Download:
pdf |
pdfPopulation Assessment of Tobacco and Health (PATH) Study (NIDA)
Attachment 21
PATH Study Interim Report
Population Assessment of
Tobacco and Health Study
Interim Report to the Office of
Management and Budget on Wave
1 and Wave 2 Data and
Biospecimen Collection
May 18, 2015
Submitted by:
Kevin P. Conway, Ph.D.
Deputy Director
Division of Epidemiology, Services, and Prevention Research
National Institute on Drug Abuse
6001 Executive Blvd., Room 5185
Rockville, MD 20852
Phone: 301-443-8755
Table of Contents
Section
1
Page
Introduction ........................................................................................................
1
1.1
1.2
1
2
Purpose of Interim Report (Terms of Clearance) ............................
Sample Design .......................................................................................
1.2.1
Overview of Sample Design for Wave 1
(Baseline Wave) .....................................................................
Overview of Sample Design for Wave 2...........................
2
3
Wave 1 ..................................................................................................................
5
2.1
Response Rates ......................................................................................
5
2.1.1
2.1.2
2.1.3
Household Screener .............................................................
Extended Interviews ............................................................
Biospecimen Collections .....................................................
5
8
12
Nonresponse Bias Analysis .................................................................
15
2.2.1
2.2.2
Method ...................................................................................
Results ....................................................................................
15
18
Statistical Approach for Addressing Nonresponse ..........................
30
2.3.1
2.3.2
Interviews...............................................................................
Biospecimens.........................................................................
30
41
Summary of Findings ...........................................................................
45
Wave 2 ..................................................................................................................
48
3.1
48
1.2.2
2
2.2
2.3
2.4
3
Response Rates ......................................................................................
3.1.1
3.1.2
3.1.3
Retention Rates for Continuing Adults and
Continuing Youth .................................................................
Recruitment Rates for Aged-up Adults and
Aged-up Youth .....................................................................
Biospecimen Collections .....................................................
iii
48
52
58
Contents (continued)
Section
Page
3.2
3.3
3.4
Nonresponse Bias Analysis .................................................................
Statistical Approach for Addressing Nonresponse ..........................
Summary of Findings ...........................................................................
60
65
66
Conclusions and Implications for the Study Going Forward ......................
69
References ...............................................................................................................................
72
4
Appendix
A
Cigarette Smoking Questions on the PATH Study and Other Surveys .....
A-1
2-1
PATH Study Wave 1 response rates by address characteristics:
Household Screener ...........................................................................................
8
2-2
PATH Study Wave 1 response rates by respondent
characteristics: Adult Extended Interview ......................................................
11
2-3
PATH Study Wave 1 response rates by respondent
characteristics: Youth Interview .......................................................................
12
2-4
PATH Study Wave 1 response rates by respondent
characteristics: Biospecimen collections..........................................................
14
2-5
Race by age distribution, based on the household enumeration:
IPS weights ..........................................................................................................
19
2-6
Distribution of male and female adults listed in the household
enumeration: IPS weights ..................................................................................
19
2-7
Distribution of household size based on households responding
to the Household Screener: IPS weights .........................................................
20
2-8
Distribution of number of adults based on households
responding to the Household Screener: IPS weights ....................................
20
Table
iv
Contents (continued)
Table
Page
2-9
Distribution of number of youth ages 12-17 based on households
responding to the Household Screener: IPS weights ....................................
20
2-10
Demographic distributions based on adults responding to the
Adult Extended Interview, and based on adults providing urine
or blood specimens: IPS weights .....................................................................
21
Comparison of education level and health insurance status based
on adults completing the Adult Extended Interview and based on
adults providing urine or blood specimens: IPS weights ..............................
22
2-12
Current cigarette smoking based on adults responding to the
Adult Extended Interview: IPS weights ..........................................................
26
2-13
Current cigarette smoking based on adults providing
biospecimens in Wave 1: IPS weights .............................................................
27
2-14
Demographic distributions based on youth ages 12-17 who
completed the Youth Interview: IPS weights .................................................
28
2-15
Cigarette smoking* based on youth ages 12-17 who completed
the Youth Interview: IPS weights ....................................................................
29
2-16
Race by age distribution, based on household enumeration ........................
34
2-17
Distribution of male and female adults listed in the household
enumeration .........................................................................................................
34
2-18
Distribution of household size based on households responding
to the Household Screener ................................................................................
35
2-19
Distribution of number of adults based on households
responding to the Household Screener ...........................................................
35
2-20
Distribution of number of youth ages 12-17 based on households
responding to the Household Screener ...........................................................
35
2-21
Demographic distributions based on adults responding to the
Adult Extended Interview .................................................................................
36
2-11
v
Contents (continued)
Table
Page
2-22
Comparison of education level and health insurance status based
on adults responding to the Adult Extended Interview ...............................
37
2-23
Current cigarette smoking based on adults responding to the
Adult Extended Interview .................................................................................
38
2-24
Demographic distributions based on youth ages 12-17 who
completed the Youth Interview........................................................................
39
2-25
Cigarette smoking* based on youth ages 12-17 who completed
the Youth Interview ...........................................................................................
40
2-26
Demographic distributions based on adults from whom urine or
blood specimens were collected .......................................................................
42
2-27
Comparison of education level and health insurance status based
on adults from whom urine or blood specimens were collected.................
43
2-28
Current cigarette smoking based on adults from whom
biospecimens were collected .............................................................................
44
2-29
Summary of PATH Study Wave 1 overall response rates ............................
45
3-1
Status of Wave 2 released cases from replicate group 1, as of
April 22, 2015 ......................................................................................................
49
3-2
PATH Study Wave 2 predicted retention rates by Wave 1
characteristics: Adult Interview (continuing adults) ......................................
53
3-3
PATH Study Wave 2 predicted retention rates by respondent
characteristics: Youth Interview (continuing youth) .....................................
54
3-4
PATH Study Wave 2 predicted recruitment rates by respondent
characteristics: Adult Interview (aged-up adults) ...........................................
56
3-5
PATH Study Wave 2 predicted recruitment rates by respondent
characteristics: Youth Interview (aged-up youth) ..........................................
57
3-6
PATH Study Wave 2 response rates by respondent
characteristics: Urine collection (continuing adults) ......................................
59
vi
Contents (continued)
Table
Page
3-7
PATH Study Wave 2 response rates by respondent
characteristics: Biospecimen collections (aged-up adults) ............................
60
3-8
Comparison of Wave 2 Continuing Adult Interview respondents
with finalized and provisional nonrespondents .............................................
62
3-9
Comparison of Wave 1 tobacco use rates for Wave 2 Continuing
Adult Interview respondents with finalized and provisional
nonrespondents...................................................................................................
63
3-10
Comparison of Wave 2 Aged-up Adult and Continuing Youth
Interview respondents with provisional nonrespondents ............................
64
3-11
Summary of PATH Study predicted response rates for Wave 2 .................
67
A-1
Question used to define “current smoking” in the PATH Study,
TUS-CPS, NHIS, NHANES, and NSDUH ..................................................
2
A-2
Questions used for youth cigarette smoking in the PATH Study,
NHANES, NSDUH, and NYTS .....................................................................
3
vii
Population Assessment of Tobacco and Health Study
Introduction
1
The National Institute on Drug Abuse (NIDA), in partnership with the Food and Drug
Administration (FDA), has prepared this report in response to the terms of clearance by the Office
of Management and Budget (OMB) in its approval of Wave 2 of the Population Assessment of
Tobacco and Health (PATH) Study.
1.1
Purpose of Interim Report (Terms of Clearance)
OMB approved the PATH Study’s Revision Request for Wave 2 on September 8, 2014 (0925-0664).
The terms of clearance of OMB's approval state: “Before submitting the information collection
request for Wave 3 to OMB, NIDA/FDA should report to OMB regarding: a) the response rates
associated with the full baseline wave, including screening, interview completion, and bio-specimen
response; b) Wave 2 retention, recruitment rates for the ‘age in to adult’ and ‘age in of shadow’
subsamples; c) the results of nonresponse analysis and statistical approach for addressing nonresponse, as well as implications for the study going forward; and d) the statistical approach to be
applied to the bio-specimen data to address potential non-response bias from lower consent and
cooperation rates with this aspect of the study.”
This report is organized in sections that correspond to OMB's terms of clearance: Section 2 presents
the Wave 1 response rates, results of a nonresponse analysis, and statistical approach for addressing
nonresponse. Section 3 presents the Wave 2 response rates (retention and recruitment rates), results
of a nonresponse analysis, and statistical approach for addressing nonresponse. Section 4
summarizes the findings and considers their implications.
The report covers the PATH Study’s Wave 1 and data collected from a probability subsample of
Wave 2 between October 23, 2014 and April 22, 2015. Response rates for Wave 1 and Wave 2 are
compared throughout this report with corresponding rates projected in the PATH Study’s respective
non-substantive change request to OMB for Wave 1 and revision request to OMB for Wave 2. For
Wave 1, the comparisons refer to the best-case and worst-case scenarios for the entire sample,
provided in “Attachment 22.” (“Attachment 22” is part of Supporting Statement B of the PATH
Study's non-substantive change request for Wave 1.)
1
Population Assessment of Tobacco and Health Study
1.2
Sample Design
This section provides an overview of the sample design for the PATH Study and a description of
replicate group 1 on which the interim report results are based for Wave 2. Information on the study
background and overall design is provided in Supporting Statement A of the PATH Study's Revision
Request for Wave 2.
1.2.1
Overview of Sample Design for Wave 1 (Baseline Wave)
The target population of the PATH Study is the civilian, non-institutionalized U.S. population
(excluding Puerto Rico) 12 years of age and older. Active duty military personnel and residents of
group quarters are excluded, with the exception of college students. For Wave 1, a four-stage
stratified area probability sample design was used, with a two-phase design for sampling the adult
cohort at the final stage. The sampling rates for adults varied by age, race, and tobacco use status. At
the first stage, a stratified sample of geographical primary sampling units (PSUs) was selected, in
which a PSU was a county or group of counties. For the second stage, within each selected PSU,
smaller geographical segments were formed and then a sample of these segments was drawn. At the
third stage, a sample of addresses was drawn from a sampling frame that consists of the residential
addresses in the U.S.; the main source of these addresses was the Postal Service (USPS)
Computerized Delivery Sequence Files (CDSFs).
The fourth stage selected persons from the sampled households. A roster of all the members in the
sampled household was constructed using the Household Screener. The Household Screener was
used to ask one adult household member (referred to as the household informant) to list members
of the household and provide demographic, and, for adults, tobacco use information about each for
use in sampling three main groups of interest:
Adults (up to two adults per household);
Children ages 12 to 17 (referred to as “youth,” generally up to two per household); and
Children ages 9 to 11 (referred to as “shadow youth,” generally up to two per
household) to be enrolled in the youth cohort in later waves of the study on reaching 12
years of age.
Two-phase sampling was used for adult selection due to potential misreporting of the tobacco use
status of other adult members of the household by the household informant. The Phase 1 sampling
2
Population Assessment of Tobacco and Health Study
depended on the age, race, and tobacco use information provided by the household informant. The
Phase 2 sampling was based on the sampled individual's self-reported age, race, and tobacco use
status, obtained by interviewing the individuals sampled at Phase 1. The sampling rates for the two
phases were designed to achieve large enough sample sizes for young adults (ages 18 to 24) and adult
tobacco users of all ages. The tobacco use status reported by the household informant is referred to
as “Phase 1 tobacco use status”; the self-reported tobacco use information obtained during Phase 2
screening is referred to as “Phase 2 tobacco use status.”
Because the full sample was selected using probability sampling methods, it is representative of the
U.S. civilian non-institutionalized population 12 years of age and older. The PATH Study Wave 1
sample was divided into four replicate groups, consisting of probability samples of approximately 20
percent, 30 percent, 30 percent, and 20 percent of the sampled segments, respectively, within each
sampled PSU. Each separate replicate group was a probability sample from the set of segments in
the frame and therefore also representative of the civilian non-institutionalized U.S. population. The
replicate groups were released to the field in a sequential manner (replicate group 1 in September
2013, replicate group 2 in November 2013, replicate group 3 in February and March 2014, and
replicate group 4 in May 2014).
1.2.2
Overview of Sample Design for Wave 2
The PATH Study is currently conducting Wave 2, which is the first follow-up wave for participants
in Wave 1. In addition, youth from the shadow youth sample established at Wave 1 who turn age 12
by Wave 2, and who are permitted by a parent or guardian to participate in the study, are asked for
assent to be interviewed for the first time at Wave 2. Similarly, persons in the youth sample at Wave
1 who reach age 18 by Wave 2 are asked to complete a Wave 1 version of the adult instrument, and
to provide urine and blood samples for the first time.
For Wave 2, the PATH Study is subsampling approximately 14,500 adults for urine collection from
adults who provided urine at Wave 1. Among these adults, at least 10,000 are expected to provide a
urine sample again at Wave 2. The Study is also collecting urine and blood samples from consenting
aged-up adults.
Results in this Interim Report for Wave 2 are based on data collected about persons in replicate
group 1 as of the cutoff date for the analyses on April 22, 2015. Wave 2 cases are released for follow
up in monthly groups with the goal of completing the Wave 2 interview as close as possible to the
3
Population Assessment of Tobacco and Health Study
one-year anniversary date of the Wave 1 interview. Cases are fielded approximately one month prior
to the anniversary date. If more than one Wave 1 respondent is from the same household, the cases
from that household are clustered for simultaneous release based on the date of the first completed
interview for a sampled adult or youth in the household. Thus, while the Wave 2 sample release is
not explicitly tied to the release of the Wave 1 replicate groups, it does correspond with the
distribution of completion dates for Wave 1 interviews. That is, a high percentage of cases in the
first Wave 2 release groups came from replicate group 1, and replicate group 1 has higher
percentages of released and finalized cases than the other replicate groups.
4
Population Assessment of Tobacco and Health Study
Wave 1
2
The PATH Study completed the Wave 1 data and biospecimen collection in December 2014. 1 This
section presents findings on the response rates for Wave 1, on the nonresponse analysis, and on the
Study’s statistical approach for addressing nonresponse.
2.1
Response Rates
This section summarizes the Wave 1 response rates for the Household Screener, Adult and Youth
Extended Interviews, and biospecimen collections.
2.1.1
Household Screener
The Household Screener combined typical screener functions (e.g., enumerating the household,
collecting demographic information about each member and some household-level data, and
selecting participants for the study) with a special purpose for the PATH Study, collecting basic
information on each adult’s tobacco use. 2 This allowed the Study to classify the adult with sufficient
validity for potential selection as a participant based on the PATH Study’s sampling strata on
tobacco use and demographic characteristics. Field interviewers conducted the Household Screener
in person using computer-assisted personal interviewing (CAPI). The Study released 167,525
addresses; of these, 141,509 Household Screener cases were finalized (i.e., were completed or were
finalized nonrespondents) during Wave 1.
1
A small number of blood collections for Wave 1 were completed in January 2015.
2
The Household Screener collected information on adult household members’ use of several different types of
tobacco products. For example, it collected information on current use of products with relatively high prevalence or
well-established use, such as cigarettes, cigars, and pipes; and on ever use of products with relatively low prevalence
or newly emerging use at the time when Wave 1 was designed, such as electronic cigarettes or e-cigarettes.
5
Population Assessment of Tobacco and Health Study
Method
The PATH Study computed the response rates presented in this report in accordance with the
response rate formula provided by OMB in its “Standards and Guidelines for Statistical Surveys”
(2006). This formula calls for calculating unweighted unit response rates (RRU) as the ratio of the
number of completed cases (or sufficient partials) to the number of in-scope sample cases. 3 The
different categories of cases that comprise the total number of in-scope cases are defined as follows:
C
R
NC
O
U
e
=
=
=
=
=
=
number of completed cases or sufficient partials;
number of refused cases;
number of noncontacted sample units known to be eligible;
number of eligible sample units not responding for reasons other than refusal;
number of sample units of unknown eligibility, not completed; and
estimated proportion of sample units of unknown eligibility that are eligible.
The unweighted unit response rate represents a composite of these components:
RRU=C/(C+R+NC+O+e(U))
In computing the response rates, refused cases, noncontacted sample units known to be eligible, and
eligible sample units not responding for reasons other than refusal were combined into nonresponse
cases (NR):
RRU=C/(C+NR+e(U))
The response rates were weighted to compensate for unequal probabilities of selection due to
planned oversampling of individuals with certain characteristics (i.e., young adults, African-American
adults, and adult tobacco users). Although the unweighted response rate measures operational
success, the weighted response rate is needed to generalize to the population. The Household
Screener inverse probability of selection (IPS) weights were calculated as the inverse of the selection
probabilities for the sampled households. (See Section 2.2 for additional information on weighting).
Note that the weights used for this report are preliminary; estimates calculated using the final
versions of the weights may differ slightly from the estimates presented in this report.
3
The sample does not have any partial completes.
6
Population Assessment of Tobacco and Health Study
Table 2-1 provides overall response rates for the Household Screener and response rates for
subgroups of sampled households that belong to Census block groups with various characteristics.
In addition to the overall row, the table includes rows on education, race, ethnicity, and poverty
status subgroups. For example, the weighted response rate for addresses in Census block groups
with “high” levels of education (>29.6% of persons ages 25 and older with Bachelor’s degrees) was
48.3 percent; it was 58.0 percent for addresses in Census block groups with “low” levels of
education. Comparing subgroups of responding and nonresponding households informed an
assessment of the extent to which the responding addresses represented all sampled addresses and,
ultimately, the population of inference. For this purpose, subgroups were defined by the
characteristics of the Census block groups in which the sampled addresses were located using
information from the 5-year (2009 to 2013) American Community Survey (ACS). 4 The “high” and
“low” subgroup categories were defined relative to the nationwide percentage of persons having the
characteristic: block groups with percentages below the national average for the characteristic were
classified as low, and those with percentages above the national average were classified as high.
Results
As indicated in Table 2-1, the weighted overall response rate for the Household Screener was 54.0
percent. The weighted response rates for subgroups defined by neighborhood characteristics
indicate the subgroups differed from one another by as much as 9.7 percentage points. The
differences among subgroups on weighted response rates were 9.7 percentage points for education,
2.7 percentage points for race, 1.1 percentage points for ethnicity, and 7.2 percentage points for
poverty status. The overall response rate for the Household Screener was lower than the projected
rate of 70 percent previously presented to OMB in Attachment 22, but it exceeded the worst-case
scenario response rate of 39.7 percent.
4
Information from the 5-year (2009 to 2013) rather than the 1-year (2013) ACS was used because 1-year ACS
estimates are not provided for smaller geographies such as Census tracts or block groups. The 5-year ACS estimates,
which are based on the accumulated sample from 2009 to 2013, are the only estimates from ACS that provide
information at the tract level and for smaller geographies (see
http://www.census.gov/acs/www/guidance_for_data_users/estimates/).
7
Population Assessment of Tobacco and Health Study
Table 2-1.
PATH Study Wave 1 response rates by address characteristics: Household Screener
Neighborhood characteristica
Overall
Education (% with Bachelor's degree)
High > 29.6%
Low <= 29.6%
Race (% Black alone or in
combination)
High > 13.8%
Low <= 13.8%
Ethnicity (% Hispanic)
High > 17.1%
Low <= 17.1%
Poverty Status
High > 15.8%
Low <= 15.8%
A:
Completed
(n)
79,198
B:
Finalized
nonresponseb
(n)
62,332
C:
Unknown
eligible
estimated
to be
eligiblec (n)
4,760
Unweighted
response
rate for
Wave 1d
(%)
54.1
Weighted
response
rate for
Wave 1 (%)
54.0
29,817
49,381
28,380
33,952
3,194
1,566
48.6
58.2
48.3
58.0
20,903
58,295
15,200
47,132
1,144
3,616
56.1
53.5
56.0
53.3
21,596
57,602
16,522
45,810
1,186
3,574
54.9
53.8
54.8
53.7
31,085
48,113
20,104
42,228
1,746
3,014
58.7
51.5
58.6
51.4
a
The information used to define the subgroups is from the 5-year (2009-2013) American Community Survey.
b
Finalized nonresponse includes refused cases, uncontacted cases known to be eligible, and other eligible cases not responding for
reasons other than refusal.
c
Product of cases of unknown eligibility and estimated proportion of cases of unknown eligibility that are eligible.
d
Response rate = A/(A+B+C).
2.1.2
Extended Interviews
The PATH Study Adult Extended Interview and Youth Extended Interview collected extensive selfreport information from the adult and youth sampled persons, respectively. The Adult Extended
Interview gathered information from adults (18 years old and older) about tobacco use behaviors,
attitudes, knowledge, and health conditions, as well as information on demographics, environmental
factors, family and peer influences, substance use, and general physical and mental health status.
Field interviewers conducted the Adult Extended Interviews in person using audio computerassisted self-interviewing (ACASI). In Wave 1, 59,088 Phase 2 Screeners were finalized, and 32,400
Adult Extended Interviews were finalized.
The Youth Interview gathered information from youth (12 to 17 years old) on topics similar to
those in the Adult Extended Interview. Sampled youth were asked about their tobacco use and
attitudes about tobacco; in addition, youth were asked for information on demographics,
8
Population Assessment of Tobacco and Health Study
environmental factors, family and peer influences, substance use, and mental health. Field
interviewers conducted the interviews in person using ACASI. The Study finalized 17,451 Youth
Extended Interviews in Wave 1.
Method
Wave 1 response rates depended on completion of the Phase 1 Household Screener. For the Adult
Extended Interview, the response rates were calculated as the product of (1) the Phase 2 or
Individual Screener5 response rate; and (2) the proportion of adults who completed the Adult
Extended Interview among those who completed the Phase 2 Screener and were selected for the
Adult Extended Interview:
RRU=(CP2/(CP2+NRP2))*(CE/(CE+NRE))
where
CP2
NRP2
CE
NRE
=
=
=
=
number of completed cases or sufficient partials for the Phase 2 Screener;
number of finalized nonresponse cases for the Phase 2 Screener;
number of completed cases or sufficient partials for the Extended Interview; and
number of finalized nonresponse cases for the Extended Interview.
All cases were eligible.
For the Youth Extended Interview, the unweighted unit response rate is as follows:
RRU=C/(C+NR),
where
C = number of completed cases or sufficient partials for the Extended Interview; and
NR = number of finalized nonresponse cases for the Extended Interview.
Again, all cases were eligible.
5
Adults selected on the basis of the Phase 1 Household Screener were asked to complete the Phase 2 or Individual
Screener. Those who completed the Phase 2 Screener were eligible for selection for the Adult Extended Interview,
subject to further subsampling to achieve the design targets for the various age, race, and tobacco use groups. Of the
adults who completed the Phase 2 Screener and were selected for the Adult Extended Interview, approximately 99.7
percent completed the Adult Extended Interview.
9
Population Assessment of Tobacco and Health Study
The adult and youth response rates were weighted to compensate for unequal probabilities of
selection due to planned oversampling of individuals with certain characteristics. Person-level
weights were computed as the product of the Household Screener IPS weights and individual IPS
weights, which were calculated as the inverse of the selection probabilities for all persons sampled
(responding and nonresponding persons). (See Section 2.2 for additional information on weighting.)
Tables 2-2 and 2-3 provide response rates for the Adult Extended Interview and Youth Extended
Interview, respectively. In addition to the overall row, these tables include rows on age, sex, race,
and ethnicity subgroups; Table 2-2 also includes a row on tobacco use status. 6 Information from the
Household Screener is used to define the demographic characteristics for the responding and
nonresponding persons. Among adults for whom the Household Screener had missing values for
information on tobacco use status, sampling at Phase 1 was based on the selection probabilities used
for tobacco users, shown in the “sampled as user” row of Table 2-2. Adults for whom the
Household Screener had missing values for other characteristics were excluded from the response
rate calculation for the particular characteristic.
Results
For the Adult Extended Interview, the weighted overall response rate was 74.0 percent. (See Table
2-2.) This overall rate was lower than the projected rate of 85 percent, but it exceeded the worst-case
scenario response rate of 58.1 percent previously provided to OMB in Attachment 22.
The weighted response rates for tobacco use status and demographic subgroups indicate the
subgroups differed from one another by as much as 10.5 percentage points. As noted, information
on the tobacco use status and demographic characteristics of eligible participants used in this table
was gathered in the Household Screener. The differences among subgroups on weighted response
rates were 3.7 percentage points for tobacco use status, 3.7 percentage points for age, 5.1 percentage
points for sex, 10.5 percentage point for race, and 2.3 percentage points for ethnicity.
For the Youth Extended Interview, the weighted overall response rate was 78.4 percent. (See
Table 2-3.) This overall rate was higher than the projected rate of 75 percent in Attachment 22. A
worst-case scenario response rate was not specified for the Youth Interview.
6
Tobacco use status is based on information obtained in the Household Screener.
10
Population Assessment of Tobacco and Health Study
Table 2-2.
PATH Study Wave 1 response rates by respondent characteristics: Adult Extended Interview
Characteristica
Overall
Tobacco use status
Sampled as user
Sampled as non-user
d
Age
18-24
25-44
45-64
65+
Sexd
Male
Female
Raced
White alone
Black alone or in combination
Other
Ethnicityd
Hispanic
Non-Hispanic
Phase 2 Screener
B:
A:
P2 Screener,
P2 Screener,
finalized
completed
nonresponseb
(n)
(n)
44,303
14,785
Adult Extended Interview
D:
C:
Adult Extended,
Adult Extended,
finalized
completed
nonresponseb
(n)
(n)
32,320
80
Unweighted
response rate
for Wave 1c
(%)
74.8
Weighted
response rate
for Wave 1
(%)
74.0
23,859
20,444
7,299
7,486
21,362
10,958
55
25
76.4
73.0
76.4
72.7
11,679
14,663
12,503
5,207
3,732
4,593
4,229
2,002
9,059
11,266
8,761
3,088
16
18
30
16
75.6
76.0
74.5
71.9
75.1
75.4
73.8
71.7
21,546
22,743
8,166
6,574
16,321
15,996
50
30
72.3
77.4
71.4
76.5
32,222
7,602
3,344
11,085
1,884
1,363
23,645
5,532
2,452
53
15
8
74.2
79.9
70.8
73.8
79.2
68.7
7,818
36,430
2,270
12,433
5,400
26,891
18
62
77.2
74.4
76.0
73.7
a
The characteristics are as sampled. That is, information on the characteristics was collected in the Household Screener.
b
Finalized nonresponse includes refused cases and other eligible cases not responding for reasons other than refusal.
c
Response rate = (A/(A+B))*(C/(C+D)).
d
The sum of counts for this category for Adult Extended, Completed (Column C) do not sum to the overall total due to missing values. The number of missing cases is 146 for age, 3 for sex,
691 for race, and 29 for ethnicity.
11
Population Assessment of Tobacco and Health Study
Table 2-3.
PATH Study Wave 1 response rates by respondent characteristics: Youth Interview
Characteristica
Overall
Aged
12-14
15-17
Sexd
Male
Female
Raced
White alone
Black alone or in combination
Other
Ethnicityd
Hispanic
Non-Hispanic
B:
Finalized
nonresponseb
(n)
3,800
A:
Completed
(n)
13,651
Unweighted
response rate for
Wave 1c
(%)
78.2
Weighted
response rate
for Wave 1
(%)
78.4
6,925
6,583
1,775
1,974
79.6
76.9
79.8
77.2
6,988
6,659
1,960
1,838
78.1
78.4
78.2
78.6
9,623
2,511
1,079
2,705
573
408
78.1
81.4
72.6
78.2
81.6
72.7
3,757
9,873
848
2,947
81.6
77.0
82.0
77.1
a
The characteristics are as sampled. That is, information on the characteristics was collected in the Household Screener.
b
Finalized nonresponse includes refused cases and other eligible cases not responding for reasons other than refusal.
c
Response rate = A/(A+B).
d
The sum of counts for this category for Completed (Column A) do not sum to the overall total due to missing values. The number of
missing cases is 143 for age, 4 for sex, 438 for race, and 21 for ethnicity.
The findings on the weighted response rates for demographic subgroups indicate the subgroups
differed from one another by as much as 8.9 percentage points. Information on the demographic
characteristics of eligible participants used in this table was gathered in the Household Screener.
Differences among subgroups on weighted response rates were 2.6 percentage points for age, 0.4
percentage points for sex, 8.9 percentage point for race, and 4.9 percentage points for ethnicity.
2.1.3
Biospecimen Collections
This section is on the method and response rates for the collection of urine and blood samples from
adults who completed Adult Extended Interviews. 7 Biospecimens are intended to provide a basis for
7
Buccal cells were collected from adult participants for approximately 8 of the 15 months of Wave 1 (through May 18,
2014, when OMB approved a change request to discontinue the collection of this biospecimen). The PATH Study
did not resume buccal cell collection in Wave 2; in addition, the Study has no plans to resume this collection in Wave
3.
12
Population Assessment of Tobacco and Health Study
the assessment of between-person differences and within-person changes in markers of tobacco
exposure, and to detect and compare indicators of conditions and related disease processes
associated with the use of tobacco products. Field interviewers collected the urine samples; on
separate visits, phlebotomists collected the blood samples. The 32,320 adults who completed the
Adult Extended Interview were eligible to provide biospecimens.
Method
Table 2-4 provides overall unweighted and weighted response rates for the biospecimen collections,
and response rates for tobacco use status and demographic subgroups. The response rates are
conditional on a completed Household Screener and a completed Adult Extended Interview. The
response rates were calculated using the following formula:
RRU = Number of samples collected/number of Adult Extended Interviews completed
This is the same formula used to compute the projected biospecimen response rates presented in
Attachment 22 for Wave 1. The denominator for the rate, the 32,320 adults who completed the
Adult Extended Interview, is the same for both of the biospecimen response rates.
In addition to the overall row, the table includes rows on tobacco use status, age, sex, race, and
ethnicity subgroups. Information from the Adult Extended Interview is used to define the tobacco
use status and demographic characteristics for the responding and nonresponding adults. Adults
with missing values for such characteristics were excluded from the response rate calculation for that
characteristic.
Results
The weighted response rate for urine was 63.6 percent; the projected response rate was 80 percent,
and the worst-case response rate was 49 percent. The differential weighted response rate for
subgroups of respondents ranges from 2.2 percentage points for sex to 11.5 percentage points for
age. The response rate for urine collection was lower than projected, but it exceeds the worst-case
scenario discussed in Attachment 22.
13
Population Assessment of Tobacco and Health Study
Table 2-4.
PATH Study Wave 1 response rates by respondent characteristics: Biospecimen collections
Characteristica
Overall
Tobacco use statusb
Sampled as user
Sampled as non-user
Ageb
18-24
25-44
45-64
65+
Sexb
Male
Female
Raceb
White alone
Black alone or in combination
Other
Ethnicityb
Hispanic
Non-Hispanic
A:
Adult Extended
Interviews
completed
(n)
32,320
Urine
B:
Collected
(n)
21,801
Unweighted
response rate
for Wave 1c
67.5
Biospecimen
Weighted
response
rate for
Wave 1
63.6
B:
Collected
(n)
14,520
Blood
Unweighted
response rate
for Wave 1c
44.9
Weighted
response rate
for Wave 1
43.0
23,084
9,234
16,116
5,683
69.8
61.5
68.9
60.3
10,764
3,754
46.6
40.7
46.4
40.8
9,110
11,269
8,818
3,110
6,457
7,742
5,725
1,873
70.9
68.7
64.9
60.2
69.0
65.8
62.2
57.5
3,884
5,004
4,191
1,438
42.6
44.4
47.5
46.2
41.4
41.0
45.4
43.9
16,309
15,982
10,763
11,025
66.0
69.0
62.4
64.6
6,920
7,594
42.4
47.5
41.1
44.6
23,242
5,538
2,726
15,531
3,911
1,801
66.8
70.6
66.1
63.2
66.6
61.4
10,637
2,381
1,152
45.8
43.0
42.3
44.2
40.3
38.4
5,536
26,288
3,870
17,633
69.9
67.1
66.4
63.1
2,452
11,885
44.3
45.2
40.5
43.7
a
The characteristics are as reported in the Adult Extended Interview.
b
The sum of counts for this category for Urine (Column B) do not sum to the overall total due to missing values. The number of missing cases is 2 for tobacco use, 4 for age, 13 for sex, 558
for race, and 298 for ethnicity. Also, the sum of counts for this category for Blood (Column B) do not sum to the overall total due to missing values. The number of missing cases is 2 for
tobacco use, 3 for age, 6 for sex, 350 for race, and 183 for ethnicity.
c
Response rate = B/A.
14
Population Assessment of Tobacco and Health Study
The weighted response rate for blood was 43.0 percent; the projected response rate was 65 percent,
and the worst-case response rate was 39 percent. The differential weighted response rate for
subgroups of respondents ranges from 3.2 percentage points for ethnicity to 5.8 percentage points
for race. The response rate for blood collection was lower than projected, but it exceeds the worstcase scenario in Attachment 22.
2.2
Nonresponse Bias Analysis
This nonresponse bias analysis investigates possible differences between estimates calculated from
Wave 1 of the PATH Study and independent estimates of those quantities from other surveys and
censuses. By so doing, the Study can assess the extent to which differential nonresponse among
population subgroups may affect estimates. Results are presented on the characteristics of
respondents to the Household Screener, Adult Extended Interview, and Youth Interview, and on
adults from whom biospecimens were collected for the PATH Study. The analysis of nonresponse
bias is based on the full set of Wave 1 data.
2.2.1
Method
The method used in the PATH Study to assess potential nonresponse bias begins by comparing
estimates of demographic counts from the Wave 1 sample 8 with corresponding estimates from the
American Community Survey (ACS). The 1-year (2013) ACS estimates, calculated from the 2013
ACS Public Use Microdata Sample (PUMS), 9 were used for comparison purposes. The estimates
calculated from the ACS PUMS excluded institutional group quarters and persons in
noninstitutional group quarters who were not college students. These exclusions correspond to
exclusions from the target population for the PATH Study.
8
The full Wave 1 sample was used for the analyses in this section; however, the weights used for this report are
preliminary. Estimates based on final releases of the Wave 1 PATH Study data may differ slightly from those in this
report.
9
The ACS PUMS files provide a probability sample of individual records from the full ACS (United States Census
Bureau, 2015), allowing data users to create custom tables that are not available through pre-tabulated ACS data
products. Using the PUMS files allowed comparison estimates to be calculated that corresponded to the target
population for the PATH Study.
15
Population Assessment of Tobacco and Health Study
The PATH Study measures a range of tobacco use behaviors; many of these variables are not
available in other studies. However, responses to the PATH Study questions on current cigarette
smoking can be compared with estimates from other surveys that ask about cigarette smoking
behavior. The following surveys were used for comparison: the Tobacco Use Supplement to the
Current Population Survey, 2010-2011 (TUS-CPS); the National Health and Nutrition Examination
Survey, 2011-2012 (NHANES); the National Health Interview Survey, 2013 (NHIS); the National
Survey on Drug Use and Health, 2013 (NSDUH); and the National Youth Tobacco Survey, 2012
(NYTS). Appendix A describes the questions used to define current smoking on each of these
surveys as well as the PATH Study, and outlines differences in target populations among these
surveys and the PATH Study.
The PATH Study oversamples young adults, African-American adults, and adult tobacco users.
Consequently, unweighted estimates of population quantities would be expected to be biased. In this
section, the inverse-probability-of-selection (IPS) weights, calculated using the probabilities of
selection, are used to estimate population quantities. Without nonresponse, estimates calculated
using the IPS weights would be expected to accord with the population counts.
The IPS weights were calculated in two stages. First, the household-level IPS weights (HHIPSWT)
were calculated for all households sampled (responding households and nonresponding households)
as the inverse of the selection probability:
𝐻𝐻𝐻𝐻𝐻𝐻𝐻𝐻𝐻𝐻𝐻𝐻𝑇𝑇𝑖𝑖𝑖𝑖𝑖𝑖 =
1
.
𝑃𝑃𝑖𝑖𝑖𝑖𝑖𝑖
Here, 𝑃𝑃𝑖𝑖𝑖𝑖𝑖𝑖 is the probability that household 𝑘𝑘 in segment 𝑗𝑗 of PSU 𝑖𝑖 is selected to be in the sample.
Addresses were sampled directly from the USPS CDSF, so that Pijk is the product of the PSU, the
segment-within-PSU, and the address-within-segment selection probabilities.
For nonresponse bias assessment purposes, the person-level IPS weights were computed using
HHIPSWT. For youth ages 12-17, these were calculated as
𝑌𝑌𝑌𝑌𝑌𝑌𝑌𝑌𝑌𝑌𝑇𝑇𝑖𝑖𝑖𝑖𝑖𝑖𝑖𝑖 = 𝐻𝐻𝐻𝐻𝐻𝐻𝐻𝐻𝐻𝐻𝐻𝐻𝑇𝑇𝑖𝑖𝑖𝑖𝑖𝑖 ×
1
.
Probability youth 𝑙𝑙 in household (𝑖𝑖𝑖𝑖𝑖𝑖)selected for sample
16
Population Assessment of Tobacco and Health Study
Most selected households had fewer than 3 youths, who were then selected with certainty, so that
for most households, the youth IPS weight (YIPSWT) is the same as the household-level IPS
weight.
Adults were selected with different probabilities according to their age, race, and tobacco use status.
The adult IPS weights (AIPSWT) were calculated as
𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝑇𝑇𝑖𝑖𝑖𝑖𝑖𝑖𝑖𝑖 = 𝐻𝐻𝐻𝐻𝐻𝐻𝐻𝐻𝐻𝐻𝐻𝐻𝑇𝑇𝑖𝑖𝑖𝑖𝑖𝑖 ×
1
.
Probability adult 𝑙𝑙 in household (𝑖𝑖𝑖𝑖𝑖𝑖)selected for sample
Sampling of adults occurred in two phases. Phase 1 selected adults based on responses to the
Household Screener. The probability that adult 𝑙𝑙 in the household was selected in Phase 1 was a
function of the number of adults in the household and of the ages, races, and tobacco use statuses
reported for those adults by the informant completing the Household Screener. Adults sampled at
Phase 1 were individually asked questions about their age, race, and tobacco use as part of the Phase
2 Screener, and were subsampled for continuation into the Adult Extended Interview on the basis of
their responses to those questions. The probability in the formula for AIPSWT is the product of the
first-phase and second-phase selection probabilities.
Note that no nonresponse adjustments were performed for the calculation of IPS weights. The
weights HHIPSWT, YIPSWT, and AIPSWT were used for all calculations employing IPS weights
reported in Section 2.2.2. In the tables presented in Section 2.2.2, the unweighted counts include
categories for missing values. Estimates of percentages calculated using weights, however, exclude
respondents with missing values for that item. The estimates calculated from other surveys that are
used for comparison purposes also exclude missing values, except where noted.
Point estimates for the PATH Study were calculated using HHIPSWT for household estimates,
AIPSWT for adult estimates, and YIPSWT for youth estimates. Replicate weights, used to calculate
variances, account for the complex sampling features of stratification and clustering. Precisions for
the estimates are reported using 95 percent confidence intervals. Estimates from TUS-CPS,
NHANES, NHIS, NSDUH, and NYTS 10 also have sampling error, so 95 percent confidence
10
Estimates from TUS-CPS were obtained from US Department of Commerce (2012) published tables. Estimates from
NHANES, NHIS, and NYTS were calculated from their respective public use files (CDC, 2014a, 2014b, 2015b).
Estimates from NSDUH were obtained from CBHSQ (2014), SAMHSA (2014), and from the Survey
Documentation and Analysis system at http://www.icpsr.umich.edu/cgibin/SDA/SAMHDA/hsda?samhda+35509-0001.
17
Population Assessment of Tobacco and Health Study
intervals are reported for the estimates from those surveys as well. 11 SAS software version 9.3 (SAS
Institute, 2011) was used to calculate all point estimates and confidence intervals.
2.2.2
Results
The first set of tables provides estimates derived from the Household Screener. Demographic
characteristics were estimated using the information on the roster of household members provided
by the household respondent. The household-level IPS weight HHIPSWT was used in Tables 2-5
through 2-9 to evaluate potential nonresponse bias. If nonresponse is not associated with
demographic characteristics, then the percentages calculated using HHIPSWT would be expected to
be close to those from the ACS. Note that the ACS estimates are for calendar year 2013 while the
PATH Study Wave 1 data were collected between September, 2013 and December, 2014. There is
therefore the possibility that differences for some rapidly changing characteristics may reflect the
later data collection for the PATH Study.
Table 2-5 presents unweighted counts and estimated population percentages for adults in the four
race/age domains used for sampling adults within households based on the enumeration of adults
from the Household Screener. Comparisons of the PATH Study estimates to the ACS estimates for
these four domains are shown in Table 2-5. The IPS-weighted estimates of percentages in the Black
18-24 and Black 25+ domains are similar to the 1-year 2013 ACS estimates. The PATH Study
estimate of the percentage in the non-Black 18-24 domain is 1.3 percentage points higher than the
corresponding ACS estimate, and the PATH Study estimate in the non-Black 25+ domain is 2.6
percentage points lower than the ACS estimate.
11
If the confidence interval from the PATH Study estimate does not overlap with the confidence interval from the
comparison study, then the results are significantly different at the 0.05 significance level. Schenker and Gentleman
(2001) show that this results in a conservative test. Confidence intervals are not reported for the ACS results because
they are based on such large sample sizes that the sampling error is negligible. For comparison with ACS results, the
PATH Study estimate significantly differs from the ACS estimate at the 0.05 significance level if the ACS percentage
is outside of the PATH Study confidence interval. In general, if a 95 percent confidence interval for percentage of
adults who are current smokers from the PATH Study includes a fixed value X, then a hypothesis test of the null
hypothesis that the percentage of adults who are current smokers equals X would have p-value > 0.05 and therefore
the difference between the PATH Study estimate and the estimate from the external survey is not statistically
significant.
18
Population Assessment of Tobacco and Health Study
Table 2-5.
Race by age distribution, based on the household enumeration: IPS weights
Race and age classification
Black* 18-24
Black* 25+
Non-Black 18-24
Non-Black 25+
Missing age or race
Total
Unweighted
count
3,669
17,934
18,058
116,957
4,003
160,621
Weighted
percentage,
using household
IPS weights
2.3%
11.4%
11.4%
74.9%
Confidence interval
using household IPS
weights
[2.0%, 2.6%]
[9.8%, 12.9%]
[10.8%, 12.0%]
[73.3%, 76.5%]
100.0%
Percentage from
ACS PUMS
2.0%
10.5%
10.1%
77.5%
100.0%
*Black alone or in combination with other race(s).
Table 2-6 compares the sex of the adults enumerated on the PATH Study household rosters with
the 1-year 2013 ACS distribution. The confidence intervals from the PATH Study include the ACS
percentages, indicating no evidence of nonresponse bias with respect to sex.
Table 2-6.
Sex
Male
Female
Missing
Total
Distribution of male and female adults listed in the household enumeration: IPS
weights
Unweighted
count
77,088
83,500
33
160,621
Weighted percentage for
adults, using household
IPS weights
48.0%
52.0%
100.0%
Confidence interval using
household IPS weights
[47.8%, 48.3%]
[51.7%, 52.2%]
Percentage from
ACS PUMS
48.2%
51.8%
100.0%
Table 2-7 compares the distribution of household size for the responding households with
corresponding estimates from the 1-year 2013 ACS. Compared with the ACS, the PATH Study
estimates fewer single- and two-person households and more households with at least three persons.
The PATH Study also estimates a lower percentage of single-adult households (Table 2-8) and,
probably related to this pattern, a slightly higher estimated percentage of households with youth ages
12-17 than found in the ACS (Table 2-9). 12 If no further weighting adjustments were performed
then to the extent that household size is associated with the PATH Study’s outcomes, those
outcomes may be affected by nonresponse bias. However, this concern is addressed by the
weighting adjustments and results described in Section 2.3.
12
Surveys commonly achieve a slightly lower percentage of one-person households because they have fewer members
available for contact. See Brault (2013), who found a similar pattern in the CPS ASEC content test.
19
Population Assessment of Tobacco and Health Study
Table 2-7.
Distribution of household size based on households responding to the Household
Screener: IPS weights
Number of
persons in
household
1
2
3
4
5+
Total
Table 2-8.
Unweighted count
17,822
26,098
13,503
11,720
10,055
79,198
Weighted
percentage,
using household
IPS weights
22.7%
33.0%
17.0%
14.7%
12.6%
100.0%
Confidence interval
using household IPS
weights
[22.1%, 23.4%]
[32.5%, 33.5%]
[16.7%, 17.3%]
[14.3%, 15.1%]
[12.0%, 13.2%]
Percentage from ACS
PUMS
27.7%
33.7%
15.7%
13.0%
9.9%
100.0%
Distribution of number of adults based on households responding to the Household
Screener: IPS weights
Number of adults in
household
0-1*
2
3+
Total
Unweighted count
21,869
40,961
16,368
79,198
Weighted
percentage,
using household
IPS weights
27.8%
51.7%
20.5%
100.0%
Confidence interval
using household
IPS weights
[27.1%, 28.5%]
[51.2%, 52.2%]
[19.8%, 21.1%]
Percentage from
ACS PUMS
33.3%
50.7%
16.0%
100.0%
*A small number of households contain only emancipated youth and hence contribute to the zero part of this category.
Table 2-9.
Distribution of number of youth ages 12-17 based on households responding to the
Household Screener: IPS weights
Number of youth
ages
12-17 in household
0
1
2+
Total
Unweighted count
65,661
9,569
3,968
79,198
Weighted
percentage,
using household
IPS weights
83.0%
12.0%
5.0%
100.0%
Confidence interval
using household
IPS weights
[82.4%, 83.5%]
[11.7%, 12.4%]
[4.7%, 5.2%]
Percentage from
ACS PUMS
84.4%
11.1%
4.5%
100.0%
Tables 2-10 and 2-11 are based on adults in the Wave 1 sample responding to the Adult Extended
Interview. As noted above, the PATH Study oversamples young adults, African-American adults,
and adult tobacco users, so estimates calculated without weights will not accord with population
estimates. The IPS-weighted estimates are calculated using the adult weight AIPSWT; if the PATH
Study had full response, it would be expected that the IPS-weighted estimates would be close to the
corresponding population quantities. Table 2-10 presents the estimated race, ethnicity, and sex/age
distributions from adults in the Wave 1 sample responding to the Adult Extended Interview.
20
Population Assessment of Tobacco and Health Study
Table 2-10.
Demographic distributions based on adults responding to the Adult Extended Interview, and based on adults providing
urine or blood specimens: IPS weights
Adult respondents to
Adult Extended Interview
Weighted
Confidence
Unpercentage,
interval using
weighted
using adult
adult IPS
count
IPS weights
weights
Sex
Male
Female
Missing
Total
16,309
15,982
29
32,320
Age group
18-24
25-44
45-64
65+
Missing
Total
9,110
11,280
8,818
3,110
2
32,320
100.0%
45.7%
54.3%
[44.9%, 46.6%]
[53.4%, 55.1%]
100.0%
13.8%
35.8%
33.7%
16.7%
[13.2%, 14.4%]
[35.0%, 36.6%]
[32.8%, 34.5%]
[16.0%, 17.4%]
Adults from whom
urine specimen is collected
Weighted
Confidence
Unpercentage,
interval using
weighted
using adult
adult IPS
count
IPS weights
weights
10,763
11,025
13
21,801
44.9%
55.1%
[43.9%, 45.9%]
[54.1%, 56.1%]
100.0%
6,457
7,744
5,725
1,873
2
21,801
14.9%
37.0%
33.0%
15.1%
100.0%
[14.2%, 15.6%]
[36.0%, 38.0%]
[32.0%, 33.9%]
[14.3%, 15.9%]
Adults from whom
blood specimen is collected
Weighted
Confidence
Unpercentage, interval using
weighted
using adult
adult IPS
count
IPS weights
weights
6,920
7,594
6
14,520
43.7%
56.3%
[42.6%, 44.8%]
[55.2%, 57.4%]
100.0%
3,884
5,005
4,191
1,438
2
14,520
13.3%
34.1%
35.6%
17.1%
100.0%
ACS PUMS
Weighted
percentage
48.1%
51.9%
100.0%
[12.5%, 14.0%]
[33.1%, 35.2%]
[34.4%, 36.7%]
[16.1%, 18.1%]
13.0%
34.3%
34.5%
18.2%
100.0%
Race
Black
alone or in
combination
White alone
Other
Missing
Total
5,538
14.1%
[12.1%, 16.1%]
3,911
14.8%
[12.6%, 17.0%]
2,381
13.2%
[11.2%, 15.2%]
12.5%
23,242
2,726
814
32,320
77.4%
8.5%
[75.1%, 79.7%]
[7.5%, 9.5%]
15,531
1,801
558
21,801
77.0%
8.2%
[74.6%, 79.4%]
[7.2%, 9.2%]
10,637
1,152
350
14,520
79.3%
7.5%
[77.1%, 81.5%]
[6.7%, 8.4%]
75.7%
11.8%
Ethnicity
Hispanic
Non-Hispanic
Missing
Total
5,536
26,288
496
32,320
100.0%
17.8%
82.2%
100.0%
[15.3%, 20.3%]
[79.7%, 84.7%]
100.0%
3,870
17,633
298
21,801
18.6%
81.4%
100.0%
21
[16.0%, 21.2%]
[78.8%, 84.0%]
2,452
11,885
183
14,520
100.0%
16.7%
83.3%
100.0%
100.0%
[13.9%, 19.5%]
[80.5%, 86.1%]
15.0%
85.0%
100.0%
Population Assessment of Tobacco and Health Study
Table 2-11.
Comparison of education level and health insurance status based on adults completing the Adult Extended Interview and
based on adults providing urine or blood specimens: IPS weights
Adult respondents to
Adult Extended Interview
Weighted
Confidence
Unpercentage,
interval using
weighted
using adult
adult IPS
count
IPS weights
weights
Adults from whom
urine specimen collected
Weighted
Confidence
Unpercentage,
interval using
weighted using adult
adult IPS
count
IPS weights
weights
Adults from whom
blood specimen collected
Weighted
Confidence
Unpercentage,
interval using
weighted using adult
adult IPS
count
IPS weights
weights
ACS PUMS
Weighted
percentage
Education
18-24
< HS, HS or GED
4,465
49.9%
[48.1%, 51.7%]
3,234
50.4%
[48.3%, 52.5%]
1,948
50.2%
[47.7%, 52.6%]
> HS
4,593
50.1%
[48.3%, 51.9%]
3,197
49.6%
[47.5%, 51.7%]
1,928
49.8%
[47.4%, 52.3%]
Subtotal
9,058
100.0%
6,431
100.0%
3,876
100.0%
43.6%
56.4%
100.0%
25+
< HS or GED
4,769
18.0%
[16.9%, 19.1%]
3,425
19.3%
[18.1%, 20.6%]
2,405
19.0%
[17.6%, 20.5%]
HS
4,763
20.0%
[19.1%, 21.0%]
3,106
19.8%
[18.6%, 20.9%]
2,071
19.0%
[17.7%, 20.2%]
23.8%
Some college,
no degree
Bachelor degree
7,650
31.3%
[30.2%, 32.3%]
5,238
32.4%
[31.3%, 33.4%]
3,721
33.4%
[31.9%, 34.8%]
29.2%
3,642
18.2%
[17.1%, 19.2%]
2,139
16.7%
[15.6%, 17.9%]
1,454
16.7%
[15.3%, 18.1%]
18.7%
> Bachelor degree
2,228
12.5%
[11.5%, 13.5%]
1,369
11.8%
[10.8%, 12.8%]
956
11.9%
[10.8%, 13.1%]
23,052
100.0%
15,277
100.0%
10,607
100.0%
Subtotal
Missing
Total
210
93
37
32,320
21,799
14,520
16.9%
11.3%
100.0%
Health insurance
Yes
No
Missing
Total
25,760
84.4%
[83.5%, 85.2%]
17,306
83.9%
[83.0%, 84.8%]
11,686
85.0%
[84.0%, 86.0%]
83.3%
6,156
15.6%
[14.8%, 16.5%]
4,335
16.1%
[15.2%, 17.0%]
2,758
15.0%
[14.0%, 16.0%]
16.7%
404
32,320
160
100.0%
21,799
76
100.0%
22
14,520
100.0%
100.0%
Population Assessment of Tobacco and Health Study
Additional columns in the table present the weighted distributions, using weight AIPSWT, for the
adults from whom urine and/or blood specimens were collected.
When using unweighted counts, males comprise more than 50 percent of the 32,320 PATH Study
respondents in Wave 1. By contrast, the IPS-weighted estimated percent of males in the population
at Wave 1 is significantly lower than the ACS estimate. This is also true for the subset of
respondents to the Adult Extended Interview who provided blood and/or urine specimens. Age
group comparisons show that the IPS-weighted estimates of the percent of adults who are ages 1844 based on the PATH Study respondents to the Adult Extended Interview, and based on the adults
providing urine specimens, are significantly higher than the ACS estimate. The nonresponseadjusted weights in Section 2.3, which calibrate to age groups and to sex, correct for these
discrepancies for the Adult Extended Interview respondents.
Table 2-10 shows that the estimated percentages in different race and ethnicity groups, calculated
using adults responding to the Adult Extended Interview, or using those who provide blood
specimens, are similar to the 1-year 2013 ACS estimates of those quantities, with the exceptions that
persons in the “other race” category are underrepresented among respondents to the Adult
Extended Interview and persons providing urine and/or blood specimens, with corresponding slight
overrepresentation of Whites among the persons providing blood specimens and of Blacks among
the persons providing urine specimens. Hispanics are slightly overrepresented among respondents to
the Adult Extended Interview and persons providing urine specimens.
Table 2-11 compares Adult Extended Interview respondents and those from whom biospecimens
were collected to the ACS with respect to education level and presence of health insurance. The
estimated distributions of educational level from adults responding to the Adult Extended Interview
and from adults providing biospecimens differ from that in the ACS. For adults responding to the
interview and for adults providing biospecimens, the estimated percentages of 18-24 year-olds with
at least some college are lower than the ACS; among older adults, the estimated percentages of
persons with a high school degree are lower than the ACS and the estimated percentages of persons
with some college but no degree are higher than the ACS. In addition, among the adults providing
biospecimens the estimated percentages of adults ages 25+ with less than a high school education
are higher than the ACS and the estimated percentages of adults ages 25+ with a bachelor’s degree
are lower than the ACS. Education level has been shown to be associated with tobacco use status
(Agaku et al., 2014); the nonresponse-adjusted weights described in Section 2.3 adjust for
educational attainment. Estimates of percentage of adults with health insurance based on
respondents to the Adult Extended Interview and on those who provided blood samples are higher
23
Population Assessment of Tobacco and Health Study
than the corresponding estimate from the 2013 ACS, although no significant difference was found
between the estimate from the PATH Study respondents providing urine samples and the estimate
from the ACS. It should be noted, however, that the ACS estimates do not account for persons who
obtained insurance in 2014 under the Affordable Care Act.
Table 2-12 presents the estimates of prevalence of current cigarette smoking 13 for adults based on
the Adult Extended Interview, for the adult population as a whole and for subgroups. These
estimates are accompanied by 95 percent confidence intervals for the percentage of current cigarette
smokers for the PATH Study estimates. The last five columns are the estimates of smoking
prevalence from TUS-CPS, NHIS, NHANES, and NSDUH, respectively, along with 95 percent
confidence intervals from those surveys. Note that these estimates exclude responses of “don’t
know” and missing values.
The estimates of current smoking prevalence differ substantially from survey to survey. Many
potential reasons can explain these disparities, including that each survey has sampling error. Beyond
that, however, the surveys differ in question order, context, design, mode of administration, and year
of most recent data collection.
In general, the TUS-CPS estimates of smoking prevalence are lower than estimates from the other
surveys, including the PATH Study. This may be related to the proxy responses used in the TUSCPS. The rotation group structure of the TUS-CPS may result in underestimates of smoking
prevalence, as smokers are more likely to drop out over the course of the panel survey (Song, 2013).
The PATH Study and NSDUH both use ACASI administration for the tobacco use questions so
that the interviewer does not see responses to the questions. By contrast, TUS-CPS, NHIS, and
NHANES have direct questioning by an interviewer: NHIS and NHANES are conducted in person,
and TUS-CPS is conducted in person and by telephone. The contexts and purposes of these surveys
also differ: CPS is a general survey on unemployment, NHIS and NHANES are general health
surveys, and NSDUH is a cross-sectional survey on substance use (including tobacco use) and
health, including mental health. Unlike the cross-sectional prevalence surveys, the PATH Study is
designed for research purposes and uses a longitudinal cohort design to assess within-person change
13
For the PATH Study, following common practice for tobacco surveys, a current smoker is someone who (1) has
smoked at least 100 cigarettes in his or her lifetime and (2) currently smokes every day or some days. The questions
used to define current smoking for each survey are provided in Appendix A.
24
Population Assessment of Tobacco and Health Study
and between-person differences in tobacco use behaviors and health over time. Other differences
among the questions used in the instruments of these different studies are outlined in Appendix A.
Table 2-12 indicates the IPS-weighted estimates of current smoking from the PATH Study are most
similar to estimates from NHIS and NHANES. All of the 95 percent confidence intervals for
percent of current cigarette smokers constructed from the PATH Study overlap with the confidence
intervals for NHIS, NHANES, or both. Estimates from TUS-CPS tend to be below the estimates
from the PATH Study, NHIS, and NHANES; estimates from NSDUH tend to be above the
estimates from the PATH Study, NHIS, and NHANES. No evidence was found to indicate
nonresponse bias in the PATH Study with respect to cigarette smoking behavior among adults,
because the PATH Study’s estimates are all within the range of estimates from comparable surveys.
Table 2-13 gives estimates of current cigarette smoking for the adults from whom urine and/or
blood specimens were collected. The IPS-weighted estimates of smoking are higher for adults who
contributed biospecimens. However, the confidence intervals for smoking among adults providing
biospecimens are in line with the estimates from NSDUH and NHANES.
Results in Tables 2-10 through 2-13 are based on adults in the full Wave 1 sample responding to the
Adult Extended Interview. Similar analyses were performed for the youth respondents. The
demographic estimates are given in Table 2-14 and estimates of two measures of cigarette smoking
commonly applied to youth are given in Table 2-15.
Table 2-14 shows that the IPS-weighted estimates of percentages of youth who are male/female and
ages 12-13/14-17 are not significantly different from the 1-year 2013 ACS percentages. The PATH
Study estimate of the percent of youth who are Hispanic, however, is approximately 6 percentage
points higher than the corresponding estimate from ACS, indicating that Hispanic youth were more
likely to respond to the PATH Study survey.
25
Table 2-12.
Population Assessment of Tobacco and Health Study
Current cigarette smoking based on adults responding to the Adult Extended Interview: IPS weights
Sample
size
32,245
PATH Study:
Unweighted
percentage
35.4%
Current smoker,
male
Current smoker,
female
Current smoker,
age 18-24
Current smoker,
age 25-44
Current smoker,
age 45-64
Current smoker,
age 65+
Current smoker,
Hispanic
Current smoker,
white non-Hispanic
Current smoker,
other non-Hispanic
Current every-day smoker
16,265
36.8%
15,952
33.9%
9,099
27.2%
11,260
41.2%
8,784
40.6%
3,100
23.3%
5,519
26.3%
19,268
38.8%
6,904
33.3%
32,245
28.0%
Current some-days smoker
32,245
7.4%
Current smoker
PATH Study: Weighted
percentage,
using adult IPS
weights
[95% confidence
interval]
18.6%
[17.7%, 19.4%]
21.2%
[20.3%, 22.2%]
16.4%
[15.4%, 17.3%]
20.1%
[18.8%, 21.5%]
22.9%
[21.8%, 24.1%]
18.7%
[17.7%, 19.8%]
7.7%
[7.0%, 8.4%]
13.4%
[12.6%, 14.2%]
19.6%
[18.5%, 20.7%]
20.4%
[19.3%, 21.5%]
14.8%
[13.9%, 15.6%]
3.8%
[3.6%, 4.1%]
Percentage
from 20102011 TUS-CPS
[95%
confidence
interval]
16.1%
[15.8%, 16.3%]
18.0%
[17.7%, 18.4%]
14.2%
[13.9%, 14.5%]
17.1%
[16.4%, 17.8%]
17.9%
[17.5%, 18.4%]
17.8%
[17.4%, 18.2%]
7.8%
[7.5%, 8.2%]
10.9%
[10.4%, 11.5%]
17.5%
[17.2%, 17.8%]
NA
12.7%
[12.4%, 12.9%]
3.4%
[3.3%, 3.5%]
Percentage
from 2013
NHIS
[95%
confidence
interval]
17.8%
[17.2%, 18.4%]
20.5%
[19.5%, 21.4%]
15.3%
[14.6%, 16.0%]
18.7%
[16.8%, 20.5%]
20.1%
[19.1%, 21.1%]
20.0%
[19.0%, 20.8%]
8.8%
[8.0%, 9.6%]
12.1%
[11.0%, 13.2%]
19.4%
[18.5%, 20.2%]
16.7%
[15.6%, 17.8%]
13.7%
[13.1%, 14.2%]
4.1%
[3.8%, 4.4%]
Percentage
from 20112012 NHANES*
[95%
confidence
interval]
19.8%
[17.5%, 22.1%]
23.9%
[20.7%, 27.1%]
16.0%
[13.5%, 18.5%]
20.4%***
[13.7%, 27.1%]
23.3%
[20.0%, 26.7%]
21.3%
[18.3%, 24.2%]
9.2%
[6.7%, 11.7%]
16.6%
[13.7%, 19.5%]
20.2%
[17.0%, 23.3%]
20.8%
[16.6%, 24.9%]
16.4%
[14.3%, 18.4%]
3.4%
[2.7%, 4.1%]
Percentage from
2013 NSDUH,
original
definition**
[95% confidence
interval]
22.9%
[22.6%, 23.2%]
25.6%
[25.1%, 26.1%]
20.4%
[20.0%, 20.8%]
NA****
Percentage from
2013 NSDUH,
modified
definition
[95% confidence
interval]
21.0%
[20.4%, 21.7%]
23.8%
[22.7%, 24.9%]
18.4%
[17.6%, 19.3%]
NA
NA
NA
NA
NA
NA
NA
18.9%
[18.1%, 19.7%]
24.1%
[23.7%, 24.5%]
21.8%
[20.2%, 23.4%]
NA
15.2%
[13.8%, 16.7%]
22.8%
[21.9%, 23.6%]
19.6%
[18.1%, 21.2%]
NA
NA
NA
*The smoking questions asked in NHANES for adults ages 20 and older differ from the questions asked for persons ages 12-19. The modes of administration also differ for the two age groups. The
NHANES estimates presented in this table are for adults ages 20 and older.
**NSDUH’s definition of a current cigarette smoker is someone who has smoked part or all of a cigarette in the past 30 days, which is more expansive than the definition used in the other surveys.
However, NSDUH contains questions on lifetime smoking and current smoking. The modified definition uses these questions to construct a measure of “current smoking” that is comparable to that of
the other surveys (Ryan et al., 2012). The construction of this variable is described in Appendix A. The estimates and confidence intervals for the NSDUH “original definition” (except for the “current
smoker, other non-Hispanic” estimate) are from the published tables (SAMHSA, 2014); the estimates and confidence intervals for the “modified definition” are calculated from the public use data
set. The estimate of current smoking for the “other non-Hispanic” group was not available from the published tables and it was also calculated from the public use data set.
***The estimate is for adults 20-24 years old.
****Detailed age information was not available in the public use file for NSDUH 2013.
26
Population Assessment of Tobacco and Health Study
Table 2-13.
Current cigarette smoking based on adults providing biospecimens in Wave 1: IPS weights
Adult respondent to
Adult Extended
Interview
Adults providing urine
Adults providing blood
Sample
size
32,245
21,757
14,493
PATH Study:
Weighted cigarette
smoking
prevalence,
using adult IPS
weights
[95% confidence
interval]
18.6%
[17.7%, 19.4%]
Percentage from
2010-2011 TUSCPS
[95% confidence
interval]
16.1%
[15.8%, 16.3%]
20.9%
[20.0%, 21.9%]
21.6%
[20.5%, 22.7%]
16.1%
[15.8%, 16.3%]
16.1%
[15.8%, 16.3%]
Percentage from
2013 NHIS
[95% confidence
interval]
17.8%
[17.2%, 18.4%]
Percentage from
2011-2012
NHANES
[95% confidence
interval]
19.8%
[17.5%, 22.1%]
Percentage from
2013 NSDUH,
original
definition*
[95% confidence
interval]
22.9%
[22.6%, 23.2%]
Percentage from
2013 NSDUH,
modified
definition
[95% confidence
interval]
21.0%
[20.4%, 21.7%]
17.8%
[17.2%, 18.4%]
17.8%
[17.2%, 18.4%]
19.8%
[17.5%, 22.1%]
19.8%
[17.5%, 22.1%]
22.9%
[22.6%, 23.2%]
22.9%
[22.6%, 23.2%]
21.0%
[20.4%, 21.7%]
21.0%
[20.4%, 21.7%]
*NSDUH’s definition of a current cigarette smoker is someone who has smoked part or all of a cigarette in the past 30 days. However, NSDUH contains questions on lifetime smoking and
current smoking. The modified definition uses these questions to construct a measure of “current smoking” that is comparable to that of the other surveys (Ryan et al., 2012). The
construction of this variable is described in Appendix A.
27
Population Assessment of Tobacco and Health Study
Table 2-14.
Demographic distributions based on youth ages 12-17 who completed the Youth
Interview: IPS weights
Unweighted
count
Sex
Male
Female
Missing
Total
Age group
12-13
14-17
Missing
Total
Race/ethnicity
Hispanic
Non-Hispanic white alone
Non-Hispanic other
Missing
Total
6,971
6,641
39
13,651
4,684
8,965
2
13,651
3,880
6,616
3,135
20
13,651
Weighted
percentage,
using youth IPS
weights
51.2%
48.8%
Confidence interval using
youth IPS weights
Percentage from
ACS PUMS
[50.3%, 52.1%]
[47.9%, 49.7%]
51.3%
48.7%
100.0%
34.3%
65.7%
100.0%
[33.5%, 35.2%]
[64.8%, 66.5%]
100.0%
28.6%
48.4%
23.0%
33.7%
66.3%
100.0%
[25.2%, 32.1%]
[45.1%, 51.7%]
[20.5%, 25.5%]
100.0%
22.3%
54.5%
23.2%
100.0%
Table 2-15 provides estimates from the PATH Study for two common measures of cigarette
smoking prevalence among youth respondents compared with estimates from NHANES, NSDUH,
and NYTS. 14 Different measures of smoking are used in this report for youth than for adults. The
primary measure of cigarette smoking among youth in this report is whether the youth has ever tried
smoking a cigarette, even one or two puffs (see Appendix A). Another measure is current smoking,
defined as having smoked at all in the past 30 days. Both are shown in Table 2-15.
14
TUS-CPS does not interview persons younger than 18 about tobacco use.
28
Population Assessment of Tobacco and Health Study
Table 2-15.
Cigarette smoking* based on youth ages 12-17 who completed the Youth Interview: IPS weights
Sample
size
PATH Study:
Unweighted
percentage
13,631
13.5%
6,959
13.9%
Ever tried smoking, female
6,634
13.1%
Ever tried smoking,
age 12-13
Ever tried smoking,
age 14-17
Have smoked in past
30 days
4,675
4.6%
8,954
18.1%
13,613
4.7%
Ever tried cigarette smoking, even
one or two puffs
Ever tried smoking, male
PATH Study:
Weighted
percentage,
using youth IPS
weights
[95% confidence
interval]
Percentage from
2011-2012
NHANES
[95% confidence
interval]
Percentage from
2013 NSDUH
[95% confidence
interval]
Percentage from
2012 NYTS
[95% confidence
interval]
13.5%
[12.6%, 14.5%]
14.0%
[13.0%, 15.1%]
13.1%
[12.0%, 14.2%]
4.6%
[3.9%, 5.4%]
18.2%
[17.1%, 19.4%]
4.7%
[4.2%, 5.1%]
20.5%
[17.5%, 23.6%]
21.1%
[15.9%, 26.3%]
20.0%
[14.6%, 25.5%]
5.6%
[1.9%, 9.4%]
28.3%
[23.5%, 33.0%]
6.9%
[4.0%, 9.8%]
15.7%
[15.4%, 16.0%]
16.3%
[15.8%, 16.8%]
15.1%
[14.6%, 15.6%]
4.0%
[3.7%, 4.3%]
21.0%
[20.0%, 22.0%]
5.6%
[5.4%, 5.8%]
25.6%
[23.6%, 27.6%]
27.2%
[25.0%, 29.3%]
24.0%
[21.8%, 26.2%]
11.8%
[10.2%, 13.4%]
32.5%
[30.0%, 34.9%]
8.7%
[7.7%, 9.8%]
*Defined as ever tried a cigarette, even one or two puffs. For comparison, an additional measure of current smoking commonly applied to youth (having smoked at all in the past 30 days) is
also included in this table.
29
Population Assessment of Tobacco and Health Study
Differences in target populations and administration among the youth surveys might lead to
differences in their estimates. In addition, the youth survey estimates have sampling error, as
demonstrated by the confidence intervals about the estimates from the comparison surveys.
Questions and their orderings also differ among the surveys, as described in Appendix A, as do the
modes of administration. The PATH Study, NHANES, and NSDUH use ACASI for the questions
about tobacco use by youth, and these are administered individually in a household or mobile
examination center setting. The NYTS is a pencil-and-paper survey administered in the classroom.
Currivan et al. (2004) found that even when telephone ACASI was used, estimates of youth smoking
prevalence were lower for a telephone survey of youth smoking than for a school-based survey of
the same population (see also Fowler and Stringfellow, 2001, for a discussion of higher smoking
rates in school-based surveys). In addition, school-based surveys often include students who are
older than 17, which is the upper age limit for youth in the PATH Study.
The PATH Study’s estimates of youth smoking are lower than comparable estimates from
NHANES and NSDUH. Part of this difference may be sampling error and part may be attributable
to differences among the survey wordings and administrations. Moreover, the comparison surveys
are from different time periods. According to SAMHSA (2013, 2014), cigarette smoking among
teens is dropping (from 2012 to 2013, the percentage of youth who had ever tried smoking dropped
by 0.8 percentage points among 12-13 year olds, 1.6 percentage points among 14-15 year olds, and
2.7 percentage points among 16-17 year olds, with similar decreases from 2011 to 2012). The lower
percentages found by the PATH Study may reflect, in part, a continuation of this trend. However,
some of the differences among the estimates of youth smoking prevalence may be attributable to
nonresponse bias or measurement error on the part of one or more of the surveys.
2.3
Statistical Approach for Addressing Nonresponse
2.3.1
Interviews
The primary approach for addressing nonresponse is to use differential weight adjustments. These
adjustments are done at the household level and at the person level. The weight adjustments
calibrate the estimates of demographic characteristics such as age, race, and sex to values calculated
from the 1-year 2013 ACS (which are considered to be highly accurate because of the large sample
size and high response rate for the ACS). They also correct for disparities among other
characteristics that might be associated with the characteristics involved in the weighting
30
Population Assessment of Tobacco and Health Study
adjustments. Among numerous sources, the handbook on household surveys by the United Nations
(2005, chapter 6) and Särndal and Lundström (2005) discuss the methods and theory of using weight
adjustments for nonresponse.
Household Nonresponse-Adjusted Weights
Household IPS weights were computed for all sampled addresses in Wave 1. The IPS weights for
responding households were adjusted to compensate for the estimated number of nonresponding
households that were eligible for the PATH Study but did not complete the Household Screener.
An eligibility adjustment was computed separately for each census region. Further adjustments were
made within weighting classes based on information available for both responding and
nonresponding households, namely the segments and blocks in which they are located. Census 2010
data were used to form weighting classes according to the percentage of occupied housing units, the
percentage of population that is Black, 15 the percentage of population that is Hispanic, and other
information related to demographics and income. Census region and the urbanicity of the PSU and
segment were also used when forming the weighting classes.
Within a weighting class, the IPS weights for the responding households were inflated
proportionately to produce the same sum as the sum of the combined IPS weights of the
responding and nonresponding households. The nonresponse-adjusted household weight is
𝐻𝐻𝐻𝐻𝐻𝐻𝐻𝐻𝐻𝐻𝑇𝑇𝑖𝑖𝑖𝑖𝑖𝑖 = 𝐻𝐻𝐻𝐻𝐻𝐻𝐻𝐻𝐻𝐻𝐻𝐻𝑇𝑇𝑖𝑖𝑖𝑖𝑖𝑖
sum of HHIPSWT for eligible sampled households in weighting class
×
sum of HHIPSWT for responding households in weighting class
The nonresponse-adjusted weights were raked to the 1-year 2013 ACS household counts by census
region and household composition. Household composition was defined by the number of nonadult persons in the household (0, 1, or 2+) and the number of adult household members (1, 2, 3+).
For raking purposes, the household composition was imputed for households missing this
information using logical imputation. 16 The final raked household weight is
𝐻𝐻𝐻𝐻𝐻𝐻𝐻𝐻𝐻𝐻𝑇𝑇𝑖𝑖𝑖𝑖𝑖𝑖 = 𝐻𝐻𝐻𝐻𝐻𝐻𝐻𝐻𝐻𝐻𝑇𝑇𝑖𝑖𝑖𝑖𝑖𝑖 × (raking adjustment).
15
Black is defined as Black alone, or in combination with other races.
16
See Lohr (2010) for a brief description of raking and imputation methods.
31
Population Assessment of Tobacco and Health Study
Person Nonresponse-Adjusted Weights
The raked household-level weight was used as the foundation for calculating the nonresponseadjusted person-level weights, for both youth and adults. The initial person-level nonresponseadjusted weight was computed as the product of the Household Screener raked weight HHRKWT
and the reciprocal of the within-household probability of selection for person 𝑙𝑙 within household 𝑘𝑘
of PSU 𝑖𝑖 and segment 𝑗𝑗, as shown in the following formulas:
1
,
Probability adult 𝑙𝑙 selected at Phase 1 from household (𝑖𝑖𝑖𝑖𝑖𝑖)
1
= 𝐻𝐻𝐻𝐻𝐻𝐻𝐻𝐻𝐻𝐻𝑇𝑇𝑖𝑖𝑖𝑖𝑖𝑖 ×
.
Probability youth 𝑙𝑙 selected from household (𝑖𝑖𝑖𝑖𝑖𝑖)
𝐴𝐴𝐴𝐴1𝐵𝐵𝐵𝐵𝑇𝑇𝑖𝑖𝑖𝑖𝑖𝑖𝑖𝑖 = 𝐻𝐻𝐻𝐻𝐻𝐻𝐻𝐻𝐻𝐻𝑇𝑇𝑖𝑖𝑖𝑖𝑖𝑖 ×
𝑌𝑌𝑌𝑌𝑌𝑌𝑇𝑇𝑖𝑖𝑖𝑖𝑖𝑖𝑖𝑖
The probability of selection differed for adults and youth, as described in Section 2.2.1. Although
shadow youth were not interviewed at Wave 1, a base weight was calculated for the shadow youth to
serve as their base weight once they age up to the youth cohort:
𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑇𝑇𝑖𝑖𝑖𝑖𝑖𝑖𝑖𝑖 = 𝐻𝐻𝐻𝐻𝐻𝐻𝐻𝐻𝐻𝐻𝑇𝑇𝑖𝑖𝑖𝑖𝑖𝑖 ×
1
.
Probability shadow youth 𝑙𝑙 selected from household (𝑖𝑖𝑖𝑖𝑖𝑖)
Similar to the adjustment for Household Screener nonresponse, a nonresponse adjustment was
performed to account for nonrespondents to the Youth Extended Interview. The weighting classes
were formed using information similar to that used for the household-level nonresponse adjustment,
and other variables from the Household Screener: age and sex of the household informant; count of
adults in the household (0, 1, 2, 3, 4, 5+); and age, sex, and race/ethnicity of the youth. The youth
weight adjusted for nonresponse is
𝑌𝑌𝑌𝑌𝑌𝑌𝑌𝑌𝑇𝑇𝑖𝑖𝑖𝑖𝑖𝑖𝑖𝑖 = 𝑌𝑌𝑌𝑌𝑌𝑌𝑇𝑇𝑖𝑖𝑖𝑖𝑖𝑖𝑖𝑖 ×
sum of YBWT for eligible sampled youth in weighting class
sum of YBWT for responding youth in weighting class
For youth, the nonresponse-adjusted weights were raked to population totals from the 1-year 2013
ACS, using census region, age, race/ethnicity, and sex as raking variables. If variables used for
nonresponse and/or raking adjustments were missing, they were imputed from the Household
Screener or by logical or hot-deck imputation. After raking, the final weights for youth are denoted
as YRKWT and were computed as
𝑌𝑌𝑌𝑌𝑌𝑌𝑌𝑌𝑇𝑇𝑖𝑖𝑖𝑖𝑖𝑖𝑖𝑖 = 𝑌𝑌𝑌𝑌𝑌𝑌𝑌𝑌𝑇𝑇𝑖𝑖𝑖𝑖𝑖𝑖𝑖𝑖 × (raking adjustment).
32
Population Assessment of Tobacco and Health Study
A similar procedure was used to create raked weights SYRKWT for shadow youth at Wave 1.
Final weights for adults were computed in three steps. First, a nonresponse adjustment was
performed using the information described above for the household-level nonresponse adjustment
(with the exception of the segment urbanicity) and other variables from the Household Screener:
count of adults in the household (0, 1, 2, 3, 4, 5+); and age, race/ethnicity, sex, and tobacco use
status of the adult. The resulting adult weight for respondents to the Phase 2 Screener after adjusting
for nonresponse between Phases 1 and 2 of the adult sampling procedure is
𝐴𝐴𝐴𝐴1𝑁𝑁𝑁𝑁𝑁𝑁𝑇𝑇𝑖𝑖𝑖𝑖𝑖𝑖𝑖𝑖 = 𝐴𝐴𝐴𝐴1𝐵𝐵𝐵𝐵𝑇𝑇𝑖𝑖𝑖𝑖𝑖𝑖𝑖𝑖
sum of AP1BWT for adults sampled at Phase 1 in weighting class
×
.
sum of AP1BWT for adults responding to Phase 2 Screener in weighting class
Second, the probability of selection at Phase 2 was used to calculate the Phase 2 weight:
𝐴𝐴𝐴𝐴2𝑊𝑊𝑇𝑇𝑖𝑖𝑖𝑖𝑖𝑖𝑖𝑖 = 𝐴𝐴𝐴𝐴1𝑁𝑁𝑁𝑁𝑁𝑁𝑇𝑇𝑖𝑖𝑖𝑖𝑖𝑖𝑖𝑖 ×
1
.
Probability adult 𝑙𝑙 from household (𝑖𝑖𝑖𝑖𝑖𝑖) selected at Phase 2
Finally, the Phase 2 adult weights were raked to independent population totals based on data from
the 1-year 2013 ACS. The raking was done using combinations of census region, age, race/ethnicity,
sex, and educational attainment. If variables used for nonresponse and/or raking adjustments were
missing, they were imputed from the Household Screener or by logical or hot-deck imputation. The
final raked weight is
𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝑇𝑇𝑖𝑖𝑖𝑖𝑖𝑖𝑖𝑖 = 𝐴𝐴𝐴𝐴2𝑊𝑊𝑇𝑇𝑖𝑖𝑖𝑖𝑖𝑖𝑖𝑖 × (raking adjustment).
Estimates calculated using the raked weights for variables of interest in the PATH Study are shown
in Tables 2-16 through 2-27, which repeat the analyses used for Tables 2-5 through 2-12, 2-14, and
2-15. The estimates calculated using IPS weights are retained in these tables to facilitate comparisons
of the estimates obtained using the two sets of weights. Confidence intervals are given for each of
the IPS-weighted and raked-weighted estimates in each of Tables 2-16 through 2-27. The confidence
intervals reported for the IPS-weighted estimates are the same as those given in the corresponding
tables in Section 2.2.
33
Population Assessment of Tobacco and Health Study
Table 2-16.
Race by age distribution, based on household enumeration
Race and age
classification
Black* 18-24
Black* 25+
Non-Black 18-24
Non-Black 25+
Missing age or race
Total
Unweighted
count
3,669
17,934
18,058
116,957
4,003
160,621
Weighted
percentage,
using
household IPS
weights
2.3%
11.4%
11.4%
74.9%
Confidence
interval using
household IPS
weights
[2.0%, 2.6%]
[9.8%, 12.9%]
[10.8%, 12.0%]
[73.3%, 76.5%]
Weighted
percentage,
using
household
raked weights
2.0%
11.1%
10.2%
76.7%
100.0%
Confidence
interval using
household
raked weights
[1.8%, 2.3%]
[9.8%, 12.5%]
[9.7%, 10.7%]
[75.2%, 78.1%]
100.0%
Percentage
from ACS
PUMS
2.0%
10.5%
10.1%
77.5%
100.0%
*Black alone or in combination with other race(s).
Table 2-17.
Sex
Male
Female
Missing
Total
Distribution of male and female adults listed in the household enumeration
Unweighted
count
77,088
83,500
33
160,621
Weighted
percentage for
adults,
using household
IPS weights
48.0%
52.0%
Confidence
interval using
household IPS
weights
[47.8%, 48.3%]
[51.7%, 52.2%]
100.0%
Weighted
percentage for
adults,
using household
raked weights
47.7%
52.3%
100.0%
Confidence
interval using
household raked
weights
[47.5%, 47.9%]
[52.1%, 52.5%]
Percentage
from ACS
PUMS
48.2%
51.8%
100.0%
The household raked weight HHRKWT adjusts the weights so that they agree with the 1-year 2013
ACS household counts by region and household size. They would therefore not be expected to
bring person-level percentages of specific demographic groups closer to the ACS values. Tables 2-16
and 2-17 compare the estimated percentage of adults in the PATH Study household rosters to the
ACS values for each race/age and sex group using the raked weights. The estimated percentages
using the raked weights are not significantly different from the ACS values for race/age, although
they are significantly different from the ACS values for sex. Table 2-21 indicates that using the raked
adult weights produces distributions for race/age and sex that are practically identical to those from
the ACS.
Tables 2-18 through 2-20 provide estimated distributions for household size and numbers of adults
and youth per household, respectively; as expected, the raked weights align the estimated
percentages with the 1-year 2013 ACS values. Confidence intervals shown in Tables 2-18 and 2-19
for the estimates computed with the raked weights are narrow because the raking constrains the
estimates to accord with the ACS on the raking dimensions.
34
Population Assessment of Tobacco and Health Study
Table 2-18.
Number of
persons in
household
1
2
3
4
5+
Total
Table 2-19.
Number of
adults in
household
0-1*
2
3+
Total
Distribution of household size based on households responding to the Household
Screener
Unweighted
count
17,822
26,098
13,503
11,720
10,055
79,198
Weighted
percentage,
using
household
IPS weights
22.7%
Confidence interval
using household IPS
weights
[22.1%, 23.4%]
33.0%
17.0%
14.7%
12.6%
100.0%
[32.5%, 33.5%]
[16.7%, 17.3%]
[14.3%, 15.1%]
[12.0%, 13.2%]
Weighted
percentage,
using
household
raked
weights
27.7%
33.7%
15.2%
12.9%
10.5%
100.0%
Confidence interval
using household
raked weights
[27.7%, 27.7%]
Percentage
from ACS
PUMS
27.7%
[33.7%, 33.7%]
[15.1%, 15.4%]
[12.7%, 13.1%]
[10.3%, 10.6%]
33.7%
15.7%
13.0%
9.9%
100.0%
Distribution of number of adults based on households responding to the Household
Screener
Unweighted
count
21,869
40,961
16,368
79,198
Weighted
percentage,
using
household IPS
weights
27.8%
51.7%
20.5%
100.0%
Confidence
interval using
household IPS
weights
[27.1%, 28.5%]
[51.2%, 52.2%]
[19.8%, 21.1%]
Weighted
percentage,
using
household
raked
weights
33.3%
50.7%
16.0%
100.0%
Confidence
interval using
household raked
weights
[33.3%, 33.4%]
[50.7%, 50.7%]
[15.9%, 16.0%]
Percentage
from ACS
PUMS
33.3%
50.7%
16.0%
100.0%
*A small number of households contain only emancipated youth and hence contribute to the zero part of this category.
Table 2-20.
Number of
youth ages
12-17 in
household
0
1
2+
Total
Distribution of number of youth ages 12-17 based on households responding to the
Household Screener
Unweighted
count
65,661
9,569
3,968
79,198
Weighted
percentage,
using
household IPS
weights
83.0%
12.0%
5.0%
100.0%
Confidence
interval using
household IPS
weights
[82.4%, 83.5%]
[11.7%, 12.4%]
[4.7%, 5.2%]
Weighted
percentage,
using
household
raked weights
84.5%
11.0%
4.4%
100.0%
Confidence
interval using
household raked
weights
[84.3%, 84.8%]
[10.8%, 11.2%]
[4.3%, 4.6%]
Percentage
from ACS
PUMS
84.4%
11.1%
4.5%
100.0%
Tables 2-21 and 2-22 present estimates of demographic characteristics, education, and health
insurance based on adult respondents in Wave 1, using the adult raked weight ARKWT. Raking
adjusted the weights to match ACS totals for combinations of sex, age, race, ethnicity, and education
categories. Raking increased the estimated percentage of adults with health insurance, however, as
35
Population Assessment of Tobacco and Health Study
noted in Section 2.2, the ACS estimates for the percentage of adults with health insurance were
based on data collected before the launch of the Affordable Care Act.
Table 2-21.
Demographic distributions based on adults responding to the Adult Extended
Interview
Adult respondents to Adult Extended Interview
Weighted
Weighted
percentage
percentage
Confidence
using adult
Confidence interval
Unweighted using adult
interval using
raked
using adult raked
count
IPS weights adult IPS weights
weights
weights
Sex
Male
Female
Missing
Total
Age group
18-24
25-44
45-64
65+
Missing
Total
Race
Black alone or in
combination
White alone
Other
Missing
Total
Ethnicity
Hispanic
Non-Hispanic
Missing
Total
16,309
15,982
29
32,320
45.7%
54.3%
[44. 9%, 46.6%]
[53.4%, 55. 1%]
100.0%
[48.1%, 48.1%]
[51.9%, 51.9%]
100.0%
[13.2%, 14.4%]
[35.0%, 36.6%]
[32.8%, 34.5%]
[16.0%, 17.4%]
100.0%
100.0%
5,538
14.1%
[12.1%, 16.1%]
13.1%
[12.9%, 13.3%]
12.5%
23,242
2,726
814
32,320
77.4%
8.5%
[75.1%, 79.7%]
[7.5%, 9.5%]
77.9%
9.0%
[77.6%, 78.1%]
[8.8%, 9.3%]
75.7%
11.8%
[13.0%, 13.0%]
[34.3%, 34.4%]
[34.5%, 34.5%]
[18.2%, 18.2%]
100.0%
100.0%
17.8%
82.2%
13.0%
34.3%
34.5%
18.2%
48.1%
51.9%
9,110
11,280
8,818
3,110
2
32,320
5,536
26,288
496
32,320
13.8%
35.8%
33.7%
16.7%
48.1%
51.9%
ACS PUMS
Percentage
100.0%
100.0%
[15.3%, 20.3%]
[79.7%, 84.7%]
100.0%
15.2%
84.8%
100.0%
36
13.0%
34.3%
34.5%
18.2%
100.0%
[15.1%, 15.3%]
[84.7%, 84.9%]
15.0%
85.0%
100.0%
Population Assessment of Tobacco and Health Study
Table 2-22.
Comparison of education level and health insurance status based on adults
responding to the Adult Extended Interview
Adult respondents to Adult Extended Interview
Weighted
Weighted
Confidence
percentage
Confidence
percentage
interval using
using adult
interval using
Unweighted using adult
adult IPS
raked
adult raked
count
IPS weights
weights
weights
weights
Education
18-24
< HS, HS or GED
> HS
Subtotal
25+
< HS or GED
HS
Some college,
no degree
Bachelor degree
> Bachelor
degree
Subtotal
Missing
Total
Health insurance
Yes
No
Missing
Total
ACS PUMS
Percentage
4,465
4,593
9,058
49.9%
50.1%
100.0%
[48.1%, 51.7%]
[48.3%, 51.9%]
43.5%
56.5%
100.0%
[43.4%, 43.6%]
[56.4%, 56.6%]
43.6%
56.4%
100.0%
4,769
4,763
18.0%
20.0%
[16.9%, 19.1%]
[19.1%, 21.0%]
16.9%
23.8%
[16.8%, 17.0%]
[23.7%, 23.9%]
16.9%
23.8%
7,650
3,642
31.3%
18.2%
[30.2%, 32.3%]
[17.1%, 19.2%]
29.2%
18.7%
[29.1%, 29.3%]
[18.7%, 18.8%]
29.2%
18.7%
2,228
23,052
210
32,320
12.5%
100.0%
[11.5%, 13.5%]
11.4%
100.0%
[11.3%, 11.4%]
11.3%
100.0%
25,760
6,156
404
32,320
84.4%
15.6%
[83.5%, 85.2%]
[14.8%, 16.5%]
85.5%
14.5%
[84.9%, 86.0%]
[14.0%, 15.1%]
83.3%
16.7%
100.0%
100.0%
100.0%
Estimates of smoking prevalence in Table 2-23 using the raked weight ARKWT are similar to the
estimates using the IPS weight AIPSWT; both are in the range of values obtained by other surveys.
The use of raked weights resulted in a slight decrease in estimated smoking prevalence for females
and non-white non-Hispanics.
Tables 2-24 and 2-25 examine the effect of the raked weight YRKWT on estimates calculated for
youth. As expected, raking corrects for the overrepresentation of Hispanics among youth in Wave 1.
The IPS-weighted estimates for youth age and sex agree with the 1-year 2013 ACS estimates.
Consequently, raking had little if any effect on estimates of these characteristics. Cigarette smoking
prevalence estimates with IPS weights and with raked weights were generally lower than estimates
from other surveys although, as noted above, the surveys took place in different time periods.
37
Table 2-23.
Population Assessment of Tobacco and Health Study
Current cigarette smoking based on adults responding to the Adult Extended Interview
Sample
size
PATH
Study:
Unweighted
percentage
PATH Study:
Weighted
percentage,
using adult IPS
weights
[95%
confidence
interval]
PATH Study:
Weighted
percentage,
using adult
raked weights
[95%
confidence
interval]
Percentage
from 20102011 TUS-CPS
[95%
confidence
interval]
Percentage
from 2013
NHIS
[95%
confidence
interval]
Percentage
from 20112012 NHANES*
[95%
confidence
interval]
Percentage from
2013 NSDUH,
original
definition**
[95% confidence
interval]
Percentage
from 2013
NSDUH,
modified
definition
[95%
confidence
interval]
Current smoker
32,245
35.4%
18.6%
[17.7%, 19.4%]
18.2%
[17.7%, 18.7%]
16.1%
[15.8%, 16.3%]
17.8%
[17.2%, 18.4%]
19.8%
[17.5%, 22.1%]
22.9%
[22.6%, 23.2%]
21.0%
[20.4%, 21.7%]
Current smoker,
male
16,265
36.8%
21.2%
[20.3%, 22.2%]
21.0%
[20.3%, 21.6%]
18.0%
[17.7%, 18.4%
20.5%
[19.5%, 21.4%]
23.9%
[20.7%, 27.1%]
25.6%
[25.1%, 26.1%]
23.8%
[22.7%, 24.9%]
Current smoker,
female
15,952
33.9%
16.4%
[15.4%, 17.3%]
15.7%
[15.0%, 16.3%]
14.2%
[13.9%, 14.5%]
15.3%
[14.6%, 16.0%]
16.0%
[13.5%, 18.5%]
20.4%
[20.0%, 20.8%]
18.4%
[17.6%, 19.3%]
Current smoker,
age 18-24
9,099
27.2%
20.1%
[18.8%, 21.5%]
19.7%
[18.5%, 20.9%]
17.1%
[16.4%, 17.8%]
18.7%
[16.8%, 20.5%]
20.4%***
[13.7%, 27.1%]
NA****
NA
Current smoker,
age 25-44
11,260
41.2%
22.9%
[21.8%, 24.1%]
23.1%
[22.2%, 23.9%]
17.9%
[17.5%, 18.4%]
20.1%
[19.1%, 21.1%]
23.3%
[20.0%, 26.7%]
NA
NA
Current smoker,
age 45-64
8,784
40.6%
18.7%
[17.7%, 19.8%]
18.4%
[17.7%, 19.2%]
17.8%
[17.4%, 18.2%]
20.0%
[19.0%, 20.8%]
21.3%
[18.3%, 24.2%]
NA
NA
Current smoker,
age 65+
3,100
23.3%
7.7%
[7.0%, 8.4%]
7.6%
[7.0%, 8.3%]
7.8%
[7.5%, 8.2%]
8.8%
[8.0%, 9.6%]
9.2%
[6.7%, 11.7%]
NA
NA
Current smoker,
Hispanic
5,519
26.3%
13.4%
[12.6%, 14.2%]
13.4%
[12.6%, 14.2%]
10.9%
[10.4%, 11.5%]
12.1%
[11.0%, 13.2%]
16.6%
[13.7%, 19.5%]
18.9%
[18.1%, 19.7%]
15.2%
[13.8%, 16.7%]
Current smoker,
white non-Hispanic
19,268
38.8%
19.6%
[18.%, 20.7%]
19.4%
[18.6%, 20.1%]
17.5%
[17.2%, 17.8%]
19.4%
[18.5%, 20.2%]
20.2%
[17.0%, 23.3%]
24.1%
[23.7%, 24.5%]
22.8%
[21.9%, 23.6%]
Current smoker,
other non-Hispanic
6,904
33.3%
20.4%
[19.3%, 21.5%]
18.3%
[17.4%, 19.2%]
NA
16.7%
[15.6%, 17.8%]
20.8%
[16.6%, 24.9%]
21.8%
[20.2%, 23.4%]
19.6%
[18.1%, 21.2%]
Current every-day
smoker
32,245
28.0%
14.8%
[13.9%, 15.6%]
14.5%
[14.0%, 15.0%]
12.7%
[12.4%, 12.9%]
13.7%
[13.1%, 14.2%]
16.4%
[14.3%, 18.4%]
NA
NA
Current some-days
smoker
32,245
7.4%
3.8%
[3.6%, 4.1%]
3.7%
[3.5%, 3.9%]
3.4%
[3.3%, 3.5%]
4.1%
[3.8%, 4.4%]
3.4%
[2.7%, 4.1%]
NA
NA
*The smoking questions asked in NHANES for adults ages 20 and older differ from the questions asked for persons ages 12-19. The modes of administration also differ for the two age groups. The
NHANES estimates presented in this table are for adults ages 20 and older.
**NSDUH’s definition of a current cigarette smoker is someone who has smoked part or all of a cigarette in the past 30 days, which is more expansive than the definition used in the other surveys.
However, NSDUH contains questions on lifetime smoking and current smoking. The modified definition uses these questions to construct a measure of “current smoking” that is comparable to that of
the other surveys (Ryan et al., 2012). The construction of this variable is described in Appendix A. The estimates and confidence intervals for the NSDUH “original definition” (except for the “current
smoker, other non-Hispanic” estimate) are from the published tables (SAMHSA, 2014); the estimates and confidence intervals for the “modified definition” are calculated from the public use data
set. The estimate of current smoking for the “other non-Hispanic” group was not available from the published tables and it was also calculated from the public use data set.
***The estimate is for adults 20-24 years old.
****Detailed age information was not available in the public use file for NSDUH 2013.
38
Table 2-24.
Population Assessment of Tobacco and Health Study
Demographic distributions based on youth ages 12-17 who completed the Youth Interview
Unweighted count
Sex
Male
Female
Missing
Total
6,971
6,641
39
13,651
Age group
12-13
14-17
Missing
Total
4,684
8,965
2
13,651
Race/ethnicity
Hispanic
Non-Hispanic white alone
Non-Hispanic other
Missing
Total
3,880
6,616
3,135
20
13,651
Youth ages 12-17 respondents to Youth Interview
Weighted
Weighted
percentage,
percentage,
using youth IPS
Confidence interval
using youth raked
weights
using youth IPS weights
weights
51.2%
48.8%
[50.3%, 52.1%]
[47.9%, 49.7%]
100.0%
34.3%
65.7%
[51.3%, 51.4%]
[48.6%, 48.7%]
100.0%
[33.5%, 35.2%]
[64.8%, 66.5%]
100.0%
28.6%
48.4%
23.0%
51.3%
48.7%
Confidence interval
using youth raked
weights
33.7%
66.3%
100.0%
22.1%
54.6%
23.3%
100.0%
39
51.3%
48.7%
100.0%
[33.7%, 33.7%]
[66.3%, 66.3%]
100.0%
[25.2%, 32.1%]
[45.1%, 51.7%]
[20.5%, 25.5%]
ACS PUMS
Percentage
33.7%
66.3%
100.0%
[22.0%, 22.1%]
[54.5%, 54.7%]
[23.3%, 23.4%]
22.3%
54.5%
23.2%
100.0%
Table 2-25.
Cigarette smoking* based on youth ages 12-17 who completed the Youth Interview
Sample
size
Ever tried cigarette smoking,
even one or two puffs
PATH Study:
Unweighted
percentage
Population Assessment of Tobacco and Health Study
PATH Study: Weighted
percentage,
using youth IPS
weights
[95% confidence
interval]
PATH Study: Weighted
percentage,
using youth raked
weights
[95% confidence
interval]
Percentage from
2011-2012 NHANES
[95% confidence
interval]
Percentage from
2013 NSDUH
[95% confidence
interval]
Percentage from
2012 NYTS
[95% confidence
interval]
13,631
13.5%
13.5%
[12.6%, 14.5%]
13.4%
[12.6%, 14.2%]
20.5%
[17.5%, 23.6%]
15.7%
[15.4%, 16.0%]
25.6%
[23.6%, 27.6%]
Ever tried smoking, male
6,959
13.9%
14.0%
[13.0%, 15.1%]
14.0%
[13.0%, 15.1%]
21.1%
[15.9%, 26.3%]
16.3%
[15.8%, 16.8%]
27.2%
[25.0%, 29.3%]
Ever tried smoking, female
6,634
13.1%
13.1%
[12.0%, 14.2%]
12.8%
[11.8%, 13.8%]
20.0%
[14.6%, 25.5%]
15.1%
[14.6%, 15.6%]
24.0%
[21.8%, 26.2%]
Ever tried smoking, age 12-13
4,675
4.6%
4.6%
[3.9%, 5.4%]
4.5%
[3.7%, 5.2%]
5.6%
[1.9%, 9.4%]
4.0%
[3.7%, 4.3%]
11.8%
[10.2%, 13.4%]
Ever tried smoking, age 14-17
8,954
18.1%
18.2%
[17.1%, 19.4%]
18.0%
[17.0%, 19.0%]
28.3%
[23.5%, 33.0%]
21.0%
[20.0%, 22.0%]
32.5%
[30.0%, 34.9%]
Have smoked in past 30 days
13,613
4.7%
4.7%
[4.2%, 5.1%]
4.6%
[4.2%, 5.0%]
5.6%
[5.4%, 5.8%]
8.7%
[7.7%, 9.8%]
6.9%
[4.0%, 9.8%]
*Defined as ever tried a cigarette, even one or two puffs. For comparison, an additional measure of current smoking commonly applied to youth (having smoked at all in the past 30 days) is also
included in this table.
40
Population Assessment of Tobacco and Health Study
2.3.2
Biospecimens
The adult raked weight ARKWT is also used for the analysis of adults in the full Wave 1 sample who
provide biospecimens. Tables 2-26 and 2-27 present estimates of demographic characteristics,
education, and health insurance for adults who provided biospecimens, using the adult raked weight
ARKWT. Although the raking was performed on the adults responding to the Adult Extended
Interview, and no additional adjustments were performed on the adults from whom biospecimens
were collected, this raking brings the estimated sex, age, race, and ethnicity distributions for adults
who provided urine specimens closer to estimates from the 1-year 2013 ACS. The raking also
resulted in estimated percentages of females, Blacks, and Hispanics from the adults who provided
blood samples that are closer to the 1-year 2013 ACS estimates, although the estimated age
distribution for the adults who provided blood samples differs from the ACS distribution. Table 227 shows that the raking decreased the estimated percentages of adults at the lowest level of
education, bringing those closer to the ACS estimates. Raking increased the estimated percentages of
adults with health insurance for each type of biospecimen.
Table 2-28 gives estimates of current cigarette smoking prevalence for the adults from whom urine
or blood specimens were collected. The estimates of cigarette smoking prevalence calculated with
the raked weights are little changed from the estimates calculated with the IPS weights. The
estimates of cigarette smoking prevalence are higher for both sets of biospecimen providers than for
the full set of respondents to the Adult Extended Interview; however, all estimates are within the
range of estimates from the comparison surveys.
The results in Tables 2-26 to 2-28 show estimates from the adults who provided urine or blood
specimens, but not all specimens collected will be analyzed in the laboratory. Approximately 6,000
blood samples and 10,000 urine samples will be chosen initially for laboratory analysis. These
biospecimens come from a probability sample of adults who are in specified tobacco use
categories. 17 If desired, an additional set of nonresponse-adjusted weights can be developed for these
adults.
17
These categories are: current exclusive established users of cigarettes, current established users of other tobacco
products, current experimental users only of any tobacco product, former established users of any product whose last
use was within the past 12 months, and never users. In other categories of tobacco use, no samples are selected for
laboratory analysis. Consequently, the samples of biospecimens will be representative of the persons in the specified
tobacco use categories, but not of the adult population as a whole.
41
Population Assessment of Tobacco and Health Study
Table 2-26.
Demographic distributions based on adults from whom urine or blood specimens were collected
Adults from whom urine specimen is collected
Adults from whom blood specimen is collected
Weighted
Weighted Confidence
Weighted Confidence Weighted Confidence
percentage Confidence percentage
interval
percentage
interval
percentage
interval
Unusing adult interval using using adult using adult
Unusing adult using adult using adult using adult
weighted
IPS
adult IPS
raked
raked
weighted
IPS
IPS
raked
raked
ACS PUMS
count
weights
weights
weights
weights
count
weights
weights
weights
weights
Percentage
Sex
Male
10,763
44.9%
Female
11,025
55.1%
Missing
Total
Age group
18-24
13
21,801
100.0%
6,457
14.9%
25-44
7,744
37.0%
45-64
5,725
33.0%
65+
1,873
15.1%
2
21,801
100.0%
3,911
14.8%
15,531
77.0%
1,801
8.2%
558
21,801
100.0%
3,870
18.6%
17,633
81.4%
298
21,801
100.0%
Missing
Total
Race
Black alone or in combination
White alone
Other
Missing
Total
Ethnicity
Hispanic
Non-Hispanic
Missing
Total
[43.9%,
45.9%]
[54.1%,
56.1%]
47.1%
[46.5%,
47.8%]
[52.2%,
53.5%]
52.9%
100.0%
[14.2%,
15.6%]
[36.0%,
38.0%]
[32.0%,
33.9%]
[14.3%,
15.9%]
14.1%
[13.8%,
14.4%]
[34.9%,
36.1%]
[33.1%,
34.3%]
[16.0%,
17.3%]
35.5%
33.7%
16.7%
100.0%
[12.6%,
17.0%]
[74.6%,
79.4%]
[7.2%, 9.2%]
13.8%
[13.3%,
14.3%]
[77.1%,
78.5%]
[7.9%,
9.0%]
77.8%
8.4%
100.0%
[16.0%,
21.2%]
[78.8%,
84.0%]
16.0%
[15.5%,
16.5%]
[83.5%,
84.5%]
84.0%
100.0%
42
6,919
43.7%
7,593
56.3%
6
14,518
100.0%
3,884
13.3%
5,005
34.1%
4,191
35.6%
1,438
17.1%
0
14,518
100.0%
2,381
13.2%
10,637
79.3%
1,152
7.5%
348
14,518
100.0%
2,451
16.7%
11,884
83.3%
183
14,518
100.0%
[42.6%,
44.8%]
[55.2%,
57.4%]
46.0%
54.0%
[45.2%,
46.9%]
[53.1%,
54.8%]
100.0%
[12.5%,
14.0%]
[33.1%,
35.2%]
[34.4%,
36.7%]
[16.1%,
18.1%]
12.5%
32.7%
36.1%
18.8%
12.2%
80.5%
7.3%
[12.0%,
12.9%]
[31.9%,
33.4%]
[35.2%,
37.0%]
[17.9%,
19.6%]
14.3%
85.7%
100.0%
13.0%
34.3%
34.5%
18.2%
100.0%
[11.6%,
12.9%]
[79.6%,
81.4%]
[6.7%,
7.9%]
100.0%
[13.9%,
19.5%]
[80.5%,
86.1%]
51.9%
100.0%
100.0%
[11.2%,
15.2%]
[77.1%,
81.5%]
[6.7%,
8.4%]
48.1%
12.5%
75.7%
11.8%
100.0%
[13.7%,
15.0%]
[85.0%,
86.3%]
15.0%
85.0%
100.0%
Population Assessment of Tobacco and Health Study
Table 2-27.
Comparison of education level and health insurance status based on adults from whom urine or blood specimens were collected
Adults from whom urine specimen collected
Adults from whom blood specimen collected
Weighted Confidence
Weighted
Confidence
Weighted Confidence Weighted
Confidence
percentage
interval
percentage
interval
percentage
interval
percentage
interval
using adult using adult using adult using adult
using adult using adult using adult using adult
Unweighted
IPS
IPS
raked
raked
Unweighted
IPS
IPS
raked
raked
count
weights
weights
weights
weights
count
weights
weights
weights
weights
Education
18-24
HS
Subtotal
25+
< HS or GED
HS
Some college, no degree
Bachelor degree
> Bachelor degree
Subtotal
Missing
Total
Health insurance
Yes
[48.3%,
52.5%]
[47.5%,
51.7%]
[18.1%,
20.6%]
[18.6%,
20.9%]
[31.3%,
33.4%]
[15.6%,
17.9%]
[10.8%,
12.8%]
[83.0%,
84.8%]
[15.2%,
17.0%]
44.3%
55.7%
100.0%
18.4%
23.5%
30.3%
17.3%
10.5%
100.0%
84.9%
15.1%
[43.4%,
45.2%]
[54.8%,
56.6%]
[17.9%,
18.9%]
[22.7%,
24.2%]
[29.7%,
30.9%]
[16.8%,
17.9%]
[10.1%,
10.9%]
[84.2%,
85.6%]
[14.4%,
15.8%]
100.0%
43
1,948
50.2%
1,928
3,876
49.8%
100.0%
2,405
19.0%
2,071
19.0%
3,721
33.4%
1,454
16.7%
956
10,607
37
14,520
11.9%
100.0%
11,686
85.0%
2,758
15.0%
76
14,520
100.0%
[47.7%,
52.6%]
[47.4%,
52.3%]
[17.6%,
20.5%]
[17.7%,
20.2%]
[31.9%,
34.8%]
[15.3%,
18.1%]
[10.8%,
13.1%]
[84.0%,
86.0%]
[14.0%,
16.0%]
43.7%
56.3%
100.0%
18.3%
22.6%
31.3%
17.2%
10.7%
100.0%
86.0%
14.0%
100%
[42.1%,
45.3%]
[54.7%,
57.9%]
[17.5%,
19.1%]
[21.5%,
23.6%]
[30.4%,
32.2%]
[16.3%,
18.1%]
[10.1%,
11.3%]
[85.1%,
86.8%]
[13.2%,
14.9%]
ACS PUMS
Percentage
43.6%
56.4%
100.0%
16.9%
23.8%
29.2%
18.7%
11.3%
83.3%
16.7%
100.0%
Population Assessment of Tobacco and Health Study
Table 2-28.
Current cigarette smoking based on adults from whom biospecimens were collected
Adult respondent to
Adult Extended
Interview
Adults providing urine
Adults providing blood
Sample
size
32,245
21,757
14,493
PATH Study:
Weighted
cigarette
smoking
prevalence,
using adult IPS
weights
[95%
confidence
interval]
18.6%
[17.7%, 19.4%]
PATH Study:
Weighted
cigarette
smoking
prevalence,
using adult
raked weights
[95%
confidence
interval]
18.2%
[17.7%, 18.7%]
Percentage from
2010-2011 TUSCPS
[95% confidence
interval]
16.1%
[15.8%, 16.3%]
Percentage from
2013 NHIS
[95% confidence
interval]
17.8%
[17.2%, 18.4%]
20.9%
[20.0%, 21.9%]
21.6%
[20.5%, 22.7%]
20.5%
[19.8%, 21.2%]
21.1%
[20.2%, 21.9%]
16.1%
[15.8%, 16.3%]
16.1%
[15.8%, 16.3%]
17.8%
[17.2%, 18.4%]
17.8%
[17.2%, 18.4%]
Percentage from
2011-2012
NHANES
[95% confidence
interval]
19.8%
[17.5%, 22.1%]
Percentage
from 2013
NSDUH, original
definition*
[95%
confidence
interval]
22.9%
[22.6%, 23.2%]
Percentage
from 2013
NSDUH,
modified
definition
[95%
confidence
interval]
21.0%
[20.4%, 21.7%]
19.8%
[17.5%, 22.1%]
19.8%
[17.5%, 22.1%]
22.9%
[22.6%, 23.2%]
22.9%
[22.6%, 23.2%]
21.0%
[20.4%, 21.7%]
21.0%
[20.4%, 21.7%]
*NSDUH’s definition of a current cigarette smoker is someone who has smoked part or all of a cigarette in the past 30 days. However, NSDUH contains questions on lifetime smoking and current
smoking. The modified definition uses these questions to construct a measure of “current smoking” that is comparable to that of the other surveys (Ryan et al., 2012). The construction of this
variable is described in Appendix A.
44
Population Assessment of Tobacco and Health Study
These weights would multiply ARKWT by an adjustment calculated using weighting adjustment cells
based on Wave 1 characteristics such as demographics, education, employment status, health
insurance, and tobacco use. These weights would not produce estimates for the full target
population of the PATH Study, however, but only for the union of the tobacco use categories from
which biospecimens were selected for analysis.
2.4
Summary of Findings
Response Rates
As reported in Section 2.1, the weighted response rates 18 for the PATH Study Household Screener
and Adult Interview and the biospecimen collections in Wave 1 are lower than projected (see Table
2-29), and the weighted response rates for all collections are higher than the worst-case scenario
rates for the full sample provided in Attachment 22. The weighted response rate for the PATH
Study Youth Interview is higher than projected.
Table 2-29.
Summary of PATH Study Wave 1 overall response rates
Collection
Household Screener
Adult Extended Interview
Youth Interview
Urine
Blood
Unweighted
response rate, based
on full Wave 1 sample
54.1%
74.8%
78.2%
67.5%
44.9%
Weighted
response rate, based
on full Wave 1 sample
54.0%
74.0%
78.4%
63.6%
43.0%
Projected
response
rate*
70%
85%
75%
80%
65%
Worst-case
scenario
response
rate*
39.7%
58.1%
-49%
39%
*Provided in the request to OMB for Wave 1 data and biospecimen collection.
The differential weighted response rates are modest for tobacco use status and demographic
subgroups (see Tables 2-1, 2-2, 2-3, and 2-4.) The largest differential weighted response rate, 11.5
percentage points, is for the age of adults who provide urine samples, which suggests a heightened
potential for nonresponse bias. Notably, the differential weighted response rates for blood
collection, ranging from 3.2 percentage points for ethnicity to 5.8 percentage points for race, were
more consistent with those of other PATH Study collections.
18
These response rates were weighted with inverse probability of selection weights.
45
Population Assessment of Tobacco and Health Study
Nonresponse Bias Analysis
Nonresponse bias analysis indicates that estimates of key demographic and tobacco use variables
calculated from the PATH Study Wave 1 sample with the inverse probability of selection weights are
comparable to those produced by national cross-sectional surveys. However, the completed
household interviews from the Wave 1 sample appear to underrepresent single- and two-person
households relative to the 1-year 2013 ACS counts. The estimated percentage of persons who are
non-Black and 25 years of age or older, from the household rosters, is also smaller than the
corresponding estimate from the ACS.
Estimated distributions of demographic characteristics for adults completing the Adult Extended
Interview are similar to those from the 1-year 2013 ACS for race (except for persons in the “other
race” category). Persons in the “other race” category are also underrepresented among the persons
providing blood or urine specimens. The estimated percentages of adults who are Hispanic are
similar to ACS values for adults who provided blood specimens, but Hispanics are overrepresented
among adults who responded to the Adult Extended Interview and those who provided urine
specimens. In addition, the estimated percentage of adults who are between 18 and 24 years old or
between 25 and 44 years old is higher for the PATH Study than for the ACS for adult respondents
as a whole and for those who provided urine specimens. Males are underrepresented among
respondents to the Adult Extended Interview, and also among the persons who provided blood or
urine specimens.
When compared to national cross-sectional surveys that measure tobacco use (TUS-CPS, NHIS,
NHANES, and NSDUH), estimates of adult cigarette smoking from the PATH Study Wave 1
sample are roughly in the middle of the range of estimates on smoking. There is no indication of
nonresponse bias with respect to this measure.
Estimates of demographic characteristics of youth in Wave 1 align with the 1-year 2013 ACS for
most demographic characteristics. However, the estimated percentage of youth who are Hispanic
youth from the PATH Study is significantly higher than the corresponding percentage estimated
from the ACS. (The nonresponse weight adjustments correct for this difference.)
PATH Study estimates of the selected youth cigarette smoking measure from the full Wave 1 sample
are at the low end of estimates in comparison with national cross-sectional surveys that measure
tobacco use (NHANES, NSDUH, and NYTS). However, estimates from the comparison surveys
46
Population Assessment of Tobacco and Health Study
are from 2011 through 2013 while those from the PATH Study are from September 2013 through
December 2014, and evidence suggests the use of traditional cigarettes is declining among youth.
The difference among surveys on time period alone is not large enough to account for the different
estimates; as indicated in Section 2.2.2, time period is one of a number of factors that may explain
the different estimates.
Statistical Approach for Addressing Nonresponse
The approach used to reduce potential nonresponse bias in the PATH Study is to adjust the weights
of respondents at the household, adult, and youth levels to account for nonrespondents. Results of
applying this approach to the full Wave 1 sample indicate the nonresponse adjustments are
successful for reducing the discrepancy between the PATH Study estimates and 1-year estimates
from the 2013 ACS with respect to demographic characteristics. Raked weights used for adults
responding to the Adult Extended Interview reduced differences between the PATH Study and ACS
for adults providing biospecimens as well, for sex and ethnicity. The raking did not reduce
differences in the age distributions for the persons providing blood specimens, however. If desired,
an additional set of nonresponse-adjusted weights could be created for the set of persons for whom
biospecimens are analyzed, using weighting adjustment cells with respect to adults’ Wave 1
characteristics.
Estimates of adult cigarette smoking using the IPS weights (before nonresponse adjustment) are in
line with estimates from other surveys; agreement in these estimates is preserved using the
nonresponse-adjusted weights. Weighting adjustments for youth corrected for the slight
overestimate of the percentage of Hispanics among youth in Wave 1 but had little effect on the
other demographic characteristics (i.e., IPS-weighted estimates already agreed with the ACS values)
and estimates of youth cigarette smoking.
47
Population Assessment of Tobacco and Health Study
Wave 2
3
Wave 2 of the PATH Study is at the approximate mid-point of data and biospecimen collection.
This section discusses response rates achieved to date for Wave 2, nonresponse analysis, and the
Study’s planned statistical approach for addressing nonresponse in Wave 2.
3.1
Response Rates
This section summarizes the three types of response rate calculations used for Wave 2 of the PATH
Study: the interim retention and recruitment rates for the interviews and the interim response rates
for the biospecimen collections. Retention for Wave 2 refers to obtaining a completed Wave 2
questionnaire from: (1) persons who completed the Adult Extended Interview in Wave 1, called
continuing adults; or (2) persons who completed the Youth Extended Interview in Wave 1 and who
are age 17 or younger at Wave 2, called continuing youth. Recruitment for Wave 2 applies to those
who have aged up, either as shadow youth who have turned age 12 and are eligible to participate in
the Wave 2 Youth Interview (called aged-up youths), or as adults who participated as youth in Wave
1 and have turned age 18, thus becoming eligible to participate as adults in Wave 2 (called aged-up
adults). Response rates for biospecimen collections refer to the percentages of persons providing
biospecimens among those who are asked to provide biospecimens.
3.1.1
Retention Rates for Continuing Adults and Continuing Youth
As stated in Section 1, the PATH Study Wave 1 sample was divided among four replicate groups.
Replicate group 1, which consisted of the addresses that were released to the field in September
2013, obtained responses from 5,951 adults and 2,698 youth, and parental consent for 1,413 shadow
youth. Approximately 98 percent of those cases had been released to the field for Wave 2 as of April
22, 2015. Replicate group 1 roughly corresponds to the earliest set of follow-ups in Wave 2, given
the effort to schedule Wave 2 interviews on or near the Wave 1 anniversary dates.
Adults from Wave 1 are asked to complete an Adult Extended Interview in Wave 2. Persons who
completed the youth questionnaire at Wave 1 are aged-up adults if they have attained age 18 by the
48
Population Assessment of Tobacco and Health Study
date of the Wave 2 interview, and are continuing youth if they are age 17 or younger on the date of
the interview. Nonrespondents and interim cases for Wave 2, however, do not have a Wave 2
interview date, so the following procedure was used to classify them as aged-up adults or continuing
youth for this report. The PATH Study field procedures call for beginning data collection contacts
for all members of a household one month prior to the first day of the anniversary month of the
household member with the earliest Wave 1 interview date. The age classification date is defined to
be four months after the beginning of data collection contacts. Nonrespondents and interim Wave 1
youth who are age 17 or younger on the age classification date are classified as continuing youth, and
all others are classified as aged-up adults. 19 A similar classification rule is used for persons who were
shadow youth at Wave 1. The Wave 1 shadow youth who completed a Youth Interview are
classified as aged-up youth, as are nonrespondent and interim Wave 1 shadow youth who attained
age 12 on or before the age classification date.
Table 3-1 displays the status of replicate group 1 cases that had been released to the field as of April
22, 2015 for the four categories of continuing adults, aged-up adults, continuing youth, and aged-up
youth. The finalized cases include respondents, finalized nonrespondents, and 36 continuing adults
known to be deceased.
Table 3-1.
Status of Wave 2 released cases from replicate group 1, as of April 22, 2015
Group
Continuing adults
Aged-up adults
Continuing youth
Aged-up youth
19
Finalized
n
4,388
382
1,765
431
%
75.3
78.4
81.7
78.5
Status of cases
Interim
n
%
1,437
24.7
105
21.6
396
18.3
118
21.5
n
5,825
487
2,161
549
Total
%
100.0
100.0
100.0
100.0
Under this rule, for retention rate calculation purposes, nonrespondent and interim youth from Wave 1 are assigned
to the category (continuing youth or aged-up adult) that would result if they completed an interview on the age
classification date. To explore the sensitivity of the response rate calculations to this definition, the response rates
were also calculated under an alternative classification that pro-rated the nonrespondents and interim cases who were
age 17 at the beginning of the data collection and age 18 on the age classification date, according to the percentage of
respondents with those ages who completed the adult questionnaire. Four persons changed age group classification
under the alternative rule, with a negligible impact on retention and recruitment rates.
49
Population Assessment of Tobacco and Health Study
Method
Consistent with the response rate calculation guidelines specified by the Office of Management and
Budget (2006), final retention rates for Wave 2 will be calculated for adults as the ratio of the
number of Wave 2 Adult Interview completed cases (or sufficient partials) to the number of cases
eligible for the Wave 2 Adult Interview. A simplified formulation will be used for this report because
the eligibility status of some interim cases is unknown. The simplified formulation corresponds to
AAPOR RR1 (AAPOR, 2011), which treats all completed cases from Wave 1, with the exception of
persons known to be deceased, as eligible for Wave 2. The projected retention rates are therefore
slightly conservative because some ineligible persons are included in the denominator.
If all the Wave 2 cases were finalized, the RR1 retention rate for continuing adults would be
calculated as (number of completes or sufficient partials at Wave 2)/(number of Wave 1 completed
cases – number of deceased persons), where the denominator can equivalently be expressed as the
sum of the respondents and finalized nonrespondents. Because the PATH Study Wave 2 data
collection is ongoing, however, the formula must consider “nonfinalized” or interim status cases as
well as finalized cases; in this sense, the retention and recruitment rates presented in the interim
report are “predicted.” Therefore, in this report, the unweighted retention rate for continuing adults
is calculated as
𝐼𝐼
𝐶𝐶𝐶𝐶
𝑝𝑝̂𝑖𝑖,𝐶𝐶𝐶𝐶 )/(CCA+NCA+ICA),
RRUCA = (CCA+∑𝑖𝑖=1
where
CCA
NCA
ICA
𝑝𝑝̂ 𝑖𝑖,𝐶𝐶𝐶𝐶
=
=
=
=
number of Wave 2 completed cases or sufficient partials among continuing adults;
number of Wave 2 finalized nonrespondents among continuing adults;
number of Wave 2 interim cases among continuing adults; and
predicted probability of interim continuing adult i becoming a respondent.
For continuing youth, the denominator of the response rate is defined using the age classification
date described earlier in this section, and the unweighted retention rate is calculated as
𝐼𝐼
𝐶𝐶𝐶𝐶
RRUCY = (CCY+∑𝑖𝑖=1
𝑝𝑝̂ 𝑖𝑖,𝐶𝐶𝐶𝐶 )/(CCY+NCY+ICY),
where
CCY = number of Wave 2 completed cases or sufficient partials among continuing youth;
50
Population Assessment of Tobacco and Health Study
NCY = number of Wave 2 finalized nonrespondents among Wave 1 Youth Extended
Interview completed cases who were age 17 or younger on the age classification
date;
ICY = number of Wave 2 interim cases among Wave 1 Youth Extended Interview
completed cases who were age 17 or younger on the age classification date; and
𝑝𝑝̂ 𝑖𝑖,𝐶𝐶𝐶𝐶 = predicted probability of interim continuing youth i becoming a respondent.
The weighted response rates are computed similarly, with each count replaced by the sum of the
Wave 1 raked weights (YRKWT or ARKWT) for individuals in that category. The sums of the
𝐼𝐼𝐶𝐶𝐶𝐶
𝐼𝐼𝐶𝐶𝐶𝐶
predicted probabilities for interim cases are replaced by ∑𝑖𝑖=1
𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝑇𝑇𝑖𝑖 𝑝𝑝̂ 𝑖𝑖,𝐶𝐶𝐶𝐶 or ∑𝑖𝑖=1
𝑌𝑌𝑌𝑌𝑌𝑌𝑌𝑌𝑇𝑇𝑖𝑖 𝑝𝑝̂ 𝑖𝑖,𝐶𝐶𝐶𝐶 ,
as appropriate.
The probability that an interim case will become a Wave 2 respondent is estimated using logistic
regression, which is commonly used to predict response propensities. Models were fit to the sets of
Wave 2 finalized and interim cases, separately for Wave 1 adults, Wave 1 youth, and Wave 1 shadow
youth, to predict the probability of an interim case becoming a respondent as a function of
respondent characteristics at Wave 1 (age, sex, race/ethnicity, tobacco use status 20), Wave 1 paradata
(ever-refused at the household, parent, or person level in Wave 1, number of contact attempts in
Wave 1, number of days in field in Wave 1), and Wave 2 paradata (ever had an interim adult or
interim parent refusal). The number of weeks in field was used to divide the cases into nine time
periods (weeks 1, 2, 3, 4-6, 7-9, 10-12, 13-15, 16-18, 19+). Logistic regression was used to estimate
the probability that a case would respond in period k given that the case did not previously respond
through a discrete time logistic survival model (Heeringa et al., 2010, Chapter 10; Berglund, 2011;
Olson and Groves, 2012), and the response propensity for an interim case in time period k was
estimated as P(respond in period 𝑘𝑘 + 1) + P(respond in period 𝑘𝑘 +2) + … P(respond in period 9),
where these probabilities were calculated as functions of the predictions from the logistic regression
model. 21
20
Tobacco use status at Wave 1 was not available for the shadow youth and was omitted from the shadow youth
model. Imputed age classification variables were used for persons missing age. Persons missing tobacco status were
considered to be nonusers for purposes of predicting response propensities: this assumption had a negligible effect
on the estimated response rates.
21
The results in the table were calculated using the conservative assumption that an interim case in time period 𝑘𝑘 would
not respond within time period 𝑘𝑘. Other models were also fit to investigate the sensitivity of the results to the model
assumptions. These included logistic regression models (Groves et al., 2008; Wagner, 2010) that predicted response
status from time in field, interim refusal status, demographic characteristics, and Wave 1 paradata: predictions from
these models varied depending on whether interim cases were included in the model-fitting. The estimated
unweighted retention rates from all models fit ranged between 82 percent and 90 percent for continuing adults.
51
Population Assessment of Tobacco and Health Study
Tables 3-2 and 3-3 provide predicted retention rates for continuing adults and continuing youth. In
addition to the overall row, each table includes rows on tobacco use status, age, sex, race, and
ethnicity subgroups based on self-reported data from the Wave 1 Extended Interviews. Persons with
missing values for these characteristics from Wave 1 are excluded from the response rate calculation
for the particular characteristic.
Results
The weighted predicted retention rates are approximately 84 percent for continuing adults and 89
percent for continuing youth. The unweighted predicted retention rates are approximately 85
percent for continuing adults and 90 percent for continuing youth. These approximately equal the
projected retention rates of 85 percent for continuing adults and 90 percent for continuing youth
provided in the Revision Request approved by OMB on 9/8/2014 for Wave 2. The predicted
retention rates are sensitive to the models used for predicting the response propensity among the
interim cases (approximately 25 percent of the continuing adults and 18 percent of the continuing
youth), and the estimates of retention rates will be more accurate as more information accrues.
The variability among predicted retention rates for subgroups is small. For continuing adults,
females appear to have slightly higher predicted retention rates than males, Blacks and Hispanics
have higher predicted retention rates than non-Blacks and non-Hispanics, and persons ages 65 and
older have lower predicted retention rates than younger persons. But these apparent differences
depend largely on the disposition of the interim cases and no definitive conclusions can be made.
The predicted retention rates appear similar for all the subgroups of continuing youth.
3.1.2
Recruitment Rates for Aged-up Adults and Aged-up Youth
This section reports recruitment rates for aged-up adults, who completed the Youth Interview at
Wave 1 and are eligible for the Adult Interview at Wave 2, and aged-up youth, who were shadow
youth at Wave 1 and are eligible for the Youth Interview at Wave 2. The Wave 2 Youth Interview is
the first interview for aged-up youth, and aged-up adults are completing the Adult Interview for the
first time. Table 3-1 displays the status of released cases for the aged-up adults and aged-up youth as
of April 22, 2015.
52
Population Assessment of Tobacco and Health Study
Table 3-2.
PATH Study Wave 2 predicted retention rates by Wave 1 characteristics: Adult Interview (continuing adults)
Characteristica
Overall
Tobacco use statusde
Current user
Current non-user
Age
18-24
25-44
45-64
65+
Sexe
Male
Female
Racee
White alone
Black alone or in
combination
Other
Ethnicitye
Hispanic
Non-Hispanic
A:
Wave 2 Adult
Interviews
completed
(n)
4,038
B:
Wave 2 Interim
likely to be
completedb
(n)
903
C:
Wave 2 Finalized
nonresponse
(n)
314
D:
Interim cases
(n)
1,437
Unweighted predicted
retention rate for
Wave 2c
(%)
85.4%
Weighted predicted
retention rate for
Wave 2c
(%)
84.3%
1,791
2,136
426
449
152
148
662
731
85.1%
85.7%
84.2%
84.9%
1,091
1,379
1,172
396
334
350
179
39
56
102
93
63
521
533
302
81
85.4%
85.8%
86.2%
80.6%
85.2%
85.3%
85.1%
79.9%
1,991
2,044
492
408
168
146
783
651
84.4%
86.3%
83.5%
84.9%
2,915
662
600
163
254
34
986
232
84.6%
88.9%
83.8%
87.6%
357
112
18
171
85.9%
84.1%
680
3,298
192
691
29
280
289
1,119
87.4%
84.9%
87.7%
83.8%
a
The characteristics are as reported in the Adult Extended Interview at Wave 1.
b
Interim likely to be completed is the sum of the predicted probabilities of an interim case becoming a respondent over all interim cases.
c
Predicted retention rate = (A+B)/(A+C+D).
d
A tobacco user is defined as someone who currently uses one or more of the tobacco products covered by the PATH Adult Interview. A tobacco non-user is someone who does not currently
use any of those tobacco products. A current user of a given tobacco product is someone who currently uses the product every day or some days and: for cigarettes, has smoked at least
100 cigarettes in their lifetime and, for any other tobacco product, has reported they ever used that product regularly. The products covered by the Adult Interview are cigarettes,
traditional cigars, cigarillos, little filtered cigars, pipes, smokeless tobacco, hookah, e-cigarettes, and dissolvable tobacco.
e
The sum of counts for this category do not sum to the overall total due to missing values. The number of missing cases is 169 for tobacco use, 6 for sex, 160 for race, and 94 for ethnicity.
53
Population Assessment of Tobacco and Health Study
Table 3-3.
PATH Study Wave 2 predicted retention rates by respondent characteristics: Youth
Interview (continuing youth)
Characteristica
Overall
Tobacco use status
Ever user
Never user
Age
12-13
14-17
Sexf
Male
Female
Racef
White alone
Black alone or in
combination
Other
Ethnicityf
Hispanic
Non-Hispanic
a The
Unweighted
predicted
retention
rate for
Wave 2d
(%)
89.8%
Weighted
predicted
retention
rate for
Wave 2d
(%)
89.2%
64
332
88.8%
90.0%
87.9%
89.5%
31
57
171
225
90.4%
89.3%
89.9%
88.7%
132
130
44
42
196
199
90.3%
89.4%
89.7%
88.8%
1,108
288
152
52
71
8
237
76
89.0%
91.4%
88.4%
91.2%
197
36
4
49
93.2%
91.8%
487
1,158
90
168
20
65
128
263
90.9%
89.2%
90.4%
88.8%
A:
Youth
Interviews
completed
(n)
1,677
B:
Interim
likely to be
completedb
(n)
263
291
1,386
42
221
20
68
723
954
113
150
870
802
C:
Finalized
nonresponsec
(n)
88
D:
Interim
cases
(n)
396
characteristics are as reported in the Youth Extended Interview at Wave 1.
Interim likely to be completed is the sum of the predicted probabilities of an interim case becoming a respondent over all interim
cases.
b
c.
Finalized nonresponse includes refused cases and all other nonresponding cases.
d Predicted
retention rate = (A+B)/(A+C+D).
e
A tobacco ‘ever user’ is someone who has ever used one or more of the tobacco products covered by the PATH Youth Interview. A
tobacco ‘never user’ is someone who has never used any of those tobacco products. Ever use of a tobacco product is defined as
having ever used the product, even one or two times. The products covered by the Youth Interview are cigarettes, traditional cigars,
cigarillos, little filtered cigars, pipes, smokeless tobacco, hookah, e-cigarettes, dissolvable tobacco, bidis, and kreteks.
f
The sum of counts for this category do not sum to the overall total due to missing values. The number of missing cases is 8 for sex, 123
for race, and 40 for ethnicity.
Method
The methods described in Section 3.1.1 for estimating the retention rates were also used to estimate
the recruitment rates for aged-up adults and aged-up youth. For aged-up adults, the unweighted
recruitment rate is
𝐼𝐼
𝐴𝐴𝐴𝐴𝐴𝐴
RRUAUA = (CAUA+∑𝑖𝑖=1
𝑝𝑝̂𝑖𝑖,𝐴𝐴𝐴𝐴𝐴𝐴 )/(CAUA+NAUA+IAUA),
54
Population Assessment of Tobacco and Health Study
where
CAUA = number of Wave 2 completed cases or sufficient partials among persons who
completed the Youth Interview at Wave 1 and the Adult Interview at Wave 2;
NAUA = number of Wave 2 finalized nonrespondents among Wave 1 Youth Extended
Interview completed cases who attained age 18 by the age classification date;
IAUA = number of Wave 2 interim cases among Wave 1 Youth Extended Interview
completed cases who attained age 18 by the age classification date; and
𝑝𝑝̂ 𝑖𝑖,𝐴𝐴𝐴𝐴𝐴𝐴 = predicted probability of interim aged-up adult i becoming a respondent.
The unweighted recruitment rate for aged-up youth is estimated by
𝐼𝐼
𝐴𝐴𝐴𝐴𝐴𝐴
RRUAUY = (CAUY+∑𝑖𝑖=1
𝑝𝑝̂ 𝑖𝑖,𝐴𝐴𝐴𝐴𝐴𝐴 )/(CAUY+NAUY+IAUY),
where
CAUY = number of Wave 2 completed cases or sufficient partials among persons who
were shadow youth at Wave 1 and were administered the youth questionnaire at
Wave 2;
NAUY = number of Wave 2 finalized nonrespondents among Wave 1 shadow youth who
were age 12 by the age classification date;
IAUY = number of Wave 2 interim cases among Wave 1 shadow youth who were age 12
by the age classification date; and
𝑝𝑝̂𝑖𝑖,𝐴𝐴𝐴𝐴𝐴𝐴 = predicted probability of interim aged-up youth i becoming a respondent.
The weighted recruitment rates were calculated by substituting the sum of Wave 1 raked weights
𝐼𝐼𝐴𝐴𝐴𝐴𝐴𝐴
(YRKWT or SYRKWT) for individuals in each category for the counts, and replacing ∑𝑖𝑖=1
𝑝𝑝̂ 𝑖𝑖,𝐴𝐴𝐴𝐴𝐴𝐴 ,
𝐼𝐼𝐴𝐴𝐴𝐴𝐴𝐴
𝐼𝐼𝐴𝐴𝐴𝐴𝐴𝐴
𝐼𝐼𝐴𝐴𝐴𝐴𝐴𝐴
and ∑𝑖𝑖=1 𝑝𝑝̂ 𝑖𝑖,𝐴𝐴𝐴𝐴𝐴𝐴 by ∑𝑖𝑖=1 𝑌𝑌𝑌𝑌𝑌𝑌𝑌𝑌𝑇𝑇𝑖𝑖 𝑝𝑝̂ 𝑖𝑖,𝐴𝐴𝐴𝐴𝐴𝐴 , and ∑𝑖𝑖=1 𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑇𝑇𝑖𝑖 𝑝𝑝̂ 𝑖𝑖,𝐴𝐴𝐴𝐴𝐴𝐴 as appropriate.
Table 3-4 provides predicted recruitment rates for the Adult Interview for aged-up adults, and Table
3-5 provides predicted recruitment rates for the Youth Interview for aged-up youth. In addition to
the overall row, each table includes rows on sex, race, and ethnicity subgroups; Table 3-4 also
includes rows on tobacco use status. There are no rows corresponding to age subgroups in Table 34 or Table 3-5, because almost all of the aged-up adults are 18 years old and almost all of the agedup youth are 12 years old; there are no rows for tobacco use status in Table 3-5, because no
information was collected about the tobacco usage of shadow youth at Wave 1. Information from
the Wave 1 Extended Youth Interview was used to define the demographic characteristics for the
aged-up adults, and information from the Wave 1 Household Screener was used to define the
55
Population Assessment of Tobacco and Health Study
demographic characteristics for the aged-up youth. Persons with missing values for these
characteristics on the Extended Youth Interview or Household Screener were excluded from the
response rate calculation for that characteristic.
Table 3-4.
PATH Study Wave 2 predicted recruitment rates by respondent characteristics:
Adult Interview (aged-up adults)
Characteristica
Overall
Tobacco use statuse
Ever user
Never user
Sex
Male
Female
Race/ethnicityf
Non-Hispanic
white alone
Other
Unweighted
predicted
recruitment
rate for
Wave 2d
(%)
86.4%
Weighted
predicted
recruitment
rate for
Wave 2d
(%)
86.4%
45
60
87.4%
85.8%
87.3%
85.7%
12
5
44
61
85.1%
87.7%
85.4%
87.2%
26
13
50
85.4%
85.5%
30
4
55
87.5%
87.3%
A:
Adult
Interviews
completed
(n)
365
B:
Interim
likely to be
completedb
(n)
56
C:
Finalized
nonresponsec
(n)
17
148
217
25
31
5
12
179
186
21
35
190
173
D:
Interim
cases
(n)
105
a
The characteristics are as reported in the Youth Extended Interview at Wave 1.
b
Interim likely to be completed is the sum of the predicted probabilities of an interim case becoming a respondent over all interim
cases.
c.
Finalized nonresponse includes refused cases and all other nonresponding cases.
d Predicted
recruitment rate = (A+B)/(A+C+D).
e
A tobacco ‘ever user’ is someone who has ever used one or more of the tobacco products covered by the PATH Youth Interview. A
tobacco ‘never user’ is someone who has never used any of those tobacco products. Ever use of a tobacco product is defined as
having ever used the product, even one or two times. The products covered by the Youth Interview are cigarettes, traditional cigars,
cigarillos, little filtered cigars, pipes, smokeless tobacco, hookah, e-cigarettes, dissolvable tobacco, bidis, and kreteks.
f
The sum of counts for this category do not sum to the overall total due to 2 missing values.
56
Population Assessment of Tobacco and Health Study
Table 3-5.
PATH Study Wave 2 predicted recruitment rates by respondent characteristics:
Youth Interview (aged-up youth)
Characteristica
Overall
Sex
Male
Female
Race/ethnicityf
Non-Hispanic
white alone
Other
a The
D:
Interim
casesd
(n)
118
Unweighted
predicted
recruitment
rate for Wave
2e
(%)
87.6%
Weighted
predicted
recruitment
rate for Wave
2e
(%)
87.5%
16
11
52
66
87.4%
87.8%
86.7%
88.2%
31
13
50
88.4%
88.2%
46
14
68
86.8%
86.5%
A:
Youth
Interviews
completed
(n)
404
B:
Interim
likely to be
completedb
(n)
77
C:
Finalized
nonresponsec
(n)
27
194
210
35
42
212
191
characteristics are as reported in the Household Screener at Wave 1.
b
Interim likely to be completes is the sum of the predicted probabilities of an interim case becoming a respondent overall all interim
cases.
c
Finalized nonresponse includes refused cases and all other nonresponding cases.
d
Total interim includes ever refused interim cases and never refused interim cases.
e
Predicted recruitment rate = (A+B)/(A+C+D).
f
The sum of counts for this category do not sum to the overall total due to 1 missing value.
Results
The predicted recruitment rate is approximately 86 percent for aged-up adults, and variability in
predicted recruitment rates is low among different subgroups of aged-up adults. This is slightly
below the projected recruitment rate of 88 percent for aged-up adults in the Wave 2 Revision
Request. The predicted recruitment rate for aged-up youth of 88 percent is higher than the projected
rate of 85 percent in the Wave 2 Revision Request.
Subgroup recruitment rates in Table 3-4 for aged-up adults are similar, with no apparent patterns.
There are also no apparent differences among recruitment rates for demographic subgroups of the
aged-up youth (Table 3-5).
57
Population Assessment of Tobacco and Health Study
3.1.3
Biospecimen Collections
The PATH Study requests a urine sample from a subsample of continuing adults in Wave 2 who
provided urine samples at Wave 1; it also requests urine and blood samples at Wave 2 from all agedup adults. As of April 22, 2015, 1,839 continuing adults and 365 aged-up adults in replicate group 1
had completed the Wave 2 Adult Interview and had been asked to provide biospecimens.
Method
The response rates were calculated using the following formula:
RRU = (Number of samples collected)/(Number of adults from whom a sample is
requested)
The urine response rate for continuing adults is based on the 1,839 adults who, as of April 22, 2015,
were asked to provide a urine specimen following their Wave 2 Adult Interview. Similarly, the urine
and blood response rates for aged-up adults are based on the 365 aged-up adults who completed the
Wave 2 Adult Interview as of this date.
Tables 3-6 and 3-7 provide predicted unweighted and weighted response rates for the biospecimen
collections. The table includes rows with response rates for tobacco use status, age, sex, race, and
ethnicity subgroups. Information from the Wave 1 Adult Interview was used to define the tobacco
use status and demographic characteristics for the continuing adults, and information from the
Wave 1 Youth Interview was used to define the tobacco use status and demographic characteristics
for the aged-up adults. Adults with missing values for a characteristic were excluded from the
response rate calculation for that characteristic. Weights ARKWT were used to calculate the
weighted response rates for Table 3-6, and weights YRKWT were used to calculate the weighted
response rates for Table 3-7.
Results
The projected response rates for biospecimen collection in the Wave 2 Revision Request were 80
percent for urine collection among continuing adults and, among aged-up adults, 69 percent for the
collection of urine and 45 percent for the collection of blood. To date, more than 96 percent of the
58
Population Assessment of Tobacco and Health Study
continuing adults asked to provide urine specimens have done so, exceeding the projected response
rate. Among the aged-up adults, the weighted response rates for urine and blood collection are 82
percent and 43 percent, respectively; the response rate to date for urine collection exceeds the
projected response rate, and the response rate to date for blood collection is slightly less than the
rate that was projected. Subgroup differences in response rates are relatively small.
Table 3-6.
PATH Study Wave 2 response rates by respondent characteristics: Urine collection
(continuing adults)
Characteristica
Overall
Tobacco use statuscd
User
Non-user
Age
18-24
25-44
45-64
65+
Sexd
Male
Female
Raced
White alone
Black alone or in combination
Other
Ethnicityd
Hispanic
Non-Hispanic
B:
Urine
collected
(n)
1,776
Urine
Unweighted
response rate
for Wave 2b
(%)
96.6%
Weighted response
rate for Wave 2b
(%)
96.1%
1,170
645
1,130
625
96.6%
96.9%
96.7%
95.8%
540
681
506
112
521
656
495
104
96.5%
96.3%
97.8%
92.9%
95.3%
96.7%
96.2%
94.6%
959
879
925
850
96.5%
96.7%
96.6%
95.6%
1,329
295
1,287
283
96.8%
95.9%
95.8%
96.4%
171
162
94.7%
97.1%
325
1,487
311
1,439
95.7%
96.8%
95.2%
96.3%
A:
Adults
requested to
provide urine
(n)
1,839
Note. This table presents results on a subsample of continuing adults (i.e., adults who completed an Adult Extended Interview in Wave 1)
who provided a urine sample in Wave 1 and were asked to provide a sample in Wave 2.
a
The characteristics are as reported in the Wave 1 Adult Extended Interview.
b
Response rate = B/A.
c
A tobacco user is defined as someone who currently uses one or more of the tobacco products covered by the PATH Adult Interview. A
tobacco non-user is someone who does not currently use any of those tobacco products. A current user of a given tobacco product is
someone who currently uses the product every day or some days and: for cigarettes, has smoked at least 100 cigarettes in their
lifetime and, for any other tobacco product, has reported they ever used that product regularly. The products covered by the Adult
Interview are cigarettes, traditional cigars, cigarillos, little filtered cigars, pipes, smokeless tobacco, hookah, e-cigarettes, and
dissolvable tobacco.
d
The sum of counts for this category do not sum to the overall total due to missing values. The number of missing cases is 24 for
tobacco use, 1 for sex, 44 for race, and 27 for ethnicity.
59
Population Assessment of Tobacco and Health Study
Table 3-7.
PATH Study Wave 2 response rates by respondent characteristics: Biospecimen
collections (aged-up adults)
Characteristica
Overall
Tobacco use
statusc
Ever user
Non-user
Sex
Male
Female
Raced
White alone
Black alone
or in
combinatio
n
Other
Ethnicityd
Hispanic
Non-Hispanic
B:
Collected
(n)
304
Urine
Unweighted
response
rate for
Wave 2b
(%)
83.3%
Weighted
response
rate for
Wave 2b
(%)
82.3%
B:
Collected
(n)
158
Blood
Unweighted
response
rate for
Wave 2b
(%)
43.3%
Weighted
response
rate for
Wave 2b
(%)
43.2%
148
217
129
175
87.2%
80.6%
86.3%
79.8%
70
88
47.3%
40.6%
47.2%
40.7%
179
186
145
159
81.0%
85.5%
80.8%
83.7%
78
80
43.6%
43.0%
44.1%
42.4%
249
57
206
50
82.7%
87.7%
82.5%
85.4%
103
26
41.4%
45.6%
42.6%
43.3%
38
31
81.6%
78.3%
19
50.0%
44.0%
96
264
83
217
86.5%
82.2%
84.2%
81.9%
44
111
45.8%
42.0%
46.3%
42.1%
A:
Adult
Interviews
completed
(n)
365
Note. This table presents results on aged-up adults (i.e., adults who completed a Youth Extended Interview in Wave 1 and became ageeligible for an Adult Interview in Wave 2).
a
The characteristics are as reported in the Youth Interview at Wave 1.
b
Response rate = B/A.
c
A tobacco ‘ever user’ is someone who has ever used one or more of the tobacco products covered by the PATH Youth Interview. A
tobacco ‘never user’ is someone who has never used any of those tobacco products. Ever use of a tobacco product is defined as
having ever used the product, even one or two times. The products covered by the Youth Interview are cigarettes, traditional cigars,
cigarillos, little filtered cigars, pipes, smokeless tobacco, hookah, e-cigarettes, dissolvable tobacco, bidis, and kreteks.
d
The sum of counts for this category do not sum to the overall total due to missing values. The number of missing cases is 21 for race
and 5 for ethnicity.
3.2
Nonresponse Bias Analysis
The standard approach for an analysis of nonresponse bias in a longitudinal cohort study such as the
PATH Study would be to compare Wave 2 respondents with Wave 2 nonrespondents with respect
to characteristics from Wave 1 (Bose and West, 2002; Javitz and Wagner, 2005; Brownstein et al.,
2009). At the mid-point of Wave 2 data collection, there are a number of interim cases yet to be
finalized as either respondents or nonrespondents. The number of finalized nonrespondents is small
and does not include interim cases that will ultimately be nonrespondents. For this analysis, Wave 2
60
Population Assessment of Tobacco and Health Study
continuing adult respondents are compared with the finalized nonrespondents. To explore the
sensitivity of results to the disposition of the interim cases, Wave 2 continuing adult respondents are
also compared with provisional nonrespondents, defined to be the set of finalized nonrespondents plus
interim refusals 22 and persons who are difficult to locate. More than a third of the interim cases
among provisional nonrespondents are expected to complete the Wave 2 interview; however, they
are more likely to require intensive contact tracing and follow-up efforts than are other interim
cases. For this reason, they are considered to be more similar to finalized nonrespondents than are
other interim cases. For continuing youth and aged-up adults, the number of finalized
nonrespondents is insufficient to permit meaningful comparisons (see Tables 3-3 and 3-4) and the
respondents are compared only with the provisional nonrespondents.
Sections 2.2 and 2.3 described the weight construction for Wave 1 of the PATH Study. The raked
weights from Wave 1 are designed to reduce the potential nonresponse bias from Wave 1. For Wave
2, the nonresponse bias analysis uses the raked weights from Wave 1: ARKWT for the continuing
adults, YRKWT for the continuing youth and aged-up adults, and SYRKWT for the aged-up youth.
Tables 3-8 and 3-9 compare Wave 1 demographic characteristics and tobacco use rates for Wave 2
continuing adult respondents with the finalized nonrespondents and with the provisional
nonrespondents. Table 3-10 presents similar comparisons for the persons who were interviewed as
youth at Wave 1; the Wave 2 aged-up adults and continuing youth are combined for this analysis
because the number of provisional nonrespondents among aged-up adults is too small to allow for
meaningful comparisons). 23 The number of provisional nonrespondents is too small to permit
comparisons of respondents and provisional nonrespondents for the aged-up youth, who do not
self-report information in Wave 1. The recruitment rate calculations in Table 3-5, however, indicate
that there are only small differences in response rates by sex and race/ethnicity.
22
23
Interim refusals for adults are cases that initially declined to participate in the Wave 2 interview but are being
followed for refusal conversion attempts. Interim refusals for youth are those for whom the parent initially declined
permission for the youth to participate in the Wave 2 interview, but are being followed for refusal conversion
attempts.
Because of the smaller sample sizes for the aged-up adults and continuing youth, the percentages using tobacco are
not broken down by demographic subgroups.
61
Table 3-8.
Population Assessment of Tobacco and Health Study
Comparison of Wave 2 Continuing Adult Interview respondents with finalized and provisional nonrespondents
Wave 2 respondents to Adult
Interview
Characteristic at
Wave 1
Unweighted
count
Weighted percentage,
using adult Wave 1
final weights
[95% confidence
interval]
Wave 2 finalized nonrespondents to
Adult Interview
Unweighted
count
Weighted percentage,
using adult Wave 1
final weights
Difference in weighted
percentages
[respondents –
finalized
nonrespondents]*
[95% confidence
interval]
[95% confidence
interval]
Wave 2 provisional nonrespondents to
Adult Interview
Unweighted
count
Weighted percentage,
using adult Wave 1 final
weights
Difference in weighted
percentages
[respondents –
provisional
nonrespondents]*
[95% confidence
interval]
[95% confidence
interval]
Sex
Male
Female
Missing
1,991
2,044
3
47.2% [45.1%, 49.4%]
52.8% [50.6%, 54.9%]
168
146
0
49.0% [41.2%, 56.8%]
51.0% [43.2%, 58.8%]
-1.8% [-10.2%, 6.6%]
1.8% [-6.6%, 10.2%]
525
418
1
52.4% [48.0%, 56.8%]
47.6% [43.2%, 52.0%]
-5.2% [-10.3%, 0.0%]
5.2% [0.0%, 10.3%]
1,091
1,379
1,172
396
0
11.8% [11.0%, 12.7%]
34.4% [32.5%, 36.3%]
35.5% [33.5%, 37.5%]
18.3% [16.6%, 19.9%]
56
102
93
63
0
7.0% [4.8%, 9.1%]
27.9% [21.9%, 33.9%]
30.4% [23.7%, 37.1%]
34.8% [27.2%, 42.4%]
4.9% [2.7%, 7.1%]
6.5% [0.3%, 12.8%]
5.1% [-2.2%, 12.5%]
-16.5% [-24.4%, -8.6%]
296
323
226
99
0
14.1% [11.7%, 16.5%]
33.5% [29.4%, 37.6%]
30.3% [25.7%, 34.8%]
22.1% [17.4%, 26.8%]
-2.3% [-4.8%, 0.2%]
0.9% [-3.6%, 5.4%]
5.2% [0.0%, 10.5%]
-3.9% [-8.9%, 1.2%]
2,438
66.2% [63.5%, 69.0%]
233
78.6% [72.3%, 84.9%]
-12.3% [-18.9%, -5.8%]
572
65.8% [61.0%, 70.6%]
0.4% [-4.1%, 5.0%]
1,534
66
33.8% [31.0%, 36.5%]
76
5
21.4% [15.1%, 27.7%]
12.3% [5.8%, 18.9%]
348
24
34.2% [29.4%, 39.0%]
-0.4% [-5.0%, 4.1%]
3,202
815
21
85.9% [84.4%, 87.3%]
14.1% [12.7%, 15.6%]
250
54
10
88.5% [84.0%, 93.0%]
11.5% [7.0%, 16.0%]
-2.6% [-7.2%, 2.0%]
2.6% [-2.0%, 7.2%]
676
247
21
79.6% [76.2%, 83.0%]
20.4% [17.0%, 23.8%]
6.3% [2.9%, 9.6%]
-6.3% [-9.6%, -2.9%]
834
858
1,434
16.7% [15.0%, 18.3%]
21.9% [20.2%, 23.7%]
31.6% [29.7%, 33.5%]
66
68
108
22.6% [16.9%, 28.4%]
25.1% [18.3%, 31.8%]
27.8% [21.9%, 33.6%]
-6.0% [-12.1%, 0.2%]
-3.1% [-9.9%, 3.7%]
3.8% [-2.5%, 10.1%]
207
221
327
22.2% [18.3%, 26.2%]
24.6% [20.7%, 28.6%]
30.7% [27.1%, 34.3%]
-5.6% [-9.8%, -1.4%]
-2.7% [-6.7%, 1.4%]
0.9% [-3.0%, 4.8%]
899
29.8% [27.3%, 32.3%]
68
24.5% [17.1%, 32.0%]
5.3% [-2.6%, 13.1%]
181
22.4% [17.9%, 27.0%]
7.4% [2.3%, 12.4%]
Age group
Age 18-24
Age 25-44
Age 45-64
Age 65+
Missing
Race/ethnicity
Non-Hispanic
white alone
Other
Missing
Health insurance
Yes
No
Missing
Education
< HS or GED
HS
Some
college, no
degree
Bachelor
degree +
Missing
13
4
8
*Due to rounding, the difference in weighted percentages may differ by 0.1 from the results of [point estimate of respondents] – [point estimate of nonrespondents].
62
Population Assessment of Tobacco and Health Study
Table 3-9.
Characteristic
at Wave 1
Overall
Sex
Male
Comparison of Wave 1 tobacco use rates for Wave 2 Continuing Adult Interview respondents with finalized and provisional
nonrespondents
Wave 2 respondents to Adult
Interview
Weighted
percentage, using
adult Wave 1 final
weights
Unweighted
[95% confidence
count**
interval]
3,927
22.7%
[21.4%, 24.1%]
1,936
27.2%
[25.1%, 29.3%]
18.8%
[17.1%, 20.5%]
161
29.1%
[25.8%, 32.3%]
26.6%
[23.8%, 29.3%]
23.3%
[21.1%, 25.5%]
10.3%
[7.8%, 12.7%]
56
2,382
23.6%
[21.8%, 25.4%]
1,487
20.9%
[18.6%, 23.2%]
Female
1,988
Age group
Age 18-24
1,070
Age 25-44
1,351
Age 45-64
1,124
Age 65+
Race/ethnicity
NonHispanic
white alone
Other
Wave 2 finalized nonrespondents
to Adult Interview
Weighted
percentage, using
adult Wave 1 final
weights
Unweighted
[95% confidence
count
interval]
300
24.5%
[19.5%, 29.4%]
382
Difference in
weighted
percentages
[respondents –
finalized
nonrespondents]*
[95% confidence
interval]
-1.7%
[-6.9%, 3.4%]
Wave 2 provisional
nonrespondents to Adult
Interview
Weighted
percentage, using
adult Wave 1 final
weights
Unweighted
[95% confidence
count
interval]
911
28.8%
[25.3%, 32.3%]
31.0%
[22.9%, 39.0%]
18.3%
[11.8%, 24.8%]
-3.8%
[-12.0%, 4.5%]
0.5%
[-6.1%, 7.1%]
505
42.8%
[29.6%, 55.9%]
32.5%
[23.3%, 41.7%]
30.7%
[19.7%, 41.8%]
8.5%
[3.1%, 13.8%]
-13.7%
[-27.9%, 0.5%]
-5.9%
[-15.2%, 3.4%]
-7.4%
[-18.9%, 4.0%]
1.8%
[-4.1%, 7.7%]
293
223
25.7%
[19.7%, 31.8%]
-2.1%
[-8.3%, 4.0%]
73
21.1%
[12.0%, 30.2%]
-0.2%
[-9.9%, 9.4%]
139
99
87
58
34.7%
[29.7%, 39.6%]
22.5%
[18.3%, 26.7%]
-7.5%
[-12.8%, -2.1%]
-3.7%
[-7.8%, 0.5%]
38.9%
[32.6%, 45.3%]
35.7%
[30.0%, 41.4%]
30.0%
[23.7%, 36.3%]
9.2%
[4.5%, 13.9%]
-9.9%
[-16.9%, -2.9%]
-9.1%
[-15.5%, -2.7%]
-6.7%
[-13.4%, -0.1%]
1.1%
[-4.0%, 6.2%]
559
32.1%
[27.0%, 37.2%]
-8.5%
[-13.7%, -3.2%]
333
23.3% [18.9%,
27.6%]
-2.4%
[-7.4%, 2.7%]
405
310
217
91
*Due to rounding, the difference in weighted percentages may differ by 0.1 from the results of [point estimate of respondents] – [point estimate of nonrespondents].
**The unweighted counts are for adults who provided information about tobacco use at Wave 1.
63
Difference in
weighted
percentages
[respondents –
provisional
nonrespondents]*
[95% confidence
interval]
-6.0%
[-9.7%, -2.4%]
Population Assessment of Tobacco and Health Study
Table 3-10.
Comparison of Wave 2 Aged-up Adult and Continuing Youth Interview respondents with provisional nonrespondents
Characteristic at Wave 1
Wave 2 respondents to Youth Interview
Weighted percentage,
using youth Wave 1 final weights
Unweighted
count
[95% confidence interval]
Wave 2 provisional nonrespondents to Youth
Interview
Weighted percentage,
using youth Wave 1 final weights
Unweighted
count
[95% confidence interval]
Difference in weighted percentages
[respondents – provisional
nonrespondents]**
[95% confidence interval]
Sex
Male
Female
Missing
1,049
988
5
51.4% [49.4%, 53.5%]
48.6% [46.5%, 50.6%]
135
120
2
53.3% [46.9%, 59.8%]
46.7% [40.2%, 53.1%]
-1.9% [-8.8%, 5.0%]
1.9% [-5.0%, 8.8%]
Age group
Age 12-13
Age 14-17
Missing
723
1,319
0
34.7% [32.6%, 36.8%]
65.3% [63.2%, 67.4%]
74
182
1
28.8% [22.9%, 34.6%]
71.2% [65.4%, 77.1%]
6.0% [-0.6%, 12.5%]
-6.0% [-12.5%, 0.6%]
Race/ethnicity
Non-Hispanic white alone
Other
Missing
996
1,006
40
56.1% [52.3%, 60.0%]
43.9% [40.0%, 47.7%]
128
125
4
57.1% [49.0%, 65.1%]
42.9% [34.9%, 51.0%]
-1.0% [-8.9%, 7.0%]
1.0% [-7.0%, 8.9%]
Tobacco use*
Ever user
Never user
Missing
439
1,603
0
21.2% [18.9%, 23.6%]
78.8% [76.4%, 81.1%]
60
197
0
23.6% [19.0%, 28.2%]
76.4% [71.8%, 81.0%]
-2.3% [-7.4%, 2.7%]
2.3% [-2.7%, 7.4%]
* A tobacco ‘ever user’ is someone who has ever used one or more of the tobacco products covered by the PATH Youth Interview. A tobacco ‘never user’ is someone who has never used any
of those tobacco products. Ever use of a tobacco product is defined as having ever used the product, even one or two times. The products covered by the Youth Interview are cigarettes,
traditional cigars, cigarillos, little filtered cigars, pipes, smokeless tobacco, hookah, e-cigarettes, dissolvable tobacco, bidis, and kreteks.
**Due to rounding, the difference in weighted percentages may differ by 0.1 from the results of [point estimate of respondents] – [point estimate of nonrespondents].
64
Population Assessment of Tobacco and Health Study
Among continuing adults, some trends differ for the comparisons with finalized nonrespondents
and the comparisons with provisional nonrespondents. The estimated percentage of persons age 1824 is higher for respondents than for finalized nonrespondents but the difference is not significant
when the respondents and provisional nonrespondents are compared. The respondents have a
higher rate of health insurance coverage than the provisional nonrespondents, but there is no
significant difference in the rate of health insurance coverage between the respondents and final
nonrespondents. The only demographic difference suggested by both sets of comparisons is for
education. The estimated percentage of adults with less than a high school education is lower for
respondents and the estimated percentage of adults with at least a college degree is higher for
respondents, although the differences are not statistically significant between respondents and
finalized nonrespondents. Respondents also have significantly lower tobacco use, overall and for
most population subgroups, compared to provisional nonrespondents, but the differences are not
significant when respondents are compared to finalized nonrespondents. Given that Wave 2 is
underway and continuing to follow up and finalize interim cases, it is advisable to interpret these
comparisons as preliminary and subject to change. With that said, the weighted predicted retention
rates for Wave1 tobacco users and non-users from Table 3-2, which predicted the response
propensity for all interim cases, are 84.2 percent and 84.9 percent, respectively.
The results in this report are based on preliminary data, and may change as more cases are finalized.
If the trends seen among provisional nonrespondents continue as the interim cases are finalized,
however, the PATH Study may experience attrition patterns that are similar to those in other
longitudinal surveys. Thompson (2015) noted that younger persons and persons with lower
educational levels are more difficult to retain in longitudinal surveys. Cunradi et al. (2005) and
Young et al. (2006) have found that smokers were less likely to be retained in subsequent waves of
surveys than nonsmokers.
Among Wave 1 youth, there are no significant differences between the respondents and the
provisional nonrespondents.
3.3
Statistical Approach for Addressing Nonresponse
Initial weights for Wave 2 respondents will be adjusted to address nonresponse at Wave 2.
Nonresponse adjustment cells will be formed using variables from Wave 1, including age, race,
ethnicity, sex, employment status, education level, tobacco use, household composition, census
65
Population Assessment of Tobacco and Health Study
block characteristics, and the type of completed Wave 1 interview (Adult, Youth, or Household
Screener).
Weight adjustments will be computed within cells formed from the cross-classification of variables
available from Wave 1. Tree-based classification software will be employed to identify cells that
distinguish between subgroups with different propensities to respond to the PATH Study (see Roth
et al., 2006 and Schouten and deNooij, 2005). SAS macros will then be used to compute and apply
the weighting adjustment factors and identify potential sources of concern in the adjustment
process, such as small cell sizes and large adjustment factors.
The procedure described in Section 2.3.1 for Wave 1 can be used to address nonresponse for adults
providing biospecimens at Wave 2.
3.4
Summary of Findings
Response Rates
Because the PATH Study Wave 2 data collection is ongoing, response rates 24 were calculated using
predicted response propensities for interim cases, as described in Section 3.1. Table 3-11 indicates
that the weighted predicted retention rates 25 for continuing adults and continuing youth are slightly
lower than the projected rates, although the results are inconclusive because only 76 percent of the
replicate group 1 cases have been finalized to date and the response status is predicted for the
remaining 24 percent using statistical models. The weighted predicted recruitment rate for aged-up
adults is slightly lower than the projection of 88 percent, and the weighted predicted recruitment rate
for aged-up youth is higher than the projected value of 85 percent.
The response rate for each biospecimen collection is calculated as the percentage of persons who
were requested to give a biospecimen who provided it. For both continuing adults and aged-up
adults in Wave 2, the response rate for urine collection is higher than projected. Blood is collected
only from aged-up adults in Wave 2, among whom the response rate for blood collection is slightly
lower than projected.
24
Response rates include retention rates for continuing adults and youth, recruitment rates for aged-up adults and
youth, and response rates for providing biospecimens.
25
These response rates were weighted using the raked weights ARKWT, YRKWT, and SYRKWT.
66
Population Assessment of Tobacco and Health Study
Table 3-11.
Summary of PATH Study predicted response rates for Wave 2
Group
Continuing adults, Adult Interview
Continuing youth, Youth Interview
Aged-up adults, Adult Interview
Aged-up youth, Youth Interview
Continuing adults, urine collection
Aged-up adults, urine collection
Aged-up adults, blood collection
Unweighted predicted
response rate
85.4%
89.8%
86.4%
87.6%
96.6%
83.3%
43.3%
Weighted predicted
response rate
84.3%
89.2%
86.4%
87.5%
96.1%
82.3%
43.2%
Projected
response rate*
85%
90%
88%
85%
80%
69%
45%
*Provided in the Revision Request to OMB for Wave 2 data and biospecimen collection.
The predicted response rates exhibit little variability across population subgroups. For continuing
adults, predicted retention rates are slightly higher for females, Blacks, Hispanics, and adults under
age 65. For continuing youth, aged-up adults, and aged-up youths, no evidence was found to
indicate the retention or recruitment rates differ across population subgroups. Response rates for
biospecimen collection are close to or exceed the projected rates for all categories.
Nonresponse Bias Analysis
The nonresponse bias analysis suggests that estimated percentages of persons with less than a high
school degree tend to be lower for the Wave 2 respondents than for the Wave 2 finalized and
provisional nonrespondents, although the difference between the respondents and finalized
nonrespondents is not statistically significant. Tobacco use rates are not significantly different
between respondents and finalized nonrespondents, although the provisional nonrespondents
exhibit higher tobacco use rates than the respondents. Provisional nonrespondents also have lower
health insurance coverage than respondents, although the difference between respondents and
finalized nonrespondents is not significant.
67
Population Assessment of Tobacco and Health Study
Statistical Approach for Addressing Nonresponse
For Wave 2, weights of respondents will be adjusted to account for nonrespondents by forming
weighting adjustment cells using Wave 1 characteristics of respondents and nonrespondents.
Consequently, nonresponse-adjusted weights of Wave 2 respondents will sum to Wave 1 population
totals. This weighting will compensate for differences between respondents and nonrespondents
with respect to sex, age, other demographic variables, and Wave 1 tobacco use status.
68
Population Assessment of Tobacco and Health Study
Conclusions and Implications for Study Going
Forward
4
This section summarizes the findings presented in this report on the PATH Study’s Wave 1 and
Wave 2 response rates, nonresponse bias analysis, and approach to addressing nonresponse. Its
conclusions are based on the full sample for Wave 1 and on the data collected from replicate group
1 during the first 6 months (out of 12) of Wave 2. The section closes with a discussion of the
implications of the conclusions for the study going forward.
Conclusions
Response rates in Wave 1 for the Household Screener and Adult Extended Interview were lower
than projected in the Non-substantive Change Request for Wave 1 of the PATH Study but higher
than the worst-case scenario. However, nonresponse bias analysis found that many characteristics of
respondents in Wave 1 align with the 1-year estimates from the 2013 ACS. Exceptions were found
for single-person households, education, and ethnicity when comparing PATH Study estimates
using IPS weights to 1-year 2013 ACS estimates. Estimates of cigarette smoking among adults in
Wave 1 are within the range of estimates found by other national health studies. Moreover, when
full sample estimates were adjusted for nonresponse using the raked weights, they more closely
approximated the ACS estimates, and adult smoking rates remained essentially the same.
The response rate for the Wave 1 Youth Interview was higher than projected. Nonresponse bias
analysis among youth found that many characteristics of respondents were consistent with the 1-year
estimates from the 2013 ACS, with the exception of ethnicity. When the full sample estimates were
adjusted for nonresponse among youth, they more closely approximated the 2013 ACS estimates,
but the ever-tried-smoking rates for youth remained lower than those found by other national
studies.
The response rates for urine and blood collections in Wave 1 were lower than projected and
exceeded the worst-case scenario response rates. Despite this, nonresponse bias analysis found that
many of the characteristics of respondents were generally aligned with estimates of these
characteristics from the 1-year 2013 ACS. In addition, when the sample estimates were adjusted for
nonresponse, they were found to approximate the ACS estimates more closely.
69
Population Assessment of Tobacco and Health Study
The unweighted retention rates for Wave 2 continuing adults and continuing youth, calculated using
predictions of response propensity for the interim cases, are close to those projected in the Revision
Request for Wave 2; the weighted predicted retention rates are less than two percentage points lower
than the projections. The estimated recruitment rate for aged-up adults is 1.6 percentage points
lower than the projected rate, and the estimated recruitment rate for aged-up youth is approximately
2.5 percentage points higher than the projected rate. The Wave 2 response rates for biospecimens
also approximately equal or exceed the projected rates. The estimated percentage of persons with at
least a college degree is higher when calculated from the respondents than from the finalized or
provisional nonrespondents; the estimated percentage of persons with less than a high school degree
is lower. There is no evidence that tobacco use differs between respondents and finalized
nonrespondents, but it is significantly lower among respondents than among provisional
nonrespondents. There is also no evidence of nonresponse bias for the continuing youth, aged-up
adults, or aged-up youth. However, as noted, it is important to regard these findings as preliminary
pending finalization of interim cases and the rest of data collection in Wave 2.
Implications for the Study Going Forward
Findings on the response rates, nonresponse bias analysis, and approach to addressing nonresponse
for Wave 1 and Wave 2 have three important implications for the PATH Study. First, the PATH
Study should continue implementing new approaches to increase response rates for Wave 2 and
subsequent waves. The lower than projected sample sizes resulting from the Wave 1 response rates
underscore the need to achieve high response rates in each of the follow-up waves. The PATH
Study is continually seeking ways to boost the response rates. For example, in Wave 2, it has
enhanced its efforts to communicate by text message and email with participants who indicated they
may be contacted in these ways. Participants continue to access information about the study on the
participant pages of the PATH Study website, which also allows adult participants to update their
contact information. The Study employs special interviewers with skills in refusal conversion and has
a cadre of experienced traveling interviewers that help to augment staffing in specific areas. For
Wave 3 and potential future waves, the PATH Study will vary the appearance of materials it
provides participants to enhance their interest and engagement, and it will provide a certificate of
appreciation to participating youth. In addition, the Study will take extra steps to interview
continuing adults who have relocated to group quarters facilities since their initial interviews.
70
Population Assessment of Tobacco and Health Study
Second, data analyses will need to consider the smaller-than-expected sample sizes for Wave 1.
Adjustments may be necessary, such as combining some subgroups in analyses. Future planning for
longitudinal analyses will need to account for the declining rates of cigarette smoking in the U.S.
population in general, notably among youth. For the PATH Study, the trends observed among
youth mean smaller sample sizes for youth smokers and therefore less statistical power for
examining within-person changes among youth cigarette smokers. At the same time, the larger
sample size of youth nonsmokers provides more power for examining the initiation of cigarette
smoking over time. The impact of the smaller sample size for youth smokers must also be
considered in the context of the recent and rapid increase of youth use of alternative tobacco
products such as e-cigarettes and hookah (CDC, 2015a).
In its review of this report, the Division of Data Policy of the Assistant Secretary for Planning and
Evaluation (ASPE), U.S. Department of Health and Human Services (HHS) asked what is the
minimum acceptable sample size for conducting a further wave and what happens if the sample size
falls below this minimum. These questions can be answered in the context of minimum sample sizes
required to detect meaningful differences in subgroups within or across data collection waves on key
measures. For example, to detect a difference between Wave 1 and Wave 2 of less than 5 percentage
points for the percent of youth who smoke menthol cigarettes, power analyses 26 for the Wave 2
Revision Request indicated that the expected sample sizes would be sufficient for 14-17 year old
youth, but not for 12-13 year olds. Cases such as this can be dealt with by combining age subgroups.
In such a case, the PATH Study would need to consider the current and projected sample sizes,
rates of attrition, and whether to replenish the affected subgroup samples in a future data collection
wave.
Third, on an ongoing and consistent basis, the PATH Study should examine the sample sizes
achieved for Wave 1 and Wave 2 to date, as well as those projected for Wave 3. In this way, it would
be possible to detect differential rates of attrition among subgroups early and make extra efforts to
retain persons in subgroups of special analytic interest. If consideration is given to replenish the
sample at some point, the replenishment could consider the subgroup sample sizes in the continuing
cohort. This may be necessary, for example, for the shadow sample, which serves as a reservoir for
aged-up youth in each wave but which will be exhausted after Wave 3.
26
The power analyses assumed power equal to 0.8, two-sided tests of significance, and alpha equal to 0.05. These
analyses made several additional assumptions, including those on the population prevalence of the behavior and the
design effects for the data collections.
71
Population Assessment of Tobacco and Health Study
References
Agaku, I.T., King, B.A., and Dube, S.R. (2014). Current cigarette smoking among adults – United
States, 2005-2012. Morbidity and Mortality Weekly Report, 63, 29-34.
American Association of Public Opinion Research (AAPOR). (2011). Standard Definitions: Final
Dispositions of Case Codes and Outcome Rates for Surveys, 7th ed.
http://www.aapor.org/AAPORKentico/AAPOR_Main/media/MainSiteFiles/StandardDefi
nitions2011_1.pdf.
Berglund, P.A. (2011). An overview of survival analysis using complex survey data. Proceedings of the
2011 SAS Global Forum, http://support.sas.com/resources/papers/proceedings11/3382011.pdf.
Bose, J., and West, J. (2002). Examining additional nonresponse bias introduced through attrition.
Proceedings of the Survey Research Methods Section, American Statistical Association, 278-283.
Brault, M.W. (2013). Non-response bias in the 2013 CPS ASEC Content Test. Paper presented at the
Federal Committee on Statistical Methodology Research Conference, Washington DC,
http://www.copafs.org/UserFiles/file/fcsm/H3_Brault_2013FCSM.pdf.
Brownstein, N., Kalsbeek, W.D., Tabor, J., Entzel, P., Daza, E., and Harris, K.M. (2009). NonResponse in Wave IV of the National Longitudinal Study of Adolescent Health.
http://www.cpc.unc.edu/projects/addhealth/data/guides/W4_nonresponse.pdf.
Center for Behavioral Health Statistics and Quality (CBHSQ). (2013). National Survey on Drug Use and
Health, 2012. ICPSR34933-v1. Ann Arbor, MI: Inter-university Consortium for Political and
Social Research [distributor], 2013-11-26. doi:10.3886/ICPSR34933.v1. Persistent URL:
http://doi.org/10.3886/ICPSR34933.v1.
Center for Behavioral Health Statistics and Quality(CBHSQ), Substance Abuse and Mental Health
Services Administration (SAMHSA), U.S. Department of Health and Human Services (HHS),
and RTI International. (2014). Results from the 2013 National Survey on Drug Use and Health:
Detailed Tables. Prevalence Estimates, Standard Errors, P Values, and Sample Sizes. Rockville, MD:
Substance Abuse and Mental Health Services Administration,
http://www.samhsa.gov/data/sites/default/files/NSDUHDetTabsPDFWHTML2013/Web/HTML/NSDUH-DetTabsTOC2013.htm.
Centers for Disease Control and Prevention (CDC). (2013). Tobacco product use among middle and
high school students — United States, 2011 and 2012. Morbidity and Mortality Weekly Report, 62,
893–897. Retrieved from www.cdc.gov/mmwr/preview/mmwrhtml/mm6245a2.htm.
72
Population Assessment of Tobacco and Health Study
Centers for Disease Control and Prevention (CDC), National Center for Health Statistics (NCHS).
(2014a). National Health and Nutrition Examination Survey Data, 2011-2012. Hyattsville, MD: U.S.
Department of Health and Human Services, Centers for Disease Control and Prevention,
http://wwwn.cdc.gov/nchs/nhanes/search/datapage.aspx?Component=Questionnaire&Cycl
eBeginYear=2011.
Centers for Disease Control and Prevention (CDC), National Center for Health Statistics (NCHS).
(2014b). National Health Interview Survey Data, 2013. Hyattsville, MD: U.S. Department of
Health and Human Services, Centers for Disease Control and Prevention,
http://www.cdc.gov/nchs/nhis/nhis_2013_data_release.htm.
Centers for Disease Control and Prevention (CDC). (2015a). E-cigarette use triples among middle and high
school students in just one year. Press Release (April 16). Atlanta, GA: Centers for Disease Control
and Prevention, National Center for Chronic Disease Prevention and Health Promotion,
Office on Smoking and Health, http://www.cdc.gov/media/releases/2015/p0416-e-cigaretteuse.html.
Centers for Disease Control and Prevention (CDC). (2015b). National Youth Tobacco Survey Data,
2013. Atlanta, GA: Centers for Disease Control and Prevention, National Center for Chronic
Disease Prevention and Health Promotion, Office on Smoking and Health,
http://www.cdc.gov/tobacco/data_statistics/surveys/nyts.
Cunradi, C. B., Moore, R., Killoran, M., and Ames, G. (2005). Survey nonresponse bias among
young adults: the role of alcohol, tobacco, and drugs. Substance Use & Misuse, 40, 171-185.
Currivan, D.B., Nyman, A.L., Turner, C.F., and Biener, L. (2004). Does telephone audio computerassisted self-interviewing improve the accuracy of prevalence estimates of youth smoking?
Evidence from the UMass Tobacco Study. Public Opinion Quarterly, 68, 542-564.
Fowler F.J., and Stringfellow, V.L. (2001). Learning from experience: Estimating teen use of alcohol,
cigarettes, and marijuana from three survey protocols. Journal of Drug Issues, 31, 643-664.
Groves, R.M., Brick, J.M., Couper, M., Kalsbeek, W., Harris-Kojetin, B., Kreuter, F., Pennell, B.,
Raghunathan, T., Schouten, B., Smith, T., Tourangeau, R., Bowers, A., Jans, M., Kennedy, C.,
Levenstein, R., Olson, K., Peytcheva, E., Ziniel, S., and Wagner J. (2008). Issues facing the
field: Alternative practical measures of representativeness of survey respondent pools. Survey
Practice, 1(3).
Heeringa, S.G., West, B.T., and Berglund, P.A. (2010). Applied Survey Data Analysis. Boca Raton, FL:
CRC Press.
Javitz, H., and Wagner, M. (2005). Analysis of potential bias in the Wave 1 and Wave 2 respondents to the
National Longitudinal Transition Study-2 (NLTS-2). Menlo Park, CA: SRI International,
http://www.nlts2.org/studymeth/nlts2_analysis_bias_respondents.pdf.
Lohr, S. (2010). Sampling: Design and Analysis, 2nd ed. Boston: Brooks/Cole.
73
Population Assessment of Tobacco and Health Study
Office of Management and Budget. (2006). Standards and Guidelines for Statistical Surveys, available at
http://www.whitehouse.gov/sites/default/files/omb/inforeg/statpolicy/standards_stat_surv
eys.pdf.
Olson, K., and Groves, R.M. (2012). An examination of within-person variation in response
propensity over the data collection field period. Journal of Official Statistics, 28, 29-51.
Roth, S., Montaquila, J., and Chapman, C. (2006). Nonresponse bias in the 2005 National
Household Education Surveys Program. Technical Report. (NCES 2007-016). U.S.
Department of Education, National Center for Education Statistics. Washington, DC: U.S.
Government Printing Office.
Ryan, H., Trosclair, A., and Gfroerer, J. (2012). Adult current smoking: Differences in definitions
and prevalence estimates—NHIS and NSDUH, 2008. Journal of Environmental and Public Health,
online.
Särndal, C.-E., and Lundström, S. (2005). Estimation in Surveys with Nonresponse. Hoboken, NJ: Wiley.
SAS Institute, Inc. (2011). SAS/STAT® 9.3 User’s Guide. Cary, NC: SAS Institute Inc.
Schenker, N., and Gentleman, J.F. (2001). On judging the significance of differences by examining
the overlap between confidence intervals. The American Statistician, 55, 182-186.
Schouten, B., and de Nooij, G. (2005). Nonresponse adjustment using classification trees. Discussion paper
05001, Statistics Netherlands. Available at www.cbs.nl.
Song, Y. (2013). Rotation group bias in smoking prevalence estimates using TUS-CPS. Paper presented at the
Federal Committee on Statistical Methodology Research Conference, Washington DC, paper
available at http://www.fcsm.gov/13papers/I3_Song_2013FCSM.pdf, and slides available at
http://www.copafs.org/UserFiles/file/fcsm/I3_Song_2013FCSM.pdf.
Substance Abuse and Mental Health Services Administration (SAMHSA). (2013). Substance Abuse
and Mental Health Services Administration, Results from the 2013 National Survey on Drug Use and
Health: Summary of National Findings, NSDUH Series H-48, HHS Publication No. (SMA) 144863. Rockville, MD: Substance Abuse and Mental Health Services Administration, 2014.
Retrieved
fromhttp://www.samhsa.gov/data/sites/default/files/NSDUHresultsPDFWHTML2013/W
eb/NSDUHresults2013.htm. The Survey Documentation and Analysis (SDA) system,
National Survey on Drug Use and Health, 2013. http://www.icpsr.umich.edu/cgibin/SDA/SAMHDA/hsda?samhda+35509-0001.
Thomson, M.E. (2015). Using longitudinal complex survey data. Annual Review of Statistics and its
Application, 2, 305-320.
United Nations. (2005). Designing Household Survey Samples: Practical Guidelines. United Nations
Publication ST/ESA/STAT/SER.F/98, New York: United Nations. Available at
http://unstats.un.org/unsd/demographic/sources/surveys/Handbook23June05.pdf.
74
Population Assessment of Tobacco and Health Study
United States Census Bureau (2015). American Community Survey 2013 ACS 1-year PUMS Files.
Retrieved from
http://www.census.gov/acs/www/Downloads/data_documentation/pums/ACS2013_PUM
S_README.pdf.
United States Department of Commerce, Census Bureau. (2012). National Cancer Institute-sponsored
Tobacco Use Supplement to the Current Population Survey (2010-11):
http://appliedresearch.cancer.gov/studies/tus-cps/.
Technical documentation: http://www.census.gov/cps/methodology/techdocs.html.
Retrieved from http://appliedresearch.cancer.gov/studies/tuscps/results/data1011/table1.html.
Wagner, J. (2010). The fraction of missing information as a monitoring tool for survey data quality.
Public Opinion Quarterly, 74(2):223-243.
Wang, K., Murphy, J., Baxter, R., and Aldworth, J. (2005). Are two feet in the door better than one? Using
process data to examine interviewer effort and nonresponse bias. Paper presented at the 2005 Federal
Committee on Statistical Methodology conference,
http://www.fcsm.gov/05papers/Wang_Aldworth_etal_VIB.pdf.
Young, A.F., Powers, J.R., and Bell, S.L. (2006). Attrition in longitudinal studies: Who do you lose?
Australian and New Zealand Journal of Public Health, 30, 353-361.
75
Population Assessment of Tobacco and Health Study
Appendix A
Cigarette Smoking Questions on the PATH Study and Other Surveys
Table A-1 lists the questions used to ask about current smoking status of adults in the PATH Study
and in the surveys used for comparison and describes the populations included in the estimates from
those surveys.
Note that although the questions used to define current cigarette smoking are similar among the
surveys, small differences could have an effect on the answers given. In the PATH Study, the
question used to establish whether an adult has smoked at least 100 cigarettes in his or her lifetime
has closed response categories:
1.
1 or more puffs but never a whole cigarette;
2.
1 to 10 cigarettes (about ½ pack total);
3.
11 to 20 cigarettes (about ½ pack to 1 pack);
4.
21 to 50 cigarettes (more than 1 pack but less than 3 packs);
5.
51 to 99 (more than 2 ½ packs but less than 5 packs); and
6.
100 or more cigarettes (5 packs or more).
In TUS-CPS, NHIS, and NHANES, however, the question “Have you smoked at least 100
cigarettes in your entire life?” calls for a yes/no response.
The positioning of the questions also differs among the surveys. In the PATH Study, the cigarette
smoking questions are near the beginning of the adult questionnaire, and the respondent knows that
the questionnaire is about tobacco use behaviors. In TUS-CPS, the smoking questions are near the
beginning of the adult questionnaire on tobacco, but the survey is administered as part of the CPS.
In NHIS, the smoking questions follow a long series of questions on health problems (breathing
problems, diabetes, hernias, hemorrhoids, etc.). These question contexts may be associated with
differences in responses.
Table A-2 lists the questions used to define youth cigarette smoking in the PATH Study, NHANES,
NSDUH, and NYTS.
A-1
Population Assessment of Tobacco and Health Study
Table A-1.
Question used to define “current smoking” in the PATH Study, TUS-CPS, NHIS, NHANES, and NSDUH
PATH Study
TUS-CPS**
NHIS
NHANES
Question to define current smoking (answers defining current smoking given in parentheses)
“Have you ever smoked a
"Have you smoked at "Have you smoked at least
"{Have you/Has SP}
cigarette, even one or two
least 100 cigarettes 100 cigarettes in your
smoked at least
puffs?” (yes) and “Do you now in your entire life?"
ENTIRE LIFE?" (yes) and "Do 100 cigarettes in
smoke cigarettes every day,
(yes) and "Do you
you NOW smoke cigarettes {your/his/her} entire
some days, or not at all?"
now smoke
every day, some days or not life?" (yes) and "{Do
you/Does SP} now
(every day or some days) and cigarettes every day, at all?" (every day or some
smoke cigarettes
“How many cigarettes have
some days, or not at days)
you smoked in your entire life? all?" (every day or
(SMQEV, SMKNOW)
every day, some
A pack usually has 20
days or not at all?"
some days)
cigarettes in it.” (100 or more (PEA1, PEA3)
(every day or some
cigarettes (5 packs or more))
days)
(SMQ020, SMQ040)
Age range included in estimate
18+
18+
18+
20+
Exclusions from population
Includes only civilian, nonIncludes only civilian, Includes only civilian
Includes only
institutionalized population.
non-institutionalized noninstitutionalized
civilian, nonExcludes residents of group
population.
population. Several
institutionalized
quarters, active military.
segments of the population population.
excluded, such as: patients
in long-term care facilities;
persons on active duty with
the Armed Forces; persons
incarcerated in the prison
system; and U.S. nationals
living in foreign countries.
Proxy responses allowed
No
Yes
Yes, for individuals physically No
or mentally incapable of
responding.
NSDUH (original
definition)
NSDUH
(modified definition)*
“Have you ever
smoked part or all
of a cigarette?”
(yes) and “During
the past 30 days,
have you smoked
part or all of a
cigarette?” (yes)
“Have you ever
smoked part or all of
a cigarette?” (yes) and
“During the past 30
days, have you
smoked part or all of
a cigarette?” (yes) and
“Have you smoked at
least 100 cigarettes
in your entire life?”
(yes)
18+
18+
Includes only
civilian, noninstitutionalized
population.
Excludes homeless
persons who do not
use shelters,
military personnel
on active duty, and
residents of
institutional group
quarters.
No
No
*The modified definition is given in Ryan et al. (2012).
** Proxies are allowed if 4th callback, the person will not return before closeout, or the household is getting irritated. See http://appliedresearch.cancer.gov/studies/tuscps/surveys/tuscps_english_2010.pdf, p3.
A-2
Population Assessment of Tobacco and Health Study
Table A-2.
Questions used for youth cigarette smoking in the PATH Study, NHANES, NSDUH, and NYTS
PATH Study
NHANES
NSDUH
Question to define ever tried cigarette smoking (answers defining ever tried cigarette smoking given in parentheses)
“Have you ever tried cigarette
“About how many cigarettes have you smoked in
CG01 Have you ever smoked
smoking, even one or two puffs?”
your entire life?” (SMQ621, values of 2-8 (more
part or all of a cigarette? (yes)
(yes)
than a puff to 100 or more cigarettes))
I have never smoked, not even a puff (1), 1 or more
puffs but never a whole cigarette (2),
1 cigarette (3),
16 to 25 cigarettes (6),
2 to 5 cigarettes (4),
26 to 99 cigarettes (7),
6 to 15 cigarettes (5), 100 or more cigarettes (8)
Questions for determining whether have smoked in past 30 days
“Have you ever tried cigarette
“During the past 30 days, on how many days did
smoking, even one or two puffs?”
you smoke cigarettes?” (SMQ640, Recoded to
(yes) and “When was the last time
SMD641 in SMQ_G file, number of day smoked,
you smoked a cigarette, even one or values of 1 through 30)
two puffs?” (Earlier today, Not today
but sometime in the past 7 days, Not
in the past 7 days but sometime in
the past 30 days)
Ages of youth in survey
12-17
12-17
Exclusions from population
Residents of group quarters
Other comments
NYTS
Have you ever tried
cigarette smoking, even
one or two puffs? (Qn7
value of 1, Yes)
CG05 [IF CG01 = 1 OR CGREF1 =
1] Now think about the past 30
days – that is, from [DATEFILL]
up to and including today. During
the past 30 days, have you
smoked part or all of a cigarette?
During the past 30 days,
on how many days did you
smoke cigarettes? (Qn13
values of 2 through 7)
12-17
12-17 year old students in
public or private schools
Includes only the U.S. civilian, noninstitutionalized
population.
Includes only the U.S. civilian,
noninstitutionalized population.
Excludes homeless persons who
do not use shelters, military
personnel on active duty, and
residents of institutional group
quarters.
Only includes youth who
attend either public or
private schools.
Those missing SMQ621 values are excluded from
the estimates.
Those with SMQ621=1, 2, 77 or 99 (never smoked,
less than 1 cigarette, RF, DK) had SMD640
recoded to 0 (0 cigarette smoked in past 30 days)
due to skip pattern.
The Center for Behavioral Health The survey is administered
Statistics and Quality (2013,
by teachers in the
2014) give estimates and the
classroom setting.
standard errors of the estimates.
A-3
| File Type | application/pdf |
| File Modified | 2015-06-19 |
| File Created | 2015-06-10 |