You are on page 1of 19

This article was downloaded by: [Eastern Michigan University]

On: 11 October 2014, At: 15:24


Publisher: Taylor & Francis
Informa Ltd Registered in England and Wales Registered Number: 1072954 Registered
office: Mortimer House, 37-41 Mortimer Street, London W1T 3JH, UK

Journal of Hunger & Environmental


Nutrition
Publication details, including instructions for authors and
subscription information:
http://www.tandfonline.com/loi/when20

Adult Caregiver Reports of Adolescents’


Food Security Do Not Agree Well With
Adolescents’ Own Reports
a b
Mark Nord & Karla Hanson
a
Economic Research Service , U.S. Department of Agriculture ,
Washington , DC , USA
b
Division of Nutritional Sciences , Cornell University , Ithaca , New
York , USA
Published online: 10 Dec 2012.

To cite this article: Mark Nord & Karla Hanson (2012) Adult Caregiver Reports of Adolescents’ Food
Security Do Not Agree Well With Adolescents’ Own Reports, Journal of Hunger & Environmental
Nutrition, 7:4, 363-380, DOI: 10.1080/19320248.2012.732926

To link to this article: http://dx.doi.org/10.1080/19320248.2012.732926

PLEASE SCROLL DOWN FOR ARTICLE

Taylor & Francis makes every effort to ensure the accuracy of all the information (the
“Content”) contained in the publications on our platform. However, Taylor & Francis,
our agents, and our licensors make no representations or warranties whatsoever as to
the accuracy, completeness, or suitability for any purpose of the Content. Any opinions
and views expressed in this publication are the opinions and views of the authors,
and are not the views of or endorsed by Taylor & Francis. The accuracy of the Content
should not be relied upon and should be independently verified with primary sources
of information. Taylor and Francis shall not be liable for any losses, actions, claims,
proceedings, demands, costs, expenses, damages, and other liabilities whatsoever or
howsoever caused arising directly or indirectly in connection with, in relation to or arising
out of the use of the Content.

This article may be used for research, teaching, and private study purposes. Any
substantial or systematic reproduction, redistribution, reselling, loan, sub-licensing,
systematic supply, or distribution in any form to anyone is expressly forbidden. Terms &
Conditions of access and use can be found at http://www.tandfonline.com/page/terms-
and-conditions
Journal of Hunger & Environmental Nutrition, 7:363–380, 2012
Copyright © Taylor & Francis Group, LLC
ISSN: 1932-0248 print/1932-0256 online
DOI: 10.1080/19320248.2012.732926

Adult Caregiver Reports of Adolescents’ Food


Security Do Not Agree Well With Adolescents’
Own Reports

MARK NORD1 and KARLA HANSON2


1
Economic Research Service, U.S. Department of Agriculture, Washington, DC, USA
Downloaded by [Eastern Michigan University] at 15:24 11 October 2014

2
Division of Nutritional Sciences, Cornell University, Ithaca, New York, USA

The accuracy of national statistics on children’s food insecurity


depends on the reliability of interview responses by adults. In this
study of 395 adolescents ages 15 to 17 in National Health and
Nutrition Examination Survey data, adolescents’ self-reported food
insecurity was more common than, and only weakly associated
with, adult proxy reports of those adolescents’ food insecurity.
Adolescents’ dietary quality as measured by the Healthy Eating
Index was best when both reports indicated food security, worst
when both indicated food insecurity, and intermediate with no sys-
tematic difference between the 2 groups when reports disagreed,
leaving unanswered the question of which report is more reliable.

KEYWORDS food security, food security measurement, food


insecurity, adolescents, NHANES, proxy report

The US Department of Agriculture (USDA) monitors the food security of


US households—meaning their consistent access to adequate food for active
healthy living for all household members—and reports the extent of food
insecurity at several levels of severity. Separate statistics are provided on
the extent and severity of food insecurity among children. Food security is
especially important for children because the nutritional content of their diets
affects not only their current health but also their physical, mental, and social
development—and thus their future health and well-being.1

Disclaimer: Views are those of the authors and may not be attributed to the Economic
Research Service, the U.S. Department of Agriculture, or Cornell University.
Address correspondence to Mark Nord, Economic Research Service, Room 5-232,
1400 Independence Ave., SW, Mail Stop 1800, Washington, DC 20250-1800. E-mail: Marknord@
ers.usda.gov

363
364 M. Nord and K. Hanson

Statistics on children’s food security published by the USDA are based


on information collected from adult respondents (parents of the children in
most cases). Concern was expressed early in the development of the cur-
rent food security measurement methods that the extent of food insecurity
among children could be underreported by parents, either because of embar-
rassment or shame or out of fear of government intervention.2 Additionally,
in-depth interviews of youth and their parents suggest that parents may not
always be aware of their children’s knowledge about food problems in the
household and the extent to which the children modify or reduce their
own food intake.3,4 In part this is because some children hide or disguise
these behaviors out of consideration for their parents or to protect younger
children.
Downloaded by [Eastern Michigan University] at 15:24 11 October 2014

This study compares self-reports of food insecure behaviors and expe-


riences by adolescents ages 15 to 17 with parents’ proxy reports of the same
behaviors and experiences, using nationally representative National Health
and Nutrition Examination Survey (NHANES) data. NHANES is the only
nationally representative data source that contains both self-reported per-
sonal food security and proxy-reported food security of adolescents. To our
knowledge, this is the first time that these data have been used to address
the important question of the accuracy of adult proxy reports of adolescents’
food security status. The study had 3 objectives:

1. To assess the extent to which adolescents and their parents agree or dis-
agree about the occurrence of specific food insecure conditions and the
severity of food insecurity experienced by the adolescents.
2. To examine whether, and to what extent, any disagreement in assessments
may systematically bias estimates of the prevalence and severity of food
insecurity among adolescents, which are currently based on adult reports.
3. In cases where self-reported and proxy-reported food insecurity of ado-
lescents disagree, to assess whether one or the other report is more
consistent with the adolescents’ diet quality.

DATA

NHANES is a program of studies conducted by the National Center for Health


Statistics designed to assess the health and nutritional status of adults and
children in the United States. The survey examines a nationally representative
sample of about 5000 sampled persons (SPs) each year. Survey compo-
nents include a family interview, generally conducted in the home of the SP,
and a personal interview and physical examination conducted shortly there-
after in a mobile examination center (MEC). The family interview includes
demographic and socioeconomic questions, including questions about food
security at the household level. For adolescent SPs, the family interview is
Adult Caregiver Reports of Food Security 365

usually completed by a parent or other adult family member. The personal


interview includes questions about health conditions, dietary intakes, and
the SP’s personal food security. The personal interview is completed by the
SP himself or herself, for both adolescent and adult SPs. In this study, self-
reported personal food security of SPs was compared with food security
status reported in the family interview (which was by an adult proxy for
adolescents).
Our sample of primary interest comprised 395 adolescent SPs aged 15 to
17 years old in NHANES 2005–2006 and 2007–2008 living in households that
were food insecure or marginally food secure based on the family survey.
Two subsamples of adult SPs also were analyzed to provide standards of
comparison for the adolescent findings. The first comprised adult SPs in
Downloaded by [Eastern Michigan University] at 15:24 11 October 2014

households with a single adult (n = 253), for whom family interview food
insecurity and personal food insecurity were known to have been reported
by, and reflect the food security of, the same person. The second comprised
adult SPs in households with more than one adult and no child (n = 617). For
some of these adult SPs, a different adult responded to the family interview,
and family interview food security may have reflected the experience of a
different adult in the family.
For our analysis of the dietary quality of adolescents by food insecurity
status, the sample was expanded to include adolescents ages 16 and 17 years
in NHANES 2003–2004 (2003–2008 total n = 574). The 2003–2004 data
were not included in the main analysis because wording differences in
the personal food security questions prior to 2005 would have complicated
question-specific analyses. Food security classification based on multiple
questions, which was the framework for the diet quality analyses, was rel-
atively robust to the questionnaire changes, so the 2003–2004 data could
be included in those analyses. However, only adolescents ages 16 and
17 could be included from the 2003–2004 data because personal food
security questions were not administered to 15-year-old SPs prior to 2005.

Measures of Household-Level Food Security


Each household’s food security was assessed in the NHANES family inter-
view using the standard 18-item US Household Food Security Survey Module
(HFSSM). The HFSSM includes 3 questions about food conditions of the
household in general, 7 questions about food conditions of adults in the
household, and, if any children ages 0 to 17 are present, 8 questions about
their food conditions. Responses from the questions can be combined to pro-
vide measures of food security for the household as a whole, for all adults
in the household, and for all children in the household. The questions were
referenced to the last 12 months. Typically, one adult household member
responded to all of these food security questions, so the responses regarding
366 M. Nord and K. Hanson

adolescents’ food security and the classifications based on them are referred
to in this article as adult proxy reports.
In households with more than one child, the questions about chil-
dren’s food security were referenced to the children as a group or to
any of the children. Parents in food insecure households typically shield
children to some extent from more severe effects of food insecurity and
generally shield younger children to a greater extent than older children.
Therefore, indications of food insecurity among children reported by the
household respondent usually reflect conditions of the oldest child or chil-
dren. We restricted our sample to adolescents aged 15 years and older so
that any reports of children’s food insecurity in the family interview were
likely to have reflected food insecurity of the sampled adolescent.
Downloaded by [Eastern Michigan University] at 15:24 11 October 2014

In households with more than one adult, most of the questions used to
measure adults’ food security were referenced to the household in general or
to “you or other adults in the household.” Three of the adult-referenced ques-
tions are referenced to the respondent (“you” rather than “you or other adults
in the household”). These are questions about internal states, such as being
hungry, which survey specialists believe cannot reliably be answered by one
adult for another. The measure of adult food security therefore reflected
some blend of the experiences of the respondent and of the most severely
food insecure adult. However, unpublished analysis by one of the authors
indicates that in most multiple-adult food insecure households, all adults
experience similar levels of severity of food insecurity.

Measures of Personal Food Insecurity


The personal food security of SPs in food insecure or marginally secure
households was assessed in the NHANES in connection with medical exami-
nation in the MEC. Personal food security was assessed by 5 (or 6, depending
on the interview year and the age of the SP) food security questions about
the SP’s personal experience of food hardship. The questions were simi-
lar in content to the more severe questions in the household module but
were referenced to the previous 30 days rather than the previous 12 months
(Table 1).5 The personal food security questions were administered only
if the family interview indicated at least some measurable level of food
insecurity (raw score 1 or higher on the 18-question module).
The personal food security questions were administered to SPs aged
12 years and older. For children aged 12 to 15, administration of the per-
sonal food security questions was via audio computer-assisted self-interview
(ACASI), which provides a greater sense of privacy for the respondent.
Connell et al suggested that self-administration, either by ACASI or pencil-
and-paper, was preferable to interviewer administration for children in this
age range.6 Administration to adolescents aged 16 and 17 and to adults was
Downloaded by [Eastern Michigan University] at 15:24 11 October 2014

TABLE 1 Questions Administered in NHANES to Assess Personal Food Security and Corresponding Questions About Adult and Child Food
Security Administered to an Adult Respondent in the Family Interview

Household food security module (family interview)

Personal food security questions


(self-reported in personal Adult questions (self-reported or Child questions (reported by
Number and label interview)a reported by another adult)b adult proxy)c

1: Cut size of meals In the last 30 days, did you cut In the last 12 months, did you In the last 12 months, did
the size of your meals because (or other adults in your you ever cut the size of
your family didn’t have enough household) ever cut the size of any of the children’s meals
money for food? (responses: your meals or skip meals because there wasn’t
often/sometimes/never) because there wasn’t enough enough money for food?
money for food? (responses: (responses: yes/no)
yes/no)
2: Skipped meals In the last 30 days, did you skip In the last 12 months, did
meals because your family any of the children ever

367
didn’t have enough money for skip meals because there
food? (responses: wasn’t enough money for
often/sometimes/never) food? (responses: yes/no)
3: Ate less than felt In the last 30 days, did you eat In the last 12 months, did you No corresponding question.
should less than you felt you should ever eat less than you felt you This question was not
because your family didn’t should because there wasn’t analyzed for adolescents in
have enough money for food? enough money to buy food? the study
(responses: (responses: yes/no)
often/sometimes/never)
4: Hungry but didn’t eat In the last 30 days, were you In the last 12 months, were you In the last 12 months, were
hungry but didn’t eat because ever hungry but didn’t eat the children ever hungry
your family didn’t have enough because you couldn’t afford but you just couldn’t afford
food? (responses: enough food? (responses: more food? (responses:
often/sometimes/never) yes/no) yes/no)
(Continued)
Downloaded by [Eastern Michigan University] at 15:24 11 October 2014

TABLE 1 (Continued)

Household food security module (family interview)

Personal food security questions


(self-reported in personal Adult questions (self-reported or Child questions (reported by
Number and label interview)a reported by another adult)b adult proxy)c

5: Lost weight In the last 30 days, did you lose In the last 12 months, did you No corresponding question.
weight because you did not lose weight because you didn’t This question was not
have enough money for food? have enough money for food? analyzed in the study
(responses yes/no) (responses: yes/no)

368
6: Did not eat for In the last 30 days, did you not In the last 12 months, did you In the last 12 months, did
whole day eat for a whole day because (or other adults in your any of the children ever
your family didn’t have enough household) ever not eat for a not eat for a whole day
money for food? (responses: whole day because there because there wasn’t
often/sometimes/never; wasn’t enough money for enough money for food?
recoded to ever/never in the food? (responses: yes/no) (responses: yes/no)
public use data)
a
Alternative wording was used as appropriate for adults and adolescents and for adults living alone. The wording shown is that administered to adolescents
aged 15 to 17.
b
Parenthetical “or other adults in the household” was read if there was more than one adult in the household.
c
In households with only one child, the child’s name was substituted for “any of the children” and conforming changes in wording were made as appropriate.
Adult Caregiver Reports of Food Security 369

by trained interviewers using computer-assisted personal interview (CAPI)


methods. This study included only adolescents aged 15 to 17 because
younger children’s food security may not have been reflected in the assess-
ment of children’s food security obtained from the family interview if there
were older children in the household. The NHANES public use file does not
include a full household roster, so the presence of older children cannot be
determined for many households.
Although the cognitive content of the personal food security questions
was generally the same as that of corresponding child-referenced food secu-
rity questions in the family interview, there were several differences that
affect comparability of the family interview and self-reports.
Downloaded by [Eastern Michigan University] at 15:24 11 October 2014

● The personal food security questions were referenced to the last 30 days,
whereas the family interview questions were referenced to the last
12 months. Food insecurity in US households is not usually chronic, so
some of the food insecure conditions that occurred during the 12-month
period covered by the family interview report would not have occurred
during the 30-day period covered by the personal food security questions.
At the national level, the prevalence of very low food security during the
30 days prior to the food security survey in mid-December 2010 was 56%
of the prevalence assessed across the entire year.7
● The response format for the personal food security questions analyzed
in this study was often/sometimes/never, whereas the response format
to the corresponding questions in the family interview was yes/no. This
has implications for the study, because the often/sometimes/never format
increases the proportion of affirmative responses (considering responses
of often or sometimes as affirmative). Some respondents for whom an
experience was rare or episodic will say “sometimes” if that is a response
option but will say “no” if the question is asked with a yes/no response.
Analysis of NHANES data (not shown) indicates that this effect is substan-
tial, with odds of responding “often” or “sometimes” 1.5 to 2.0 times that
of responding “yes.”
● The personal food security question, “In the last 30 days, did you eat
less than you felt you should because your family didn’t have enough
money for food?” was consistent with the corresponding adult-referenced
question in the family interview but had no equivalent for children in the
family interview. Therefore, this question was analyzed for adult SPs but
not for adolescent SPs.
● The personal food security module included separate questions about cut-
ting the size of meals and skipping meals. This was consistent with the
child items in the family interview. However, in the adult section of the
family interview, the 2 behaviors were combined in a single question about
cutting the size of meals or skipping meals. To compare the consistency of
adult responses in the family and personal interviews, therefore, responses
370 M. Nord and K. Hanson

to the 2 questions in the personal interview were combined to form a


single “cut or skipped” item.

Diet Quality Information


Dietary quality measures were calculated from 24-hour dietary recall
data obtained in-person from adolescent respondents using the USDA’s
Automated Multiple Pass Method. Dietary quality was measured using the
Healthy Eating Index 2005 (HEI-2005), a reflection of how well dietary
intake complied with Dietary Guidelines for Americans.8 Respondents’ indi-
vidual foods were aggregated into groups that corresponded to the My
Pyramid food groups, using the MyPyramid Equivalents Database9 and an
Downloaded by [Eastern Michigan University] at 15:24 11 October 2014

addendum.10 Dietary densities were then calculated from these MyPyramid


groups and daily calories consumed and assessed against HEI-2005 density
standards. Nine of the HEI-2005 components reflect the adequacy of nutrient
intake via consumption of total fruit (and whole fruit), total vegetables (and
dark green and orange vegetables), total grains (and whole grains), milk,
meat and beans, and oils. The other 3 components reflect moderation stan-
dards for saturated fat, sodium, and “extra” calories (from solid fat, alcohol,
and added sugar [SoFAAS] combined). Ratings for each of the 12 compo-
nents were assessed individually and also summed into total HEI, which
could range from 0 to 100.

Demographic and Household Information


Gender and age of the SP and the number of persons in the SP’s household
were provided by the NHANES demographic data files. The public use files
do not include a complete household roster.

METHODS

Statistics assessing the concordance between adult proxy responses and ado-
lescents’ self-reports were compared with the same statistics assessing the
concordance of reports in the family and personal interviews by adults aged
18 to 64 in single-adult households. In these single-adult households, the
same adult reported his or her own food security in both interviews. The
extent of concordance between those 2 self-reports provides a baseline of
what the expected concordance would be for adolescents if the adult proxy
responded to questions about the adolescent’s food security in the family
interview exactly as the adolescent would have responded if reporting for
himself or herself.
Statistics assessing concordance were also calculated for adult SPs aged
18 to 64 in households with 2 or more adults and no child. The presence or
Adult Caregiver Reports of Food Security 371

absence of children was inferred by whether the child-referenced food secu-


rity questions were administered in the family interview. Concordance of
these responses is expected to be somewhat lower than for adults in single-
adult households, both because of lack of perfect knowledge by the family
respondent about the experiences of other household members and because
actual behaviors and experiences of food insecurity may differ somewhat
between adults in the same household. Adult SPs in households with chil-
dren were omitted from this analysis sample because it was not possible to
determine whether there were adults other than the SP in the household.
The following statistics were calculated for each question based on
cross-tabulation of the responses in the family interview and the self-reported
responses in the MEC interview.
Downloaded by [Eastern Michigan University] at 15:24 11 October 2014

● Percentage of SPs who affirmed the item about their personal food security
given that the equivalent item was affirmed with regard to them in the
family interview (expected to be high if reports are concordant).
● Percentage of SPs who affirmed the item about their personal food security
given that the equivalent item was denied with regard to them at the
household level (expected to be low if reports are concordant).
● Percentage of family interview respondents who affirmed the item with
reference to the SP or the SP’s age group, given that the SP affirmed the
equivalent personal item (expected to be high if reports are concordant).
● Pearson correlation between household-level and personal-level response
to equivalent items (expected to be high if reports are concordant). For 2
× 2 tables, the Pearson correlation is equivalent to Cramer’s V .
● Ratio of the number of SPs who affirmed a personal-level item to the
number of SPs for whom the equivalent item was affirmed at the household
level. This is not only a measure of concordance but also a basis for
assessing whether adult proxy reports generally overstate or understate
the extent of food insecurity among adolescents. The ratio would be near
unity if there were no overstatement or understatement and if questions
were asked with reference to the same time period and with the same
response set. The effects of the difference in time period and response set
were expected to be partially offsetting, but the extent and balance of the
offsetting effects was not known a priori.

The same statistics were calculated from cross-tabulations of food security


classifications based on multiple questions. Personal food security classifica-
tion was assigned based on whether the SP affirmed any of the personal
food security items that had equivalent items in the family interview.
Corresponding classifications (proxy reported for adolescents) were based
on the equivalent items in the family interview. Statistics based on cross-
tabulations are not reported for 3 items for adolescents and one item for
372 M. Nord and K. Hanson

adults because there were fewer than 10 persons with reports of these
conditions in both the family interview and personal interview. Ratios of
affirmative responses in the personal and family interview were calculated
for these questions, and responses to the questions were included in the
calculation of the classification variable.
The level of severity of food insecurity differentiated by these multiple-
question classifications is about midway between the thresholds for low and
very low food security in the standard classification schemes on which the
USDA publishes food security statistics. For ease of communication, the terms
food secure and food insecure are used hereafter to refer to these classifica-
tions, even though the criterion is at a more severe level than indicated by
those terms in the USDA’s food security reports.
Downloaded by [Eastern Michigan University] at 15:24 11 October 2014

Interpretation of the concordance of adolescent self-reports with adult


proxy reports is not straightforward. Differences in reference period (30 days
versus 12 months) and response options (often/sometimes/never versus
yes/no) between the food security questions in the family and MEC interview
were expected to be partially offsetting, but the relative size of their effects
was not known. Therefore, rather than interpreting concordance statistics for
adolescents directly, we compared them with the same statistics for adult SPs
in households with no other adults. The difference in reference period and
response options should have had the same effects on adolescent and adult
responses, so comparing concordance statistics for adolescents with those
for adults effectively controls for those 2 differences. With rare exceptions,
the adult SP in households with no other adults would have responded with
regard to his or her own food security in the family interview as well as in the
MEC interview, so the concordance between their responses in the 2 inter-
views is a standard for judging the accuracy of adults’ proxy reports of the
food security of adolescents. The comparison with adult SPs in households
with 2 or more adults and no children provided a second standard against
which to assess concordance for adolescents. For many of these adult SPs,
the family interview report was provided by a different adult in the family.
Mean HEI total and component scores were compared across 4 groups
of adolescents identified by cross-classification by their self-reported and
proxy-reported food insecurity. Self-reported personal food insecurity was
identified by a raw score of 1 or more on the 5 personal food security
items, as described above. An approximately equal level of severity based
on the adult proxy report—identified by comparing the cognitive content of
the items in the 2 modules—was a raw score of 3 or more on the 8 child-
referenced items in the standard 8-item child module.
Response analyses were weighted by the MEC 2-year weights. HEI anal-
yses were weighted by the 1-day dietary recall data weights. Standard errors
for the Pearson correlations in the response analysis and the HEI means in
the dietary quality analysis were not adjusted for the complex sample design.
Such adjustment would have been difficult and uncertain given the small
Adult Caregiver Reports of Food Security 373

sample sizes and, in particular, given the small cell sizes in many analyses.
The difference in response consistency between adolescents and adults was
so large that there was little question that considerable difference exists in
the population. In the diet quality analyses, however, marginally significant
results should be interpreted cautiously.

FINDINGS

Responses by adolescents and by adult proxies to the question about cutting


the size of meals were only weakly associated. Among adolescent SPs in
households in which an adult proxy reported that a child had cut the size
Downloaded by [Eastern Michigan University] at 15:24 11 October 2014

of meals in the previous 12 months, only 25.6% of adolescents reported


that they had sometimes of often done so in the previous 30 days (Table 2,
measure 1). By comparison, the corresponding statistic was 69.4% for adult
SPs in households with no other adults and 59.9% for adult SPs in households
with 2 or more adults and no child. This statistic for the 2 items, “ate less
than felt should” and “hungry but didn’t eat” (calculated for adults only) was
also higher than for cut size of meals for adolescents.
Associations between multiple-item classifications based on equivalent
items in the family interview and personal interview were also weaker for
adolescents than for adults. In households in which the adult proxy reported
any of the 4 food insecure conditions, 41.2% of adolescent SPs affirmed one
or more of the equivalent questions. The corresponding statistic was 66.0%
for adult SPs in households with no other adults and 60.1% for adult SPs in
households with 2 or more adults and no child.
The obverse response pattern, affirming an item at the personal level
that was denied at the household level, should be near zero if responses
were perfectly objective and the questions were exactly equivalent. In these
data, however, there were 2 situations in which these seemingly inconsistent
responses may have occurred without indicating misreporting in either inter-
view. First, the condition or behavior could have occurred in the time period
between the family interview and the personal interview. This was likely to
have been rare, because the time period between the 2 interviews was typi-
cally only a few days. Second, a condition may have occurred so rarely that
it was denied in the family interview, in which the respondent had to choose
between “yes” and “no” but reported as having occurred “sometimes” in the
personal interview where that response was an option. On this statistic, ado-
lescent SP responses did not differ greatly from those of adult SPs (Table 2,
measure 2).
The third measure of concordance assessed was the extent to which a
problem reported in the personal interview was also reported in the fam-
ily interview. With the 2 exceptions described in the previous paragraph,
this statistic should be 100% if conditions were reported without error. For
Downloaded by [Eastern Michigan University] at 15:24 11 October 2014

TABLE 2 Measures of Concordance Between Reports of Equivalent Food Insecure Conditions in the Personal and Family Interviewsa

Measure 1 Measure 2 Measure 3 Measure 4 Measure 5

Family Correlation of
interview personal responses
Personal affirmative with family interview Ratio of
affirmative given responses affirmative
Personal affirmative given negative affirmative responses in
Personal food security given affirmative response in response in Pearson personal and
respondent and food response in family family personal correla- family interview
security questionb interview (%) interview (%) interview (%) tion P (ratio)

Adolescents aged 15–17 (N = 395)

374
Cut size of meals 25.6 20.6 7.3 0.03 .284 3.5
Skipped meals NA NA NA NA NA 2.1
Hungry but didn’t eat NA NA NA NA NA 1.2
Did not eat for whole day NA NA NA NA NA 1.4
Any of the questionsc 41.2 22.3 17.3 0.13 .004 2.4
Adult in households with only one adult (N = 253)
Cut size of meals or 69.4 22.9 78.4 0.46 <.001 .9
skipped meals
Ate less than felt should 61.9 11.4 83.1 0.53 <.001 .7
Hungry but didn’t eat 58.2 9.8 69.1 0.51 <.001 .8
Did not eat for whole day NA NA NA NA NA 1.2
Any of the questionsc 66.0 21.9 82.7 0.43 <.001 .8
Downloaded by [Eastern Michigan University] at 15:24 11 October 2014

Adult in households with more than one adult and no child (N = 617)
Cut size of meals or 59.9 24.6 54.6 0.35 <.001 1.1
skipped meals
Ate less than felt should 42.2 15.2 55.1 0.30 <.001 .8
Hungry but didn’t eat 41.7 11.3 46.9 0.32 <.001 .9
Did not eat for whole day NA NA NA NA NA .8
Any of the questionsc 60.1 23.8 63.8 0.37 <.001 .9

375
Source: Calculated by the authors using data from the National Health and Nutrition Examination Survey, 2005–2006 and 2007–2008.
NA = Not reported because there were fewer than 10 cases with affirmative response in both family interview and self-report.
a
Calculations were weighted using the MEC 2-year weights.
b
For complete question wording, see Table 1.
c
Classification based on “any of the items” included responses to questions that had too few affirmative responses to be analyzed individually: “skipped meals,”
and “hungry but didn’t eat” for adolescents and “did not eat for whole day” for both adolescents and adults.
376 M. Nord and K. Hanson

adolescent SPs, this statistic was only 7.3% for cut size of meals and 17.3% for
food insecurity classification based on the equivalent items (Table 2, mea-
sure 3). By contrast, for adult SPs in households with no other adults, these
statistics ranged from 69% to 83% and for adult SPs in households with 2 or
more adults from 47% to 64%.
Concordance between reports in the personal and family interviews
as measured by Pearson correlations was much weaker for adolescent SPs
than for adult SPs in households with no other adult (Table 2, measure 4).
Correlations for adult SPs in multiple-adult households with no child were
intermediate.
On average, adult proxies underreported the extent and severity of
food insecurity compared with adolescent self-reports. Adolescent SPs were
Downloaded by [Eastern Michigan University] at 15:24 11 October 2014

3.5 times as likely to affirm cutting the size of a meal as the adult proxy was
to affirm the equivalent item referenced to any children in the household
(Table 2, measure 5). The corresponding ratio was lower for the other items,
ranging from 1.2 to 2.1, and was 2.4 for classification based on the equiv-
alent items. By comparison, the numbers of adult SPs who reported each
condition were near, or slightly smaller than, the percentage for whom the
equivalent items were reported in the family interview.
Diet quality as represented by mean HEI scores was positively associ-
ated with reported food security but did not differ significantly between the
2 groups of adolescents whose self-reported food security differed from their
adult proxy-reported food security. Mean overall HEI was higher for adoles-
cents who were not insecure by either self- or proxy report (47.31) than
for those who were insecure by both (41.00; P < .01; Table 3). Mean HEI
scores were intermediate for those who only self-reported food insecurity
(46.17) and those for whom only the adult proxy reported food insecurity
(44.81). In the sample, mean HEI was lower in the group that was reported
food insecure only by the adult proxy, but the difference was not nearly
statistically significant (P = .50).
Patterns for 4 of the HEI components were similar to that of overall
HEI scores. Mean scores for whole fruit, total vegetables, saturated fat, and
SoFAAS were significantly higher (better diet quality) for adolescents who
were not insecure by either self- or proxy report than for those who were
insecure by both. Except for total vegetables, the in-sample means for these
components were lower (poorer diet quality) for adolescents who were
reported food insecure only by the adult proxy than for those who were
food insecure based on their self-report but not based on the proxy report.
However, only the difference for whole fruit was even marginally statistically
significant (P = .09).
Differences in mean HEI between the 2 groups with discordant food
security reports were statistically significant for 2 components, whole grains
and dark green and orange vegetables and legumes (lower mean for self-
reported food insecurity), and marginally significant for total fruit (lower
Downloaded by [Eastern Michigan University] at 15:24 11 October 2014

TABLE 3 Mean Healthy Eating Index (HEI) Scores by Adolescents’ Food Security Status as Self-Reported by Adolescents Aged 15 to 17 and
Proxy Reported by a Parent or Other Adult in the Householda

Group number 1 2 3 4 Difference, group 4 less


Difference, group 1 less group 3 (only
group 2 (both reported proxy-reported insecure
secure less both reported less only self-reported
Adolescent self-report Secure Insecure Insecure Secure insecure) insecure)

Proxy report Secure Insecure Secure Insecure HEI points Pb HEI points Pb

HEI-2005 total score 47.31 41.00 46.17 44.81 6.30 <.01 −1.36 .47
Total fruit 1.85 1.64 1.79 1.22 0.21 .52 −0.57 .06
Whole fruit 1.60 0.59 1.20 0.70 1.01 <.01 −0.50 .09
Total grains 4.20 4.15 3.97 4.18 0.05 .79 0.22 .26
Whole grains 0.66 0.59 0.36 0.85 0.07 .64 0.48 .00
Milk 5.30 5.04 4.98 5.39 0.26 .58 0.41 .48

377
Meat and beans 6.55 6.86 7.54 6.99 −0.32 .55 −0.55 .27
Total vegetables 2.16 1.37 1.93 2.03 0.79 <.01 0.10 .68
Dark green and orange 0.61 0.59 0.34 0.75 0.01 .95 0.41 .03
vegetables and legumes
Oils 5.30 5.28 4.92 4.91 0.01 .98 −0.02 .98
Saturated fat 5.99 3.88 6.07 5.84 2.11 <.01 −0.23 .67
Sodium 4.65 5.30 4.83 4.39 −0.65 .18 −0.44 .37
SoFAAS 8.44 5.70 8.24 7.57 2.74 <.01 −0.68 .51
Number of cases 347 43 63 121
Source: Calculated by the authors using data from the National Health and Nutrition Examination Survey, 2003–2004, 2005–2006, and 2007–2008.
a
Calculations were weighted using the one-day dietary recall data weights. Self-reported food insecurity was assessed at a level of severity about midway
between the thresholds for low food security and very low food security. Specifically, adolescent self-reported food insecurity was identified by a raw score
of 1 or more on questions 1–4 and 6 in Table 1, and proxy-reported adolescent food insecurity was identified by a raw score of 3 or more on the standard
8 child-referenced items administered in the family interview. Labels food secure and food insecure refer to these classifications.
b
Probability that the means do not differ in the population. These are random sample–based probabilities that do not adjust for the complex sample design of
the NHANES. Error variance may be underestimated.
378 M. Nord and K. Hanson

mean for proxy-reported food insecurity). However, for these 3 components,


differences were small and not statistically significant between the 2 groups
of adolescents for whom proxy and self-reports agreed.
The distribution of male and female adolescents across the 4 groups
in Table 3 did not differ greatly. Separate analysis by gender (not shown)
suggested that the patterns of differences in dietary quality described above
were more prominent among boys than girls, but sample sizes by gender
were too small to support reliable statistics.
Sensitivity of the HEI results to the level of severity at which food secu-
rity is assessed was tested by replicating the analyses described above with
the proxy-reported cutoff at raw score ≥ 4, near the severity level of very
low food security among children, and that for self-reported food insecurity
Downloaded by [Eastern Michigan University] at 15:24 11 October 2014

at an approximately comparable level of raw score ≥ 2. The results (not


shown) were generally similar to those in Table 3 even though only a small
number of adolescents (n = 13) remained in the category that was insecure
by both proxy and self-report. Key analyses were also repeated without sam-
pling weights. The findings did not differ in any important ways from those
reported.

DISCUSSION

Self-reported food insecurity by adolescents aged 15 to 17 was only weakly


associated with reports by parents or other adult proxies. Correlation
between adolescent and adult proxy responses was 0.13 for food security
classification based on 4 equivalent items. The corresponding statistic for
adult SPs in households with no other adult members was 0.43.
The results suggest that the prevalence of food insecurity among chil-
dren as a subpopulation may be underrepresented by adult proxy responses.
Adolescents aged 15 to 17 self-reported an incidence of food insecurity
2.4 times that reported by their adult proxies. It is likely that proxy reports
for younger children are more accurate, because caregivers have more direct
knowledge of the dietary intake of younger children. If so, the bias in esti-
mating the prevalence of food insecurity for children of all ages would be
smaller than for the older adolescents assessed in this study. Still, extrapo-
lating to the USDA’s 2010 national food security statistics, if adult proxies
underreported the incidence of food insecurity among adolescents to the
extent suggested by these analyses and reported the food security of chil-
dren aged 0 to 14 accurately, the number of households with food insecure
children would have been underreported by 1.8 million households, and the
true prevalence of food insecurity among children would have been 14.3%
rather than the 9.8% reported by the USDA.11
However, assessment of the diet quality of adolescents suggests that
adolescent self-reports were not necessarily more accurate than adult proxy
Adult Caregiver Reports of Food Security 379

reports. Average diet quality was best among adolescents who were food
secure by both self- and proxy report and worst among adolescents who
were food insecure by both reports. There was little systematic difference
in diet quality between adolescents who only self-reported food insecurity
and those who were food insecure only by proxy report. It may be that the
food security status of these adolescents was inconsistently reported because
many of them were on the edge of food insecurity—in a borderland neither
fully food secure nor severely insecure.
Several limitations should be kept in mind when interpreting these
results. Personal food security questions were only administered to individ-
uals in households that registered at least some indication of food insecurity
in the family interview. It seems unlikely that any substantial proportion of
Downloaded by [Eastern Michigan University] at 15:24 11 October 2014

adolescents would have been food insecure at the level of severity mea-
sured by the personal food security items if not even one of the least
severe adult/household or child items was affirmed in the family interview.
However, given the weak association between proxy and self-reports, the
possibility cannot be ruled out by these data. Any personal food insecurity
among adolescents in households classified as fully food secure based on
the family interview would increase the extent to which proxy-reported food
insecurity understates food insecurity that would be reported by the adoles-
cent. A second limitation is that severity levels were not directly comparable
between proxy-reported and self-reported items and classifications. Due to
differences in reference period (30 days versus 12 months) and response
formats, even questions with near-identical wording may not be directly
comparable, and raw score–based classification may only be approximately
comparable.
Further research could extend these analyses using the restricted
NHANES data. The household roster information in the restricted data would
permit analysis of all children aged 12 and older if they were the oldest child
in the household. Household composition data would also allow for inclu-
sion of additional adults for the comparison analyses because the number of
adults in all households would be known. Finally, the restricted data would
support comparison of personal food insecurity reported by adolescents and
adults in the same household, which could provide a richer picture of intra-
household food security relationships and an estimate of the extent to which
adults shield adolescents from levels of food insecurity experienced by adults
in the household.

REFERENCES

1. Nord M, Parker L. How adequately are food needs of children in low-income


households being met? Child Youth Serv Rev. 2010;32:1175–1185.
2. Hamilton WL, Cook JT, Thompson WW, et al. Household Food Security in the
United States in 1995: Technical Report. Alexandria, Va: Food and Consumer
Service, US Department of Agriculture; 1997.
380 M. Nord and K. Hanson

3. Frongillo, EA, Fram MS, Jones SJ. Children are not fully buffered from
experiencing food insecurity. FASEB J. 2010;24:104.3.
4. Fram MS, Frongillo EA, Jones SJ, et al. Children are aware of food insecurity and
take responsibility for managing food resources. J Nutr. 2011;141:1114–1119.
5. National Center for Health Statistics. National Health and Nutrition Examination
Survey 2005–2006: Documentation, Codebook, and Frequencies: Food Security.
Hyattsville, Md: National Center for Health Statistics, Center for Disease Control
and Prevention, Department of Health and Human Services; 2008. Available
at: http://www.cdc.gov/nchs/nhanes/nhanes2005-2006/FSQ_D.htm. Accessed
March 13, 2012.
6. Connell C, Nord M, Lofton K, Rehner T, Yadrick K. Food security of older
children can be assessed by using a standardized survey instrument. J Nutr.
2004;134:2566–2572.
Downloaded by [Eastern Michigan University] at 15:24 11 October 2014

7. Coleman-Jensen A, Nord M, Andrews M, Carlson S. Household Food Security


in the United States in 2010: Statistical Supplement. Washington, DC: Economic
Research Service, US Department of Agriculture; 2011. Administrative Publication
AP-057. Available at: http://www.ers.usda.gov/publications/ap/ap057. Accessed
March 13, 2012.
8. Guenther PM, Reedy J, Krebs-Smith SM, Reeve BB, Basiotis PP. Development
and Evaluation of the Healthy Eating Index—2005: Technical Report.
Alexandria, Va: Center for Nutrition Policy and Promotion, US Department of
Agriculture; 2007. Available at: http://www.cnpp.usda.gov/HealthyEatingIndex.
htm. Accessed March 13, 2012.
9. Bowman SA, Friday JE, Moshfegh A. MyPyramid Equivalents Database, 2.0 for
USDA Survey Foods, 2003–2004. Beltsville, Md: Food Surveys Research Group,
Beltsville Human Nutrition Research Center, Agricultural Research Service, US
Department of Agriculture; 2008. Available at: http://www.ars.usda.gov/ba/
bhnrc/fsrg. Accessed March 13, 2012.
10. Koegel KL, Kuczynski KJ. Center for Nutrition Policy and Promotion Addendum
to the MyPyramid Equivalents Database 2.0 [CNPP Addendum to MPED 2.0].
Alexandria, Va: Nutrition Guidance and Analysis Division, Center for Nutrition
Policy and Promotion, US Department of Agriculture; 2011. Available at: http://
www.cnpp.usda.gov/OtherProjects.htm. Accessed March 13, 2012.
11. Coleman-Jensen A, Nord M, Andrews M, Carlson S. Household Food Security
in the United States in 2010. Washington, DC: Economic Research Service,
US Department of Agriculture; 2011. Economic Research Report no. ERR-125.
Available at: http://www.ers.usda.gov/publications/err125. Accessed March 13,
2012.

You might also like