Professional Documents
Culture Documents
Food insecurity occurs “whenever the availability of nutri- al. 1999, Hamilton et al. 1997a). It was based on earlier
tionally adequate and safe foods or the ability to acquire research completed by the Radimer/Cornell (RC) research
acceptable foods in socially-acceptable ways is limited or un- team and by leaders of the Community Childhood Hunger
certain” (Life Sciences Research Office 1990). Previous re- Identification Project (CCHIP) (Radimer 1990, Radimer et al.
searchers have defined food insecurity as an experience of 1992, Wehler et al. 1992). The original data set used to
severe economizing of food resources (Bickel et al. 1996, develop the CFSM was the April 1995 food security supple-
Radimer 1990, Wehler et al. 1992). To effectively ameliorate ment of the Current Population Survey (CPS). As indicated
hunger and related health problems, policymakers depend on in Table 1, the CFSM contains 18 items, of which 8 pertain
surveillance or monitoring measures for valid and reliable only to households with children. The order of the items
information (Nord et al. 1990). If the Core Food Security represents the “CFSM modal response pattern” of respondents
Module (CFSM)4 monitoring tool for food insecurity is unre- who completed the 1995 food security supplement. As indi-
liable or inaccurate, vulnerable segments of the population cated in the first column of the table, categorization of house-
may be adversely affected. holds is based on the sum of affirmative responses. For exam-
The national food security measure called the CFSM was ple, responses to at least three items were required for
created by a team of experts to measure the extent and severity classification as “food insecure.” As explained by Carlson
of household food insecurity during a 12-mo period (Carlson et
(1999), “Determination of the initial threshold of each desig-
nated severity range was done by identifying the second or
1
Findings were reported as an oral presentation at the 2000 ASNS Annual
third items in the modal response pattern sequence that con-
Meeting. ceptually indicates the continuous characterizing of the cate-
2
Supported in part by a grant from the Institute for Research on Poverty, gory.” According to Bickel (1999), face validity of categoriza-
University of Wisconsin, Madison.
3
To whom correspondence should be addressed. tion was not a priority.
E-mail: laniwai4@pixi.com Previous work by the author confirmed the overall face
4
Abbreviations used: CCHIP, Community Childhood Hunger Identification validity (Derrickson and Anderson 2000) and “goodness-of-
Project, CFSM, Core Food Security Module, CPS, Current Population Survey,
HHS, Hawaiı̀ Health Survey, Q, Question, RC, Radimer/Cornell, RAI, Resource fit” of the CFSM items with Asians and Pacific Islanders in
Augmentation Index. Hawaiı̀ (Derrickson et al. 2000a). Limited resource audiences
749
750 DERRICKSON ET AL.
TABLE 1
Operationalized framework of the Core Food Security Module (CFSM) for households with children1
Food secure 2. Worried food would run out Diminished household food resources force economizing in food
Sum: ⬍34 3. Food bought didn’t last spending: running short of money, substituting cheaper, dietary
monotony
Food insecure 4. Family couldn’t afford balanced meals3 Food insecurity short of actual hunger: extreme food acquisition and
Sum: 3–7 5. Relied on a few low cost foods management coping strategies, use of socially non-normative
food resources, nutritional quality of diets and health affected
8. Cut size of meals/skip meals: adults
6. Could not afford balanced meal: child
in Hawaiı̀ (Caucasians, Hawaiians and part Hawaiians, Filipi- sure and the FVFSM. The measures evaluated are all measures
nos and Samoans) consistently reported that their experience that defined food security status categories, not to be confused
of “hunger” meant a cyclical pattern of inadequate intake (i.e., with the CFSM scale measure (Hamilton et al. 1997a). It was
“When you don’t know when your next meal is coming, or hypothesized that one of the alternative food security measures
where it’s coming from and/or how”). Similar to Radimer would be more credible to use in Hawaiı̀ than the CFSM.
(1990), the Face Valid Food Security Measure (FVFSM) cre- Until this time, no research team has independently evaluated
ated by the author was developed to be true to the “grounded the CFSM measure or compared the CFSM measure with
experience” reported by low-income Hawaiı̀ residents (Der- other food security measures (Radimer et al. 1992, Wehler et
rickson 1999, Glaser and Strauss 1967). As indicated in Table al. 1992).
2, similar to the RC and CCHIP measures, in the FVFSM a Credibility assessment was operationalized to include an
household with any affirmative response is categorized as “at assessment of reliability and validity. “Reliability refers to the
risk of hunger.” However, unlike all three other categorical consistency or reproducibility of an operational definition;
measures, hunger categorization in the CFSM requires specific validity refers to the goodness of fit between an operational
affirmation of hunger items. Adult hunger is determined by an definition and the concept it is purported to measure” (Sin-
affirmative response to either question (Q)10, “Respondent gleton et al. 1993). Concurrent validity was also defined by
hunger,” or Q12, “Any adult did not eat for a whole day.” Singleton et al. (1993) as “the ability of a measure to indicate
Child hunger is determined by an affirmative response to Q14, an individual’s present standing on the criterion variable.”
“Child hunger.” In summary, the FVFSM accepts the respon-
dent’s declaration of their experience as an adequate threshold
indicator of three food insecurity categories (at risk of hunger, MATERIALS AND METHODS
adult hunger and child hunger). In contrast, the CFSM mea- Samples. Three samples were surveyed for a total sample size of
sure categorizes various levels of severity of household food 1664, which were derived from the following:
insecurity through a pattern of multiple indicators, regardless 1. A convenience sample of 144 food pantry recipients
of the content of the items. 2. Sixty-one of the 144 respondents who completed the survey a
This study represents one part of a larger effort to determine second time; a mean of 11 d apart
the most effective food security monitoring tool to use in 3. A statewide sample of 1469 respondents who completed the
Hawaiı̀ Health Survey (HHS) in 1998
Hawaiı̀. The study objectives were to assess the credibility of Before data collection, all participants confirmed verbal consent as
applying the CFSM in Hawaiı̀ and to compare the concurrent required by a university human subjects review committee. This is the
validity of the CFSM with three alternative food security same data set we used to assess the validity and reliability of the
measures: an adapted RC measure, an adapted CCHIP mea- CFSM scale measure (Derrickson et al. 2000a). The population
HOUSEHOLD FOOD SECURITY MEASURES IN HAWAIÌ 751
TABLE 2
Comparison of four household food security measures by food security category
Adapted Community
Childhood Hunger
Category Core Food Security Module Face Valid Food Security Measure Adapted Radimer/Cornell Identification Project
Food secure ⬍3 affirmative responses 0 affirmative responses 0 affirmative responses 0 affirmative responses
“Only worried” Affirmative to Q2 “Worried
food would run out”
Food insecure/at Adults only: 3–6 affirmative Any affirmative response to food Any affirmative response to Any affirmative response to
risk of hunger responses; with 1⫹ child: insecurity indicators food insecurity indicators food insecurity indicators
3–7 affirmative responses except Q2
Severe/child Adults only: 9–10 Affirmative response to Q14 only Affirmative response to any Affirmative response to any
hunger affirmative responses; one of the following: Q7 one of the following: Q7
with 1⫹ children: 13⫹ or 13–16 or 13–16
affirmative responses
Summary Categorization based on Based on responses to specific Based on responses to Based on responses to
sum of responses indicators specific indicators specific indicators
Hunger categories may Separate hunger categories for Separate hunger categories Separate hunger categories
include both adult and adults and children for adults and children for adults and children
child hunger
Conservative, particularly in More liberal food insecurity More liberal classification More liberal food insecurity
classifying food classification but most strict of all categories and child hunger: more
insecurity hunger classification strict adult classification
sampled was representative of the ethnicity of Hawaiı̀ residents: 909 Specifically, the four resource augmentation questions queried
(54.6%) indicated they most identified with an Asian or a Pacific whether the following coping behaviors were used to enhance either
Islander ethnic group (Department of Business, Economic Develop- the household food supply or money for food: 1)using charitable food
ment and Tourism 1997). assistance, 2) delaying bill payments, 3) borrowing money for food,
Data collection. All data were collected in Hawaiı̀ between June and 4) sending children over to someone else’s house (only for
and November 1998. Data collection began with a pilot study of food households with children). These questions were used by the CFSM
pantry respondents who completed the survey once and 61 (80%) research team (Hamilton et al. 1997a). In addition, four follow-up
who also completed the survey a second time at an average of 11 d questions (“How often did you [resource augmentation behavior]”)
later. Data were gathered by interviewers who were experienced in were asked if the initial response was affirmative. Follow-up questions
calling limited resource households (Derrickson et al. 1995, SMS were asked to determine how reliant a household was on each
Research and Marketing Service, Inc. 1992) using standard telephone particular coping behavior. Response format to the follow-up ques-
survey methods to enhance response rates and minimize interviewer tions was consistent with that asked of the follow-up questions in the
bias (Lavarakas 1988, SMS Research and Marketing Service, Inc.
CFSM (almost every month, some months but not every month or
1998). Preliminary findings indicated that the CFSM was likely to be
reliable and valid to use in Hawaiı̀ (Derrickson 1999). only 1 or 2 mo). Thus, a maximum of eight resource augmentation
The final data set came from food security questions that were questions were asked of households with children, and six were asked
included in the HHS between September and November 1998. HHS of households without children. Affirmative responses were summed
is a telephone interview survey of at least 3500 households each year. to create the Resource Augmentation Index (RAI).
The survey is modeled after the National Household Interview Sur- Previous research indicated that in Hawaiı̀, food insecurity often
vey conducted by the National Center for Health Statistics (SMS led to compromised vegetable intake and increased reliance on an
Research and Marketing Service, Inc. 1998). Households were ran- inexpensive high-fat dried noodle product, locally called Saimin or
domly chosen from local telephone books. Once a household was Ramen, to stretch food resources (Derrickson and Anderson 2000).
chosen, the household was sent a letter from the director of the Survey respondents were specifically asked the following questions:
Department of Health encouraging survey participation. Data collec- 1. “Not counting salad or potatoes, how many servings of veg-
tion was administered through telephone interview using a computer- etables do you usually eat a day? Count 1/2 cup (120 mL), like
assisted telephone interviewing system (SMS Research and Market- the size of a pudding cup, as one serving.”
ing Service, Inc. 1998). A complete description of the data collection 2. “How many times last month did you or the child/children
methods used in the HHS is given elsewhere (SMS Research and (whoever ate more) eat Saimin that was purchased dried, not
Marketing Service, Inc. 1998). frozen?”
Survey instrument. The survey instrument included the follow- Analysis of responses to the vegetable frequency question was
ing: based on the serving sizes of the Food Guide Pyramid (U.S. Depart-
● Basic demographic questions (age, household composition and ment of Agriculture 1999). The vegetable frequency question was
ethnicity): “With what ethnic group do you identify most?” validated with a 24-h vegetable recall (Derrickson 1999). Values of
● The 18 CFSM questions preceded by the four-part food insuf- the two vegetable indices were highly correlated (r ⫽ 0.81, P
ficiency question (Rose et al. 1995) ⬍ 0.001). Mean values were not significantly different between
● Eight resource augmentation questions measures (1.61 ⫾ 1.15 for recall and 1.54 ⫾ 1.21 for vegetable
● Two diet-quality indicators frequency; t ⫽ ⫺0.7, P ⫽ 0.40). Responses to the Saimin question
752 DERRICKSON ET AL.
TABLE 3
Comparison of the household food security sum vs. item response for Hawaiı̀ data (n ⫽ 1664)
were all converted to a monthly frequency by multiplying weekly Data analysis. Data analysis can be broken into two parts
responses by 4. corresponding to our objectives:
Food security measures. The food security data were collected 1. An evaluation of the validity and reliability of applying the
and analyzed according to the “Guidelines for Using the Core Food CFSM categorical measure in Hawaiı̀
Security Module” (Price et al. 1997). The CFSM scale measures and 2. A concurrent validity comparisons of the four food security
item calibration values were created using the Rasch FACETS soft- categorical measures, which included the following:
ware program (Derrickson et al. 2000a, Linacre 1986, Rasch 1966,
Wright and Masters 1982, Wright and Stone 1979). As indicated in a. Comparisons of food security status prevalence estimates
Table 1, an “item calibration value” represents the position of the b. Comparison of dietary measures, the RAI and the CFSM
item along the constructed food insecurity scale. For example, Q16, scale measure
with a high positive item calibration value of 4.82, indicates a very The evaluation of the reliability and validity of the CFSM
high degree of insecurity and hunger, whereas an item with a low measure was initiated by an assessment of modal response patterns.
negative calibration, such as Q2 (⫺4.99), is indicative of mild food Specifically, the Hawaiı̀ modal response pattern was compared
insecurity (Hamilton et al. 1997). Similarly, household “scale mea- with the original CFSM modal response pattern (Hamilton et al.
sures” indicate the severity of household food insecurity reported by 1997a) to assess impact on categorization. To clarify, as outlined
the respondent. A higher number of affirmative responses result in a in Table 1, the CFSM modal response pattern is the sequence of
higher positive placement on the food insecurity scale. Scale mea- items ranging from Q2 to Q16. It is expected that if a respondent
sures ranged from a low of ⫺4.5, which is indicative of less severe responded in the affirmative to an item (i.e., Q10, “Respondent
food insecurity, to a high of ⫹4.5, which is indicative of severe hungry”), then she or he also responded in the affirmative to all of
hunger (Derrickson et al. 2000a). the preceding items (Q2 to Q9) (Hamilton et al. 1997a). If there
The total number of affirmative responses was called the “re- were notable differences in the item hierarchy between samples,
spondent food security sum.” The algorithms outlined in Table 2 then the CFSM categorical algorithm, which is based on a set sum
were applied to create the four food security categorical variables. of affirmative responses, would not be expected to consistently
Because both the CCHIP and RC measures do not contain all of categorize households.
the CFSM items, these measures were “adapted” to the CFSM To study the effect of modal pattern response rates on the CFSM
items. They are referred to as the adapted RC and the adapted categorical measure, a “Hawaiı̀ modal pattern” variable was created.
CCHIP measures. Most important, because the adapted CCHIP Specifically, those who followed the Hawaiı̀ modal response pattern
measure does not include Q2, “Worried food would run out,” a outlined in Table 3 were given a score of 1, and those who did not
new category of food insecurity was created. The “Only worried follow this modal pattern were given a score of 0. Adherence to the
about food insecurity” category was created within the adapted Hawaiı̀ modal response pattern was explored across the four CFSM
CCHIP measure to study the responses of those who responded food security categories. Finally, an investigation of content validity
affirmatively to Q2 but not to any other items. Creation of the of the CFSM was completed by comparing affirmative responses with
“Only worried about food insecurity” category is an applied re- key food security items (Q2, Q3, Q4, Q8, Q9, Q10, Q12 and Q14)
search artifact, not in anyway meant to discredit or alter the across the four CFSM food security categories.
original CCHIP measure designed to measure hunger among chil- To assess concurrent validity, the four food security measures were
dren (Wehler et al. 1992). compared across indices of resource augmentation and vegetable
HOUSEHOLD FOOD SECURITY MEASURES IN HAWAIÌ 753
TABLE 4
Affirmative responses to CFSM items by the CFSM household categorical measure1
n n (%)2
Q2. “Worried food would run out” 291 61 (4.3) 137 (86.7) 62 (96.9) 31 (100)
Q3. “Food bought didn’t last” 258 40 (2.8) 126 (79.7) 61 (95.3) 31 (100)
Q4. “Unable to eat balanced meals” 237 30 (2.1) 119 (75.3) 59 (93.7) 29 (96.7)
Q7. “Child/ren not eating enough” 60 1 (1.7) 15 (12.5) 20 (50.0) 24 (88.9)
Q8. “Adult/s cut size or skip meals” 127 1 (0.9) 40 (25.8) 56 (87.5) 30 (96.8)
intake and with mean values of the household scale measures (Der- respondents with only one affirmative response constituted
rickson et al. 2000a). To assess the relationship of the four categorical 25% of the respondents who followed the Hawaiı̀ modal re-
measures to the CFSM household scale measure, mean household sponse pattern. Twenty-four (39%) of these 62 respondents
scale measures were compared across each category of the four mea-
sures. Specifically, the mean household scale measures were compared
affirmatively answered either Q3 (12), “Food bought didn’t
with the closest corresponding item calibration values (Derrickson et last,” or Q4 (12), “Could not afford to eat balanced meals,” not
al. 2000a). Next, the item with the closest corresponding item cali- Q2, “Worried about food.” Only eight (10.5%) of the 76
bration value was compared with categorical algorithms outlined in households with eight or more affirmative responses followed
Table 2. Statistical analysis was completed with SPSS (Version 6.2; the Hawaiı̀ modal response pattern. Four of the 31 (12.9%)
SPSS, Chicago, IL). One-way ANOVA and Tukey’s post hoc tests households classified as experiencing severe hunger by the
were also used to assess differences in the mean vegetable intake, scale CFSM measure followed the Hawaiı̀ modal response pattern.
measures and the resource augmentation index between measures. There were no statistically significant differences in modal
The ␣ value was set at P ⫽ 0.05 for all ANOVA tests. pattern response by household family status (Pearson 2 ⫽
0.83, P ⫽ 0.36). Because of high measurement variability,
RESULTS these findings question the reliability of using a modal response
Assessment of the credibility of the CFSM measure. pattern as the basis of food security categorization.
Findings outlined in Table 3 indicated that the modal response The potential issue of “misclassification” of the CFSM, or
patterns or item hierarchy differed between the national 1995 validity of applying the CFSM categorical measure to Hawaiı̀
CPS sample and the Hawaiı̀ sample. When differences in data, was further elucidated through comparisons of affirma-
modal response pattern sequencing involves a threshold item, tive response rates to selected items across the four CFSM
the reproducibility of the CFSM categorical algorithm is af- categories (Table 4). Although many findings are noteworthy,
fected. For instance, the fifth through eighth items in both the most important findings disputing the face validity of the
modal response patterns (Q6, Q8 and Q9) and the 15th CFSM categories with Hawaiı̀ data are listed below:
through 17th (Q12a through Q15a) items were not in the Food secure. Given that 1300 of the 1411 households
same sequence. However, because no “threshold items” (Q4, categorized as food secure had no affirmative responses, a
Q8a and Q12) were affected, there was no apparent effect on relatively high percentage of the remaining 111 had an unex-
CFSM categorization. Q14, “Child hungry,” is 14th in the pected response pattern: 30 (27%) responded affirmatively to
CFSM modal response pattern but 12th in the Hawaiı̀ modal Q4, “Unable to afford balanced meals,” and 7 (6.3%) re-
response pattern (Hamilton et al. 1997a). Findings imply that sponded affirmatively to Q9, “Adult not eating enough.”
households with children in Hawaiı̀ who followed the Hawaiı̀ Food insecure. Seventeen of the 158 (10.9%) households
modal response up to Q14 were classified as experiencing classified as food insecure responded affirmatively to Q10,
“moderate hunger” rather than “severe hunger.” As indicated “Respondent hungry,” 5 (3.2%) to Q12, “Adults did not eat
in Table 1, “Severe hunger” is the category that is supposed to for a whole day” and 4 (2.5%) to Q14, “Child hungry.”
capture hunger among children. This issue would be more Moderate hunger. Thirty-two of 64 (50%) households clas-
troublesome if hungry households did not follow the Hawaiı̀ sified as experiencing moderate hunger responded affirmatively
modal response pattern. to Q10, “Respondent hungry,” and 14 (22%) responded affir-
Response rates to the Hawaiı̀ modal response pattern were matively to Q12, “Adults not eat for a whole day,” a key
next explored to further clarify the effect on food security indicator of severe hunger (Hamilton et al. 1997a).
categorization. Overall, only 129 (36%) of the 364 respon- Severe hunger. Twenty-one of the 31 (67%) households
dents with one or more affirmative responses followed the classified as experiencing severe hunger responded affirma-
Hawaiı̀ modal response pattern: 100 (77%) of the 129 had five tively to Q12. Twelve of the 33 (36%) households admitting
or fewer affirmative responses. Also, 32 (52%) of the 62 to experiencing hunger among children (Q14) were classified
754 DERRICKSON ET AL.
TABLE 5
Household food security status prevalence rates by the CFSM categorical measure1
Food security measures Overall Adults only ⱖ1 Child HHS1 Food pantry Retest
n 1664 957 (57.5%)2 707 (42.5%)2 1459 (87.7%)2 144 (8.7%)2 61 (3.7%)2
n (%)3
CFSM1
Secure 1411 (84.8) 894 (93.3) 517 (73.2) 1360 (93.2) 36 (25.0) 15 (24.6)
Insecure 158 (9.5) 36 (3.8) 122 (17.3) 72 (4.9) 60 (41.7) 26 (42.6)
Moderate 64 (3.8) 24 (2.5) 40 (5.7) 20 (1.4) 30 (20.8) 14 (23.0)
1 CFSM, Core Food Security Module; CCHIP, Community Childhood Hunger Identification Project; HHS, Hawaiı̀ Health Survey.
2 Percentages indicate row totals.
3 Percentages indicate column totals.
as experiencing food insecurity without hunger (n ⫽ 4) or food food secure respondents reported a mean daily vegetable
insecurity with moderate hunger (n ⫽ 8). intake of 2.0 servings of vegetables, whereas respondents in
Comparison of various four categorical measures. Com- households with hungry children averaged only a single
parisons of prevalence estimates between samples are pre- serving. The “Only worried about food insecurity” respon-
sented in Table 5. Although not unexpected, the 6.7% dif- dents (with only an affirmative response to Q2) reported
ference in the overall rate or percentage classified as food significantly higher use of resource augmentation behaviors
secure between the CFSM categorical measure and other mea- (mean of 1.2 versus 0.2), lower mean vegetable intake (1.4
sures in the total sample is perhaps most important. Also, both versus 2.0 servings) and significantly greater reliance on
the adapted CCHIP and adapted RC measures classified 73 Saimin (10.4 versus 3.7 times a month) than respondents
(4.4%) of the sample as experiencing hunger among children. who were responded affirmatively to no CFSM items.
This is more than double the 31 (1.9%) categorized as “se- The results of the previous “concurrent validity assess-
verely hungry” by the CFSM or the 33 (2.0%) categorized as ments” were similar between measures. However, the compar-
experiencing child hunger with FVFSM. The food security isons of mean CFSM scale measures were quite different be-
measure reporting the highest prevalence of hunger was the tween measures. Notably, the household scale measure could
adapted RC method; 119 (7.2%) were categorized as experi- be calculated for only 362 respondents who responded affir-
encing adult hunger. matively to one or more items (Derrickson et al. 2000a,
Table 6 depicts comparisons of the mean values of the Wright and Stone 1980). Therefore, only the CFSM had
RAI, respondent daily vegetable intake, monthly Saimin household scale measures for respondents categorized as food
intake and the CFSM scale measures for each categorical secure. The mean household scale measure was consistent with
measure. All four categorical measures supported the fol- the item calibration of Q2, “Worried food would run out.”
lowing findings: as the severity of food insecurity worsened, Household scale measures of the other CFSM categories cor-
there was an increased utilization of resource augmentation responded well to their categorical algorithms outlined in
behaviors, a general decrease in vegetable intake and an Table 2 (Hamilton et al. 1997a). However, in the adult hunger
increased dependence on Saimin. The mean resource aug- category of both the adapted RC and adapted CCHIP mea-
mentation index values for all measures were significantly sures, mean household scale measures were equivalent to Q9,
different between food security categories. The range gen- “Respondent did not eat enough.” Q9 is much less severe than
erally extended from 0.25 for food secure respondents to 4.1 Q10, in which the respondent indicated she or he had per-
for respondents classified as experiencing hunger among sonally experienced hunger (Hamilton et al. 1997a, Derrick-
children. The categorical distinction in reduced vegetable son et al. 2001). Similarly, the mean household scale measures
intake was most pronounced with the FVFSM in which for those classified as experiencing hunger among children in
HOUSEHOLD FOOD SECURITY MEASURES IN HAWAIÌ 755
TABLE 6
One-way ANOVA of selected variables by household food security categorical measures (n ⫽ 1603)1
Daily
Measures and vegetable Monthly Scale Item with item calibration value
categories RAI2 intake3 saimin4 measure5 closest to mean scale measure5
CFSM6
Food secure 0.2a 2.00a 3.9a ⫺4.05a Q2. Worried food would run out
Food insecure 2.3b 1.72a 9.5b ⫺1.96b Q5. Reliance on low-cost foods
Moderate hunger 3.7c 1.18b 7.7b 0.53c Q8a. Adults skip/cut size often
Severe hunger 5.2d 1.29a 14.2c 2.52d Q12a. Adults not eat 1⫻ day often
Face valid
Food secure 0.2a 2.02a 3.7a — —
⫺2.84a
1 The sample size of 1603 includes all respondents from the Hawaiı̀ Health Survey and all food pantry respondents but not those who completed
the survey a second time (n ⫽ 61). All tests were conducted with Tukey’s post hoc test after significant F tests. Subscript letters indicate that mean
values in a column were not statistically significantly different from mean values with the same letter.
2 The range of resource augmentation index (RAI) values are from 0 to 8 for households with children and a maximum of 6 for households without
children.
3 Represents average daily vegetable intake of respondent.
4 Indicates the maximum saimin intake of anyone in the household in the last month.
5 Sample size ⫽ 362.
6 CFSM, Core Food Security Module; CCHIP, Community Child Hunger Identification Project.
both adapted measures were consistent with Q7, “Children 1. The questionable reliability of the modal response pat-
not eating enough.” Again, Q7 is less severe item than Q14, tern (the basis for the CFSM categorical measure) was
“Child hunger” (Derrickson et al. 2001, Hamilton et al. also reported in the original CFSM work (Hamilton et
1997a). The FVFSM did yield mean scale measures for each al. 1997a). In the 1995 CPS study, the authors report
food security category that were consistent with the concep- that 82% of households without children followed the
tual basis of the CFSM measure as outlined in Table 1. It was CFSM modal response pattern. However, 65% of these
the only measure with hunger categories that consistently households responded “no” to all items. Of the house-
corresponded with a respondent’s report of experiencing “hun- holds responding to at least one item affirmatively, only
ger” (Q10 and Q14). 49% followed the modal pattern (Hamilton et al.
1997a). Moreover, a majority of these respondents affir-
DISCUSSION matively answered less than four items. No data were
reported on the reliability of the modal response pattern
This study marks the first independent evaluation of the for households with children.
CFSM categorical food security measure. 2. Radimer (1999) questioned the credibility of requiring
Findings confirm our hypothesis that in Hawaiı̀, the CFSM three or more affirmative responses. Her field experience
categorical measure does not reliably or accurately categorize with food insecure households was similar to our own
food insecure households based on the face validity of affir- (Derrickson and Anderson 2000, Radimer 1990). Her
mative responses. The CFSM modal response pattern was not conceptual work contributed immensely to the CFSM
reproducible. Most households did not follow a modal response operationalized framework outlined in Table 1 (Bickel
pattern. These results may be due to differences in ethnic and et al. 1996).
geographical composition between samples (Hamilton et al. 3. The questionable credibility of categorization arises
1997b, State of Hawaiı̀, Department of Business, Economic when a comparison of the content of affirmative re-
Development and Tourism 1997). Thus, caution should be sponses confirms the observations of Bavier (1999). He
used in extrapolating these findings. However, like the CFSM inquired why the content of responses (i.e., Q10, “Re-
scale measure (Derrickson and Anderson 2000a), the identi- spondent hungry”) in national data sets do not more
fied weaknesses found with the CFSM categorical measure in consistently align with the CFSM categorization (Mod-
Hawaiı̀ are likely to exist across many samples. For instance: erate hungry).
756 DERRICKSON ET AL.
Some probable causes of the uncertain aspects of the CFSM ministration of food and nutrition assistance programs
categorical measure are as follows: designed to enhance food and nutrition security.
1. High measurement variability affects categorization. If an unreliable or inaccurate food security categorical mea-
The experience of household food insecurity reported by sure is used as a risk criterion for assistance, as a screening tool
respondents varies. Households do not uniformly reply or even in comparative surveillance efforts, vulnerable seg-
to food insecurity indicators in the same manner because ments of our population may be adversely affected.
their experience of household food insecurity varies. Previously, Derrickson (1999) reported that for reasons of
2. The dependence on a modal response pattern for cate- respondent fatigue and lower cost of administration and to
gorization is a fundamental problem. Because food inse- minimize the loss of dignity of the respondent over sensitive
curity is not reported in a highly consistent manner, the questions, a smaller set of food security questions should be
content of questions to which a household responds used). Embretson (1996) has argued that “shorter tests can be
affirmatively to varies across households. Thus, face or more reliable than longer tests.” One hesitation over using a
content validity of the CFSM will always be an issue shorter set of questions, or requiring only one affirmative
unless a high percentage of food insecure households response to categorize households, may be the fear of increased
ACKNOWLEDGMENTS Glaser, B. G. & Strauss, A. (1967) The Discovery of Grounded Theory: Strat-
egies for Qualitative Research. Aldine, Chicago, IL.
We are indebted to Gary Bickel, the Hawaiı̀ State Department of Hamilton, W. L., Cook, J. T., Thompson, W. W., Buron, L. F., Frongillo, E. A., Jr.,
Health, Office of Health Status Monitoring, SMS Research and Olson, C. M. & Wehler, C. A. (1997a) Household Food Security in the
United States in 1995: Technical Report of the Food Security Measurement
Marketing Service, Inc. staff, George Chee, Dwayne Makalena, var-
Project. Report prepared for the U.S. Department of Agriculture, Food Con-
ious Salvation Army staff members on O’ahu and Scott Derrickson sumer Service, Alexandria, VA.
for their constructive suggestions and editorial assistance. Hamilton, W. L., Cook, J. T., Thompson, W. W., Buron, L. F., Frongillo, E. A., Jr.,
Olson, C. M. & Wehler, C. A. (1997b) Household Food Security in the
United States in 1995: Summary Report of the Food Security Measurement
LITERATURE CITED Project. Report prepared for the U.S. Department of Agriculture, Food Con-
Bavier, G. (1999) Proceedings of the Second Food Security Measurement and sumer Service, Alexandria, VA.
Research Conference (Sponsored by the Economic Research Service, U.S. Lavarkas, P. J. (1988) Telephone Survey Methods: Sampling, Selection and
Department of Agriculture), February 22–24, 1999, Alexandria, VA. Supervision. Sage Publications, Newbury Park, CA.
Bickel, G. (1999) Toward a Research Agenda: Next Steps. Proceedings of the Life Sciences Research Office (Anderson, S. A., ed.). (1990) Core items of
Second Food Security Measurement and Research Conference (Sponsored nutritional state for difficult-to-sample populations. J. Nutr. 120:1557S–
by the Economic Research Service, U.S. Department of Agriculture), February 1600S.
22–24, 1999, Alexandria, VA. Linacre. J. (1986 –94) FACETS. MESA Press, Chicago, IL.