. 2014 Jun 25;18(5):877–892. doi: 10.1017/S1368980014001050

Table 2.

Psychometric evidence of the epidemiological instruments measuring household food insecurity

Instrument	Place and year	No. of items*	Sample size	Application method (recall period)	Psychometric results (type of study in italics)
CFSM/HFSSM and its linguistic and semantic variants	USA, 2000⁽ ⁹⁸ ⁾	18	1664	Telephone (12 months)	∙ Test–retest reliability: Pearson correlation coefficient: r=0·75 (P<0·01) ∙ Structural validity: Partly corroborated→factorial analysis via the Rasch model. Findings pertaining to GOF of the respondents indicated an acceptable rate of misfit. This did not vary with family status or with any ethnic group except the Samoans
	USA, 2001⁽ ⁹⁹ ⁾	18	1664	Telephone (12 months)	∙ Face validity: Uncorroborated→face validity was not presented for the population of Hawaii ∙ Construct validity: Corroborated→respondents with only one affirmative response to item ‘Only worried about FI’ exhibited behaviours consistent with FI (decreased vegetable intake, greater reliance on alternative food resources and low-cost foods)
	USA, 2003⁽ ¹⁰⁰ ⁾	18	25 477	–	∙ Structural validity: Partly corroborated→factorial analysis via GLM and Rasch model. FI questions with a large number of interactions indicate that these are interpreted differently by households with different characteristics
	USA, 2004⁽ ¹⁰¹ ⁾	15	35 555	–	∙ Structural validity: Factorial analysis via Rasch model. Partly corroborated→severity calibrations were significantly different for families with and without children. Rasch model cannot be used to equate the food-security status of households with and without children
	Brazil, 2004⁽ ¹⁰² ⁾	15	847	Interview (3 months)	∙ Internal consistency: α=0·91 ∙ Face validity: Corroborated. ∙ Construct validity: Corroborated→FI severity level had a strong dose–response with income strata and the probability of daily intake of fruits, vegetables, meat/fish and dairy
	Mexico, 2005⁽ ¹⁰³ ⁾	15(?)	99	Interview (3 months)	∙ Construct validity: Corroborated→FI was significantly and inversely correlated with the number of food items in the household, animal-source foods, dairy products, processed foods, fruits and vegetables. FI was also associated with low dietary variety
	Trinidad and Tobago, 2006⁽ ⁸⁰ ⁾	15	3858	Self-administered –	∙ Internal consistency: α=0·915 for the adult-referenced items and α=0·818 for the child-referenced items ∙ Structural validity: Corroborated→factorial analysis via IRT. Fitting either a 2PL model, which allowed discrimination parameters to vary between items, or a differential item functioning model, which allowed item calibrations to vary between ethnic groups, had little influence on interpretation ∙ Construct validity: Corroborated→in both children and adults FI showed a strong, graded association with lower monthly household income
	USA, 2007⁽ ¹⁰³ ⁾	15	122	Interview (30 d)	∙ Test–retest reliability: κ=0·66
	Brazil, 2007⁽ ¹⁰⁴ ⁾	15	49	Interview (30 d)	∙ Internal consistency: α=0·93
	Brazil, 2007⁽ ¹⁰⁵ ⁾	15	403	Interview (3 months)	∙ Internal consistency: α=0·95, both in urban and rural areas ∙ Construct validity: Corroborated→FI was significantly and inversely correlated with the income strata. FI was associated with different patterns of daily food consumption. FI severe was associated with daily consumption of milk, milk derivatives, eggs and beans
	Ecuador, 2007⁽ ⁷⁵ ⁾	15	52	– –	∙ Structural validity: Corroborated→Factorial analysis via Rasch model. Majority of the items presented adequate GOF values ∙ Construct validity: Corroborated→relationships was statistically significant between HFI and total food supply, as well as the supply of meat, vegetables, processed products, drinks and snacks
	Brazil, 2008⁽ ¹⁰⁶ ⁾	15	108 606	Interview (3 months)	∙ Structural validity: Corroborated→factorial analysis via Rasch model. Scale presented good fitness and most item GOF values were within adequate range, being practically identical when comparing female and male responses. None of the items presented substantial DIF
	Brazil, 2008⁽ ¹⁰⁷ ⁾	15	847	Interview (3 months)	∙ Structural validity: Corroborated→factorial analysis via the Rasch model. Except for one item, GOF statistics were within a range considered adequate, indicating a common phenomenon being measured with approximately equal discrimination
	Iran, 2009⁽ ⁸³ ⁾	15	2004	– (12 months)	∙ Structural validity: Corroborated→factorial analysis via the Rasch model. Most Item GOF statistics were near unity, and none exceeded 1·20
	Brazil, 2009⁽ ¹⁰⁸ ⁾	15(?)	458	Interview (3 months)	∙ Internal consistency: α=0·91
	USA, 2009⁽ ¹⁰⁹ ⁾	15	242	Interview (12 months before and during pregnancy)	∙ Internal consistency: α=0·91 (before pregnancy) and α=0·88 (during pregnancy) ∙ Structural validity: Corroborated→factorial analysis via the Rasch model. Two of 14 questions examined did not fit within the acceptable range. Severity level of 12 of the 14 questions was similar across two time periods, before pregnancy and the past month of pregnancy
	Peru, 2010⁽ ⁷⁸ ⁾	15	300	Interview (12 months)	∙ Internal consistency: α=0·86 (overall); α=0·76 (anxiety about food supply); α=0·86 (anxiety about food quality); α=0·84 (food ingestion by adults); α=0·75 (food ingestion by children) ∙ Construct validity: Corroborated→descriptive results confirmed the parallelism of responses to the items of scale for the variables: farm ownership, family size and use of communal kitchens
	USA, 2011⁽ ¹¹⁰ ⁾	15	63	Self-administered (12 months)	∙ Internal consistency: α=0·73
SPHFSS and its linguistic and semantic variants	Venezuela, 1999⁽ ²⁵ ⁾	12	238	Interview (6 months)	∙ Internal consistency: α=0·82 ∙ Structural validity: Corroborated→factorial analysis via EFA. Solution with two factors. Factor 1 explained 37·2 % and factor 2, 32·1 % of total variance ∙ Construct validity: Corroborated→HFI varied in an expected manner with the predictor variables (poverty level, monthly income per capital, predictor of energy availability and social class)
	Colombia, 2005⁽ ⁶⁸ ⁾	12	193	Interview (6 months)	∙ Internal consistency: α=0·90 ∙ Structural validity: Corroborated→factorial analysis via EFA. Solution with one factor. Factor 1 explained 67·1 % of total variance ∙ Construct validity: Corroborated→FI was inversely correlated with the schooling and goods index. FI was associated with income, nutritional state and malnutrition
	Colombia, 2005⁽ ⁶⁹ ⁾	12	1624	Interview (6 months)	∙ Internal consistency: α=0·95 for the first factor and α=0·89 for the second ∙ Structural validity: Corroborated→factorial analysis via EFA and Rasch model. Solution with two factors. All items showed GOF values within acceptable range ∙ Construct validity: Corroborated→scale correlated significantly with food availability, begging, children’s labour, household size and occupation of the head of household
	Venezuela, 2007⁽ ¹¹¹ ⁾	12	924	Interview (6 months)	∙ Internal consistency: α=0·89 ∙ Structural validity: Corroborated→factorial analysis via EFA. Solution with two factors. Factor 1 explained 29·3 % and factor 2, 27·9 % of total variance
	Colombia, 2008⁽ ¹¹² ⁾	12	1319	Interview (30 d)	∙ Structural validity: Corroborated→factorial analysis via EFA and Rasch model. Most CHFSS items presented good fitness with most GOF values within the adequate range. Only two adult items exhibited DIF between urban and rural households
	Colombia, 2009⁽ ¹¹³ ⁾	12	2784	Interview (30 d)	∙ Construct validity: Corroborated→statistically significant associations were found between household food insecurity and diagnoses of children’s diarrhoea, respiratory infections and parasitosis. Risk for child stunting and underweight increased in a dose–response way as FI became more severe
	Colombia, 2009⁽ ¹¹⁴ ⁾	12	108	Interview (30 d)	∙ Internal consistency: α=0·92 in the 1st application and 0·94 in the 2nd ∙ Test–retest reliability: r=0·98 (95 % CI 0·98, 0·99) ∙ Structural validity: Corroborated→factorial analysis via EFA and SEM. Solution with two factors. Total variance explained was 85 %
HFSSM-6SF and its linguistic and semantic variants	USA, 1999⁽ ¹⁴ ⁾	6	44 647	– –	∙ Criterion validity: In order to determine the HFI in general, in households without children and in households with children, Sens=92·0 %; 99·7 % and 85·9 % and Spec=99·4 %, 99·3 % and 99·5 %, respectively. To determine the FI with hunger, Sens=84·7 %, 90·3 % and 78·4 %, and Spec=99·6 %, 99·9 % and 99·2 %, respectively
	Trinidad and Tobago, 2004⁽ ¹¹⁵ ⁾	6	531	Interview (12 months)	∙ Internal consistency: α=0·87 ∙ Structural validity: Corroborated→factorial analysis via Rasch model. Item-to-score correlations were consistent, in general, among the ethnic groups ∙ Construct validity: Corroborated→measurement of FI was related in an expected manner with the consumption of green vegetables and salads
	Trinidad and Tobago, 2005⁽ ⁸⁸ ⁾	6	1903	Self-administered (12 months)	∙ Internal consistency: α=0·77 ∙ Structural validity: Corroborated→factorial analysis via Rasch model. Relative severity of majority of the items was very similar in each ethnic group and in boys and girls ∙ Construct validity: Corroborated→after adjustment for socio-economic variables, FI was associated with consumption of fruit, fish, biscuits and cakes
	Iran, 2007⁽ ⁹¹ ⁾	6	300	Interview (12 months)	∙ Criterion validity: using hunger screening as a reference standard→Sens=98·7 %, Spec=85·5 %. Using hidden hunger screening as a reference standard→Sens=23·5 %, Spec=96·9 %
	Australia, 2009⁽ ¹¹⁶ ⁾	6	49	Self-administered –	∙ Internal consistency: α=0·90
R/CSm_a	USA, 1995⁽ ¹³ ⁾	10	193	Interview –	∙ Internal consistency: α=0·84 (household level/factor); α=0·86 (individual level/factor) and α=0·85 (child level/factor) ∙ Structural validity: Corroborated→factorial analysis via EFA. Confirmed that hunger and FI are experienced differently at family level and at individual level. There are distinct aspects related to the child and the adult ∙ Construct validity: Corroborated→FI was strongly associated with the demographic characteristics of the subjects and inversely associated to income
R/CSm_b	Indonesia, 2001⁽ ³² ⁾	9	1356	Interview (varied)	∙ Face validity: Corroborated
R/CSm_d	USA, 2004⁽ ¹⁹ ⁾	8	381	Telephone/interview (12 months)	∙ Internal consistency: α=0·94
R/CSm_d	Tanzania, 2008⁽ ¹¹⁷ ⁾	8	530	Interview –	∙ Internal consistency: α=0·85 (5 items – family level) and α=0·78 (3 items – child level). ∙ Structural validity: Corroborated→factorial analysis via EFA. Solution with two factors. Both accounted for 66·2 % of the total variance ∙ Construct validity: Corroborated→FI measure showed significant associations with selected sociodemographic factors (e.g. mother’s age, mother’s marital status, mother’s schooling) in the expected directions
R/CSm_e	Iran, 2007⁽ ³⁰ ⁾	16	250	Interview (12 months)	∙ Internal consistency: α=0·90 (factor: family insecurity); α=0·82 (individual insecurity) and α=0·796 (child hunger scales) ∙ Structural validity: Corroborated→factorial analysis via EFA. Solution with three factors. In most cases, factor loadings of the items were as expected and similar to those reported in previous studies ∙ Construct validity: Partly corroborated→individual insecurity and child hunger were inversely correlated with monthly per capita income, father’s education, mother’s education and father’s occupational status, and positively correlated with household size, as expected. However, household insecurity did not follow the same pattern
CCHIP	USA, 1992⁽ ³⁵ ⁾	8	377	Interview (varied)	∙ Internal consistency: α=0·86 ∙ Face validity: Corroborated ∙ Structural validity: Corroborated→factorial analysis via EFA. Solution with one factor, accounting for 52 % of the variance ∙ Construct validity: Corroborated→FI was strongly associated in the expected direction with economic and socio-demographic variables, with reliance on strategies to cope with food shortages and with health problems of the children
CCHIP	USA, 1997⁽ ¹¹⁸ ⁾	8 (?)	193	Interview (30 d)	∙ Criterion validity: Sens=86 %; Spec=73 %
R/CS and its linguistic and semantic variants	USA, 1992⁽ ³⁷ ⁾	12	193	Interview –	∙ Internal consistency: α=0·91 (factor 1); α=0·89 (factor 2); α=0·92 (factor 3) ∙ Structural validity: Corroborated→factorial analysis via EFA. Solution with three factors accounting for 62 % of the variance. Household factor alone accounted for 53 % of the variance, indicating the strong presence of a single factor ∙ Construct validity: Corroborated→relationships between the R/CS and other scales (CCHIP and NHANES) occurred as expected
R/CS and its linguistic and semantic variants	USA, 1997⁽ ¹¹⁸ ⁾	12(?)	1993	Interview –	∙ Criterion validity: Sens=89 %; Spec=63 %
HFIAS and its linguistic and semantic variants	Ethiopia, 2009⁽ ⁷⁰ ⁾	9	99	Interview (30 d)	∙ Internal consistency: α=0·85 (1st application); α=0·84 (2nd) and α=0·83 (3rd) ∙ Face validity: Corroborated ∙ Construct validity: Corroborated→household per capita income was inversely associated with FI score. Dose–response trends between FI level and likelihood of previous-day consumption of various foods were observed for meats (not including fish), vegetables and fruits, with minor deviations between moderate and severe levels of FI
HFIAS and its linguistic and semantic variants	Tanzania, 2010⁽ ⁷¹ ⁾	9	237	Interview (30 d)	∙ Internal consistency: α=0·90 (overall); α=0·83 (food quality subscale); α=0·89 (food quantity subscale) ∙ Structural validity: Corroborated→factorial analysis via EFA. Solution with two factors. Both factors explained 69 % of the total variance ∙ Construct validity: Corroborated→FI was positively associated with maternal education, husband’s education, household wealth status, being of an agricultural rather than pastoral tribe and animal-source food consumption; it was negatively associated with maternal age and household size
ELCSA	Colombia, 2010⁽ ¹¹⁹ ⁾	17	150	Interview (30 d)	∙ Internal consistency: α=0·93 (10 items for families with adults only); α=0·96 (17 items for families with adults, youths and children) ∙ Structural validity: Corroborated→factorial analysis via EFA. Solution with three factors, accounting for 75 % of the variance
EbMHFI	Burkina Faso, 2006⁽ ¹²⁰ ⁾	9	126	Interview (varied)	∙ Internal consistency: α varied from 0·81 to 0·85 in the five applications of the scale ∙ Construct validity: Corroborated→HFI was associated with predictor variables (dietary intake, nutritional status and economic status) as expected
MHFI	Costa Rica, 2008⁽ ²⁷ ⁾	14	213	Interview (12 months)	∙ Internal consistency: α=0·89 (three response categories) and α=0·87 (dichotomized responses) ∙ Face validity: Corroborated ∙ Structural validity: Corroborated→factorial analysis via EFA. Solution with one factor that explained 46 % of the variation ∙ Construct validity: Corroborated→there was an association between the levels of FI and poverty line method, as well as unsatisfied basic needs method and integrated poverty classification
IFI	Kenya, 2011⁽ ²⁹ ⁾	4	6795	– (30 d)	∙ Internal consistency: α=0·73 ∙ Structural validity: Corroborated→factorial analysis via IRT. Likelihood ratio test rejected the one-parameter model in favour of the two-parameter one. Loevinger coefficient H=0·88 ∙ Construct validity: Corroborated→FI has a significant and negative association with household income level
TAPFI	USA, 2006⁽ ²⁰ ⁾	7	85	Interview –	∙ Internal consistency: α=0·84 ∙ Construct validity: Corroborated→past FI was significant correlation with maternal education, crowding in the mother’s childhood household and past food insufficiency. Foreign-born Latino mothers reported significantly greater levels of past food insecurity than US-born mothers, demonstrating discriminant validity
FIE	USA, 2003⁽ ⁶⁷ ⁾	14			∙ Criterion validity: Sens=89 %; Spec=69 %
NHANES III_ FIQ	USA, 1997⁽ ¹¹⁸ ⁾	12(?)	193	Interview –	∙ Criterion validity: Sens=90 %; Spec=32 %
SAFSSM	USA, 2004⁽ ¹⁸ ⁾	9	345	Self-administered (12 months)	∙ Structural validity: Partly corroborated→factorial analysis via the Rasch model. Poorer model fit. Response patterns differed between younger (<12 years) and older (≥12 years) ∙ Criterion validity: Simulation of precision of classification based on a population with ‘true’ FI distributed approximately as observed in the sample, gave the following results: on the threshold of FI (gross score≥2), Sens=0·89, Spec=0·93 and PPV=0·82; on the threshold of FI with hunger (gross score≥6), Sens=0·77, Spec=0·98 and PPV=0·71

CFSM/HFSSM, Core Food Security Measurement/Household Food Security Survey Module; SPHFSS, Self-Perceived Household Food Security Scale; HFSSM-6SF, HFSSM Six-Item Short Form; R/CSm_a, Modified Radimer/Cornell Scale (a); R/CSm_b, Modified Radimer/Cornell Scale (b); R/CSm_d, Modified Radimer/Cornell Scale (d); R/CSm_e, Modified Radimer/Cornell Scale (e); CCHIP, Community Childhood Hunger Identification Project; R/CS, Radimer/Cornell Scale; HFIAS, Household Food Insecurity Access Scale; ELCSA, Latin American Food Security Measurement Scale; EbMHFI, Experience-Based Measurement of Household Food Insecurity; MHFI, Measurement of Household Food Insecurity; IFI, Items of Food Insecurity; TAPFI, Tool to Assess Past Food Insecurity; FIE, Food Insecurity by Elders; NHANES III_ FIQ. Food Insecurity Questions of NHANES III; SAFSSM, Self-Administered Food Security Survey Module for Children Aged 12 and Older; GOF, goodness of fit; FI, food insecurity; GLM, general linear model; IRT, Item Response Theory; 2PL, the two-parameter model; HFI, household food insecurity; DIF, differential item functioning; EFA, exploratory factor analysis; CHFSS, adapted Colombian Household Food Security Survey; SEM, structural equation model; Sens, sensitivity; Spec, specificity; PPV, positive predictive value.

*‘?’ sign means that the study did not explicitly discriminate the number of items on the scale used. The same number of items as in the original version was thus considered.