Seasonal prevalence and determinants of food insecurity in Iqaluit, Nunavut

Background Food insecurity is an ongoing problem in the Canadian Arctic. Although most studies have focused on smaller communities, little is known about food insecurity in larger centres. Objectives This study aimed to estimate the prevalence of food insecurity during 2 different seasons in Iqaluit, the territorial capital of Nunavut, as well as identify associated risk factors. Designs A modified United States Department of Agriculture Food Security Survey was applied to 532 randomly selected households in September 2012 and 523 in May 2013. Chi-square tests and multivariable logistic regression were used to examine potential associations between food security and 9 risk factors identified in the literature. Results In September 2012, 28.7% of surveyed households in Iqaluit were food insecure, a rate 3 times higher than the national average, but lower than smaller Inuit communities in Nunavut. Prevalence of food insecurity in September 2012 was not significantly different in May 2013 (27.2%). When aggregating results from Inuit households from both seasons (May and September), food insecurity was associated with poor quality housing and reliance on income support (p<0.01). Unemployment and younger age of the person in charge of food preparation were also significantly associated with food insecurity. In contrast to previous research among Arctic communities, gender and consumption of country food were not positively associated with food security. These results are consistent with research describing high food insecurity across the Canadian Arctic. Conclusion The factors associated with food insecurity in Iqaluit differed from those identified in smaller communities, suggesting that experiences with, and processes of, food insecurity may differ between small communities and larger commercial centres. These results suggest that country food consumption, traditional knowledge and sharing networks may play a less important role in larger Inuit communities.

conducted as part of the Healthy Foods North programme) or have examined the prevalence of food insecurity at a regional scale (e.g. the Inuit Health Survey) (4,7). While these studies have substantially contributed to our understanding of food insecurity, limited research has been conducted in larger centres of the North (15,16). These rapidly growing settlements are home to almost one fifth of all Inuit people in Canada, and differ in socialeconomic-demographic structure from smaller communities where research has been primarily conducted (17). As such, it is unknown if predictors of food insecurity identified in the literature are applicable to larger regional centres, where identified protective factors, including sharing networks, employment, education, and participation in traditional harvesting activities, may differ.
As pointed out by the Expert Panel on the State of Knowledge of Food Security in Northern Canada, there has also been limited research conducted on the seasonality of food insecurity (18). Scholarship on food security in the Arctic, including the 2007Á2008 Inuit Health Survey, is mainly cross-sectional (6,9,19). These methodological choices matter because the timing of ice break-up and freeze-up, and weather conditions influence the distribution and accessibility of harvesting sites, ultimately affecting the type and quantity of food consumed (20,21). For instance, data on country food harvest in Iqaluit indicate seasonal variation in caribou, ringed seal and Arctic char harvesting rates (22). Only a few studies, however, have assessed the composition of the seasonal diet of Indigenous people in Canada. Kuhnlein et al. (23) conducted dietary assessments in 44 communities (Inuit, Dene/Métis and Yukon First Nations) during a season of high and low traditional food availability. Other studies conducted in Baffin Island, Yukon, and the Inuvialuit Settlement Region found significant seasonal variation in availability or consumption of traditional food (13,24).
In this study, we report on a seasonal analysis of household food insecurity and associated determinants in Iqaluit, Nunavut (population of 6,699) (25). As the capital and largest city of Nunavut, Iqaluit is the seat of many governmental agencies and Inuit organizations. Compared to small villages in the Canadian Arctic, Iqaluit has a strong wage economy and attracts a large number of external workers, both Inuit from other Arctic communities and  non-Indigenous migrants from southern regions (26). Despite its size and function, there are little data on food insecurity in Iqaluit specifically, with previous research either focusing on particular groups (e.g. community food programmes) or on Nunavut as a whole (7,15,16). The goal of this study was therefore to estimate the prevalence of food insecurity in Iqaluit, as well as the prevalence and predictors of food insecurity for Inuit households in 2 different seasons.

Methods
Data were collected using a repeated randomized crosssectional household survey conducted in Iqaluit from September 15th to October 5th 2012 and from May 18th to May 31st 2013 (Fig. 1). The open water boating season in Iqaluit typically runs from late June/early July to November (21). Outside of this period, the use of a snowmobile to go out on ''the land'' is preferred due to stable ice conditions and extensive snow cover (21). The 2 survey periods were thus chosen to reflect a period of potentially low and high accessibility to country food harvest, respectively. May is part of the ''shoulder season,'' a period of heightened vulnerability due to spring ice break-up (21). Transient food insecurity can occur during this period, as access to hunting grounds (both by boat and snowmobile) becomes limited (12). September/October was chosen as a period of high accessibility to harvesting areas as it occurred during the boating season, before the winter ice freeze-up (21).
During both seasons, the City of Iqaluit's House Number Atlas was used to select households by dividing the city area into 4 neighbourhoods, based on shared characteristics and geographical location. Each neighbourhood was then further divided into map components (or ''blocks''), for a total of 18. ''Blocks'' within each neighbourhood were proportionally and randomly selected for surveyors to subsequently visit. The process of random selection was done separately for each season and was thus independent. We surveyed all households in the block sample, both Inuit and non-Inuit. An individual from each household was randomly selected, based on the person with the most recent birthday, to answer questions about country food consumption. These questions were part of a larger survey on acute gastrointestinal illnesses, for which a random sample was required (27). Questions Overall, 532 and 523 respondents were interviewed in September and May, respectively. The crude response rate was 75% in September and 55% in May. The lower response rate in May might possibly be due to more people going out on the land in the weeks preceding the spring break-up. Surveys were predominantly conducted face-to-face by the survey team, which consisted primarily of local Inuit research assistants, and some southernbased university students, healthcare practitioners, and academics. Respondents were given the choice of completing the questionnaire in English, Inuktitut, or French. A small number of questionnaires were completed via telephone if the respondent was unavailable to answer the questionnaire in person at the time and requested a telephone survey (9.2% in September and 17.5% in May). During both seasons, surveys were performed using an iPad-based application, iSurvey (version 2.8.3). A $20 Canadian dollar (CAD) gift card for local food stores or gas stations was provided as compensation to respondents along with a ticket for a larger prize draw, as per guidelines for conducting research in northern settings.
To estimate food security status, we used a food security questionnaire with a modified recall period based on the United States Department of Agriculture (USDA) Food Security Survey Module (FSSM) (29). The FSSM is a validated and widely used tool to measure food security (30). The module contains 10 standard questions and an additional 8 questions asked if children (B18 years old) are living in the household. In contrast to previous northern food security research that used a 12-month recall period (5,19,31), but consistent with other studies (32,33), we employed a recall period of 1 month. This shorter recall period allowed for repeated sampling and assessment of seasonality. More importantly, discussion with local residents and decision makers revealed concerns over asking questions based on a 12-month recall period, which was believed to be too long. This recall period is similar to the methodology of other studies, such as work done in Toronto, as well as in the USDA Reports on Household Food Security (32,33). Several Arctic studies and the Canadian Community Health Survey use a 12-month recall period, however, and must be considered when comparing this study with previous work.
Food security status was determined using the USDA classification (Appendix A). Each household was given a score that represented the number of affirmative (positive) responses from the FSSM (29). Positive answers were coded 1 and negative answers were coded 0 and were then totalled (Appendix B). The household score was then converted to a code ranging from 0 to 3, based on a scale of severity of food insecurity. A food secure household was coded as 0 or 1, while a food insecure household was coded as 2 or 3 (Appendix A).
Pearson's chi-squared test of independence was used to verify the association between food security status and 9 respondent characteristics previously identified in the literature as food security risk factors in the Canadian Arctic (Table I). For each test, food security was defined as a dichotomous variable reflecting food secure versus food insecure, with the food secure category including 2 levels of food security (''high'' and ''marginal''), and the food insecure category including 2 levels of food insecurity (''low'' and ''very low''). Multivariable logistic regression was used to identify significant associations between food insecurity and potential risk factors. Three multivariable models were built: (a) September 2012 model, (b) May 2013 model, and (c) a model with aggregated September and May data. For each model, a best-fit model was built using a manual iterative backward stepwise elimination procedure to identify key predictor variables. Selection used a preliminary significance of a !0.20. Spearman's rank correlation was used to determine correlation between predictor variables to avoid collinearity problems in the models. Predictor variables that were strongly correlated with each other were removed and the more significant variable was retained. Results were considered statistically significant at a 00.05. Lowess smoothing was used to visually verify linearity between age and food security. Variance inflation factors (VIF) for each model were examined, and Akaike's and Bayesian information criteria were used to build our best-fit model. We graphically assessed standardized and Pearson residuals, influence by delta-beta, and leverage using scatterplots. Any covariate patterns showing unusual values were noted, and models were rerun without these covariate patterns to assess any changes in the coefficients and p values in the model to ensure that the models were appropriately specified and well-fit. Data were imputed in Microsoft Excel (Version 12.0) and tests conducted using Stata/SE (Version 13.0).

September 2012
Forty-two of the 532 September household questionnaires were removed because the respondent in charge of food preparation respondent did not answer the food security  (Table II). The proportion of food insecure households is more than 3 times higher than the Canadian average (8.3%) (34). The level of food insecurity in Iqaluit were, however, lower compared to the Nunavut average (36Á69%), including Igloolik (64%) and Kugaaruk (83%) (3,6,9,34). Prevalence of food insecurity in our study was also lower compared to the Canadian Arctic average as measured by the Inuit Health Survey (63%) (7).

May 2013
For the May 2013 data, 33 of the 523 questionnaires were removed, as the food security section was not completed. Thirty-seven additional questionnaires were excluded as they had incomplete responses that could not be imputed, leaving 453 participants responses (26% of all) for analysis (adjusted response rate of 49%) (25).
Among the 453 participants who answered the food security section of the questionnaire, 275 (61%) were female and 178 (39%) were male. Two hundred fifty-four (56%) self-identified as being Inuit and 199 (44%) as non-Inuit. Prevalence of food insecurity was similar to the September results and lower compared to other Arctic communities: 330 (72.8%) of households (Inuit and non-Inuit) were considered food secure and 123 households were categorized as food insecure (27.2%) (Table II) (34).

Characteristics of food insecurity in Iqaluit
Food insecurity in general and more severe food insecurity in particular were more prevalent among Inuit households completing the survey than non-Inuit households (pB0.05, Tables II). Approximately 45% of Inuit households surveyed in both September and May were considered food insecure compared to only 5% of non-Inuit households in September and 4% in May (Fig. 2). Due to the low number of non-Inuit respondents categorized as food insecure (n08 in September, n 07 in May), and also our expectation that processes and predictors of food insecurity may be patterned by ethnicity, we restricted univariate and multivariable analysis to Inuit respondents only (Table III and IV).

Risk factors for food insecurity among Inuit households
The prevalence of food insecurity and the risk factors associated with food insecurity did not vary between September and May. As such, the model that aggregated the September and May data is presented. In the best-fit model, food insecurity was associated with households living in poor quality housing and relying on income support (Table V). Increased age of respondents in charge of food preparation (aged 41 and older) was associated with lower odds of food insecurity than respondents in younger age groups (0Á20 and 21Á40 years old). Employment status of the person in charge of food preparation was also associated with lower odds of food insecurity. Season was not significantly associated with food insecurity and did not change or confound the other variable coefficients or p values in the model. However, we forced the season variable into the model to reflect the structure of our dataset. Post-estimation diagnostics indicated that the model was a good fit for the data.

Discussion
As one of the first published studies to examine the prevalence of food insecurity specifically in Iqaluit, Nunavut, this paper contributes to a nascent scholarship focusing on food insecurity in the larger, rapidly growing settlements of the Canadian Arctic. While the magnitude of food insecurity documented here is lower than results presented in previous work focusing on smaller communities and from regional studies, the prevalence of food insecurity is still higher than in southern Canada (Table VI). This result indicates that food insecurity remains a problem even in large Canadian northern communities. Food insecurity was strongly influenced by ethnic origin, with the percentage of food insecure Inuit households being 9 times higher than non-Inuit households in September 2012 and 11 times higher in May 2013. Indeed, when Inuit households alone were included in the analysis, the prevalence of food insecurity was closer to that documented elsewhere in Nunavut. This prevalence of food insecurity was particularly high given the strong economic growth in Iqaluit associated with resource development, government and associated services, and was consistent with research on Iqaluit food programmes, which has identified a chronically food insecure subset of Iqaluit's population who has been unable to benefit from economic development (16). While we do not examine the underlying causes of such trends here, other work has identified acculturative stresses associated with community relocation, environmental dispossession, and often, forced cultural assimilation (e.g. through residential schools), as important underlying stresses facing contemporary Inuit settlements, and which frames low rates of educational attainment, higher unemployment and food security challenges (11). Indeed, as Wakegjijig et al. (14) note, to  achieve any kind of success in the North, food policy must take into account these broader determinants.
Poor quality housing, unemployment and reliance on income support were associated with food insecurity for Inuit households. These factors usually indicate a lower socio-economic status, reducing households' ability to afford fresh nutritious food (7). The Inuit Health Survey also found a higher prevalence of food insecurity in individuals living in a house requiring major repairs (7). Additionally, work done in the Canadian Arctic and across Canada indicated that households relying on income support exhibited higher prevalence of food insecurity (7,31). Unemployment is often associated with household food insecurity, although this has only been shown in communities outside of the Canadian Arctic (35). Older age was also significantly associated with food security, which might reflect better budget management skills (8).
We did not find gender or country food consumption to be associated with improved food security among Inuit respondents, contrary to previous work done elsewhere in the North (9). The lack of association between gender and food security status could potentially be explained by the format of the surveys. The questionnaire inquired about food security at the household level rather than at the individual level, which may or may not represent the food security status of the person in charge of food preparation.
The finding that consumption of country food was not associated with a food secure status supports the idea of a ''nutritional transition'' taking place in the Canadian Arctic from traditional foods to store-bought foods, especially in younger generations and in the larger Inuit settlements (36). Moreover, given a high degree of transience in residence, in-migration from other communities, and the size of Iqaluit, it has been argued that food sharing is practiced less often compared to smaller settlements, such that traditional foods are not necessarily available when people do not have the funds to access store-bought foods (15,16), an important and widely documented coping mechanism in smaller communities. A third explanation is that despite the continued importance of hunting and fishing in Iqaluit, engagement in these activities is proportionally lower than in smaller communities due to the strong wage economy (37). Nonetheless, wage workers in Iqaluit are still able to access country food through purchase rather than direct harvesting, which reflects yet another difference with smaller communities (38). Finally, food security prevalence and associated risk factors in Iqaluit did not differ by season. Again, unlike smaller communities, Iqaluit's economy is primarily wage-based and less dependent on hunting and other harvesting activities, which are heavily influenced by climatic conditions (37). The results provide timely insights for food policy in Nunavut, which has been identified as a priority by different levels of government, communities, and activists (14,39) and emphasizes the unique needs of, and differences faced by, larger settlements.
Food insecurity remains a critical issue in Iqaluit. Future policies and programmes need to consider identified risk factors, taking into account the distinct needs and challenges faced by urban food insecure households. Specifically, interventions to improve financial accessibility to food, such as the Nunavut Food Security Coalition's Nunavut Food Security Strategy and Action Plan 2014Á 2016, have specific importance. Indeed, two of the Coalition's goals are to ''help align income assistance food allowances with the cost of living in Nunavut'' and ''further develop the priority of instilling self-reliance among Public Housing tenants'' (39). The Government of Nunavut's new mandate, Sivumut Abluqta, also aims to improve public housing and income support to Nunavummiut (40). Such efforts have particular relevance for food security as they take place within the broader context of reducing poverty and rely on collaboration between different governmental agencies, not just public health. As the Coalition states ''[food insecurity] is larger than the mandate of any one organization. A collaborative approach is essential'' (39).
There are several limitations to this study. First, the food security questions were household-specific, while some predictor variables (such as ethnic origin) were measured primarily Á and necessarily in most cases Á at the individual level. As such, the socio-economic characteristics of the person in charge of food preparation might underestimate that of the household, as that person is more likely to stay at home and not be formally employed, while other household members might be. This bias is a challenge faced not only by this study, but also by observational studies in general using similar household-level standardized surveys. Moreover, predictor variables that evaluated food quality or attributes of traditional food systems such as food sharing were not included, which might underestimate food insecurity.

Conclusion
The prevalence of food insecurity in Iqaluit was more than 3 times higher than the Canadian average and was strongly patterned by ethnic origin. These results highlight the persistence of socio-ethnic gradients in food insecurity in the Northern Canada and suggest that the factors that affect the vulnerability of households to food insecurity may be different between large and small Inuit communities in Canada.

Authors' contributions
Yang Guo co-analyzed data, interpreted results, prepared the primary manuscript draft, and led the editing. Dr. Lea Berrang-Ford conceptualized the project, co-designed the questionnaires, co-supervised analyses, contributed to interpretation of results, and contributed to manuscript writing. Dr. James Ford conceptualized the project, contributed to community-based research collaborations, contributed to interpretation of results, and contributed to manuscript writing. Marie-Pierre Lardeau contributed to analyses and interpretation of results, as well as manuscript editing. Dr. Victoria Edge contributed to community-based research collaborations, contributed to planning and facilitation of data collection, and contributed to manuscript editing. Kaitlin Patterson contributed to data cleaning and analyses. Dr. Sherilee Harper co-designed the questionnaires, contributed to community-based research collaborations, led data collection, co-supervised data analyses and contributed to manuscript editing.