See all ›
7 CitationsNon-response in Wave IV of the Add Health Study
Abstract
Non-response is a potential threat to the accuracy of estimates obtained from sample surveys and can be particularly difficult to avoid in longitudinal studies. The objective of this report is to investigate non-response and consequent bias in estimates for Wave IV of the National Longitudinal Study of Adolescent Health (Add Health). The Survey Research Unit at the University of North Carolina at Chapel Hill previously analyzed the non-response rates for the first three waves of Add Health. As shown in Chantala, Kalsbeek and Andraca, 2005, the total bias in Waves I, II, and III for 13 measures of health and risk behaviors rarely exceed 1%, which is small relative to the 20% to 80% prevalence rates for most of these measures. Results are similar for Wave IV.
In this paper, first, we outline the Wave IV sampling design and results of the field work. Second, we characterize the non-response rates overall and stratified by a number of demographic variables. Next, we use data on the health risk measures reported by Wave IV responders and non-responders during their Wave I In-home interview to estimate total and relative bias due to non-response in Wave IV. We conclude with a discussion of Wave IV bias due to non-response.
- 1 -
Non-Response in Wave IV of the National Longitudinal Study of Adolescent Health
Naomi Brownstein
University of North Carolina at Chapel Hill
William D. Kalsbeek
Survey Research Unit
University of North Carolina at Chapel Hill
Joyce Tabor
Carolina Population Center
University of North Carolina at Chapel Hill
Pamela Entzel
Carolina Population Center
University of North Carolina at Chapel Hill
Eric Daza
University of North Carolina at Chapel Hill
Kathleen Mullan Harris
Carolina Population Center
University of North Carolina at Chapel Hill
INTRODUCTION
Non-response is a potential threat to the accuracy of estimates obtained from sample surveys and
can be particularly difficult to avoid in longitudinal studies. The objective of this report is to
investigate non-response and consequent bias in estimates for Wave IV of the National
Longitudinal Study of Adolescent Health (Add Health). The Survey Research Unit at the
University of North Carolina at Chapel Hill previously analyzed the non-response rates for the
first three waves of Add Health. As shown in Chantala, Kalsbeek and Andraca, 2005, the total
bias in Waves I, II, and III for 13 measures of health and risk behaviors rarely exceed 1%, which
is small relative to the 20% to 80% prevalence rates for most of these measures. Results are
similar for Wave IV.
In this paper, first, we outline the Wave IV sampling design and results of the field work.
Second, we characterize the non-response rates overall and stratified by a number of
demographic variables. Next, we use data on the health risk measures reported by Wave IV
responders and non-responders during their Wave I In-home interview to estimate total and
relative bias due to non-response in Wave IV. We conclude with a discussion of Wave IV bias
due to non-response.
THE WAVE IV SAMPLE AND FIELD WORK RESULTS
Add Health Wave IV was designed as a follow-up interview with all original Wave I in-home
respondents (n=20,745) (Harris, Halpern, Whitsel, Hussey, Tabor, Entzel and Udry, 2009). The
- 2 -
final disposition status for these cases is shown in Figure 1. At Wave III, 96 Wave I respondents
were deceased and 687 were deemed ineligible because they were not part of the probability
sample or the genetic sample (Chantala et al., 2005), leaving 19,962 cases to be fielded at Wave
IV. During Wave IV field work, 402 additional cases were determined to be ineligible for follow
up because the participant was either deceased, out of the country during the data collection
period, or on active military duty and inaccessible to the field interviewers. This left a total of
19,560 original Wave I participants eligible for Wave IV.1
Wave IV interviewers established contact with 18,036 cases, and completed – in whole or in part
– a total of 15,701 interviews. Table 1 provides frequencies and descriptions of the final status
codes in each category depicted in Figure 1. The “Not solicited” group consists of eligible
sample members with whom the interviewer was unable to establish contact. In most of these
cases, the field contractor, RTI, was unable to locate the sample member. The “Solicited, but
unable” group encompasses sample members who were located but (1) unavailable to
participate; (2) physically, linguistically, or mentally incapable of completing the interview; or
(3) unable to participate due to a language barrier. The “Solicited, but unwilling” group is
comprised of sample members who refused to participate in the Wave IV interview. The
“Other” group consists of 55 people who do not fit into the aforementioned four groups.
WAVE IV NON-RESPONSE
Table 2 lists Wave IV response rates, both weighted and unweighted. Wave IV yielded 15,701
completed interviews for an overall unweighted response rate of 80.27% for the full sample of
19,560 eligible cases. Weighted estimates were calculated for the 18,467 eligible respondents
who had known sample weights in Wave I (determined by the variable gswgt1) and known
disposition codes for Wave IV (determined by the variable wave4dsp). The refusals (Unwilling)
were the most common type of non-responders, followed by those who were not contacted (Not
Solicited) and those who were unable to participate in the interview (Unable). The “Other”
group comprises less than 1% of the total non-response.
Survey process rates, including response rates, contact rates, and refusal rates are stratified by
biological sex, race and other demographic variables in Tables 3 - 11. Females were more likely
than males to be contacted and to respond to the survey (Table 3). Whites were more likely to be
contacted than any other racial group (Tables 4,5). Over 95% of white sample members were
contacted, while contact rates for other races ranged from 85% to just over 90%. Whites had the
highest response rate, 83.3%, but they also had high refusal rates. Asians and Pacific Islanders
had the highest refusal rates (13.7%). Native Americans and blacks had the lowest refusal rates,
5.7% and 6.0%, respectively. About 75.1%, a relatively low rate, of Hispanics (any race)
responded. The lowest response rate, 70.7%, was among those whose race fell into the “Other”
group.
Response varied by urban or rural status, region of the country, parental education, immigration
status, and genetic relatedness. Urban respondents were more likely than rural to respond (Table
1 Note that this eligibility classification differs from the approach taken in Wave III. In the Chantala et al. report,
individuals who were inaccessible to the field interviewer were classified as eligible for creating the Wave III final
sample weights; at Wave IV these cases were classified as ineligible for weighting purposes.
- 3 -
6). Survey process rates were most favorable for individuals from the Midwest (response and
contact rates were highest and refusal rates were lowest), followed by those from the South, then
the West, and finally from the Northeast (Table 7). These rates change monotonically and are
least favorable in the Northeast, where the response rate is only 72.1% and the refusal rate is
12.0%. Rates are less favorable for participants whose parents have very little education and are
most favorable for participants whose parents have some college education (Table 8). However,
participants whose parents graduated college were more likely to refuse to participate in the
Wave IV survey than were any of the other groups.
Socioeconomic status of respondents was also associated with contact, response, and refusal. As
socioeconomic status increased, all three types of rates increased (Table 9). Response rates and
contact rates were as low as 73.7% and 84.8%, respectively, in the lowest socioeconomic stratum
and as high as 83.4% and 95.1% in the highest stratum. Refusal rates were lowest for the lowest
socioeconomic stratum, at 7.5%, and highest at the second highest stratum, at 10.9%.
With increasing generation in the U.S., response rates and contact rates also increased, and
refusal rates decreased. First generation immigrants (i.e., foreign-born to foreign-born parents)
were least likely to respond and most likely to refuse, while third generation and higher
Americans (i.e., native-born to native-born parents) were most likely to respond and least likely
to refuse to participate in the survey (Table 10). These differences are marked. About 67% of
first generation immigrants, 77% of second generation participants (i.e., native-born to foreign-
born parents), and over 82% of third or higher generation participants responded. Moreover,
13.5% of first generation immigrants refused to participate, while only 8.3% of third generation
participants refused. For participants in the Wave I genetic sample (Harris et al., 2009), non-
related participants had the lowest response and contact rates (Table 11). Other related
individuals had similar survey process rates regardless of the type of relatedness.
EFFECT OF NON-RESPONSE ON STUDY ESTIMATES
In this section, we quantify the total and component bias related to non-response for the Wave IV
sample, overall and stratified by gender and race. Both respondents and non-respondents in
Wave IV completed the survey in Wave I. Therefore, we use the known answers from Wave I to
evaluate bias in Wave IV. We calculate weighted estimates of the prevalence of health risk
outcomes using the grand sample weight from Wave I (gswgt1) and examine total, component
and relative bias. Total bias is the bias due to any form of non-response. Component bias is bias
due to an individual category of non-response. The four components are “No Contact,”
“Unable,” “Refusal,” and “Other.” Components are additive in that the sum of the four
component biases equals the total bias. Relative bias is defined as the total bias for a particular
measure (e.g., smoking) divided by the prevalence of that measure. We analyze Wave I
characteristics that are similar to those examined in the Wave III non-response analysis,
including demographic characteristics, school experiences, health attitudes and physical
activities, substance abuse, violence, and delinquency.
For 13 measures of health risk, we also compare bias rates for males and females. We defined
biological sex of the respondent by the most recent available response. That is, we defined
biological sex using the Wave IV variable, bio_sex4, if available; if bio_sex4 was missing, we
defined biological sex using the Wave III variable, bio_sex3 and if bio_sex4 and bio_sex3 were
- 4 -
missing, we defined biological sex by bio_sex2. If biological sex was only recorded at Wave I,
then we defined biological sex by the Wave I variable, bio_sex.
All analyses were completed using procedures and macros in SAS version 9.2.
Methods
We calculated bias using sample weights from Wave I (gswgt1) in the full eligible Wave I
sample of 18,467 respondents with known weights. The Wave I variables that reveal potential
bias are indicators for whether a particular behavior is present, so the estimated outcomes are
probabilities. Non-response bias remaining was computed by weighting the difference in
prevalence between responders and non-responders by the non-response rate:
BIASREMAINING = (1-RR4)(PR – PNR)
where:
PR = the weighted prevalence estimate for all respondents (N=14,800)
PNR= the weighted prevalence estimate for all non-respondents (N=3667)
RR4= the weighted response rate using AAPOR definition 42
We also conducted t-tests to determine if the bias remaining is significantly different from zero.
There were 232 tests, total, so we used a Bonferroni adjustment for multiple comparisons.
By dividing the bias by the estimate for all eligible cases, we calculated the relative bias, given in
Tables 12-17. Bias and relative bias are both reported in percentages.
BIASRELATIVE = (BIASREMAINING / PALL)*100
where:
PALL = the weighted prevalence estimate for all eligible cases (N=18,467)
Variables of interest may be compared by estimating relative bias percentages.
Results
Bias remaining in variables measuring health and physical activities is shown in Table 12. The
first column lists the Wave I variable measured as indicated… The second column shows the
2 Response rate is defined as
)()()(
4UOUHeONCRPI PI
RR
where I = completed interview,
P = Partial Interview, R = Refusal and Break-off, NC = No Contact, O = Other, UH = unknown if
household/occupied HU, UO = unknown other, and e = estimated proportion of unknown cases that are eligible.
- 5 -
prevalence of each indicator variable among all those eligible for the Wave IV interview (i.e.,
both Wave IV respondents and non-respondents). The third column lists the percent bias
remaining, and the fourth column, the percent relative bias remaining for each indicator variable.
The results show bias remaining to be less than 1 percentage point in absolute value. These
measures include access to medical care, self-assessment of overall health, and obesity. Both the
highest bias and the highest relative bias were for those lacking current health insurance in 1995,
comprising 12.3% of eligible participants. However, for all measures in this table, the bias due to
non-response was not statistically significant.
Table 13a shows that biases remaining in estimates of use of individual substances are small in
magnitude and statistically non-significant. Table 13b reports similar results based on the
substance use index, which is an aggregated measure based on answers to the questions reported
in Table 13a. Table 14a compares reports of individual acts of violence and delinquency. Table
14b characterizes the bias within delinquency and violence indices. The bias was not
significantly different from zero for any measure in these tables.
Information about family structure is in Table 15. Responders were significantly more likely to
have had two biological parents at Wave I. The relative bias is also notably high for the “other”
category, which could mean either that significantly more non-responders than responders did in
fact have other guardians at Wave I, or that the large and statistically significant relative bias is
just a statistical artifact of the low prevalence of all subjects (i.e., 6% of responders) in this
category.
Table 16 displays information on hearing vocabulary, used as a proxy for cognitive performance.
The AHPVT is a modified version of the Peabody Picture Vocabulary Test (PPVT; Dunn, 1982);
it includes 87 items that ask the respondent to match words (read aloud by the interviewer) with
pictorial representations. Scores were age-standardized to a mean of 100 and a standard
deviation of 15. There is a statistically significant trend across three of the four AHPVT score
categories. Responders were more likely than non-responders to have scores above 110, while
non-responders were more likely than responders to have scores between 70 and 90. The bias is
also significant for the “< 70” category, which could mean either that significantly more non-
responders than responders did in fact have very low AHPVT scores at Wave I, or that the large
and statistically significant relative bias is a statistical artifact of the low prevalence of subjects
(i.e., 2.5% of responders) in this category.3
In Table 17, we selected 13 health risk measures for further analysis of bias according to non-
response components (Table 18) and by biological sex (Table 19) and race (Table 20). These
were chosen in order to compare results with previous non-response analysis (Kalsbeek et al,
2001). Responders were significantly more likely than non-responders to lack an appetite. No
other health risk measures had bias statistically different from zero.
In Table 18, bias in the 13 health risk measures is broken down into its components – No
Contact, Unable, Refusal, and Other. All bias measures were less than 1% in magnitude. Very
3 Due to IRB concerns, there was more lost to follow up on those who had confirmed or suspected cognitive
impairment. Although some of these individuals completed the survey, they are listed in the “Other” category of
non-response.
- 6 -
few of the bias measures due to “Other” reasons for not responding to the survey were
significant. The only variable with significant bias for Refusals and Other reasons was lacking
an appetite. Only one of the risk measures, smoking, had significant bias due to those unable to
respond to the survey, but the magnitude of the bias was less than 0.25%. Fighting and skipping
school both had negative bias for those not contacted. This means that people who were not
locatable in Wave IV were significantly more likely than responders to have skipped school and
to have participated in a fight.
Bias in the 13 health risk measures is broken down by biological sex in Table 19. Only two
measures, skipping school and lacking an appetite, were significant for males. Among males,
non-responders were more likely than respondents to skip school, and responders were more
likely than non-responders to lack an appetite. No measures had significant bias for females.
Finally, bias is presented by race in Table 20. For comparison, whites, blacks, and Hispanics
each had only one significant bias measure. Among whites, non-respondents were more likely
than respondents to skip school. Among blacks, non-respondents were less likely than
respondents to lack an appetite. Among Hispanics, respondents were more likely than non-
respondents to lie to their parents. No bias due to non-response was statistically significant for
Asians/Pacific Islanders, Native Americans or other races.
CONCLUSION
This report presented Wave IV response rates by demographic characteristics and analyzed bias
remaining due to Wave IV non-response using characteristics from Wave I. Females were more
likely to respond than males, and whites were more likely to respond than other races. Response
rates also increased as parental education and socioeconomic levels increased.
Bias and relative bias were small in magnitude for nearly all measures. Moreover, only a few
variables had bias significantly different from zero. Consequently, the differences in
measurements between non-respondents and respondents are most likely due to random
variation, and so do not reflect appreciable non-response bias. For example, according to the
delinquency index, there is little statistical evidence of differences in delinquency levels between
non-responders and responders.
However, there were a few significant results. The highest relative bias measure was the 35%
relative bias due to hearing vocabulary for the lowest group with APHVT scores less than 70.
While this may signify that significantly more non-responders than responders did in fact have
very low AHPVT scores at Wave I, the large and statistically significant relative bias may be
merely a statistical artifact of the low prevalence of all respondents (i.e., 2.5% of responders) in
this category. Similarly, although the relative bias of 8% for an individual from a family
structure of “other” was statistically significant, this again most likely resulted from the low
prevalence of all such participants (6% of responders); note the small magnitude of the bias
(0.47%). All other variables had less than 6% relative bias. That is, while taking into account the
proportion of eligible Wave I subjects with a particular health risk outcome, the adjusted
difference in prevalence of this outcome between responders and non-responders typically does
not exceed 6%.
- 7 -
Note, however, that our use of a Bonferroni adjustment for multiple comparisons results in an
extreme loss of power for each t-test. We use this adjustment to adequately control the overall
probability of a type 1 error, meaning we have only a 5% probability of incorrectly concluding
that a bias measure is statistically significant from zero (i.e., that bias exists). On the other hand,
this safeguard also means we may have low power to conclude that a particular bias measure
differs from zero when the bias that actually exists is either small in magnitude or has a relatively
high standard error.
In conclusion, with the few aforementioned exceptions, Wave IV non-response bias is negligible
and the Wave IV sample adequately represents the same population surveyed at Wave I.
ACKNOWLEDGMENTS
This research uses data from Add Health, a program project directed by Kathleen Mullan Harris
and designed by J. Richard Udry, Peter S. Bearman, and Kathleen Mullan Harris at the
University of North Carolina at Chapel Hill, and funded by grant P01-HD31921 from the Eunice
Kennedy Shriver National Institute of Child Health and Human Development, with cooperative
funding from 23 other federal agencies and foundations. Special acknowledgment is due Ronald
R. Rindfuss and Barbara Entwisle for assistance in the original design. Information on how to
obtain the Add Health data files is available on the Add Health website
(http://www.cpc.unc.edu/addhealth).
The data analysis for this paper was generated using SAS software, Version 9.2 of the SAS
System for Windows. Copyright © 2002-2008 SAS Institute Inc. SAS and all other SAS Institute
Inc. product or service names are registered trademarks or trademarks of SAS Institute Inc.,
Cary, NC, USA.
REFERENCES
The American Association for Public Opinion Research. 2008. Standard Definitions:
Final Dispositions of Case Codes and Outcome Rates for Surveys. 5th edition. Lenexa,
Kansas: AAPOR.
Biemer, Paul P., Aragon-Logan, Elvessa D. National Longitudinal Study of Adolescent Health.
Wave IV Weights. Research Triangle Park, NC: RTI International.
Chantala, Kim, Kalsbeek, William D., and Andraca, Eugenio. (2005), “Non-response in Wave III
of the Add Health Study”
Harris, K.M., C.T. Halpern, E. Whitsel, J. Hussey, J. Tabor, P. Entzel, and J.R. Udry. 2009. The
National Longitudinal Study of Adolescent Health: Research Design [WWW document]. URL:
http://www.cpc.unc.edu/projects/addhealth/design.
Kalsbeek, William D., Yang, Juan, and Agans, Robert P. (2002), “Predictors of nonresponse in a
longitudinal survey of adolescents”, ASA Proceedings of the Joint Statistical Meetings 2002.
- 8 -
Kalsbeek, William D, Morris, Carolyn, B., and Vaughn, Benjamin J. (2001) “Effects of
nonresponse on the mean squared error of estimates from longitudinal study”, ASA Proceedings
of the Joint Statistical Meetings, 2001.
- 9 -
15,701
Completed Wave IV
Interview
18,036
Contacted
19,560
Eligible for Wave IV
783
Ineligible for Wave IV
1,524
Not Solicited
502
Solicited, but unable
1,778
Solicited but unwilling
Figure 1. Wave IV disposition status of Add Health cases from Wave I.
20,745
Interviewed at Wave I
19,962
Fielded at Wave IV
402
Field Determined Ineligible
for Wave IV
55
Other non-interview
10
Table 1. Wave IV Final Disposition of the 20,745 Cases Fielded at the Wave I Interview
Description
Disposition Category
N
Not fielded for Wave
IV
N/A
783
Ineligible Cases
Deceased
131
(N=402)
Out of country for duration of study
184
Active Duty Military – Unavailable for Duration
87
Eligible, Interviewed
Interview finished, break-off/partial interview
7
Retained (N=15,701)
Interview finished
15694
Eligible, Not
Interviewed
Not solicited
(N=1,524)
Access Denied
16
No one home after repeated attempts
8
Incarcerated – final
110
Institutionalized – final
15
Unlocatable
1348
Moved beyond interviewing area
3
Wrong person interviewed
24
Solicited, but unable
Unavailable after repeated attempts
418
(N=502)
Unavailable for duration of field period
3
Language barrier Spanish
4
Language barrier Other (specify)
2
Physically/mentally incapable (specify)
75
Solicited but
unwilling
Final Refusal by Sample Member
1587
(N=1,778)
Final refusal by other
191
Other (N=55)
Interview Completed – Mentally Challenged Case – Mental Capacity
Inadequate
5
Interview Completed – Mentally Challenged Case – Unable to
Determine Mental Capacity
10
Interview Completed – Prison Case – Data Deleted
1
Other non-interview (specify)
39
11
Table 2. Response Rates for Add Health Wave IV
Final Response
Category
Total Number
of Respondents
Unweighted %
Respondents
with Weights
Weighted %
Interviewed
15,701
80.27
14,800
80.54
Not solicited
1524
7.79
1430
7.46
Unable
502
2.57
479
2.60
Unwilling
1778
9.09
1711
9.06
Other
55
0.28
47
0.33
Total Eligible
19,560
100.00
18,467
100.0
Table 3. Survey Process Rates by Biological Sex, 2 Add Health Wave IV
Gender
Males
Females
Rate
Weighted %
Unweighted %
Weighted %
Unweighted %
Response
78.2
77.6
83.0
82.8
Contact1
90.9
90.4
94.2
93.9
Refusal
9.2
9.3
8.9
8.9
Total Eligible
8958
9477
9509
10,083
1 Contact Rate is defined as the number contacted divided by the number assigned
2 Gender responses are defined by the sample member’s most recent available response
Table 4. Unweighted Survey Process Rates by Race and Ethnicity, Add Health Wave IV
Race
Rate
White
Black
Asian/Pacific
Islander
Native
American
Hispanic
Other
Response
83.3
80.5
71.2
80.2
75.1
70.7
Contact
95.2
89.7
90.3
87.4
88.2
88.8
Refusal
9.6
6.0
13.7
5.7
9.9
13.2
Total
Eligible1
9952
4343
1368
348
3325
205
1There were 19 eligibles of unknown race
12
Table 5. Weighted Survey Process Rates by Race and Ethnicity, Add Health Wave IV
Race
Rate
White
Black
Asian/Pacific
Islander
Native
American
Hispanic
Other
Response
83.4
78.0
69.2
79.4
73.6
69.0
Contact
95.3
88.0
88.5
87.4
86.8
84.7
Refusal
9.4
6.4
15.1
5.8
9.5
11.7
Total
Eligible1
9450
4013
1318
328
3151
189
1 There were 18 eligibles of unknown race
Table 6. Unweighted Survey Process Rates by Urban/Rural, Add Health Wave IV
Rate
Urban
Rural
Response
84.2
77.4
Contact
94.6
90.4
Refusal
8.5
9.5
Total Eligible1
8331
11051
1 There were 178 eligibles whose urban/rural status was unknown.
Table 7. Unweighted Survey Process Rates by Region, Add Health Wave IV
Rate
West
Midwest
South
Northeast
Response
76.5
85.7
82.3
72.1
Contact
89.8
95.0
93.2
89.4
Refusal
9.8
7.4
8.8
12.0
Total Eligible1
4654
4547
7152
2769
1 There were 438 eligibles whose region was unknown.
13
Table 8. Unweighted Survey Process Rates by Parental Education, Add Health Wave IV
Rate
Less than High School
High School
Some College
College Grad
Response
76.8
80.0
82.7
81.7
Contact
88.3
92.0
93.6
94.1
Refusal
8.2
9.1
8.7
9.7
Total Eligible1
2308
5586
3886
6546
1 There were 1234 eligibles with unknown parental education.
Table 9. Unweighted Survey Process Rates By Socioeconomic Status Scale1
SES Stratum
2
3
4
5
6
7
8
9
10
Response
73.7
78
77.5
80.9
81.2
82.2
82.8
79.1
83.4
Refusal
7.5
8.4
9.6
8.8
9
8.8
9
10.9
9.3
Contact
84.8
90.4
90.2
91.6
93.2
93
94.6
93
95.1
Total Eligible2
981
1730
2140
1712
2500
2318
1765
2076
3359
1 Socioeconomic status is measured on an ordinal scale from lowest (2) to highest (10) that
measures a participant’s socioeconomic status at Wave I based on parent education and
occupation.
2 There were 979 eligible respondents of unknown socioeconomic status.
Table 10. Unweighted Survey Process Rates by Immigrant Generation, Add Health Wave IV
Rate
First Generation1
Second Generation2
Third+ Generation3
Response
67.0
77.2
82.3
Contact
85.5
91.7
93.1
Refusal
13.5
10.8
8.3
Total Eligible4
1552
2830
14923
1 foreign-born to foreign-born parents
2 native-born to foreign-born parents
3 native-born to native-born parents
4 There were 255 eligibles whose immigration generation was unknown
14
Table 11. Unweighted Survey Process Rates by Genetic Relatedness, Add Health Wave IV
Rate
Twin
Full Sibling
Half Sibling
Non-Related
Response
85.8
86.0
82.2
77.6
Contact1
96.1
96.0
92.9
88.2
Refusal
8.5
7.9
6.6
7.9
Total Eligible1
1531
2145
708
953
1There were 14,223 eligibles who were not in the genetic sample.
Table 12. Bias Remaining in Estimated Health and Physical Activities Reported at the Wave I
In-home Interview.
Variable from Wave I In-home interview
Prevalence1
(%)
% Bias
remaining
% Relative Bias
remaining
Lacking current health insurance
12.4
-0.62
-5.02
Needed, did not get
medical care 1
18.3
0.46
2.54
Reported poor to fair health
6.8
0.12
1.71
Participated in team sports
at least weekly
68.1
0.45
0.66
Participated in aerobic activity at least
weekly
82.2
0.44
0.53
Obese using self-report BMI
8.8
0.39
4.39
Physically disabled
2.4
-0.08
-3.34
Emotionally disabled
4.1
0.03
0.79
1 Prevalence is percent of all eligible Wave IV respondents (i.e., all Wave I respondents eligible to
participate in Wave IV) who meet the indication of the variable from the Wave I interview.
15
Table 13a. Bias Remaining in Substance Use Reported at the Wave I In-home Interview.
Variable from Wave I In-home interview
Prevalence
(%)
% Bias
remaining
% Relative Bias
remaining
ever tried marijuana
29.6
-0.23
-0.79
ever used hard drugs
11.7
0.13
1.11
ever smoked cigarettes daily
20.1
0.15
0.73
smoke cigarettes daily during the last
month
8.0
0.02
0.21
drink alcohol without family
38.1
0.25
0.65
get drunk once a month or more
16.9
-0.01
-0.04
Table 13b. Bias Remaining in Substance Use Index1 Reported at the Wave I In-home Interview.
Variable from Wave I In-home interview
Prevalence
(%)
% Bias
remaining
% Relative Bias
remaining
(0) never used substances
34.9
-0.522
-1.50
(1) tried smoking or drink alcohol once
a month or more
27.9
0.47
1.68
(2) regular smoker, or get drunk one or
more times a month, and no use of
marijuana or hard drugs
14.1
0.13
0.94
(3) used marijuana in the last month,
smoked or drank alcohol but no use of
hard drugs
8.1
-0.20
-2.44
(4) used hard drugs in any combination
with other substances
11.7
0.12
1.04
1The substance use index is an ordinal scale that measures the severity of risk involved with
specific or multiple substances: 0=never used substances; 1=tried smoking or drink once a
month or more; 2=regular smoker or get drunk one or more a month and no use of marijuana or
hard drugs; 3=used marijuana in the last month, smoked or drank alcohol but no use of hard
drugs; and 4=used hard drugs in any combination with other substances.
2A negative percentage indicates non-respondents are higher in the listed characteristic.
16
Table 14a. Bias Remaining in Violence and Delinquency Reported at the Wave I In-home
Interview.
Variable from Wave I In-home interview
Prevalence
(%)
% Bias
Remaining
% Relative Bias
Remaining
saw shooting or stabbing
11.3
-0.392
-3.49
threatened someone with a knife or gun
4.7
-0.36
-7.62
paint graffiti 1
9.2
-0.14
-1.50
damage property 1
18.9
0.33
1.74
shoplift 1
23.1
0.00
-0.02
in a serious physical fight 1
31.8
-0.49
-1.53
seriously injure someone 1
18.6
-0.46
-2.48
run away from home 1
8.5
-0.20
-2.34
steal a car 1
10.2
-0.25
-2.47
steal goods worth $50 or more1
4.8
-0.22
-4.62
burglarize a building 1
4.6
0.03
0.72
use or threaten others with a weapon 1
4.1
-0.11
-2.72
sell drugs 1
7.8
-0.14
-1.73
steal goods worth less than $50 1
18.6
0.32
1.70
take part in a group fight 1
19.2
-0.42
-2.21
1 Reports are for past year, (1994-1995)
2A negative percentage indicates non-respondents are higher in the listed characteristic.
17
Table 14b. Bias Remaining in Violence and Delinquency Indices Reported at the Wave I In-
home Interview.
Variable from Wave I In-home interview
Prevalence
(%)
% Bias
Remaining
% Relative
Bias
Remaining
Delinquency Index1
0
59.9
-0.053
-0.09
1
19.8
0.06
0.32
2
10.0
0.06
0.62
3+
18.1
-0.07
-0.40
Violence Index2
0
52.2
0.59
1.12
1
13.8
0.09
0.65
2
12.2
-0.22
-1.79
3+
20.4
-0.46
-2.23
1 Higher values indicate greater delinquency. The delinquency index is created from nine behaviors
reported at Wave I including paint graffiti, damage property, shoplift, runaway from home, steal a car,
sell drugs, and burglary. The count of delinquent acts is expressed as a proportion of all possible and
non-missing responses multiplied by 9.
2 Higher values indicate greater violence. The violence index is created from nine behaviors reported at
Wave I including such items as fighting, pulled a knife or gun on someone, shot or stabbed someone, and
used a weapon in a fight. The count of violent acts is expressed as a proportion of all possible and non-
missing responses multiplied by 9.
3A negative percentage indicates non-respondents are higher in the listed characteristics.
18
Table 15. Bias Remaining in Family Structure Reported at the Wave I In-home Interview.
Variable from Wave I In-home interview
Prevalence
(%)
% Bias
remaining
% Relative Bias
remaining
Family Structure
2 biological parents
52.7
1.43*1
2.72
2 parents
17.8
-0.242
-1.35
single mom
19.4
-0.65
-3.35
single dad
3
-0.07
-2.3
other
6
-0.47*
-7.88
1* Denotes that the bias is significantly different from zero.
2A negative percentage indicates non-respondents are higher in the listed characteristics.
Table 16. Bias Remaining in Hearing Vocabulary (AHPVT)1 Measured at the Wave I In-home
Interview.
Variable from Wave I In-home interview
Prevalence
(%)
% Bias
remaining
% Relative Bias
remaining
Hearing Vocabulary
< 70
2.5
-0.86*2
-34.953
(AHPVT)
70 – 90
19.7
-1.16*
-5.96
91 - 110
49.8
0.45
0.93
> 110
28
1.58*
5.69
1The AHPVT is standardized to a mean of 100 and a standard deviation of 15.
2* Denotes that the bias is significantly different from zero.
3A negative percentage indicates non-respondents are higher in the listed characteristics.
19
Table 17. Prevalence, Bias, and Relative Bias Remaining in 13 Selected Health Risk Measures
Reported at the Wave I In-home Interview.
Variable from Wave I In-home
interview
Prevalence (%)
% Bias
remaining
% Relative Bias
remaining
Inactive1
5.5
-0.235
-4.22
Smoked 2
27.5
0.32
1.16
Drink 2
47
0.29
0.63
Drunk 2
28.8
0.08
0.27
Fought 2
32.5
-0.62
-1.91
Skipped School 2
28.8
-0.85
-3.05
Lied to Parents 2
51.6
0.72
1.39
No appetite 3
35.4
0.74*6
2.1
Felt Depressed 3
38.5
0.21
0.55
Felt Tired 3
56.7
0.41
0.73
Felt Isolated
26.7
-0.13
-0.47
Felt Unhappy at School
33.7
-0.01
-0.03
Felt Unsafe at School
29.5
-0.08
-0.26
* Denotes that the bias is significantly different from zero.
A negative percentage indicates non-respondents are higher in the listed characteristics.
1 Does not exercise at least once on normal weeks
2 Reports are for experiencing the attitude or feeling during the past 12 months.
3 Reports are for experiencing the attitude or feeling most or all of the time during the past week.
4 Prevalence is percent of all eligible Wave IV respondents (i.e., all Wave I respondents eligible to
participate in Wave IV), who meet the indication of the variable from the Wave I interview.
5* Denotes that the bias is significantly different from zero.
6A negative percentage indicates non-respondents are higher in the listed characteristics.
20
Table 18. Total and Component Bias Remaining in 13 Health Risk Measures Reported at the
Wave I In-Home Interview.
Health Risk Indicator
% Total
% No Contact
% Unable
% Refusal
% Other
Inactive 1
-0.23
-0.09
-0.09
-0.03
-0.03
Smoked
0.32
-0.07
0.23*2
0.11
0.05
Drink
0.29
0.24
0.19
-0.22
0.09
Drunk
0.08
0.23
0.12
-0.31
0.04
Fought
-0.62
-0.91*
0.11
0.18
0.01
Skipped School
-0.85
-0.60*
0.01
-0.28
0.02
Lying to Parents
0.72
0.25
0.18
0.26
0.03
No appetite 1
0.74*
0.07
0.09
0.50*
0.09*
Felt Depressed 1
0.21
-0.31
0.09
0.39
0.05
Felt Tired 1
0.41
-0.24
0.14
0.48
0.03
Felt Isolated
-0.13
-0.28
0.04
0.08
0.03
Felt Unhappy at School
-0.01
-0.35
0.15
0.13
0.06
Felt Unsafe at School
-0.08
-0.28
-0.04
0.25
0.00
1 Reports are for experiencing the attitude or feeling most or all of the time during the past week.
2* Denotes that the bias is significantly different from zero.
21
Table 19. Biological sex Breakdown of Total and Component Bias Remaining in 13 Selected
Health Risk Measures at the Wave I In-home Interview.
Health Risk Indicator
% Total Bias
% Bias for Males
% Bias for Females
Inactive 1
-0.23
-0.37
-0.16
Smoked
0.32
-0.07
0.67
Drink
0.29
0.05
0.53
Drunk
0.08
-0.03
0.24
Fought
-0.62
-0.55
-0.12
Skipped School
-0.85
-0.84*2
-0.72
Lying to Parents
0.72
0.61
0.69
No appetite 1
0.74*
0.41*
0.61
Felt Depressed 1
0.21
0.06
-0.03
Felt Tired 1
0.41
0.19
0.51
Felt Isolated
-0.13
-0.29
0.05
Felt Unhappy at School
-0.01
-0.38
0.3
Felt Unsafe at School
-0.08
-0.29
0.09
1 Reports are for experiencing the attitude or feeling most or all of the time during the past week.
2* Denotes that the bias is significantly different from zero.
22
Table 20. Racial Breakdown of Total and Component Bias Remaining in 13 Selected Health
Risk Measures at the Wave I In-home Interview.
Health Risk
Indicator
% Total
Bias
% Bias
for
Whites
% Bias
for
Blacks
% Bias
for
Asians
% Bias
for Nat.
Am.
% Bias
for
Hispanics
%
Bias
for
Other
Races
Inactive 1
-0.23
-0.2
-0.02
-0.38
0.14
-0.67
0.71
Smoked
0.32
0.17
-1.05
1.29
-0.25
-0.13
3.49
Drink
0.29
-0.38
0.4
0.55
-2.48
2.01
6.14
Drunk
0.08
-0.49
-0.01
2.51
-3.15
1.08
0.88
Fought
-0.62
-0.44
-1.23
1.12
-1.6
-0.14
0.56
Skipped School
-0.85
-0.94*2
-0.5
2.11
-4.08
-0.26
0.41
Lying to Parents
0.72
0.3
1.05
2.47
1.6
2.66*
-0.62
No appetite 1
0.74*
0.44
1.37*
1.71
2.05
1.69
1.7
Felt Depressed 1
0.21
0.3
0.58
1.03
-0.36
0.03
1.24
Felt Tired 1
0.41
0.61
1.05
0.24
0.59
-0.23
-0.16
Felt Isolated
-0.13
-0.03
-0.25
2.27
-4.39
0.08
1.18
Felt Unhappy at
School
-0.01
-0.04
-0.34
-0.4
-4.23
0.97
-1.68
Felt Unsafe at
School
-0.08
0.08
-0.29
0.3
-1.75
0.97
2.82
1 Reports are for experiencing the attitude or feeling most or all of the time during the past week.
2* Denotes that the bias is significantly different from zero.
- CitationsCitations7
- ReferencesReferences0
- This represents 73% of the baseline sample, when they were young adults aged 18–26 in 2001–2002. Analyses of Add Health data suggest that nonresponse from Wave I to Wave III does not introduce significant bias to estimates obtained from the sample (Chantala, Kalsbeek, and Andraca 2005; Kalsbeek, Morris, and Vaughn 2001; Kalsbeek, Yang, and Agans 2002). The Add Health data-set is ideal for our research purposes as it includes an oversample of minority groups (see the section Analytic Plan).
[Show abstract] [Hide abstract] ABSTRACT: Objective: The objectives of this study are to examine racial and ethnic differences in suicidal behaviour, its main risk factors, and the effect of the risk factors on suicidal behaviour in young adults in the United States. Design: Using nationally representative data (n=10,585) from Add Health, we calculate the prevalence of suicidal behavior and associated risk factors for non-Hispanic White, non-Hispanic Black, and Hispanic youth (aged 18-26) using logistic regression models of suicidal ideation stratified by race. Results: Non-Hispanic White and Hispanic young adults have higher rates of suicidal ideation than their non-Hispanic Black counterparts, but racial/ethnic differences in attempts are not statistically significant. Non-Hispanic Whites and Hispanic young adults are more likely to possess key risk factors for suicide. With the exception of substance use variables (i.e. alcohol and marijuana use) which appear to be more conducive to suicidal ideation in non-Hispanic Black than in non-Hispanic White young adults, the effects of risk factors appear to be similar across race/ethnicity. Conclusion: The higher prevalence of suicidal ideation in non-Hispanic White and Hispanic young adults may be driven by their greater exposure to risk factors, as opposed to differences in the effects of these risk factors. More research is needed to uncover why non-Hispanic White and Hispanic young adults have higher rates of suicidal ideation than their non-Hispanic Black counterparts; yet, rates of suicide attempts are comparable and non-Hispanic White young adults have the highest rate of completed suicides.- [Show abstract] [Hide abstract] ABSTRACT: Although exposure to peer and family violence is a documented risk factor for adolescent dating violence, less is known about the relationship between violent crime exposure and dating violence victimisation. Participants in the National Longitudinal Study of Adolescent Health (n = 4794) aged 13-17 years self-reported witnessing violent crime (someone being shot or stabbed) in the 12 months prior to Wave I interview (1994-95), physical partner violence victimisation within the 18 months prior to Wave II interview (1995-96), and physical and sexual partner violence victimisation within the 18 months prior to Wave III interview (2001). Twelve per cent of respondents reported dating violence victimisation at Wave II. Witnessing violent crime was positively associated with victimisation in crude (OR = 2.11, 95% CI 1.56 to 2.86) and adjusted (AOR = 1.53, 95% CI 1.09 to 2.15) analyses. Of the adolescent partner violence victims (n = 549), 32% reported continued victimisation into early adulthood; after adjusting for gender, age, urbanicity and childhood maltreatment history, witnessing violent crime in adolescence was negatively associated with having non-violent relationships in early adulthood (AOR = 0.40, 95% CI 0.19 to 0.84). In cross-sectional and longitudinal analyses, associations between violent crime exposure and victimisation did not vary by age, gender or race/ethnicity. Adolescents exposed to violent crime experience an increased risk of partner violence victimisation in adolescence and continuing victimisation into adulthood. Targeting dating violence prevention and intervention programmes to geographic areas with high levels of violent crime may be an efficient strategy to reach higher risk adolescents. Reducing community violent crime may also have spillover effects on partner violence.
- [Show abstract] [Hide abstract] ABSTRACT: To determine the prevalence of patterns of intimate partner violence (IPV) victimization from adolescence to young adulthood, and document associations with selected sociodemographic and experiential factors. We used prospective data from the National Longitudinal Study of Adolescent Health to group 4134 respondents reporting only opposite-sex romantic or sexual relationships in adolescence and young adulthood into four victimization patterns: no IPV victimization, adolescent-limited IPV victimization, young adult onset IPV victimization, and adolescent-young adult persistent IPV victimization. Forty percent of respondents reported physical or sexual victimization by young adulthood. Eight percent experienced IPV only in adolescence, 25% only in young adulthood, and 7% showed persistent victimization. Female sex, Hispanic and non-Hispanic black race/ethnicity, an atypical family structure (something other than two biologic parents, step-family, single parent), more romantic partners, experiencing childhood abuse, and early sexual debut (before age 16) were each associated with one or more patterns of victimization versus none. Number of romantic partners and early sexual debut were the most consistent predictors of violence, its timing of onset, and whether victimization persisted across developmental periods. These associations did not vary by biological sex. Substantial numbers of young adults have experienced physical or sexual IPV victimization. More research is needed to understand the developmental and experiential mechanisms underlying timing of onset of victimization, whether victimization persists across time and relationships, and whether etiology and temporal patterns vary by type of violence. These additional distinctions would inform the timing, content, and targeting of violence prevention efforts.
- [Show abstract] [Hide abstract] ABSTRACT: No longitudinal analyses using national data have evaluated the increase in obesity from adolescence into early adulthood. We examined obesity incidence, persistence, and reversal in a nationally representative cohort of US teens followed into their early 30s, using measured height and weight data, in individuals enrolled in wave II (1996; 12-21 years), wave III (2001; 17-26 years), and wave IV (2008 early release data; 24-32 years) of the National Longitudinal Study of Adolescent Health (N = 8,675). Obesity was defined as a BMI >or=95th percentile of the 2000 Centers for Disease Control/National Center for Health Statistics growth charts or >or=30 kg/m(2) for individuals <20 years and >or=30 kg/m(2) in individuals >or=20 years. In 1996, 13.3% of adolescents were obese. By 2008, obesity prevalence increased to 36.1%, and was highest among non-Hispanic black females (54.8%). Ninety percent of the obese adolescents remained obese in 2008. While annual obesity incidence did not decline in the total sample across the two study intervals (2.3% per year 1996-2001 vs. 2.2% per year 2001-2008), rates among white females declined (2.7 to 1.9% per year) and were highest among non-Hispanic black and Hispanic females (3.8 and 2.7% per year, 1996-2001 vs. 3.0 and 2.6% per year, 2002-2008, respectively). Obesity prevalence doubled from adolescence to the early 20s, and doubled again from the early to late 20s or early 30s, with strong tracking from adolescence into adulthood. This trend is likely to continue owing to high rates of pediatric obesity. Effective preventive and treatment efforts are critically needed.
- [Show abstract] [Hide abstract] ABSTRACT: To compare the sexual behaviors of young people in South Africa (SA) and the United States (US) with the aim to better understand the potential role of sexual behavior in HIV transmission in these two countries that have strikingly different HIV epidemics. Nationally representative, population-based surveys of young people aged 18-24 years from SA (n = 7,548) and the US (n = 13,451) were used for the present study. The prevalence of HIV was 10.2% in SA and <1% in the US. Young women and men in the US reported an earlier age of first sex than those in SA (mean age of coital debut for women: US [16.5], SA [17.4]; for men: US [16.4], SA [16.7]). The median number of lifetime partners is higher in the US than in SA: women: US (4), SA (2); men: US (4), SA (3). The use of condom at last sex is reported to be lower in the US than in SA: women: US (36.1%), SA (45.4%); men: US (48%), SA (58%). On average, young women in SA report greater age differences with their sex partners than young women in the US. Young people in the US report riskier sexual behaviors than young people in SA, despite the much higher prevalence of HIV infection in SA. Factors above and beyond sexual behavior likely play a key role in the ongoing transmission of HIV in South African youth, and thus should be urgently uncovered to develop maximally effective prevention strategies.
- [Show abstract] [Hide abstract] ABSTRACT: Based on the stage environment and the person environment fit perspectives, the current study examined the relation between school disciplinary policies and offending from adolescence into young adulthood. Using Waves I and III of the National Longitudinal Study of Adolescent Health (a.k.a., Add Health), hierarchical multinomial logistic regression models were utilized to test whether school disciplinary policies were related to offending patterns during adolescence and young adulthood. Descriptive results suggest that, overall, severe school policies were not associated with the course of offending. However, relations between individual characteristics (i.e., inattention and impulsivity) and offending patterns did appear to differ depending on the severity of disciplinary policies. Within schools with more severe policies, adolescents scoring higher on inattention were more likely to be in the adolescent-limited offender group over the persistent offender group. On the other hand, adolescents with high levels of impulsivity were more likely to be in the persistent group over the non-offender group within schools with more severe policies. The results suggest that severe policies may not be effective for all students and the policies, alone, may not be promising avenues for the prevention of offending during adolescence and young adulthood.
Data provided are for informational purposes only. Although carefully collected, accuracy cannot be guaranteed. Publisher conditions are provided by RoMEO. Differing provisions from the publisher's actual policy or licence agreement may be applicable.
This publication is from a journal that may support self archiving.
Learn more