACE_R in
2019 to follow the standards of the Office of Management and Budget (OMB). 13 Additional differences
between the 2016, 2017, 2018, and 2019 NSCH questionnaires are noted in the NSCH crosswalk:
https://census.gov/programs-surveys/nsch/technical-documentation/codebooks.html
Data Collection Instruments
The data collection design focuses on efforts to increase response rates. Respondents have multiple
ways to respond to the survey and receive assistance including:
• Web Instrument (English and Spanish)
• Paper Instrument (English and Spanish)
• Telephone Questionnaire Assistance (TQA) (available in several languages)
• Email Questionnaire Assistance (EQA)
• Spanish Language Translation
Web Instrument
The web survey was programmed using the U.S. Census Bureau’s Centurion system for internet data
collection. This software presented the questionnaire on a computer screen or other electronic device,
e.g., tablet or cell phone. The interview was self-administered by the respondent. The respondent
logged in to the web survey by accessing the URL provided on the mailed invitation and entering their
unique 8-digit login ID.
Respondents were asked to verify their address. If the respondent answered that the address selected
for the sample (and displayed on screen) did not match their own, the survey was concluded and the
address was removed from further mailings.
If the listed address matched the respondent’s residence, the case was assigned a PIN that the
respondent would need to log back in to the survey. Alternatively, the respondent could create a new
PIN by correctly answering a security question, which the respondent previously provided during the
original PIN creation process.
After setting up the PIN, the respondent reported the number of children (0-17 years of age) that usually
reside at that address. If there were no children that usually reside at the address, the survey was
concluded and the address removed from further mailings. If there were children that usually reside at
the address, the respondent was then directed through the rest of screener questionnaire.
There were two hard edits programmed into the web instrument which required respondents to provide
a valid answer before continuing. These answers were necessary for subsampling: child’s first name,
13
The NSCH survey uses the race categories defined in the 1997 OMB Standards for the Classification of Federal
Data on Race and Ethnicity.
2019 National Survey of Children’s Health
U.S. Census Bureau
16
initials, or nickname; and age. Respondents were able to skip all other questions and continue the
survey.
After the respondent completed the screener questionnaire, the web instrument applied the
subsampling methodology to select one child from each household to be the subject of the topical
questionnaire. At this point in the survey process, content from the screener portion of the
questionnaire was locked.
The name and sex of the selected child was then prefilled into the topical survey questions, and the web
instrument guided respondents through skip patterns. Some response fields only accepted responses
that represented legitimate values; other fields offered a “pick list” of response categories. There were
soft edits for some questions that prompted respondents to provide an answer or revise an existing
answer, but respondents were able to continue past these edits without changing their answers.
Respondents could review and edit any answers before submitting. Once the survey was submitted, a
submission confirmation screen appeared with the date and time of completion. The instrument was
then locked and the respondent was only able to view the submission confirmation screen if they logged
back in. Submitted responses were saved in the output data file.
Respondents from households without children completed the web instrument in an average of 1
minute, 26 seconds. Respondents from households with children completed the screener portion of the
instrument in 5 minutes, 32 seconds; the web topical portion in 31 minutes, 22 seconds; and the entire
web instrument in 36 minutes, 54 seconds, on average. Online help screens and text were also available
in the instrument to aid respondents.
Table 1. Web Submission Times (in minutes)
With Children
No Children
Mean
Median
Mean
Median
Screener
5.5
4.3
1.4
0.8
Topical
31.4
26.7
Total
36.9
31.7
1.4
0.8
Paper Instrument
The second mode of data collection was a two-phase, self-administered mail survey using paper
questionnaires. The paper questionnaires were created using Amgraf One Form Plus. They were printed,
trimmed, and stitched through an in-house print on-demand process using a Docuprint system that
allowed personalization to each respondent.
In the first phase of this mode of data collection, paper screener questionnaires were mailed to High
Paper addresses with the initial invitation, and to all other addresses (High Web) with the second nonresponse follow-up mailing. 14 Respondents completed a screener questionnaire to determine if there
More information on the High Paper/High Web group assignments is covered in the Mailout Content and
Schedule section.
14
2019 National Survey of Children’s Health
U.S. Census Bureau
17
were any children 17 years of age or younger who usually lived or stayed at the address. Resident
children were rostered in the screener instrument. Detailed information was collected for up to four
children, while basic information (name, age, sex) was collected for an additional six children.
If the respondent mailed back the screener, it was then processed to determine if eligible children
usually reside at the address. Returned forms were processed by iCADE to capture responses through
OMR (optical mark recognition), OCR (optical character recognition), and KFI (keying from image). If the
respondent answered that the address selected for the sample did not match their own or that there
were no children that usually reside at the address, the survey was concluded and the household was
removed from further mailings. If the respondent listed children that usually reside at the address,
Census Bureau staff applied the subsampling methodology to select one child from the household roster
to be the subject of the topical questionnaire.
In the second phase, households that reported eligible children were mailed one of the three age-based
topical questionnaires requesting more information about one selected child living at the address. In
order to ensure respondents answered the topical questions for the selected child, Docuprint systems
printed the selected child’s first name, initials, or nickname, age, and sex provided on the screener
questionnaire onto the topical questionnaires and survey invitation letters.
The paper and web instruments were designed to be as similar as possible to minimize the influence of
mode on responses. While automatic skips and soft edits could not be implemented in the paper
instrument, the questionnaire did include skip instructions within the question wording to mimic the
web instrument.
Telephone Questionnaire Assistance (TQA)
The National Processing Center call center in Tucson, Arizona provided telephone questionnaire
assistance (TQA) for the 2019 NSCH. Respondents could call a toll-free telephone line if they had
questions about the survey, wanted to complete the interview over the phone using the web
instrument, or submit feedback. All mail content and the web instrument listed this toll-free number.
Interviewers were trained to use the Automated Tracking and Control (ATAC) system to report call-ins
using one of the following TQA purpose codes seen in Table 2.
18
2019 National Survey of Children’s Health
U.S. Census Bureau
Table 2. TQA Purpose Codes used in ATAC System
TQA Purpose Codes
Definitions
01
Internet questionnaire completed over the phone
02
Refusal to participate
03
Wrong address – paper respondent
04
Wrong address – web respondent
06
Paper questionnaire status
09
Out of scope (vacant, business, not a full-time residence)
12
Child listed on questionnaire moved or doesn’t live at this
residence most of the time
20
Questions about monetary incentive
21
Questions about prior mail never received
30
Request replacement survey (English)
31
Request Spanish language questionnaire
32
Trouble filling out the paper questionnaire
33
Child listed on questionnaire is deceased
49
Respondent requested PIN
50
Respondent requested Login ID
51
Problem logging into Internet instrument
52
Other instrument issues
53
PIN/security question reset request
54
Screener card test case only (0 children)
55
Screener card test case only (1 child)
56
Screener card test case only (2 children)
57
Screener card test case only (3 children)
58
Screener card test case only (4+ children)
60
Question regarding the survey (General FAQ)
80
Comments
Call monitoring sessions of recorded TQA calls were scheduled throughout data collection. If any
changes were needed to the ATAC TQA instrument based on comments received from interviewers, the
survey team coordinated programming updates. All updates to procedures were communicated to the
TQA interviewers. Incoming call volumes were also monitored throughout data collection and
interviewer schedules were adjusted accordingly.
In 2019, approximately 2,500 TQA cases were recorded in ATAC. The most common cases included
internet questionnaire completed over the phone (~1,400), comments (~500), and out of scope (~150).
19
2019 National Survey of Children’s Health
U.S. Census Bureau
Email Questionnaire Assistance (EQA)
In addition to the toll-free telephone line, respondents were able to interact with Census Bureau staff
via email. An email address ([email protected]) was listed on all invitation letters and on the
Centurion login page. Emails were answered by call center staff in Tucson, Arizona. Staff checked the
email inbox daily and replied to respondents’ messages within 2 business days when possible. Emails
were logged in a tracking spreadsheet and cases were assigned purpose codes similar to the TQA
purpose codes in Table 2.
EQA agents employed scripted responses for common concerns and questions. These scripts ensured
consistent and accurate information. When replying to the messages, agents removed any information
in the response email that could be considered personally identifiable (e.g., address, phone number,
name).
Spanish Language Translation
The NSCH paper and web instruments were available in both English and Spanish. The Census Bureau
reviewed and verified text from the 2018 Spanish-language questionnaires and provided new
translations where necessary for the 2019 questionnaires. Respondents were provided instructions to
request a Spanish language questionnaire by calling TQA. Spanish-speaking respondents that called the
TQA line were placed in a Spanish language calling queue; a trained Spanish language agent then
answered any questions or administered the Spanish language web instrument over the phone. The
agent flagged the case if a Spanish paper questionnaire was requested and informed the respondent
that a questionnaire would arrive in the mail within three weeks.
If a respondent returned a Spanish language paper screener questionnaire indicating the presence of
children in the household, the Spanish language topical questionnaire was subsequently mailed to the
household. The web instrument included a toggle on every page that allowed respondents to switch
between the English or Spanish language version of the instrument.
Mailout Content and Schedule
Respondent contact strategies and letters were carefully designed to capture the attention of the
respondent and pique interest in the subject matter. Cash incentives, follow-up mailings, reminder
postcards, toll-free telephone numbers, and translated questionnaires were used to maximize response.
Data collection for the 2019 NSCH involved a series of mailings and nonresponse follow-up activities,
emphasizing questionnaire completion. Mailouts began Friday June 28, 2019 and continued until the
survey closeout on Friday January 17, 2020. The approach to data collection and nonresponse follow-up
was based on previous project experience and recommendations made by Dillman and colleagues
(2009): 15
Dillman DA, Smyth JD, Christian LM. 2009. Internet, Mail and Mixed-Mode Surveys: The Tailored Design Method,
3rd edition. Hoboken, NJ: John Wiley & Sons.
15
20
2019 National Survey of Children’s Health
U.S. Census Bureau
•
•
Invitation letter. An initial invitation letter was mailed to all potential respondents providing
details about the study, a web URL with the login ID for accessing the web version of the
questionnaire (which combined the screener and topical into a consolidated instrument), and a
toll-free number and email address for individuals to contact if there were questions or
comments.
Additional mailings. Subsequent to the first invitation mailing, the Census Bureau sent all
remaining non-respondents additional invitations. Addresses also received reminder postcards
after the first two mailings.
The production mailing schedule for the 2019 NSCH is summarized below in Table 3.
Table 3. Production Mailout Schedule
Date
Friday, June 28, 2019
Friday, July 05, 2019
Friday, July 26, 2019
Friday, August 02, 2019
Event
Initial Screener Invitation
High Web - Reminder Postcard
High Paper - Reminder Postcard
Low Paper – 1st Follow-Up Invitation
High Paper - 1st Follow-Up Invitation
Low Paper - Reminder Postcard
Friday, August 9, 2019
High Paper - Reminder Postcard
Friday, August 16, 2019
1st Topical Mailing
Low Paper - 2nd Follow-Up Invitation
Friday, August 23, 2019
Topical Reminder Postcard – Group A
Friday, August 30, 2019
2nd Topical Mailing
High Paper - 2nd Follow-Up Invitation
Friday, September 06, 2019
Topical Reminder Postcard – Group B
Friday, September 13, 2019
3rd Topical Mailing
Friday, September 20, 2019
Topical Reminder Postcard – Group C
Friday, September 27, 2019
Friday, October 4, 2019
Friday, October 11, 2019
4th Topical Mailing
Low Paper - 3rd Follow-Up Invitation
Topical Reminder Postcard – Group D
High Paper - 3rd Follow-Up Invitation
5th Topical Mailing
21
2019 National Survey of Children’s Health
U.S. Census Bureau
Date
Event
Friday, October 18, 2019
Topical Reminder Postcard – Group E
Friday, October 25, 2019
6th Topical Mailing
Friday, November 01, 2019
Topical Reminder Postcard – Group F
Friday, November 08, 2019
7th Topical Mailing
Friday, November 15, 2019
Topical Reminder Postcard – Group G
Friday, November 22, 2019
8th Topical Mailing
Friday, November 29, 2019
Topical Reminder Postcard – Group H
Friday, December 06, 2019
9th Topical Mailing
Friday, December 13, 2019
Topical Reminder Postcard – Group I
Friday, January 17, 2020
Survey Closeout
Initial Screener Invitation
The initial mailing included the following treatment groups:
• High Paper (30% of addresses): A paper questionnaire was included with the initial invitation in
place of the standard web-push strategy
• Cash incentive (90% of addresses): 45% of addresses received a $2 bill, 45% received a $5 bill
• Envelope redesign (50% of addresses): The initial invitation was packaged in a redesigned
envelope with simple colors and graphics in place of the traditional envelope
• Screener card (4,000 addresses): In place of the traditional invitation letter, a small test group
was selected to receive the new screener card instrument
High Paper, cash incentive and envelope redesign assignment were independent of one another, so nonscreener card test cases received one of eight package types. The screener card test cases were assigned
to one of three treatments that varied the sequence of mailings. In total, there were eleven package
types used for the initial mailing.
Screener Reminder Postcards
Pressure-sealed reminder postcards were sent 5 to 7 days after the initial and first non-response followup mailing. The pressure-sealed reminder postcard included the necessary details for the respondent to
complete the survey by web.
22
2019 National Survey of Children’s Health
U.S. Census Bureau
Screener Non-response Follow-up Mailings
The screener data collection strategy included three non-response follow-up attempts 16. With the
second follow-up mailing, High Web addresses received their first paper screener. All nonresponding
addresses received paper screeners in this and all subsequent follow-up mailings.
Topical Questionnaire
Respondents that returned a complete paper screener with eligible children and did not submit a web
survey were assigned to one of nine topical mailing groups (A through I). Group assignments depended
on the date that the National Processing Center received the paper screener questionnaire and the next
mailing date of topical questionnaires.
The topical questionnaire and accompanying cover letter were personalized with the sample child’s
name and other identifying information to ensure that the survey was completed for the correct child. In
their first topical mailing, 90% of addresses received an unconditional $5 bill as a token of appreciation
for participating in the survey. Following on the success of the reminder postcards with the screener
mailings in previous survey administrations, the 2019 NSCH also mailed a reminder postcard one week
after the initial topical mailing. Household could receive of up four topical mailings. 17
Screener Card Test
Cases selected for the screener card treatment received invitations based on one of three mailout
schedules. These groups were designed to evaluate different data collection priorities: Group B
prioritized screener card response, Group D prioritized web response, and Group C represented a more
balanced approach. The schedules included initial mailings and follow-up content as shown in Table 4.
Addresses stopped receiving mailings if the residents submitted a web survey, returned a complete paper
screener, explicitly refused to participate, or if the address was out-of-scope (i.e., not an occupied residence). The
address also received fewer mailings if the USPS determined the address to be undeliverable as addressed.
16
Respondents received fewer packages if they returned a topical form or explicitly refused to participate, the
selected child no longer resided at the address when the topical form was received, or the household was assigned
to a later topical group.
17
23
2019 National Survey of Children’s Health
U.S. Census Bureau
Table 4. Screener Test Card Mailout Schedule and Content
Date
Friday June 28, 2019
Friday July 05, 2019
Friday July 26, 2019
Friday August 02, 2019
Event
Initial Mailings
1st Follow-up
2nd Follow-up
3rd Follow-up
Group Web Push letter with
B
screener card &
return envelope
Pressure sealed
reminder postcard
Web Push letter with Pressure sealed
screener card &
reminder postcard
return envelope
Group Web Push letter that
C
mentions a screener
card is forthcoming
Web Push letter with Web Push letter with Pressure sealed
screener card &
screener card &
reminder postcard
return envelope
return envelope
Group Web Push letter that
D
does not mention a
screener card is
forthcoming
Pressure sealed
reminder postcard
that mentions a
screener card is
coming
Web Push letter with Pressure sealed
screener card &
reminder postcard
return envelope
24
2019 National Survey of Children’s Health
U.S. Census Bureau
Response Analysis
Response Rates
Table 5 provides a summary of the survey completion counts. 68,500 18 households completed a
screener portion of the survey. Of those, 36,196 reported children and are included on the Screener
data file.
Complete and sufficient partial topical questionnaires are included on the Topical Public Use File. Of the
36,196 screened households with children, 29,433 returned a complete or sufficient partial topical
survey. In 2019, 79.2% of respondents completed the survey using the web instrument and 20.8% of
respondents completed the survey using the paper instruments.
Table 5. 2019 Final Dispositions (Unweighted)
Final Disposition
Count a
Total Cases
180,000a
Occupied Households (Estimated)
144,000 a
Households with Children (Estimated)
81,000 a
Completed Screeners
68,500 b
Screeners with Children
36,196a
Completed Topicals
29,433a
a
Rounded to the nearest thousand
b
Rounded to the nearest five hundred
For the purposes of calculating response rates, all sampled addresses were assigned screener and
topical outcomes codes. These outcomes can generally be categorized as not eligible, eligible but not
complete, or complete.
For some addresses, we did not receive sufficient correspondence to determine if the address was
eligible to complete the screener or topical questionnaires. These addresses were classified as
unresolved. Among these addresses, we estimated the share that were occupied residences using the
Household Rate, which is the proportion of resolved addresses that are occupied residences. 19 We also
estimated the Child Rate, which is the share of those households that include children, based on the
proportion of households that have children by state and stratum in the 2017 American Community
Survey (ACS). The product of the Household Rate and Child Rate is the Eligibility Rate (e), the estimated
proportion of unresolved addresses that are households with children. Using this approach, we
Rounded to the nearest five hundred based on the U.S. Census Bureau disclosure avoidance practices.
Specifically, we used the midpoint between the Household Rate including undeliverable addresses (the
proportion of all resolved addresses that are occupied residences) and the Household Rate excluding undeliverable
addresses (UAAs) by state and stratum. Because UAAs are identified by the U. S. Postal Service, it is assumed that
UAAs are identified at a higher rate than other noneligible addresses (businesses, vacant residences, etc.) that
must be self-identified. The midpoint assumes that there are some UAAs still unresolved but at a lower rate than
they appear among the resolved addresses.
18
19
25
2019 National Survey of Children’s Health
U.S. Census Bureau
estimated that 86% (weighted) of unresolved addresses were households and 41% (weighted) of those
households with children.
𝑒𝑒 = 𝐻𝐻𝐻𝐻𝐻𝐻𝐻𝐻𝐻𝐻ℎ𝑜𝑜𝑜𝑜𝑜𝑜 𝑅𝑅𝑅𝑅𝑅𝑅𝑅𝑅 ∗ 𝐶𝐶ℎ𝑖𝑖𝑖𝑖𝑖𝑖 𝑅𝑅𝑅𝑅𝑅𝑅𝑅𝑅
Three different response rates were calculated based on the estimated proportion of eligible addresses
that completed the screener and topical questionnaires. Definitions of completion and calculation of
these three response rates are detailed below.
Screener Completion Rate
The Screener Completion Rate (SCR) is the estimated proportion of households (occupied residences)
that completed a screener. A completed screener had to 1) be returned from a sampled address, and 2)
indicate that there were no children present or provide a valid age for at least one child. The
denominator includes both screened households and the number of unresolved addresses that are
estimated to be households.
𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶 𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆
𝑆𝑆𝑆𝑆𝑆𝑆 =
𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆 𝐻𝐻𝐻𝐻𝐻𝐻 + (𝑈𝑈𝑈𝑈𝑈𝑈𝑈𝑈𝑈𝑈𝑈𝑈𝑈𝑈𝑈𝑈𝑈𝑈𝑈𝑈 𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴 ∗ 𝐻𝐻𝐻𝐻𝐻𝐻𝐻𝐻𝐻𝐻ℎ𝑜𝑜𝑜𝑜𝑜𝑜 𝑅𝑅𝑅𝑅𝑅𝑅𝑅𝑅)
Topical Completion Rate
The Topical Completion Rate (TCR) is the estimated proportion of households with children that
submitted a topical questionnaire, either complete or sufficient partial. Completed topical
questionnaires have valid answers for at least 40 of 50 test questions and at least one item in
Section K (family income, household and family count), or the respondent submitted the topical
portion of the web instrument. Sufficient partial topical questionnaires have valid answers for at
least 25 of 50 test questions and at least one item in Section H or beyond, or the respondent
submitted the topical portion of the web instrument. The denominator includes both screened
households with children and the number of unresolved addresses that are estimated to be
households with children (Unresolved Addresses * e).
𝑇𝑇𝑇𝑇𝑇𝑇 =
𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶 𝑇𝑇𝑇𝑇𝑇𝑇𝑇𝑇𝑇𝑇𝑇𝑇𝑇𝑇𝑇𝑇
𝑆𝑆𝑆𝑆𝑆𝑆𝑒𝑒𝑒𝑒𝑒𝑒𝑒𝑒𝑒𝑒 𝐻𝐻𝐻𝐻𝐻𝐻 𝑤𝑤𝑤𝑤𝑤𝑤ℎ 𝐶𝐶ℎ𝑖𝑖𝑖𝑖𝑖𝑖𝑖𝑖𝑖𝑖𝑖𝑖 + (𝑈𝑈𝑈𝑈𝑈𝑈𝑈𝑈𝑈𝑈𝑈𝑈𝑈𝑈𝑈𝑈𝑈𝑈𝑈𝑈 𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴 ∗ 𝑒𝑒)
Interview Completion Rate and Overall Response Rate
The Interview Completion Rate (ICR) and Overall Response Rate (ORR) account for the multi-stage
design of the NSCH. They are the products of two (for ICR) or three (for ORR) response rate metrics that
are each consistent with the American Association for Public Opinion Research (AAPOR) standards 20.
The ICR is the probability a household progresses through the screener and topical portions of the
survey.
𝐼𝐼𝐼𝐼𝐼𝐼 = 𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆 𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶 𝑅𝑅𝑅𝑅𝑅𝑅𝑅𝑅 ∗ 𝑇𝑇𝑇𝑇𝑇𝑇𝑇𝑇𝑇𝑇𝑇𝑇𝑇𝑇 𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶 𝑅𝑅𝑅𝑅𝑅𝑅𝑅𝑅
The American Association for Public Opinion Research. 2016. Standard Definitions: Final Dispositions of Case
Codes and Outcome Rates for Surveys. 9th edition. AAPOR.
20
2019 National Survey of Children’s Health
U.S. Census Bureau
26
The Screener Conversation Rate is the proportion of resolved households that completed the screener.
𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆 𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶 𝑅𝑅𝑅𝑅𝑅𝑅𝑅𝑅 =
𝐶𝐶𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜 𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆
𝑅𝑅𝑅𝑅𝑅𝑅𝑅𝑅𝑅𝑅𝑅𝑅𝑅𝑅𝑅𝑅 𝐻𝐻𝐻𝐻𝐻𝐻𝐻𝐻𝐻𝐻ℎ𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜
The Topical Conversation Rate is the proportion of screened households with children that completed a
topical.
𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶 𝑇𝑇𝑜𝑜𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝
𝑇𝑇𝑇𝑇𝑇𝑇𝑇𝑇𝑇𝑇𝑇𝑇𝑇𝑇 𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶 𝑅𝑅𝑅𝑅𝑅𝑅𝑅𝑅 =
𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆 𝐻𝐻𝐻𝐻𝐻𝐻𝐻𝐻𝐻𝐻ℎ𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜 𝑤𝑤𝑤𝑤𝑤𝑤ℎ 𝐶𝐶ℎ𝑖𝑖𝑖𝑖𝑖𝑖𝑖𝑖𝑖𝑖𝑖𝑖
The Overall Response Rate (ORR) is the probability an address progresses from resolution to screener
complete to topical complete and is given by the equation below,
𝑂𝑂𝑂𝑂𝑂𝑂 = 𝑅𝑅𝑒𝑒𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠 𝑅𝑅𝑅𝑅𝑅𝑅𝑅𝑅 ∗ 𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆 𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶 𝑅𝑅𝑅𝑅𝑅𝑅𝑅𝑅 ∗ 𝑇𝑇𝑇𝑇𝑇𝑇𝑇𝑇𝑇𝑇𝑇𝑇𝑇𝑇 𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶 𝑅𝑅𝑅𝑅𝑅𝑅𝑅𝑅
where the Resolution Rate is the proportion of addresses in sample that were resolved as occupied
households. In 2019, the weighted Resolution Rate was 52.2%.
𝑅𝑅𝑅𝑅𝑅𝑅𝑅𝑅𝑅𝑅𝑅𝑅𝑅𝑅𝑅𝑅𝑅𝑅𝑅𝑅 𝑅𝑅𝑅𝑅𝑅𝑅𝑅𝑅 =
𝑅𝑅𝑅𝑅𝑅𝑅𝑅𝑅𝑅𝑅𝑅𝑅𝑅𝑅𝑅𝑅 𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴
𝑇𝑇𝑇𝑇𝑇𝑇𝑇𝑇𝑇𝑇 𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴
Table 6 lists the weighted rate for each of the four response metrics discussed above. A breakdown of
the response rates by state is provided in attachment D.
Table 6. 2019 NSCH Weighted Response Rates
Metric
Screener Completion Rate
Topical Completion Rate
Interview Completion Rate
Overall Response Rate
Rate
47.8%
35.3%
79.5%
42.4%
Item-Level Response
The item response rate is the proportion of item-eligible respondents that provided a valid answer to a
particular item. Many items were applicable to a subset of survey respondents only; for example, some
questions were applicable to children in a specific age range. In that case, the denominator for the item
response rate is the count of children in the eligible age range, and the numerator is the count of those
children with valid responses.
In some cases, it is uncertain if the child was eligible for an item due to nonresponse on a preceding
item. For example, before asking about the severity of certain conditions, we asked if the child currently
had the condition. The severity item was applicable if the child currently had the condition, and it was
not applicable if the child did not currently have the condition. If the respondent chose to skip the
current condition filter item, we cannot know definitively if the severity item was applicable or not.
27
2019 National Survey of Children’s Health
U.S. Census Bureau
We account for this situation in the item response rate by assigning eligibility to cases with unknown
eligibility equal to the proportion of cases that were eligible when eligibility was known. For example, if
10% of respondents reported that the child did have the condition currently, and so were eligible for the
severity follow-up question, the denominator for the severity item response rate becomes
# 𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸 + (# 𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸 𝑈𝑈𝑈𝑈𝑈𝑈𝑈𝑈𝑈𝑈𝑈𝑈𝑈𝑈 ∗ .1)
Across all survey items, more than 98% of response opportunities produced a valid response. Items that
require a write-in response, that require respondents to follow a skip pattern, and are near the end of
the survey tend to have higher nonresponse. Table 7 lists the 35 variables with the lowest item response
rates. The list predominantly reflects items that are at the end of a skip pattern and are on-path for few
respondents (e.g., DOWNSYN_DESC which captures the severity of a reported diagnosis of current Down
Syndrome), items that require a write-in response (e.g., A2_LIVEUSA which captures how long a second
caregiver has lived in the US), and items near the end of the survey (e.g., A2_PHYSHEALTH which
captures the reported health status a second caregiver). 21
Table 7. 35 Lowest Item Response Rates
Response
Rate
On-Path
(%)
Down Syndrome Severity Description
89.4%
0.2%
GENETIC_DESC
Genetic Condition Severity Description
90.5%
4.7%
GENETIC_SCREEN
Genetic Condition Newborn Screening
90.9%
4.7%
K2Q35A_1_YEARS
Autism ASD - First Told Age in Years
91.8%
3.0%
SUBABUSE_CURR
Substance Use Disorder Currently
92.7%
0.2%
A2_LIVEUSA
Adult 2 - Come to Live in the United States (Year)
93.5%
12.6%
SLEEPPOS
Position Most Often Lay Your Baby Down to Sleep
94.4%
3.2%
K2Q38B
Tourette Syndrome Currently
95.6%
0.3%
BIRTHWT
Birth Weight Status
95.7%
100.0%
BIRTHWT_L
Birth Weight is Low (<2500g)
95.7%
100.0%
BIRTHWT_VL
Birth Weight is Very Low (<1500g)
95.7%
100.0%
BIRTHWT_OZ_S
Standardized Birth Weight, Ounces
95.7%
100.0%
K2Q35D
Autism ASD - First Told Doctor Type
95.8%
3.0%
A2_DEPLSTAT
Adult 2 - Deployment Status
95.8%
5.9%
A1_LIVEUSA
Adult 1 - Come to Live in the United States (Year)
95.8%
12.9%
ARTHRITIS_CURR
Arthritis Currently
96.0%
0.4%
CERPALS_DESC
Cerebral Palsy Severity Description
96.1%
0.3%
A2_BORN
Adult 2 - Where Born
96.1%
87.0%
Variable
Description
DOWNSYN_DESC
21
This table does not include the six poverty status implicates (FPL1-FPL6). Values are derived from several survey items, and
partial responses are used to inform the multiple imputation. For comparison, 17.6% of respondents provide incomplete or
inconsistent responses to those survey items used to derive FPL.
28
2019 National Survey of Children’s Health
U.S. Census Bureau
A2_PHYSHEALTH
Adult 2 - Physical Health
96.1%
87.0%
K5Q22
Arrange Or Coordinate As Much Help As Wanted
96.1%
4.6%
K12Q01_G
Reason Not Covered - Other
96.2%
6.0%
K12Q01_F
Reason Not Covered - Application/Renewal Problems
96.2%
6.0%
K12Q01_E
Reason Not Covered - Inadequate Providers
96.2%
6.0%
K12Q01_D
Reason Not Covered - Inadequate Benefits
96.2%
6.0%
K12Q01_C
Reason Not Covered - Unaffordable
96.2%
6.0%
K12Q01_B
Reason Not Covered - Cancellation Overdue Premiums
96.2%
6.0%
K12Q01_A
Reason Not Covered - Change in Employer/Employment
96.2%
6.0%
A2_MENTHEALTH
Adult 2 - Mental or Emotional Health
96.2%
87.0%
LIVEUSA_MO
How Long Living in the United States - Months
96.2%
3.1%
LIVEUSA_YR
How Long Living in the United States - Years
96.2%
3.1%
A2_K11Q50_R
Adult 2 - Employed 50 Out Of Last 52 Weeks
96.2%
87.0%
A2_MARITAL
Adult 2 - Marital Status
96.2%
87.0%
A2_GRADE
Adult 2 - Highest Completed Year of School
96.2%
87.0%
A2_ACTIVE
Adult 2 - Active Duty
96.3%
87.0%
A2_AGE
Adult 2 - Age in Years
96.4%
87.0%
Treatment Groups and Response
This section reviews response patterns based on the treatment group assignments:
•
•
•
•
Screener cash incentive
Topical cash incentive
Envelope redesign
High Web vs. High Paper
Screener Cash Incentive
In the 2019 NSCH, sampled addresses received either a $2 bill, a $5 bill, or they were part of the control
group that did not receive a cash incentive in the initial screener mailing. The treatment groups
represented 90% of the sample, with 45% of addresses receiving a $2 bill and 45% receiving a $5 bill.
The remaining 10% of addresses made up the control group and received no incentive.
The screener cash incentives results in Table 8 show the average cost per completed screener
questionnaires, the average cost per completed topical questionnaires, the percent of eligible
households who completed a screener questionnaire, and the percent of eligible households who
completed a topical questionnaire. The results in Table 8 show that providing an unconditional screener
incentive in the initial mailing was an effective strategy for encouraging response. Eligible households
that received a $5 incentive were more likely to complete the Screener and Topical questionnaires than
households that received a $2 incentive or no incentive. Eligible households that received a $2 incentive
29
2019 National Survey of Children’s Health
U.S. Census Bureau
were more likely to complete the Screener questionnaire and Topical questionnaire than households
that received no incentive.
Table 8. Average Cost per Completed Questionnaires and Percent of Eligible Households that Completed
Questionnaires by Screener Cash Incentive Group
Screener Cash
Incentive
Group
Total
No Incentive
$2 Incentive
$5 Incentive
Average Cost
per Completed
Screener
$21.70
$16.34
$19.43
$24.86
Percent of Eligible
Households that
Completed Screeners
46.9%
41.2%
46.3%
48.9%
Average Cost
per Completed
Topical
$56.78
$45.67
$52.26
$62.91
Percent of Eligible
Households that
Completed Topicals
36.5%
31.0%
35.6%
38.7%
Topical Cash Incentive
The 2019 NSCH also used a $5 cash incentive in the initial topical mailing. Approximately 10% of cases
were assigned to the control group (no incentive), with the remaining cases receiving the $5 incentive.
Table 9. Average Cost per Completed Topical Questionnaires and Completion Rate by Topical Cash
Incentive Group
Incentive
Average Cost per
Group
Completed Topical
Completion Rate*
Total
$46.23
58.2%
No Incentive
$49.21
47.0%
$5 Incentive
$45.97
59.5%
*Percent of households that were mailed a paper topical invitation that
subsequently completed a topical interview.
Households that received a $5 incentive were more likely to complete the topical questionnaire than
households that received no incentive (see Table 9). Also, by reducing the number of non-response
follow-up mailings, the topical incentive reduced the cost of data collection versus the control group.
Cash incentives were also relatively more effective among groups that were otherwise less likely to
respond. For example, Table 10 shows the incentive effect on the probability of screener and topical
response by education, race and poverty status. The cash incentive increased screener and topical
response for all groups. But the effect was larger for less educated households, Black households
(compared to White households), and households in poverty; the difference between these groups is
statistically significant in all cases except for Black households compared to White households receiving
the $2 screener incentive. Because less educated, Black and poor households are generally less likely to
respond to the survey, incentives may reduce nonresponse bias.
30
2019 National Survey of Children’s Health
U.S. Census Bureau
Table 10. Screener and Topical Response Probability (Incentive versus No Incentive) by Education, Race,
and Poverty Status
Screener
Topical
P($2 incentive) /
P(control)
113.5% *
111.4% *
+2.1% †
P($5 incentive) /
P(control)
120.3% *
117.1% *
+3.2% †
P($5 incentive) /
P(control)
130.7% *
126.2% *
+4.6% †
Race
Black alone
White alone
Black vs. White
115.9% *
112.2% *
+3.7%
124.8% *
118.1% *
+6.8% †
149.3% *
126.6% *
+22.7% †
Poverty Status
Poverty
Not Poverty
Poverty vs. Not
115.7% *
112.3% *
+3.4% †
122.7% *
118.4% *
+4.2% †
139.2% *
127.6% *
+11.6% †
Education
High School or less
College or more
HS vs. College
* H0: P(incentive)/P(control) <= 1, p<0.05; † H0: Difference <= 0, p<0.05
High Paper/High Web
Prior to mailing, all NSCH addresses were assigned a score based on the probability the address would
respond by mail and would not respond by web. The 30% of addresses with the highest scores were
assigned to the High Paper group. High Paper addresses were mailed a paper screener questionnaire
with all mailed invitations. The other 70% of addresses were assigned to the High Web group. High Web
addresses received their first paper screener in the second nonresponse follow-up.
Screener response was about 26% higher for the High Paper group. The mode of response correlated
with our expectations: the High Paper group was more likely to respond by paper but less likely to
respond by web compared to the High Web group.
The additional screener response from the High Paper group came at a cost, approximately $4.17 per
case. The majority of this cost came from the paper topical follow-up effort to the High Paper addresses
that returned a paper screener. High Web addresses were more likely to complete the web screener and
topical in a single session. Also, households with children using the paper questionnaire were less likely
to complete the topical questionnaire than those that used the web questionnaire. Consequently, the
High Paper strategy was an efficient approach for collecting screener responses but less effective for
collecting topical interviews.
Envelope Redesign
The envelope redesign did not significantly impact response probabilities. Addresses assigned to the
envelope redesign group were not more or less likely to complete the screener or topical.
31
2019 National Survey of Children’s Health
U.S. Census Bureau
Data Processing
Data were processed and edited to ensure data quality and respondent confidentiality.
Unduplication
All nonresponding households were offered two modes, web and paper, for completing the survey. In
some cases, respondents utilized both options. In these cases, we selected one response, web or paper,
to include in the data file. We chose the response to include based on the type of return and the level of
completeness. Completed web returns were always chosen over completed paper returns. However,
completed paper returns were chosen over partial web survey returns. The web/paper unduplication
hierarchy is detailed in Table 11.
Table 11. Unduplication Criteria for both Web and Paper Returns
Order Chosen
Type of Return
1
Completed web survey - Household with children
2
Completed paper screener and topical
3
Completed web survey - Household w/o children
4
Completed paper screener - Household w/o children
5
Partially completed web survey
6
Out of scope paper return
7
Refusal paper return, Hard Refusal
8
Incomplete, Duplicate
9
Blank, Soft Refusal
10
Deceased
11
Undeliverable address (UAA) with address correction –
mail forwarded, UAA with address correction
12
UAAs, Forwarding Order Expired, Moved out of U.S.
13
Default
14
Blank form
Multiple follow-up mailings included paper questionnaires, so it was also possible that respondents
received and returned more than one questionnaire. In these cases, one return was selected to
represent that case in the data file. The paper/paper unduplication hierarchy is detailed in Table 12.
Table 12. Unduplication Criteria for Two Paper Returns
Order Chosen
Type of Return
1
Completed paper screener/topical - Household with
children
2
Completed paper screener - Household w/o children
32
2019 National Survey of Children’s Health
U.S. Census Bureau
Order Chosen
Type of Return
3
Out of scope paper return
4
Refusal paper return, Hard Refusal
5
Incomplete, Duplicate
6
Blank, Soft Refusal
7
Deceased
8
UAA with address correction – mail forwarded, UAA with
address correction
9
UAAs, Forwarding Order Expired, Moved out of U.S.
10
Default
11
Blank form
Paper to Web Standardization
Responses were standardized across web and paper so they could be appended in a single data file.
Although the majority of the survey questions had the same valid values for the paper and web
instruments, sometimes the values output for the paper questionnaire did not match the output from
the web survey instrument. For instance, any questions that included a list of checkboxes where the
respondent was instructed to “mark (X) ONE box” differed between paper and web. The difference in
output between the two modes was due to the fact that the web had the ability to prevent the selection
of more than one checkbox via a radio button, whereas a paper respondent could mark more than one
box even if the question explicitly said not to. Since all data from the paper instruments is captured for
processing, each of the response option checkboxes have their own associated output variable.
Therefore, prior to appending web and paper responses into a single data file, paper responses were
reformatted to the proper valid values.
Data Edits
The 2019 NSCH raw output was processed to manage inconsistent and invalid responses in nine
sequential steps:
•
Stop Process Edit. A case is removed from the data file if the case fails address verification (the
respondent indicates that their address does not match the address on file), the respondent
indicates that there are no children in the household, or the respondent does not complete a
screener for a household with children. The cases are not eligible to be included on a NSCH data
file, so are removed from processing.
•
Not in Universe Edit. An item is not in universe if it is not included in the instrument the
respondent received. Some items are unique to web or paper, and others are specific to a
version of the topical instrument, T1, T2, or T3. The value for an item that is not in universe is
set to ‘.N’.
33
2019 National Survey of Children’s Health
U.S. Census Bureau
•
Range Edit. If a value falls outside the bounds of a defined minimum and maximum for that
item, the value is replaced with an indicator that the response is missing. The minimum and
maximum are selected to represent a reasonable range of possible responses to the item.
•
Backfill Edit. The backfill edit imputes values to some items based on responses to subsequent
items that necessarily indicate the correct response to the edited item. Backfill edits apply
almost exclusively to paper questionnaires, which cannot prevent a respondent from skipping a
root item but answering follow-up questions. For example, INCWAGES is a binary item that
filters respondents on whether the family did (INCWAGES=1) or did not (INCWAGES=2) receive
wage or salary income. If a respondent does not answer INCWAGES, but provides a valid and
non-zero value for INCWAGES_AMT, the dollar amount of wage and salary income, then it is
necessarily correct that INCWAGES=1.
•
Yes/No Edit. The NSCH includes several series that ask respondents to select all applicable items
from a list. These series may or may not allow the respondent to answer in the negative,
indicating that the item is not applicable. In most cases, if a respondent answers in the
affirmative (=1) to at least one item in the series, it is assumed that all other items in the series
do not apply (=2) unless otherwise noted. If a respondent is only able to respond in the
affirmative, and the items in the series are not comprehensive (e.g., they do not include an
“Other” option), then it is assumed that all unanswered items do not apply (=2) without
imposing the requirement that at least one item is answered in the affirmative.
•
Consistency Edit. If responses to two items in the survey are fundamentally inconsistent, one
response is maintained and the other is removed and changed to missing. Most consistency
edits require that a child does not experience a life event at an age greater than her current age.
Because the instrument generally trends from more general, fundamental information to more
specific, priority is given to the item that appears first in the instrument.
•
Legitimate Skip Edit. Unlike the ‘Not in Universe Edit’, the legitimate skip edit applies to items
that are on the respondent’s instrument, but not on path. The value for an item that is in
universe but not on path is set to ‘.L’.
•
Missing in Error Edit. If an item is in universe (does not equal .N), is on path (does not equal .L),
but does not hold a valid value, that item is missing in error, identified as ‘.M’.
•
Disclosure Edit. Some survey responses, if published, could compromise a respondent’s
confidentiality. Disclosure edits involve removing entire items (e.g., child’s name) or suppressing
rare or unique values (e.g., top codes on the family poverty ratio). Census disclosure avoidance
standards make reference to weighted and unweighted cell counts (i.e., the number of children
with a characteristic or set of characteristics), the size of the underlying population (e.g., the
number of children in Kentucky Metropolitan Statistical Areas), and the existence of outside
34
2019 National Survey of Children’s Health
U.S. Census Bureau
data sources that could be matched to the NSCH (e.g., a registry of children diagnosed with
Cerebral Palsy).
Edits were applied in two stages. In the first stage, edits for screener items were applied to completed
screeners with children. When these edits were completed, cases that did not return a completed
topical were removed from edits, and the second stage edits to topical items were applied.
Recoded and Standardized Variables
Standardized Variables
Several questions in the 2019 NSCH allowed respondents to provide an answer using more than one unit
(e.g., years and months) and to choose from two systems of units (e.g., imperial or metric). In these
cases, we provide standardized variables that convert responses across units and systems to a single
unit. See Table 13 for a list and description of these variables.
Table 13. List of Standardized Variables
Variable
Description
Units
BIRTHWT_OZ_S
Child birth weight
Ounces
BREASTFEDEND_DAY_S
Stopped breastfeeding
Days
BREASTFEDEND_WK_S
Stopped breastfeeding
Weeks
BREASTFEDEND_MO_S
Stopped breastfeeding
Months
FRSTFORMULA_DAY_S
First fed formula
Days
FRSTFORMULA_WK_S
First fed formula
Weeks
FRSTFORMULA_MO_S
First fed formula
Months
FRSTSOLIDS_DAY_S
First fed solids
Days
FRSTSOLIDS_WK_S
First fed solids
Weeks
FRSTSOLIDS_MO_S
First fed solids
Months
Derived and Recoded Variables
A number of variables on the public use data files are derived from a set of items on the survey or a
recoded version of a single item. These variables are listed in Table 14.
Table 14. List of Derived and Recoded Variables
Variable
Description
Derived from
AGEPOS4
Birth position of the selected child relative to
other children in household
C_AGE_YEARS
C_AGE_MONTHS
TOTMALE
Count of male children in household
C_SEX
TOTFEMALE
Count of female children in household
C_SEX
C_CSHCN
Special Health Care Needs (SHCN) status
C_K2Q10 - C_K2Q23
35
2019 National Survey of Children’s Health
U.S. Census Bureau
Variable
Description
Derived from
SC_CSHCN
SHCN status of selected child
C_CSHCN
TOTCSHCN
Count of children with SHCN
CSHCN
TOTNONSHCN
Count of children that do not have SHCN
C_K2Q10 - C_K2Q23
TOTAGE_0_5
Count of children 0 to 5 years old in household
C_AGE_YEARS
TOTAGE_6_11
Count of children 6 to 11 years old in household
C_AGE_YEARS
TOTAGE_12_17
Count of children 12 to 17 years old in household
C_AGE_YEARS
SC_AGE_LT4
Age of selected child (less than 4 months)
SC_AGE_YEARS
SC_AGE_MONTHS
SC_AGE_LT6
Age of selected child (less than 6 months)
SC_AGE_YEARS
SC_AGE_MONTHS
SC_AGE_LT9
Age of selected child (less than 9 months)
SC_AGE_YEARS
SC_AGE_MONTHS
SC_AGE_LT10
Age of selected child (less than 10 months)
SC_AGE_YEARS
SC_AGE_MONTHS
C_RACER
Race of child
C_RACE_R
C_RACEASIA
Asian race category is included for the following
states: CA, HI, MA, MD, MN, NJ, NV, NY, VA, WA
C_RACE_R
C_RACEAIAN
American Indian/Alaska Native race category is
C_RACE_R
included for the following states: AK, AZ, NM, MT,
ND, OK, SD
C_HISPANIC_R
Hispanic origin
C_HISPANIC
SC_RACER
Race of selected child
SC_RACE_R
SC_RACEASIA
Asian race category is included for the following
states: CA, HI, MA, MD, MN, NJ, NV, NY, VA, WA
(Selected Child)
SC_RACE_R
SC_RACEAIAN
American Indian/Alaska Native race category is
SC_RACE_R
included for the following states: AK, AZ, NM, MT,
ND, OK, SD (Selected Child)
SC_HISPANIC_R
Hispanic origin of selected child
SC_HISPANIC
HOUSE_GEN
Parental nativity
BORNUSA
A1_RELATION
A1_BORN
A2_RELATION
A2_BORN
36
2019 National Survey of Children’s Health
U.S. Census Bureau
Variable
Description
Derived from
FAMILY_R
Family structure
A1_RELATION
A2_RELATION
A1_MARITAL
A2_MARITAL
A1_SEX
A2_SEX
CURRINS
Current health insurance coverage status
K3Q04_R
CURRCOV
K12Q03, K12Q04,
K12Q12, TRICARE,
HCCOVOTH, K11Q03R
INSTYPE
Type of insurance
CURRINS
K12Q03, K12Q04,
K12Q12, TRICARE,
HCCOVOTH, K11Q03R
INSGAP
Health insurance coverage over the past 12
months
K3Q04_R, CURRINS
FPL_I1-FPL_I6
Family poverty ratio
FAMCOUNT
TOTINCOME
HIGRADE
Highest level of education for reported adults
(three categories)
A1_GRADE
A2_GRADE
HIGRADE_TVIS
Highest level of education for reported adults
(four categories)
A1_GRADE
A2_GRADE
BIRTHWT
Birth weight status
BIRTHWT_OZ_S
BIRTHWT_L
Low birth weight (<2500g)
BIRTHWT_OZ_S
BIRTHWT_VL
Very low birth weight (<1500g)
BIRTHWT_OZ_S
BMICLASS
Body Mass Index
WEIGHT_*
HEIGHT_*
Specifications of Select Derived Variables
The 2019 NSCH reports several derived variables that include information on the child’s family status.
This includes Family Poverty Ratio (FPL), Household Nativity (HOUSE_GEN), and Family Structure
(FAMILY_R).
•
Family Poverty Ratio (FPL) - The family poverty ratio is calculated as the ratio of total family
income to the family poverty threshold, and reported as a rounded percentage. Respondents
reported total family income in item K4 on the paper instrument: “The following question is
about your 2018 income. Think about your total combined family income IN THE LAST
CALENDAR YEAR for all members of the family. What is that amount before taxes?” Additional
text instructed respondents to include all money incomes, for example, social security,
dividends, and child support. Responses to K4 were edited for consistency against answers in K3,
37
2019 National Survey of Children’s Health
U.S. Census Bureau
a series of questions about specific sources of income. Finally, missing or invalid responses were
replaced with multiply imputed values.
The family poverty threshold is derived from the Census Bureau’s poverty thresholds.
Thresholds vary by family size and the number of related children under 18 years old. They do
not vary across geographies. Family size was reported in K2 of the paper instrument. Missing or
invalid values were imputed. The number of related children was determined by the number of
children reported in the screener.
To protect the confidentiality of respondents, only FPL is reported in the Public Use File; total
family income and the family poverty threshold are not included. Further, FPL is top and bottom
coded. Reported values range from 50 (total family income is 50% of the family poverty
threshold) to 400 (total family income is 400% of the family poverty threshold). Values beyond
this range are reported as 50 or 400, respectively.
•
Household Nativity (HOUSE_GEN) - Household nativity is determined by the birth location of the
child (BORNUSA) and parents (A1_BORN and A2_BORN). If the child was born outside of the U.S.
and all reported parents were born outside of the U.S., the household is reported as a 1st
generation household. Second generation households have members born both inside and
outside of the U.S. For example, the child was born in the U.S. and at least one parent was born
outside of the U.S., or the child was born outside of the U.S. and one of two parents was born in
the U.S. Finally, in 3rd+ generation households, all parents were born in the U.S. The fourth
category, “Other”, captures households with insufficient information about the nativity of the
parents.
•
Family Structure (FAMILY_R) - A family structure variable uses the reported information on the
child’s primary caregivers to organize households into common types. Notably, the NSCH
collects information on only two adults in the household and requires only that the two adults
be primary caregivers of the child. As a result, in multigenerational households, this can mean
that a biological, adoptive, or step parent is not reported.
Further, respondents do not report their relationship to other adult members of the household,
only to the child; consequently, we may know that the two reported adults are married, but we
do not know if they are married to each other. Instead of making assumptions about the
relationship of the reported adults with each other, the family structure variable depends only
on the number of adults, their relationship to the child, and their individual marital statuses. For
example, a reported value of 1 for FAMILY means that the two reported adults are
biological/adoptive parents of the child and they are currently married; one may assume that
they are married to each other, but in some cases that will not be true.
Two family structure categories (FAMILY_R=5 and 6) are also defined by the sex of the
respondent. In these cases, it is specified that the responding caregiver is female (5) or male (6)
and that no other parents (biological, adoptive, or step) are in the household.
38
2019 National Survey of Children’s Health
U.S. Census Bureau
The 2019 NSCH reports several variables that include information on the child’s health insurance status
and insurance type. We strongly recommend that data users interested in current health insurance
status and insurance type use the derived variables CURRINS (Currently Insured), INSGAP (Gaps in
Coverage), and INSTYPE (Insurance Type) in their analyses.
•
Currently Covered (CURRINS) - CURRINS is derived primarily from the respondent-reported
values in K3Q04_R (Health Insurance Coverage – Past 12 Months) and CURRCOV (Health
Insurance Coverage – Currently Covered). We indicate that the child is currently insured
(CURRINS=1) if the respondent reported that the child had coverage for all of the last 12 months
(K3Q04_R=1) or reported that the child is currently covered (CURRCOV=1), but with an
important caveat. If the respondent reported that the child is currently insured but reported
only Indian Health Service or health care sharing ministry as the type of coverage, we indicate
that the child does not have current insurance coverage (CURRINS=2). Consequently, a
respondent may report that a child is insured, but we consider that the child is not insured.
•
Gaps in Coverage (INSGAP) - INSGAP is derived primarily from the respondent reported values in
K3Q04_R (Health Insurance Coverage – Past 12 Months) and CURRCOV (Health Insurance
Coverage – Currently Covered). We indicate that the child had consistent coverage (INSGAP=1) if
the respondent reported that the child had coverage for all of the last 12 months (K3Q04_R=1)
but with an important caveat. If the respondent reported that the child is currently insured but
reported only Indian Health Service or health care sharing ministry as the type of coverage, we
indicate that information as to the consistency of the child’s coverage is missing (INSGAP=.M).
•
Insurance Type (INSTYPE) - INSTYPE is derived from CURRINS (Currently Insured) and respondent
answers to questions on the coverage type: K12Q03 (Current/Former Employer or Union),
K12Q04 (Directly Purchased), K12Q12 (Government Assistance Plan), TRICARE (TRICARE or other
military health care), K11Q03 (Indian Health Service), and HCCOVOTH_WRITEIN (Other Type,
Write-in). Any insurance reported as coming from an employer or union, directly purchased,
TRICARE or other military health care, or the Affordable Care Act is considered private. Coverage
from any government assistance plan is considered public. Both the private and public coverage
categories reflect a single reported source of coverage; a combined category for children with
both public and private coverage is also included.
In addition, Health Insurance write-in (HCCOVOTH_WRITEIN) responses were back-coded to flag public
and private insurance types, religious health care sharing ministry, and Indian Health Service coverage.
These flags were used in the derivation of CURRINS and INSTYPE. To protect respondent confidentiality,
answers to HCCOVOTH_WRITEIN are not reported in the Public Use File.
Missing Values and Imputation
For most variables in the public data files, missing values are coded to identify the type of missing data.
These include
39
2019 National Survey of Children’s Health
U.S. Census Bureau
•
•
•
•
(.L) Legitimate Skip – The item is not applicable to the respondent, as determined by a previous
answer to a root question.
(.M) Missing in Error – The value is missing due to respondent or system errors, or the
respondent did not provide a valid answer.
(.N) Not in Universe – The item was not included on the respondent’s age-appropriate version of
the topical questionnaire.
(.D) Suppressed for Confidentiality – The value is suppressed in order to protect respondent
confidentiality.
However, variables use for raking during weighting procedures require imputation. Table 15 lists the
2019 variables that are imputed and includes the imputation flag variables to indicate records with
imputed values. Tenure, sex, race, and Hispanic origin were imputed using hot-deck imputation. Adult 1
education, household size, and poverty ratio were imputed using sequential regression imputation
methods. 22
Table 155. List of Imputed Variables and their Imputation Flags
Variable
Public Use
File
Variable
Missing Rates
Imputation Flag Variable
Household tenure
(TENURE)
Screener
0.63%
Flag for Household Tenure
(TENURE_IF)
Child’s sex
(C_SEX)
Screener
0.19%
Flag for child’s sex
(C_SEX_IF)
Child’s race
(C_RACE_R)
Screener
0.96%
Flag for child’s race
(C_RACE_R_IF)
Child’s Hispanic origin
(C_HISPANIC_R)
Screener
0.83%
Flag for child’s Hispanic origin
(C_HISPANIC_R_IF)
Selected child’s sex
(SC_SEX)
Topical
0.14%
Flag for selected child’s sex
(SC_SEX_IF)
Selected child’s race
(SC_RACE_R)
Topical
0.68%
Flag for selected child’s race
(SC_RACE_R_IF)
Selected child’s Hispanic
origin (SC_HISPANIC_R)
Topical
0.57%
Flag for selected child’s Hispanic origin
(SC_HISPANIC_R_IF)
Adult 1’s highest
completed year of school
(A1_GRADE)
Topical
2.40%
Flag for adult 1’s highest completed
year of school
(A1_GRADE_IF)
22
For more information on data analysis using imputed values, see https://www2.census.gov/programssurveys/nsch/technical-documentation/methodology/NSCH-Analysis-with-Imputed-Data-Guide.pdf
2019 National Survey of Children’s Health
U.S. Census Bureau
40
Household size
(HHCOUNT)
Topical
2.48%
Flag for household size
(HHCOUNT_IF)
Family poverty ratio
(FPL)
Topical
17.60%
Flag for family poverty ratio
(FPL_IF)
Multiple Imputation
Using sequential regression imputation methods, FPL is multiply imputed and contains six versions or
implicates. The public use file includes all six imputed values for FPL [FPL_I1-FPL_I6]. The primary
motivation for the multiple imputation is to allow interested researchers to appropriately account for
uncertainty in estimates using FPL that is hidden when using a single implicate. 23 FPL input includes
imputed values for family income (not included in the public use file) and number of people that are
family members (FAMCOUNT). An estimated family count (FAMCOUNT) was derived from HHCOUNT
and other household information when FAMCOUNT was not reported by the household. The imputation
was executed by sequential regression modeling imputation 24 using IVEWare. 25
Suppressed Variables
A number of variables had range caps or suppressed values to protect respondent confidentiality
consistent with U.S. Census Bureau protocols. For example, a reported value must represent at least
10,000 children (weighted estimate). These variables are listed in Table 16.
Table 166. List of Suppressed Variables
Variable
Description
Valid Values
TOTKIDS_R
Number of children living in the household
1=1
2=2
3=3
4 = 4+
MOMAGE
Age of mother when child was born
18 = 18 years or younger
45 = 45 years or older
K2Q35A_1_YEARS
Age of child when first diagnosed with
autism
1 = 1 year or younger
15 = 15 years or older
BIRTHWT_OZ_S
Birth weight
72 = 72 oz. or less
155 = 155 oz. or more
K11Q43R
Number of time the child has moved to a
new address
13 = 13 or 14 times
15 = 15 or more times
A1_AGE
Age of Adult 1
75 = 75 years or older
Schaefer JL, Graham JW. 2002. “Missing Data: Our View of State of the Art”. Psychological Methods, 7(2): 147-77.
Raghunathan TE, Lepkowski JM, Hoewyk JV, Solenberger PW. 2001. “A Multivariate Technique for Multiply
Imputing Missing Values using a Sequence of Regression Models”. Survey Methodology, 27: 85–95.
25
Raghunathan TE, Solenberger PW, Hoewyk JV. 2016. IVEware: Imputation and Variance Estimation Software
User’s Guide (Version 0.3). Ann Arbor, MI: Institute for Social Research, University of Michigan.
23
24
41
2019 National Survey of Children’s Health
U.S. Census Bureau
Variable
Description
Valid Values
A2_AGE
Age of Adult 2
75 = 75 years or older
A1_LIVEUSA
When Adult 1 came to live in the U.S.
1970 = Before or in 1970
A2_LIVEUSA
When Adult 2 came to live in the U.S.
1970 = Before or in 1970
BREASTFEDEND_DAY_S
Stopped breastfeeding, age in days
Suppressed if > 5
BREASTFEDEND_WK_S
Stopped breastfeeding, age in weeks
Suppressed if > 8
BREASTFEDEND_MO_S
Stopped breastfeeding, age in months
29 = 29 or more
FRSTFORMULA_DAY_S
First fed formula, age in days
Suppressed if > 6
FRSTFORMULA_WK_S
First fed formula, age in weeks
Suppressed if > 9
FRSTFORMULA_MO_S
First fed formula, age in months
12 = 12 or more
FRSTSOLIDS_DAY_S
First fed solids, age in days
Suppressed if > 1
FRSTSOLIDS_WK_S
First fed solids, age in weeks
Suppressed if > 4
FRSTSOLIDS_MO_S
First fed solids, age in months
15 = 15 or more
FPL
Family poverty ratio
50 = 50% or less
400 = 400% or more
FAMCOUNT
Family Count
8 = 8 or more
HHCOUNT
Household Count
10 = 10 or more
K4Q37
Received Special Services - Age in Years
15 = 15 or more
SESPLANYR
Special Education Plan - Age in Years
16 = 16 or more
Geography Variables
The 2019 NSCH includes four geography variables on the Public Use File
•
•
•
•
FIPSST (State of Residence)
CBSAFP_YN (Core-Based Statistical Area Status)
METRO_YN (Metropolitan Statistical Area Status)
MPC_YN (Metropolitan Principal City Status)
Table 17 provides a general description of the geography variables and their valid values. To protect
respondent confidentiality, CBSAFP_YN, METRO_YN, and MPC_YN are not reported in some states. If a
variable or intersection of variables could be used to identify a geographic area within a state with a
child population under 100,000, reported values for that variable were replaced with ".D", indicating
"Suppressed for Confidentiality".
42
2019 National Survey of Children’s Health
U.S. Census Bureau
Table 177. List of Geography Variables
Variable
Description
FIPSST
State of Residence
CBSAFP_YN Core Based Statistical Area (CBSA): County or counties
associated with at least one core (urbanized area or urban
cluster) of at least 10,000 population, plus adjacent
counties having a high degree of social and economic
integration with the core as measured through commuting
ties.
METRO_YN Metropolitan Statistical Area (MSA): County or counties
associated with at least one urbanized area of at least
50,000 population, plus adjacent counties having a high
degree of social and economic integration with the core as
measured through commuting ties.
MPC_YN
Metropolitan Principal City: An incorporated place or
census designated place in a Metropolitan Statistical Area
that meets specific population and workforce
requirements.
Valid Values
[FIPS code]
.D = Suppressed for confidentiality
1 = In a CBSA
2 = Not in a CBSA
.D = Suppressed for confidentiality
1 = In a MSA
2 = Not in a MSA
.D = Suppressed for confidentiality
1 = In a Metropolitan Principal City
2 = Not in a Metropolitan Principal
City
Additional geographies are identified through the intersection of CBSAFP_YN, METRO_YN, and MPC_YN
shown in Table 18.
Table 188. Geographies Identified at the Intersections
Intersection
CBSAFP_YN =1 and
METRO_YN =2
METRO_YN =1 and
MPC_YN=2
Additional Geography Level
Micropolitan Statistical Area: County or counties (or equivalent
entities) associated with at least one urban cluster of at least
10,000 but less than 50,000 population, plus adjacent counties
having a high degree of social and economic integration with the
core as measured through commuting ties
In an MSA, but not a Metropolitan Principal City: County or counties
associated with at least one urbanized area of at least 50,000
population, plus adjacent counties having a high degree of social
and economic integration with the core as measured through
commuting ties, but is not incorporated place or census designated
place within the Metropolitan Statistical Area.
Alternative and lower level geographic identifiers are not included with the public use data file. Access
to these variables is restricted to the Federal Statistical Research Data Centers (RDCs). Researchers can
apply for RDC access; proposed projects must demonstrate scientific merit, require non-public data, be
feasible, pose no risk to respondent confidentiality, and provide benefit to Census Bureau programs. The
43
2019 National Survey of Children’s Health
U.S. Census Bureau
currently open RDCs are listed at https://www.census.gov/about/adrm/fsrdc/locations.html, and
additional information on the RDC application process is available at https://www.census.gov/programssurveys/ces/data/restricted-use-data/apply-for-access.html.
44
2019 National Survey of Children’s Health
U.S. Census Bureau
Weighting Specifications
Overview
The 2019 NSCH uses case- and household-level weights for population-based estimates. These include
•
•
•
Household Weight-Screener (FWH)
Child Weight-Screener (FWS)
Selected Child Weight-Topical (FWC)
Each weight is the product of the base sampling weight, nonresponse adjustment factors, and raking to
population controls. The Selected Child Weight also includes a subsampling adjustment. Population
controls are derived from the 2018 American Community Survey (ACS).
Base Sampling Weights
The weighting process began with the base sampling weight for each sample household. The base
weight (i.e., sampling interval) for each sample housing unit was the inverse of its probability of
selection for the screener. Base weights were calculated separately for each of the two strata and each
state, including the District of Columbia. If there had been no nonresponse and the survey frame was
complete, using this weight would give unbiased estimates for the survey population.
Adjustment for Screener Nonresponse
Following the base weight, an adjustment for screener nonresponse was implemented to increase the
weights of the households that responded to the screener in order to account for all of the households
that did not respond to the screener. Households were put into one of sixteen cells defined by stratum,
a block group poverty measure variable (yes or no) indicating the proportion of households with income
less than 150% of the federal poverty level, web group (High Paper or High Web), and Metropolitan
Statistical Area Status. The screener nonresponse adjustment factor was calculated within each cell
using the following formula:
weighted sum of screener interviews + weighted number of screener non-interviews
�
�
weighted sum of screener interviews
where the number of screener non-interviews =
weighted sum of screener interviews
�
�
weighted sum of screener interviews + weighted sum of screener ineligible households
×
(weighted sum of households with unknown screener eligibility)
In other words, the count of screener non-interviews was an estimate of the expected number of
eligible households from those cases for which nothing was returned. The term “eligible” here refers to
the address belonging to an occupied, residential household. The expected number of eligible cases was
estimated by taking the eligibility rate among the known cases and applying it to the unknown cases.
45
2019 National Survey of Children’s Health
U.S. Census Bureau
The screener nonresponse adjustment was the last step of the weight processing that included the
households for which there was no screener interview and the screener-interviewed households that
indicated no eligible children.
Adjustment to Population Controls at the Household Level
All households with children that completed a screener were given a household-level weight. In addition
to the base weight and screener nonresponse adjustment, a household post-stratification adjustment
was applied in order to achieve the final household screener weight. This factor consisted of ratio
adjustments to population controls at the household level obtained from the 2018 ACS data.
Households were put into one of 255 cells defined by state, race of the child selected for the topical, and
Hispanic origin (yes or no) of the selected child if the selected child’s race was White. Within each cell,
the household post-stratification adjustment was calculated as the ACS population count for the cell
divided by the cell’s weighted total. The product of the base weight, screener nonresponse adjustment,
and this household post-stratification adjustment constituted the final household screener weight.
First Raking to Population Controls: All Screener Children
All eligible children (four at most) from completed screener interviews were given a child-level screener
weight. The weights of children from completed screener interviews were adjusted to match the 2018
ACS estimates for the following characteristics:
•
•
•
Dimension #1 – State by Child’s Race (White alone, Black alone, Asian alone, Other)
Dimension #2 – State by Child’s Ethnicity (Hispanic, Non-Hispanic)
Dimension #3 – State by Child’s Sex by Child’s Age Group (0-5, 6-11, 12-17 years)
Each iteration of this process consisted of calculating three ratio adjustments, one for each dimension,
sequentially. The adjustment factor calculated for Dimension 1 was applied to the weights accordingly
and this newly adjusted weight went into the calculation of the adjustment factor for Dimension 2. This
iterative raking process continued until the difference between the sum of the weights and the control
total associated with each cell was less than 1% of the control. The resulting weight from this process
was the final child-level screener weight for each eligible child. Only the children selected for the topical
continued in the weighting process to eventually receive a final interviewed child weight.
Adjustment for Households with More than One Child
In households with multiple children, the selected child represented all eligible children in their
household. Thus, a within-household subsampling factor was applied to account for the selection of a
single child, as well as the oversampling for young children and children with special health care needs
(CSHCN). The value of this adjustment was the inverse of the probability of selection for the selected
child. Probabilities varied by the number of children in the household, the presence of children aged 0-5,
and the presence of CSHCN.
Adjustment for Topical Nonresponse
Similar to the screener nonresponse adjustment, the weights of the households responding to the
topical needed to be increased to account for all of the households not responding to the topical. The
46
2019 National Survey of Children’s Health
U.S. Census Bureau
adjustment considered all topical interviews (complete and sufficient partial) defined by questionnaires
with valid answers for at least 25 of 50 test questions, and at least one item in Section H or beyond or the
respondent submitted the topical portion of the web instrument. Returned topical that did not meet the
criteria were considered a topical non-interview.
All topical-eligible households were put into one of sixteen cells depending on imputed poverty/non-poverty
status, web group (High Paper vs. High Web), tenure (owner occupied or not), and presence of CSHCN. The
topical nonresponse adjustment was calculated within each of the sixteen cells as:
weighted sum of topical interviews + weighted sum of topical non-interviews
�
�
weighted sum of topical interviews
After this adjustment, the selected children from topical non-interview households were no longer
involved in the weighting process and only interviewed children continued to the last steps.
Second Raking to Population Controls: Topical Interviewed Children
The final step of the weighting was accomplished through a second iterative raking process to ACS
population controls. The process was equivalent to that of the child-level screener weight, with the
exception of additional and different dimensions as well as a trimming step. The following eight
analytical domains of interest were used:
•
•
•
•
•
•
•
•
Dimension #1 – State by Family Poverty Ratio (≤100%, 101-200%, >200%)
Dimension #2 – State by Household Size (≤3, 4, >4)
Dimension #3 – State Groupings by Respondent’s Education (High
School)
Dimension #4 – State by Selected Child’s Race (White, Black, Asian, Other)
Dimension #5 – State by Selected Child’s Ethnicity (Hispanic, Non-Hispanic)
Dimension #6 – State by Selected Child’s Special Health Care Needs Status
Dimension #7 – Selected Child’s Race by Ethnicity (at the national level)
Dimension #8 – Selected Child’s Sex by Single Age (at the national level)
For Dimension #3, some states needed to be grouped due to the low number of respondents in each
state with less than a high school degree. States were grouped with others that had similar education
distributions based on ACS data. The states were first sorted by the ACS-derived percent of children in
households where the respondent has less than a high school degree, followed by an additional sort by
the percent of children in households where the respondent has a high school degree. State groupings
were made with the intent of keeping these distributions similar within each group. The result was 16
state groupings and 8 stand-alone states. The following were the resulting groupings:
•
•
•
•
Group 1: Maine, New Hampshire, North Dakota, and Vermont
Group 2: Minnesota, Utah, and Virginia
Group 3: Hawaii, Iowa, Montana, and Wyoming
Group 4: Massachusetts and Colorado
47
2019 National Survey of Children’s Health
U.S. Census Bureau
•
•
•
•
•
•
•
•
•
•
•
•
•
Group 5: Connecticut, Nebraska, New Jersey, and South Dakota
Group 6: DC and Maryland
Group 7: Illinois, Oregon, and Rhode Island
Group 8: Michigan, Ohio, and Wisconsin
Group 9: Idaho, Kansas, and Washington
Group 10: Missouri and South Carolina
Group 11: Delaware, Kentucky, and Pennsylvania
Group 12: Alaska and West Virginia
Group 13: New York and North Carolina
Group 14: Alabama and Florida
Group 15: Arkansas and Louisiana
Group 16: Arizona and Texas
Stand-alone states: California, Georgia, Indiana, Mississippi, Nevada, New Mexico, Oklahoma,
and Tennessee
Trimming of Large Weights
The resulting weights from each iteration of the raking process were checked for extreme values in
order to prevent a small number of cases with large weights from having undue influence on estimates
and increasing the variance. An extreme value was determined to be a weight that exceeded the median
weight plus six times the interquartile range (IQR) of the weights in each state. These extreme weights
were truncated to this cutoff (median plus six times the IQR of weights in that state) and the weights
were checked for convergence to the controls. Convergence required the weighted total of each cell to
be within 1% of the control for the cell. If convergence was not met for every cell, another iteration of
the raking process was applied again. This process of raking and trimming was reiterated until
convergence was met and there were few extreme weights left. In general, the remaining extreme
weights were observed to be very close to the cutoff. The remaining extreme weights were truncated a
final time to the median plus six times the IQR in the state and the process was complete.
Population Controls
Population controls used throughout the weighting were derived from the 2018 ACS one-year estimates.
By using the 2018 ACS data, the weighted totals were ensured to match the most up-to-date population
control totals available for key demographic variables for children and households in the U.S. The
controls were used in the household post-stratification adjustment, the raking to attain the child-level
screener weights, and the raking to attain the final topical interviewed children weights. Almost all
controls used were at the state level, with the exception of the last two dimensions where national-level
controls were used in the second raking process.
For the household post-stratification adjustment, the NSCH household weights were adjusted so that
the sum of the weights equaled the 2018 ACS estimates for the number of households in each state by
race (White, Black, Asian, Other) and by Hispanic origin (yes or no) if the selected child’s race was White.
In the first raking process, up to four children from each screener received adjustments so that the sum
48
2019 National Survey of Children’s Health
U.S. Census Bureau
of the weights of all children listed on screeners equaled the ACS estimates for the number of children in
each state by race, state by Hispanic origin, and state by sex by age group (0-5, 6-11, 12-17 years).
Finally, in the second raking process, the weights of the NSCH topical interviewed children were
adjusted so that the sum of their weights equaled the ACS estimates for each state by family poverty
ratio (≤100%, 101-200%, >200%), household size ( ≤3, 4, >4), respondent’s highest level of education
(High School), race, Hispanic origin, and special health care needs status, as
well as race by ethnicity and sex by age in years at the national level.
Limitations
In order to minimize the variability of the weights caused by large adjustment factors, cells having fewer
than 30 cases were collapsed with a neighboring cell. The adjustment factors were then calculated for
the merged cells by combining the population controls and the sample cases for the two cells. Since the
individual cells were combined, and only one adjustment factor was created per cell, only the weighted
total for the combined cell will match the control following the raking procedure. Consequently, the
weighted totals for the individual cells will most likely not match the population controls for the original
individual cells. As shown in Table 19, cells were collapsed in two of the dimensions in the last raking
step.
Table 19. Collapsed Dimensions of Final Raking and Affected States
Collapse
Dimension Collapsed
Affected States
Black collapsed with
Other in 26 states
Dimension #4 - State by Selected Child’s AK, AZ, CA, CO, CT, HI, ID, IA, KS, ME, MA, MN,
Race (White, Black, Asian, Other)
MT, NE, NH, NM, ND, OK, OR, SD, UT, VT, WA,
WV, WI, WY
Asian collapsed with
Other in 35 states
Dimension #4 - State by Selected Child’s AL, AZ, AR, CO, DC, FL, ID, IN, IA, KS, KY, LA, ME,
Race (White, Black, Asian, Other)
MI, MS, MO, MT, NE, NH, NM, NC, ND, OH, OK,
OR, PA, RI, SC, SD, TN, UT, VT, WV, WI, WY
Hispanic and NonDimension #5 - State by Selected Child’s AL, KY, ME, MI, MS, MO, ND, SD, VT, WV
Hispanic collapsed in 10 Ethnicity (Hispanic, Non-Hispanic)
states
49
2019 National Survey of Children’s Health
U.S. Census Bureau
Estimation, Hypothesis Testing, and Data Use Guidelines
Variance Estimation
When survey weights are used, the resulting estimates from the 2019 NSCH are representative of all
non-institutionalized children aged 0 to 17 years in the U.S. and in each state and the District of
Columbia who live in housing units. These weighted estimates do not generalize to the population of
parents, mothers, or pediatric health care providers. Analysts are advised to avoid statements such as
“the percent of parents”.
Two stratum identifiers should be used to estimate variance: FIPSST (state of residence) and STRATUM
(identifies households flagged with children). Each record in the data file is assigned a unique household
identifier, HHID. Some analysts may be using statistical programs that only permit the specification of a
single stratum variable. These users should define a new variable with 102 levels by crossing STRATUM (2
levels) with FIPSST (51 levels). This new variable can then be used as the stratum variable. For example,
Stata users can specify only one variable in the strata() option of svyset. This new variable (named here
as STRATACROSS) can be created using the following statement:
•
EGEN STRATACROSS = GROUP (FIPSST STRATUM)
SUDAAN users can identify both FIPSST and STRATUM in the NEST statement. However, SUDAAN users
should note that the first variable listed after the word NEST is assumed to be the stratum variable, and
the second variable listed is assumed to be the PSU. To properly identify the PSU variable, the PSULEV
option must be invoked in the NEST statement as shown here:
•
NEST FIPSST STRATUM HHID / PSULEV = 3;
In both individual year and multi-year analyses, the NSCH sample size may be limited for smaller
populations (e.g., American Indian or Alaska Native) and state-level subgroups or rare outcomes (e.g.,
adolescent CSHCN or autism in a particular state). Small sample sizes may produce unstable
estimates. To minimize misinterpretation, we recommend only presenting statistics with a sample
size or unweighted denominator of 30 or more. Further, if the 95% confidence interval width exceeds
20 percentage points or 1.2 times the estimate (≈ relative standard error >30%), we recommend
flagging for poor reliability and/or presenting a measure of statistical reliability (e.g., confidence
intervals or statistical significance testing) to promote appropriate interpretation.
State-level estimates may be compared to national estimates using a nested z-test to identify
significant differences at a given alpha or Type 1 error level (e.g., 0.05, 0.01). The formula for this is as
follows:
𝑍𝑍 =
𝑋𝑋�𝑖𝑖 − 𝑋𝑋�𝑗𝑗
�𝑆𝑆𝐸𝐸𝑖𝑖2 + 𝑆𝑆𝐸𝐸𝑗𝑗2 − 2𝑃𝑃 ∗ 𝑆𝑆𝐸𝐸𝑗𝑗2
50
2019 National Survey of Children’s Health
U.S. Census Bureau
Where j is a subset of i (e.g., Alabama as part of the Total US), 𝑋𝑋� is the mean or proportion, SE is the
standard error, and P is the proportion of the weighted denominator for a given indicator that is
specific to j (e.g., Alabama weighted denominator divided by the Total US weighted denominator). A
simple independent Z-test would be a more conservative test that may increase Type II error—the
probability of failing to reject the null of no difference when there is a difference.
Combining Data across Survey Years
Data across multiple years of the redesigned NSCH (2016 and later) can be combined to increase the
analytic sample size. By leveraging a larger sample, data users can analyze smaller population groups
and rare outcomes that are not sufficiently represented in a single year sample and produce national
and state-level estimates with smaller standard errors. Guidance for producing multi-year estimates is
available at https://www2.census.gov/programs-surveys/nsch/technicaldocumentation/methodology/NSCH-Guide-to-Multi-Year-Estimates.pdf.
Confidentiality
Participation in the 2019 NSCH was voluntary, and all data collected that could potentially identify an
individual person are confidential. Data are kept private in accordance with applicable law. Respondents
are assured of the confidentiality of their replies in accordance with 13 U.S.C. Section 9. All access to
Title 13 data from this survey is restricted to Census Bureau employees and those holding Census
Bureau Special Sworn Status pursuant to 13 U.S.C. Section 23(c). In compliance with this law, all data
released to the public are only in a statistical format. No information that could personally identify a
respondent or household may be released. The Screener and Topical public use data files went through
a thorough disclosure review process and were approved by the Census Disclosure Review Board prior
to release.
Guidelines for Data Use
The U.S. Census Bureau is conducting the NSCH on the behalf of the Health Resources and Services
Administration’s Maternal and Child Health Bureau (HRSA MCHB) within the U.S. Department of Health
and Human Services (HHS) under Title 13, United States Code, Section 8(b), which allows the Census
Bureau to conduct surveys on behalf of other agencies. Title 42 U.S.C. Section 701(a)(2) allows HHS to
collect information for the purpose of understanding the health and well-being of children in the U.S.
The data collected under this agreement are confidential under 13 U.S.C. Section 9. All access to Title 13
data from this survey is restricted to Census Bureau employees and those holding Census Bureau Special
Sworn Status pursuant to 13 U.S.C. Section 23(c).
Any effort to determine the identity of any reported case is prohibited. The Census Bureau and HRSA
MCHB take extraordinary measures to assure that the identity of survey subjects cannot be disclosed. All
direct identifiers, as well as characteristics that might lead to identification, have been omitted from the
data set. Any intentional identification or disclosure of a person or establishment violates the assurances
of confidentiality given to the providers of the information. Therefore, users must:
•
Use the data in this data set for statistical reporting and analysis only
51
2019 National Survey of Children’s Health
U.S. Census Bureau
•
•
Make no use of the identity of any person discovered, inadvertently or otherwise
Not link this data set with individually identifiable data from any other Census Bureau or nonCensus Bureau data sets
Use of the data set signifies users’ agreement to comply with the previously stated statutory-based
requirements. Before releasing any statistics to the public, the Census Bureau reviews them to make
sure none of the information or characteristics could identify someone. For more information about the
Census Bureau’s privacy and confidentiality protections, contact the Policy Coordination Office toll-free
at 1-800-923-8282.
52
2019 National Survey of Children’s Health
U.S. Census Bureau
Supporting Material
References
Bethell CD, Read D, Neff J, Blumberg SJ, Stein RE, Sharp V, Newacheck PW. 2002. “Comparison of the
Children with Special Health Care Needs Screener to the Questionnaire for Identifying Children with
Chronic Conditions—Revised.” Ambulatory Pediatrics, Jan-Feb 2(1): 49-57.
Blumberg SJ, Luke JV. 2010. Wireless Substitution: Early Release of Estimates from the National Health
Interview Survey, January–June 2010. National Center for Health Statistics. Available from:
http://www.cdc.gov/nchs/nhis.htm
Blumberg SJ, Foster EB, Frasier AM, et al. 2012. Design and Operation of the National Survey of
Children’s Health, 2007. National Center for Health Statistics. Vital Health Stat, 1(55). Available from:
http://www.cdc.gov/nchs/data/series/sr_01/sr01_055.pdf
Bramlett MD, Blumberg SJ, Ormson AE, et al. 2014. Design and Operation of the National Survey of
Children with Special Health Care Needs, 2009–2010. National Center for Health Statistics. Vital Health
Stat, 1(57). Available from: http://www.cdc.gov/nchs/data/series/sr_01/sr01_057.pdf
Brick JM, Williams D, Montaquila JM. 2011. “Address-Based Sampling for Subpopulation Surveys.” Public
Opinion Quarterly, 75(3): 409-28.
Dillman DA, Smyth JD, Christian LM. 2009. Internet, Mail and Mixed-Mode Surveys: The Tailored Design
Method, 3rd edition. Hoboken, NJ: John Wiley & Sons.
Foster EB, Frasier AM, Morrison HM, O’Connor KS, Blumberg SJ. 2010. All Things Incentive: Exploring the
Best Combination of Incentive Conditions. Paper presented at the American Association for Public
Opinion Research annual conference, Chicago, IL.
Raghunathan TE, Lepkowski JM, Hoewyk JV, Solenberger PW. 2001. “A Multivariate Technique for
Multiply Imputing Missing Values using a Sequence of Regression Models.” Survey Methodology, 27: 8595.
Raghunathan TE, Solenberger PW, Hoewyk JV. 2016. IVEware: Imputation and Variance Estimation
Software User’s Guide (Version 0.3). Ann Arbor, MI: Institute for Social Research, University of Michigan.
Schaefer JL, Graham JW. 2002. “Missing Data: Our View of State of the Art”. Psychological Methods,
7(2): 147-77.
53
2019 National Survey of Children’s Health
U.S. Census Bureau
Attachment A: Estimated State-Level Production Sample Sizes for the 2019
National Survey of Children’s Health
State sample sizes by stratum were determined using the following criteria. First, the Stratum 1
oversampling rates for each state were maximized such that the variance did not far exceed that of a
design that sampled equally in the two strata. Second, the target number of topical interviews per state
was adjusted until the total sample size was at the desired size, which for 2019 is approximately 180,000
addresses yielding 515 topical interviews per state.
Table A-1: Address Sample Size and Strata Distribution by State
State
Alabama
Alaska
Arizona
Arkansas
California
Colorado
Connecticut
Delaware
District of Columbia
Florida
Georgia
Hawaii
Idaho
Illinois
Indiana
Iowa
Kansas
Kentucky
Louisiana
Maine
Maryland
Massachusetts
Michigan
Minnesota
Mississippi
Missouri
Montana
Nebraska
Nevada
Total Sample
(approx.)
4,100
5,200
4,200
5,100
3,300
2,900
3,000
3,600
3,500
4,400
4,300
3,800
2,900
3,200
3,200
2,700
2,900
3,900
5,300
3,100
2,800
2,700
2,600
2,000
5,200
3,200
3,800
2,900
4,100
Stratum 1
Stratum 2A
61.0%
43.3%
55.9%
54.2%
67.5%
60.7%
63.5%
63.3%
67.2%
59.6%
62.0%
41.8%
58.9%
60.7%
63.0%
64.6%
66.6%
58.1%
58.1%
62.3%
65.8%
66.7%
68.1%
68.9%
60.1%
65.0%
51.6%
62.9%
60.4%
39.0%
56.7%
44.1%
45.8%
32.5%
39.3%
36.5%
36.7%
32.8%
40.4%
38.0%
58.2%
41.1%
39.3%
37.0%
35.4%
33.4%
41.9%
41.9%
37.7%
34.2%
33.3%
31.9%
31.1%
39.9%
35.0%
48.4%
37.1%
39.6%
54
2019 National Survey of Children’s Health
U.S. Census Bureau
New Hampshire
New Jersey
New Mexico
New York
North Carolina
North Dakota
Ohio
Oklahoma
Oregon
Pennsylvania
Rhode Island
South Carolina
South Dakota
Tennessee
Texas
Utah
Vermont
Virginia
Washington
West Virginia
Wisconsin
Wyoming
3,100
2,900
5,400
3,800
3,500
3,100
2,900
4,700
2,800
2,800
3,400
4,000
3,200
3,400
4,100
2,200
3,200
2,800
2,900
4,900
2,300
4,300
64.9%
64.1%
48.9%
57.3%
65.4%
60.6%
68.0%
55.1%
64.8%
66.3%
63.9%
63.4%
56.9%
66.0%
63.6%
70.2%
57.0%
64.7%
60.4%
46.9%
67.0%
52.5%
35.1%
35.9%
51.1%
42.7%
34.6%
39.4%
32.0%
44.9%
35.2%
33.7%
36.1%
36.6%
43.1%
34.0%
36.4%
29.8%
43.0%
35.3%
39.6%
53.1%
33.0%
47.5%
55
2019 National Survey of Children’s Health
U.S. Census Bureau
Attachment B: Traditional and Redesigned Envelope Designs
For the initial mailing, 50% of the sample received traditional envelopes and 50% received the
redesigned envelopes. The redesigned envelope moved the Census logo to the upper left corner, added
a red banner with white call-out text, and added icons. The two “Low Paper” envelopes were business
standard size envelopes and contained a Web-push invite to the survey. The two “High Paper” initial
mailing envelopes were 9"x11.5" flat mail envelopes and contained a mixed-mode invite, both a web
survey invite and a paper screener questionnaire.
Traditional Low Paper Envelope
Redesigned Low Paper Envelope
56
2019 National Survey of Children’s Health
U.S. Census Bureau
Traditional High Paper Envelope
57
2019 National Survey of Children’s Health
U.S. Census Bureau
Redesigned High Paper Envelope
58
2019 National Survey of Children’s Health
U.S. Census Bureau
Attachment C: Determined Household Type from the 2019 National Survey of
Children’s Health
Households are given a type based on the created variables from the screener responses:
•
•
•
•
•
TOTKIDS_E Total number of eligible children
CHILDY0_5 Indicator child in 0-5 years old
CHILDN0_5 Indicator child is not 0-5 years old
TOTCSHCN Total number of special needs children
TOTNON Total number of not special needs children
Table B-1 shows the value for TYPE (1,2,3A,3B,4,5A,5B,5C,6,7,8) based on from the values of the five
screener variables indicated in columns 2-6. Each record picks up the selected child number value of
(0,1,2,3,4) from the corresponding HHTYP_1 - HHTYP_8 variable. SC_CHILD receives the value of the
selected child number for that household, which is the number of the child after sorting.
Table B-1: Household Type Assignment from the Values of Five Screener Variables
Household
Type
TYPE=1
→
HHTYP_1
TYPE=2
→
HHTYP_2
TYPE=3A
→
HHTYP_3A
TYPE=3B
→
HHTYP_3B
TYPE=4
→
HHTYP_4
TYPE=5A
→
HHTYP_5A
TYPE=5B
→
HHTYP_5B
TYPE=5C
→
HHTYP_5C
Screener Variables
TOTKIDS_E
CHILDY0_5
CHILDN0_5
TOTCSHCN
TOTNON
0 or blank
n/a
n/a
n/a
n/a
1
n/a
n/a
n/a
n/a
2
0
2
0
2
0
2
50% chance of selection
0
2
0
2
0
2
1
1
2
0
0
2
2
n/a
n/a
1
1
Selected child is 0 to 5 years
old probability: 62%
Over 5 years: 38%
Select CSHCN probability: 64%
Select Non CSHCN: 36%
3
0
3
0
3
0
3
All three children have 33%
chance of selection
0
3
0
3
0
3
1
2
3
0
0
3
Child 0 through 5: 44%
Child greater than 5: 28%
3
2
1
3
0
0
3
2
3
No child selected
Only child was selected
Child 0 through 5: 38%
Child greater than 5: 24%
59
2019 National Survey of Children’s Health
U.S. Census Bureau
Household
Type
TYPE=6
→
HHTYP_6
TYPE=7
→
HHTYP_7
TYPE=8
→
HHTYP_8
Screener Variables
TOTKIDS_E
CHILDY0_5
CHILDN0_5
TOTCSHCN
TOTNON
3
n/a
n/a
1
2
Select CSHCN probability: 48%
Select Non CSHCN: 26%
3
n/a
n/a
2
1
Select CSHCN probability: 39%
Select Non CSHCN: 22%
≥4
n/a
n/a
n/a
n/a
All four children have 25%
chance of selection
60
2019 National Survey of Children’s Health
U.S. Census Bureau
Attachment D: Weighted Response Rates by State
State
United States
Alabama
Alaska
Arizona
Arkansas
California
Colorado
Connecticut
Delaware
District of Columbia
Florida
Georgia
Hawaii
Idaho
Illinois
Indiana
Iowa
Kansas
Kentucky
Louisiana
Maine
Maryland
Massachusetts
Michigan
Minnesota
Resolution
Rate
53.4%
51.8%
67.1%
55.0%
55.0%
45.9%
55.1%
51.3%
51.2%
51.2%
50.4%
47.3%
60.3%
63.4%
52.6%
55.1%
64.9%
57.8%
58.1%
49.0%
64.6%
54.7%
54.5%
56.2%
63.5%
Screener
Conversion
Rate
99.0%
98.8%
99.3%
98.8%
99.3%
98.7%
99.3%
98.5%
99.7%
99.3%
98.7%
99.1%
99.3%
99.4%
98.8%
99.4%
99.7%
99.4%
99.8%
98.8%
99.8%
99.0%
98.7%
99.3%
99.4%
Screener
Completion
Rate
47.8%
44.9%
56.7%
46.9%
46.9%
41.7%
49.5%
47.3%
46.3%
46.8%
43.3%
40.7%
53.5%
57.3%
47.7%
49.3%
61.9%
52.1%
52.8%
40.2%
56.5%
50.7%
50.6%
51.2%
59.5%
Topical
Conversion
Rate
80.3%
78.6%
80.5%
80.1%
77.9%
77.9%
81.2%
74.6%
78.8%
78.3%
77.4%
79.2%
78.2%
80.6%
79.7%
78.6%
78.9%
81.4%
77.8%
79.5%
76.7%
78.2%
77.9%
78.2%
82.4%
Topical
Completion
Rate
35.3%
33.3%
39.0%
36.7%
33.3%
31.9%
38.1%
36.6%
32.8%
37.4%
33.4%
31.8%
38.1%
43.9%
37.3%
37.2%
40.2%
39.5%
33.8%
31.0%
36.9%
38.6%
38.1%
37.2%
45.5%
Interview
Completion
Rate
79.5%
77.6%
79.9%
79.1%
77.3%
76.9%
80.6%
73.5%
78.6%
77.8%
76.4%
78.5%
77.7%
80.1%
78.7%
78.1%
78.7%
80.9%
77.7%
78.6%
76.5%
77.5%
76.9%
77.7%
81.9%
Overall
Response
Rate
42.4%
40.2%
53.7%
43.5%
42.5%
35.3%
44.4%
37.7%
40.2%
39.9%
38.5%
37.1%
46.8%
50.7%
41.4%
43.0%
51.1%
46.7%
45.1%
38.5%
49.4%
42.4%
41.9%
43.6%
52.0%
61
2019 National Survey of Children’s Health
U.S. Census Bureau
Mississippi
Missouri
Montana
Nebraska
Nevada
New Hampshire
New Jersey
New Mexico
New York
North Carolina
North Dakota
Ohio
Oklahoma
Oregon
Pennsylvania
Rhode Island
South Carolina
South Dakota
Tennessee
Texas
Utah
Vermont
Virginia
Washington
West Virginia
Wisconsin
Wyoming
51.9%
58.0%
68.9%
61.5%
49.5%
63.8%
50.1%
62.6%
50.1%
53.0%
66.1%
55.6%
56.7%
57.3%
57.8%
50.7%
53.8%
67.0%
50.6%
47.5%
56.7%
71.1%
55.0%
59.2%
63.2%
63.3%
66.9%
99.4%
99.6%
99.9%
99.4%
99.3%
99.7%
99.2%
99.5%
98.1%
99.2%
99.6%
99.5%
99.0%
99.0%
99.4%
99.0%
98.7%
99.8%
99.3%
98.8%
99.0%
99.8%
98.7%
98.7%
99.7%
99.1%
99.6%
44.0%
52.3%
63.1%
57.6%
43.8%
56.9%
46.6%
54.9%
45.7%
46.3%
61.2%
50.7%
49.2%
53.7%
52.6%
46.4%
46.1%
62.9%
45.8%
40.6%
51.9%
60.3%
50.2%
54.3%
55.5%
58.5%
57.1%
74.1%
81.0%
79.2%
80.1%
78.0%
80.3%
78.9%
78.7%
73.5%
78.0%
77.1%
74.5%
77.8%
81.7%
81.1%
78.3%
74.4%
82.1%
77.4%
74.5%
82.1%
81.9%
81.0%
80.4%
78.6%
80.3%
81.3%
30.9%
40.4%
42.7%
38.5%
31.3%
41.2%
35.2%
37.1%
32.8%
34.2%
40.8%
35.3%
35.2%
41.9%
41.7%
33.8%
32.0%
42.7%
33.3%
28.4%
42.6%
42.7%
38.6%
41.2%
36.2%
43.8%
40.3%
73.6%
80.7%
79.1%
79.6%
77.4%
80.0%
78.2%
78.3%
72.1%
77.4%
76.8%
74.2%
77.0%
80.9%
80.6%
77.5%
73.5%
82.0%
76.9%
73.6%
81.3%
81.7%
79.9%
79.4%
78.3%
79.6%
80.9%
38.2%
46.8%
54.5%
48.9%
38.4%
51.1%
39.2%
49.0%
36.2%
41.0%
50.7%
41.2%
43.7%
46.3%
46.6%
39.3%
39.5%
54.9%
38.9%
35.0%
46.1%
58.1%
44.0%
47.0%
49.5%
50.4%
54.1%
62
2019 National Survey of Children’s Health
U.S. Census Bureau
| File Type | application/pdf |
| File Title | 2019 National Survey of Children’s Health |
| Subject | Methodology Report |
| Author | U.S. Census Bureau |
| File Modified | 2021-01-13 |
| File Created | 2020-09-29 |