Utilising a multi‐item questionnaire to assess household food security in Australia

Abstract Issue addressed Currently, two food sufficiency questions are utilised as a proxy measure of national food security status in Australia. These questions do not capture all dimensions of food security and have been attributed to underreporting of the problem. The purpose of this study was to investigate food security using the short form of the US Household Food Security Survey Module (HFSSM) within an Australian context; and explore the relationship between food security status and multiple socio‐demographic variables. Methods Two online surveys were completed by 2334 Australian participants from November 2014 to February 2015. Surveys contained the short form of the HFSSM and twelve socio‐demographic questions. Cross‐tabulations chi‐square tests and a multinomial logistic regression model were employed to analyse the survey data. Results Food security status of the respondents was classified accordingly: High or Marginal (64%, n = 1495), Low (20%, n = 460) or Very Low (16%, n = 379). Significant independent predictors of food security were age (P < 0.001), marital status (P = 0.005), household income (P < 0.001) and education (P < 0.001). Conclusion Findings suggest food insecurity is an important issue across Australia and that certain groups, regardless of income, are particularly vulnerable. So what? Government policy and health promotion interventions that specifically target “at risk” groups may assist to more effectively address the problem. Additionally, the use of a multi‐item measure is worth considering as a national indicator of food security in Australia.

for the provision of targeted support services. In addition, a precise understanding of the nature and size of the problem may assist in driving government policy direction, social service provision and evaluation of the impact of food insecurity.
Food sufficiency and food security are often used interchangeably in the literature. Specifically, food sufficiency is a physical concept, referring to the adequate intake of food to meet requirements. 4 Food security (FS) is a broader concept and is inclusive of food sufficiency, but also acknowledges psychological, social and cultural factors. 4 The Australian Bureau of Statistics (ABS) utilises food sufficiency questions, rather than a dedicated FS instrument, as a proxy to monitor FS nationally. 5 Prior to 2011, the National Nutrition Survey used a single question to estimate the proportion of the population that was food insecure: "In the past 12 months, have you or anyone in your household run out of food and not had enough money to purchase more?". 5 The 2011-2012 National Nutrition and Physical Activity Survey used a similarly worded question and an additional question. The survey read: "In the past 12 months was there any time when you or members of your household ran out of food and couldn't afford to buy more?" and "When this happened, did you or members of your household go without food?". 6 Responses to food sufficiency questions define individuals as food sufficient or insufficient; however, they do not reveal the severity of insufficiency.
The failure of ABS food sufficiency questions to capture the full spectrum of FI has been attributed to underreporting of the problem in Australia. 7 These questions measure food deprivation, which only encompasses the severest form of FI. The focus of this type of FS measure is food shortage due to financial constraints, and more diverse contributing factors such as social and psychological dimensions are missed. 8 Mild-to-moderate FI is often overlooked by food sufficiency questions meaning the true extent of the problem cannot be ascertained. 8 Food insufficiency can be interpreted as a proxy for "FI with hunger" or "Very Low" FS. 8 An alternative FS surveillance instrument to the ones commonly used in Australia is the 18-item questionnaire Household Food Security Survey Module (HFSSM). 9 The HFSSM is administered annually in the United States, and adaptations of this instrument have been used in several other countries, such as Canada, Mexico and Brazil. Changes in reported food intake, as a consequence of declining household resources, and the severity of FI are indicated by the HFSSM. These aspects are missed by food sufficiency questions. 9 Validation of the HFSSM involved 44 647 household interviews derived from the April 1995 U.S. Current Population Survey. 10 A more concise six-item version of the HFSSM was developed in 1999 using data from the same survey. The short form questionnaire has comparable accuracy to the longer form (it correctly identified 97.7% of households as food secure), whilst having the advantage of reduced respondent burden. 11 It is widely accepted in Australia and internationally that there is interplay between socio-demographic characteristics and FS status. 12,13 Income has been prolifically studied and is often considered the most significant factor in predicting FS. [12][13][14] However, many other demographic characteristics, such as age, gender, marital status and education, may have an additional impact, above and beyond, income. 12,15,16 To our knowledge, no previous studies have utilised a standardised multi-item FS instrument and reviewed multiple socio-demographic factors at a general population level in multiple Australian states. Therefore, the purposes of this study were to investigate FS using the short form of the HFSSM within an Australian sample population; and explore the relationship between FS status and the multiple social, economic and physical dimensions encapsulated within the definition of FS.

| Selection and description of participants
The surveys were administered through Qualtrics (Provo, Utah, USA) and disseminated online to registered panel respondents through a commercial research marketing company. Inclusion criteria were that the respondent was the main grocery purchaser in the household, resided in one of five states of Australia (New South Wales (NSW), Victoria (VIC), Western Australia (WA), South Australia (SA) and Queensland (QLD)), was aged between 18 and 84 years, and had computer ownership or access. No additional exclusion criteria were applied.
Quotas were set to ensure sufficient representation terms of gender, age groups and location. A minimum proportion of male representation was set at 30%. The quotas for were set to align with national Australian population demographics for age: 10% were aged 19-24 years; 20% 25-34 years; 19% 35-44 years; 18% 45-54 years; 15% 55-64 years; 18% 65-84 years, and for location: 34% NSW; 26% VIC; 11% WA; 7% SA; 22% QLD. 17 De-identified data were available for analysis. Household income was reduced from 13 categories to six and based on the 2016/2017 Australian Tax Office income brackets. The categories for household income are as follows: refused to answer, very low (<$18 000), low ($18 001-37 000), middle ($37 001-87 000), high ($87 001-180 000) and very high (>$180 000). 20 The number of children (classified as <18 years of age) residing in the household variable was reduced from seven to four categories (0, 1, 2 and 3 or more). Likewise, the number of adults in the household was reduced from seven to three categories (1, 2 and 3 or more).

| Statistics
Highest education status achieved was recoded as either secondary or less, vocational or university and based on the ABS categories. 21 Responses to the six HFSSM questions (Q1-Q6, see Table 2) were coded and assessed in accordance with the US HFSSM user notes. 22 In brief, any affirmative responses, including "yes," "sometimes true" and "often true," were assigned a score of one. The sum of the affirmative responses (range 0-6) from the six questions was calculated to classify the household into three levels of FS:

| Ethics approval
All procedures performed in studies involving human participants were in accordance with the ethical standards of the institutional and/or national research committee and with the 1964 Helsinki Declaration and its later amendments or comparable ethical standards.
Informed consent was obtained from all individual participants included in the study. Ethics approval has been granted by Edith Cowan University's Human Research Ethics Committee.

| Sample demographics
Characteristics of the survey respondents are given in Table 1. In terms of income, the greatest representation was in the low bracket ($18 001-37 000, 31%) followed by middle ($37 001-87 000) and high ($87 001-180 000) income earners equally (24%). Approximately 3 in 5 (61%) of the respondents were married or in a de facto relationship. Almost three-quarters of the respondents (73%) had attained some form of post-secondary education (
Those classified as having Low or Very Low FS were most likely to respond in the affirmative to the six questions ( Table 2). For instance, 76% and 97% of those in the Low and Very Low FS classifications gave an affirmative response to Q1: "The food that (I/we) bought just didn't last, and (I/we) didn't have money to get more," with almost 2 in 5 (37%) in the latter category indicating that the situation is often true for them. In contrast, less than 1 in 10 (7.3%) of those with High or Marginal FS, considered the above as an issue. This pattern of response continued for the remaining HFSSM questions.
One particular question was polarising: Q3 "In the last 12 months, did (you/you or other adults in your household) ever cut the size of your meals or skip meals because there wasn't enough money for food?" All of the High or Marginal FS respondents disagreed with this question, whilst all of the Very Low FS respondents agreed with the statement.
Other notable disparities were observed in Q5 and Q6 where 96% and 81%, respectively, of those in the Very Low FS category responded "yes" to having to eat less or not eat at all because there was not enough money for food, whereas for those with Low FS, only 28% and 19%, respectively, responded in the same way. In comparison, over 95% of the High or Marginal FS respondents did not have to reduce their food consumption in the past 12 months due to insufficient funds.

| Determinants of FS status
No significant association was observed individually between FS status and respondents' gender (P = 0.448), location (P = 0.151) or being a first-, second-or third-generation migrant (P = 0.346). Conversely, age, household income, marital status, number of adults and children in the household, occupation, immigration and highest level of education attainment were shown to have a significant association with FS status (Table 1). When analysed using a multivariable multinomial logistic regression model, the only significant independent predictors of increasing risk of FI were younger age (P < 0.001), divorce or separation (P = 0.005), lower household income (P < 0.001) and lower educational attainment (P < 0.001) (  (Table 3).
Respondents who were divorced or separated were 2.3 times more likely to be categorised with Very Low FS status than those who were single (P < 0.001, Table 3). Of those who reported their income and were not retired, 54% of respondents who were either divorced or separated were in the low-or very low-income bracket, and only 14% were in the high-or very high-income brackets. In comparison with those who were married (or in a de facto relationship), 17% were in the low-or very low-income bracket and 54% were in the high-or very high-income bracket. Respondents with secondary level education or less were 89% and 73% more Marital status also appears to be linked to FS. Divorced or separated people appear to be at greatest risk of FI, whilst being married or widowed are protective both in the findings of this study and others. 16,32 The division of income across households, the cost of child support and lack of social support have been suggested as causal mechanisms for the relationship between separation and FI. 32 There is minimal research in developed nations investigating the relationship between education attainment and FS. University level education was shown to be a protective factor in our study.

| 13
American consensus that financial restraints prevent the uptake of healthy diet and limit the individual's ability to successfully employ food literacy skills. 33 Further research is warranted in the Australian context to get a clearer depiction of the impact of food literacy skills on FS status.

| Application of findings and limitations
The reference period of our study included the Christmas or festive season. All of the short form HFSSM questions, however, refer to "in the last 12 months" suggesting that the festive season should not have impacted the results. It is possible that respondents may have been experiencing higher than normal financial stress at this time, bringing the concept of deprivation to front of mind and potentially skewing the findings.
The study population was similar, but not identical to the Australian general population. Sixty per cent of the respondents were females, which was approximately 10% more than that of the general Australian population. The distributions across the age groups and locations of the study population were within 3% and 2%, respectively, of the general Australian population. 17 Excluding those who refused to answer, the proportion of respondents in the very low-income brackets (3.5%) was severely underrepresented when compared to the Australian population (18.8%); conversely, there was an overrepresentation of high (26%)-and very high (9.3%)-income earners in the study population as compared to the Australian population (high 17.2%; very high 3%). 20 The majority (61%) of the respondents in this study were married or in a de facto relationship, which is similar to the overall Australian population where 54% were married and 11% were in a de facto relationship. 6 The post-secondary educational attainment of the study population (73%) was 16% higher when compared to the overall Australian population (57%). 17   significant association between location and FS would have been apparent. This may be important as both ATSI status and geographical location are both considered significant social determinants of FS in Australia. This is a limitation of the study and requires further research to ascertain the impact of location in relation to urban, rural and remote populations. Generalising these results across all populations is therefore cautioned.
The ABS data collection format is by face-to-face interview. The research presented in this study utilised an online form of data collection; however, this in itself can be considered a limitation. Online surveys as a mode of data collection preclude access of those without access to an Internet connection and people who cannot read written English. These are groups for whom the risk of FI has been shown to be higher than the general population. 14 Nevertheless, it has been established that these vulnerable, hidden populations are hard to reach in general, even with traditional means of research such as interviews or paper-based surveys employed by the ABS. 34

| CONCLUSION
Food insecurity remains an important issue across Australia. Our results indicate that certain groups, regardless of income, are particularly vulnerable to FI; these include younger Australians (particularly those aged 25-34 years old), those with lower educational attainment and divorced or separated individuals. Government policy and community interventions that specifically target these "at risk" groups may assist to more effectively address the problem. Additionally, the use of the multi-item measure is certainly worth considering in future studies and as a national indicator of FS in the Australian context.