Nutritional diversity and community perceptions of health and importance of foods in Kiribati: a case study

Over the twentieth century, small island states in the Pacific experienced a dietary shift from traditional foods characteristic of the region to imported foods of poor nutritional quality. In Kiribati, noncommunicable diseases including diabetes and hypertension are now the largest burden of disease in the country and the primary cause of premature mortality. Food and nutritional security in Kiribati are now a major public health challenge. Climate change is likely to negatively affect traditional food production systems. In this study, twelve gender-segregated focus groups in six communities were carried out across Kiribati. Food charts were obtained from each focus group, where all food items mentioned by communities were listed and scored for health and importance. These were analysed using Excel and Stata. Health scores were on average 42.4% lower for imported food compared to local food with strongly significant findings at a p value of <0.001. This significance however disappears entirely for importance scores, with a p value of 0.943, indicating that while local foods are perceived as healthier, they are not seen as more important. Nutritional diversity is greater in rural communities compared to urban communities in Kiribati. Taking local perceptions of food into account, this study informs policymakers on new strategies to improve food and nutritional security and help island communities become more climate resilient through a renewed focus on local agricultural production, which would also improve health outcomes. Food access and availability issues in Kiribati mean that public health policies should factor these in future policymaking.


Introduction
The Republic of Kiribati is a Western and Central Pacific country consisting of 32 coral atolls and one raised atoll straddling the equator and the 180th meridian. The islands are split into three groups with the Gilbert Islands to the East, the Phoenix Islands and the Line Islands to the East, with a total land area of around 800 km 2 ; they are dispersed over a large area of ocean roughly equivalent to the area of India at 3.5 million km 2 (Fig. 1). It is home to more than 110,000 people with most inhabitants concentrated in the Gilbert Islands group and more than half the population living in the southern part of the Tarawa atoll, which is also the capital of Kiribati (Kiribati National Statistics Office 2016).
Kiribati is the country with the lowest GDP per capita in the Pacific (Government of Kiribati and MFED 2016) and experiences inherent climate vulnerability with limited land area, population pressure, low-elevation of islands and water insecurity exacerbating public health problems . Atoll socioeconomic activities can be said to be almost entirely coastal, given the topography characteristic of an atoll island. In recent years, Kiribati has often made it to news headlines in relation to its vulnerability to climate change, with a narrative of islands being submerged this century due to sea level rise (Bowers 2017;Doherty 2017;Ives 2016;UN News 2019;Warne 2015).
Food and Nutrition Security (FNS) plays an increasing role in the burden of disease in Kiribati, with Non-Communicable diseases (NCDs) such as diabetes and cardiovascular disease now the primary causes of premature death in the country World Health Organization 2016). The transition towards urbanisation in Kiribati plays a dominant role in changes in food dietary intake, with a shift from traditional foods characterised by legumes, tubers, fresh fish, seafood and green leafy vegetables to one heavily dependent on non-nutritious, cheap, imported food consisting of oils, fatty and processed meats, confectionary and dominated by rice as a staple food (Campbell 2015;Farrand et al. 2017;Houghton and Brown 2017;Sahal Estimé et al. 2014;Santos et al. 2019;Snowdon et al. 2013). Malnutrition, vitamin and mineral deficiencies continue to be a major public health threat in Kiribati because of this dietary shift (Eme et al. 2019).

Food security and climate resilience
Much has been written about the problems facing Kiribati and other Pacific Island Countries and Territories (PICTs) regarding food security in lieu of the growing challenges of climate change (Barnett 2011;Hanna and McIver 2014;McIver et al. 2015;Russell 2011; SPC Secretariat of the Pacific Community and CSIRO 2011; Weir et al. 2017). Food security is a key component of climate resilience (FAO et al. 2018;Tyler et al. 2013) and Kiribati, with a heavy dependence on imported food long-recognised by international organisations (Connell 2015;Food and Agriculture Organisation 2008;King et al. 1984;Sahal Estimé et al. 2014; World Food Programme and Pacific Community 2018) is particularly vulnerable to problems in international shipping or rises in food prices, which will be explored in greater detail below (Rimon 2011).
In a review by Savage et al. (2019) four dominant pathways between climate change, FNS and NCDs were identified: the impact of climate change on agriculture, fisheries, migration and humanitarian food assistance. These pathways, if not addressed, are likely to lead to impaired FNS and an increased burden of NCDs in PICTs. In the context of Kiribati, a complex picture emerges where cultural, socioeconomic and environmental problems collide to make an increasingly urgent situation that needs addressing. Most international reports, however, fail to recognise the nuances and particular circumstances of FNS in a geographically spread-out island nation, with little focus given to the importance of community perspectives on why FNS is becoming increasingly problematic (Cvitanovic et al. 2016). Academically, this study is important as there is an identified lack of research in vulnerable communities in small island developing states (Moncada et al. 2018); the inclusion of community perspectives, often neglected in climate change adaptation research is increasingly important as they are on the frontline of climate change and require urgent attention. Over the course of the twentieth century, Kiribati experienced a nutrition shift as globalisation increasingly impacted its food markets and the dietary preferences of its citizens. In the early 1930s, the renowned anthropologist of Kiribati culture, Sir Arthur Grimble described the people of Kiribati as 'Pandanus People' and wrote extensively about the food practices at the time which were still heavily dominated by traditional foods (Grimble 1933a(Grimble , b, 1934. Traditional food in Kiribati consisted mainly of plants foods in the form of roots and tubers, with accompanying pieces of fish or coconut. Terry Coyne describes the traditional diet as '… in general, moderate to high in energy, moderate to low in fat, moderate to low in protein, high in complex carbohydrates and fibre, possibly high in antioxidants, potassium and trace minerals and low in simple carbohydrates (sugar) and salt' (Coyne 2000).
According to P. Alofa, (personal communication, 12th May 2019) and D. Ieete (personal communication, 15th May 2019), rice was introduced into Kiribati during the colonial period when workers on Banaba island were provided rice as a staple: on their return to islands such as Tarawa, rice acquired a status symbol, and demand for its provision increased to the point of becoming a staple in the local diet, replacing other food items such as breadfruit which takes longer to boil and process for consumption. This pattern of introduction of rice is similar to that of other Pacific island countries in the post-war period, such as the Solomon Islands (Jourdan 2010) and Fiji (Pollock 2017). This narrative is matched by published reports on dietary consumption in Kiribati. Following World War II, the Colonial government produced reports on the changes occurring in Kiribati society over the following decades, including nutritional and dietary changes. A study was carried out in 1953 by Holmes that showed that while the diet was still mostly traditional, rice and sugar were making inroads in Kiribati diets (Holmes 1953). There was at least a 40% increase in consumption of flour, white rice and sugar in Kiribati between 1948(Holmes 1953Turbott 1949) and 1968(Willmott 1968). In 1968, Wilmott also found that consumption of flour, white rice and sugar was higher in the urban area of Betio than the rural atoll of Maiana (Willmott 1968). Dental caries and Vitamin A deficiency was also reported, with greater prevalence in urban areas (Danks et al. 1992).
Studies carried out in the 1970s and 1980s Thompson 1978;Pargeler et al. 1982) showed that imported food comprised an ever increasing proportion of food consumption, especially in urban areas, with significant differences compared to rural areas. By 1984, the prevalence of diabetes in urban South Tarawa was reported at around 9% as compared to 3% in rural North Tabiteuea and was described as alarming. The dependence on rice in South Tarawa was such that they were described as 'rice eaters' .
Consequently, there has been a strong decline in consumption of starchy vegetables from the 1960s to the 2010s (Food and Agriculture Organisation 2019a; Lewis 1988). These are an important source of dietary fibre, pointing to a significant dietary shift. This has been accompanied by a rise in canned produce and imported poor-quality meats that is severely impacting the health of the people in Kiribati following a pattern across many Pacific Island nations (Hughes and Lawrence 2005;Sahal Estimé et al. 2014; World Health Organization 2018a).

The prevalence of NCDs in Kiribati
The situation since has only worsened. Diabetes prevalence figures vary, but are now in the 22% -25% range and rising, as compared to the 9% prevalence for the WHO Pacific region (Government of Kiribati and MFED 2016;World Health Organization 2015, 2018b. Cardiovascular disease and diabetes are now considered the two highest causes of premature death in Kiribati (Institute for Health Metrics and Evaluation (IHME) 2018). Lower limb amputation due to diabetes and smoking is the most rapidly increasing reason for surgical admission in the country (Department of Foreign Affairs and Trade 2014; Government of Kiribati and MFED 2016).
Given the trends described above, policies need to effectively address the problem of poor FNS by strengthening local traditional food production. In doing so, Kiribati would improve its climate resilience, food security and nutritional diversity, while taking on the considerable non-communicable diseases burden that challenges the health of its people (Mbow et al. 2019).

Rationale for study
The rationale for this research was to assess community perceptions of importance and health in relation to their food. While there is a wide expanse of literature on the problems facing Kiribati in terms of FNS, there have been few efforts in trying to explore community perceptions of this problem (Cvitanovic et al. 2016). The social construct of individualism and Western modes of thinking often fail to recognise the rich community perceptions of Kiribati culture (Kuruppu 2009).
The aim of this study therefore is to investigate how communities value their food in terms of importance and health, which can give us insights into how it plays a role in FNS and decision-making processes on the household level. (Kuruppu 2009) Recently, some studies focusing on maternal, infant and young child nutrition have found multiple levels of influence on household food decision-making in Kiribati: improving knowledge on the health of food items was not enough, and problems relating to the food system play a strong role in decisions surrounding food selection (Kodish et al. 2019a;b). This study further explores these findings, focusing on foods consumed by communities in Kiribati leading to FNSrelated NCDs. This study was also carried out on both Rural and Urban islands as defined by the Government of Kiribati (Kiribati National Statistics Office 2016), which differ in food availability and access to both imported and traditional foods, and can be seen as having different food systems, though increasingly connected.
Importantly, this paper follows recent literature outlining environmental effects being experienced at present (Cauchi et al. 2020), and increasing in magnitude due to climate change. An integrated climate change adaptation policy for Kiribati should include FNS, and this study could strongly contribute to effective strategies in strengthening food security as it puts community perceptions at the heart of its rationale.

Methods
This article provides a descriptive analysis of a mixed methods approach, where qualitative data in the form of focus groups employed quantitative methods of analysis, including regression modelling. The data was collected over three months between February and May 2019.

Study location
Four islands in the Gilbert Islands group of Kiribati were selected for this study: South Tarawa, Butaritari, North Tarawa and Nonouti (see Fig. 2 below). South Tarawa represents the urban population of Kiribati, where approximately half the population of Kiribati population is located, whereas Butaritari (Northern Gilberts), North Tarawa (Central Gilberts) and Nonouti (Southern Gilberts) represent rural islands corresponding to a decrease in annual precipitation prevalence from North to South. This selection was important to represent the diversity of traditional foods and associated traditions, which on atoll islands with poor soils would be heavily dependent on rainfall. Butaritari has the highest rainfall in the entire Gilbert islands group, approaching 3000 mm of annual rainfall, with rich soils (Office of Te Beretitenti 2012a). Rainfall is significantly less on North Tarawa and is down to around 1500 mm of annual rainfall in Nonouti in the Southern Gilbert islands, which often suffers from prolonged droughts (Government of Kiribati 2019; Office of Te Beretitenti 2012b). The selection of these islands was also discussed with the Department of Public Health at the Ministry of Health and Medical Services (MHMS) in Kiribati, who gave their recommendations and support for this study.

Study design and methods
The qualitative data were collected during 12 focus group sessions (FGs) in the form of food charts. The research was carried out with the help of two partner organisations in this research, the Kiribati Health Retreat Association (KHRA) and the Kiribati Climate Action Network (KiriCAN) who have been carrying out outreach activities over the years related to climate change and food security, and are well-known by certain communities on South Tarawa. The language of people in Kiribati is i-Kiribati (Gilbertese) which is a Micronesian language, with a large proportion of the population also fluent in English. As sole on-site researcher of the study not fluent in i-Kiribati, the first author was accompanied for translation and facilitation purposes by a member of KHRA whose primary language is i-Kiribati and is also fully fluent in English and has accompanied previous researchers in Kiribati in the past.

Urban South Tarawa
On South Tarawa, communities were selected based on a diversity principle in order to reflect the difference in communities on the island. Communities on South Tarawa exhibit diverse characteristics, going from the densely urbanised Betio to the west to the increasingly rural east past the town of Bikenibeau. Betio was selected as a 'Dense Urban' community due to its very high population density at around 11,000 people per square kilometre (Asian Development Bank 2014; Kiribati National Statistics Office 2016). Teaoraereke was selected as a fishing community to reflect what probably would be a greater nutritional diversity based on a more direct access to seafoods than other communities would. Tamoa community is at the easternmost flank of the atoll where it bends to become North Tarawa. The population density decreases here such that it can be defined as a community with relatively easy access to both agricultural produce and imported food diets typical of Betio. It is therefore described as 'Agricultural Urban', having characteristics of both Agricultural and Urban characteristics in terms of food access. All three communities were familiar with KHRA or KiriCAN. This was an important factor as familiarity of the community with the accompanying NGO staff made it easier for the setting up of focus groups.

Rural Outer Islands
Butaritari in the Northern Gilberts was selected as a rural outer island due to its rainfall and its food abundance, whereas Nonouti in the Southern Gilberts was selected as a rural 'dry' island on the recommendation of the MHMS. On these islands communities were selected on the recommendation of the local island mayor. North Tarawa was selected as a rural island that had easier access to imported food than other outer islands due to its proximity to South Tarawa and the main distribution centres of Betio and Bairiki. North Tarawa can be defined as 'Proximal Rural' for the purposes of this study as the closest Outer Island to the main distribution centres of Betio and Bairiki on South Tarawa.

Focus groups
The focus group activities were adapted from the Cristal Food Security Manual, a specialized version of the Communitybased Risk Screening Tool -Adaptation and Livelihoods (CRiSTAL), (IISD 2014) developed by Tyler et al. (2013). This is a tool to systematically assess the risks to the key factors that underpin FNS issues posed by climate variability and change. The complexity of food systems can often be overwhelming, and by using participatory rural appraisal including food charts and mapping exercises one can obtain a community's own subjective experience of food security and socioecological challenges to its health and wellbeing. This is a tool targeted at local-level decision makers, and thus it was thought suitable for the community-level focus of this research. Table 1 describes the communities selected for the focus groups, characterising them per community type.
Community elders were first approached by the partner NGO or a village local councillor. The selection of participants was carried out by leaders of the community, who invited as diverse a group of people as possible in terms of age, employment and social standing. Seven to ten men/women were invited to participate in gender-segregated focus groups: this was done with the rationale of accommodating cultural traditions while ensuring that participants felt free to express their opinions and outlook. Focus groups were carried out in the community's meeting house (maneaba), with the exception of Nonouti for practical reasons. Each focus group took between 90 min and 150 min. The mean provided age of all focus group participants was 36.8 years (n = 114, SD = 15.73); for men mean age was 33.5 years (n = 55, SD = 13.07) and for women mean age was 40.0 years (n = 59, SD = 17.38).
All output was written in i-Kiribati by the participants themselves whenever possible. No audio-recordings were carried out given the language barrier and the nature of the desired output in the form of maps and food chart. Questions asked at the end of the focus group were written down in English following on-site translation. This ensured that as many topics as possible were explored and any salient gaps in knowledge addressed. Subsequent translations of all written output were carried out in the months following data collection.

Food charts
A main output of the focus group was a food chart. The questions asked for this output were as follows: a) Please list all the food items you eat in your community b) How important is each food item for your community?
Score this importance from 0 to 3 to each food item, 0 being 'not important at all' and 3 being 'very important to us'. c) What do you think of each food item in terms of health?
Give a health score to each food item listen, 0 being 'very unhealthy' and 3 being 'very healthy'. d) What problems are there, if any, in obtaining the food item listed?
This activity therefore led to a community-generated food chart listing all the food items in no specific order, and then led to the community scoring each food item.

Data analysis
Data were inserted into a Microsoft Excel (Microsoft Corporation 2018) listing all food items along with the scores given in each community. Translation of each food item listed was carefully carried out and spelling differences between islands were considered so that any duplicates were removed. For the purposes of this research any drinks mentioned as food items, such as coffee, cordial and tea were also included. This decision was made in order to respect cultural perceptions of what constitutes a food, while also being recognised as a food under the FAO's own list of food categories (Kennedy et al. 2011). Each distinct food item was then categorised by food type and its frequency of mention counted, then multiplied by the average importance score given across all communities to generate an 'importance index'. The rationale behind including both counts and importance scores to measure importance is that the frequency of mention of a food item would indicate how important it is to Kiribati communities across the spectrum. The maximum score possible was 36 (12 communities * maximum average value of 3). These scores were then transformed into values between 0 and 1, 1 being the maximum possible score. Health scores were transformed similarly, with average health scores divided by 3 (maximum score possible). An example of these transformations is provided in Box 1.
Box 1 Generation of the importance score for a food item The native fig te Bero (Ficus tinctoria), a local food, was mentioned in 10 communities and on average an importance score of 2.0 was given. 10 mentions (counts) * 2.0 (importance score average)=20 20/36 (maximum score possible)= 0.56 (importance score) The same food received an average health score of 3.0 across the 10 communities. Health score of 3.00 / 3 (maximum possible score)= 1.00 (health score) Te Bero therefore got an importance score of 0.56 and a health score of 1.00, overall.
This score was generated for each of the food items listed first altogether, and then for each of the two main community categoriesurban, and rural. This therefore made it possible to compare which foods are given more importance in the different island categories.

Nutritional diversity scoring
While by no means exhaustive or quantitative, the dataset obtained proved useful to analyse nutritional diversity in terms of how many items were mentioned in rural and urban communities. Certain communities mentioned more food items than others. By grouping the number of food items mentioned per category, a picture emerged of a contrast between sources of nutrition mentioned in urban communities compared to rural. When further subdivided into smaller community types (see Table 1) an interesting picture emerged, giving indications as to which communities have greater access to a more diverse food.
For the 53 common food items mentioned in both urban and rural communities, importance and health score difference indices were created by subtracting rural scores from urban scores. This therefore gave a picture of how urban perceptions of food differ from that of urban communities.

Statistical analysis
Statistical analysis was carried out on the data obtained using Stata 15 (StataCorp 2017). The food groups (imported vs local) and subcategories (fresh vs processed) were analysed using linear regression analysis against the importance and health scores in order to collectively assess whether there is a difference in scoring for different categories.

Descriptive analysis
A total of 98 distinct food items were mentioned across 12 focus groups. Following consultation with our research partner KHRA three food items were merged into others as they were deemed simply to be the same item spelt differently: 'te bwan' and 'te ban', 'te aai' and 'te uaii', 'te kumara' and 'te koumara'. One food item mentioned by just one community, 'Katawawa' was removed from the list as nobody was able to translate it and it was important to remove any potential duplicates. 94 food items were then considered for analysis. The following table shows the different food items mentioned categorised first simply as 'imported' vs 'local', and then subcategorised as a) imported processed b) local processed c) imported fresh and d) local fresh. This therefore distinguished the different food items in terms of source and processing, which has important bearings on health potential of foods (Farrand et al. 2017;McLennan and Ulijaszek 2015;Sahal Estimé et al. 2014;Snowdon et al. 2013). Table 2 shows the prevalence of food types mentioned during the focus groups. There is almost no difference in proportion of food items for the main categories of imported food and local food in urban and rural communities. There was a greater number of food items mentioned in rural communities which might reflect a greater nutritional diversity. A ratio (called 'Diversity Score') was subsequently created to reflect this difference, where the urban (n = 57) and rural (n = 78) total foods mentioned were divided by the total number of foods mentioned (n = 94). Table 3 above summarizes these diversity scores, and indicates that rural communities visited had a greater diversity score compared to urban communities, possibly indicating a greater diversity of foods available. Table 4 summarizes these scores on a more individual level (communities), where a pattern emerged where rural communities on average scored a higher diversity score.
The health and importance difference score was created and plotted on a graph (see Fig. 3). This was done by subtracting rural scores from urban scores, thereby giving a picture of how urban communities perceive food items in terms of health and importance. Note that, in order to avoid graph clutter, all health difference scores of zero were excluded, so this graph only demonstrates food items where health scores were different (n = 18).

Regression analysis
Regression analysis was carried out comparing local food and imported food. Health scores were on average 42.4% lower for imported food compared to local food with strongly significant findings at a p value of <0.001 (Table 5). This significance however disappears entirely for importance scores, with a p value of 0.943. This might indicate that while local foods are perceived as healthier, importance given to local food doesn't differ from importance given to imported food.
This analysis was repeated for urban and rural groups respectively, to assess whether there is a difference in assessment of health and importance between the different communities, shown here in Table 6: The results of this community type analysis (Table 6) shows similar results, with local foods perceived as important as imported foods in both community types (urban and rural).
Finally, regression analysis was then carried out on a combination of food categories which were a) imported processed b) imported fresh c) local processed and d) local fresh (see Table 7). This was done in order to assess subtypes of food categories in the same vein. 'Imported Processed' food was considered as baseline, as it was assumed to be considered the least healthy food category.
Using 'Imported Processed' food as baseline, health scores were compared. As can be seen here local food consistently scored higher health scores with strong significance. Local processed food and local fresh foods had on average 34.0% higher and 47.8% higher health scores, respectively. Local foods are perceived as healthier foods, the fresher, the better. However, there is no difference across all food categories in terms of importance.

The urban and the rural: Two lifestyles, linked food systems
Approximately half the people of Kiribati live in what the government defines as 'urban' areas, a convenient comparative to 'rural' areas (Kiribati National Statistics Office 2016; Storey and Hunter 2010a). Urban areas tend to have a cashbased economy, with people employed and receiving a wage while in rural areas subsistence lifestyle holds sway. With land increasingly limited in urban South Tarawa, people have little to no space for agriculture, limiting their access to local foods (Office of Te Beretitenti 2012c). Rural communities on the other hand have difficulties in obtaining imported food due to irregular shipments and the extra cost of fuel involved (Ministry of Environment Lands and Agricultural Development 2020; Rimon 2011). Very little published information is available on domestic markets, however these seem to be few in number (Rimon 2011). Similarly, a freight subsidy for locally grown produce only includes certain items (eg. banana, pumpkin) while excluding others such as giant swamp taro (babai), local figs, breadfruit, coconuts and marine products (Rimon 2011). Local knowledge on traditional preparation and preservation of food is largely unknown to young people today, leading to a decline of consumption of local food especially off season (Ministry of Environment Lands and Agricultural Development 2020).
Health perceptions of local food remains strongly positive, and this is reflected in the research outcomes shown above. The implications of differences in health and importance scoring, however, is that while local food is perceived as healthier in general, it doesn't make it more important. This therefore implies that health perceptions of food may not play a significant role in the decision-making process for food choices in Kiribati. Factors such as availability, convenience and taste may play a more significant role in community food choices than knowledge or perceptions of health in relation to food (Rimon 2011). Interestingly, these findings corroborate recent research by Kodisch et al. (Kodish et al. 2019a;b) also emerging from Kiribati. With large families often living in clustered units, households in Kiribati, practical choices encompassing affordability and convenience of foods might take precedence over health. Consequently, while some foods are deemed 'less healthy' they are also 'more important' to the decisions makers in a household, and given more prominence. Rice, for example, is easily bought in large amounts and needs to be boiled for a shorter time than breadfruit, so it is more costeffective both in terms of time (no need to collect breadfruit) and affordability (less fuel, cheaper to buy especially on urban islands). This is important for policy purposes in the public health domain, which perhaps needs a new approach in its strategies for tackling the problem of unhealthy food in Kiribati given the scale of the growing NCD problem in Kiribati.

Perceptions of imported and local foods in Kiribati
In 2017, food items accounted for over 30% of Kiribati's import bill (Observatory of Economic Complexity 2018) and food imports account for as much as 35% of the available dietary energy supply (World Bank 2011). In the latest Household Income Expenditure Survey (HIES) carried out in 2006, food accounted for 46% of all household income expenditure (Kiribati National Statistics Office 2006). External economic shocks, particularly the food and fuel prices spikes of 2008, hit Kiribati harder than almost any other country in the world (World Bank 2011). Worst hit were the people living in the outer islands of Kiribati who have little access to cash (Rimon 2011;World Bank 2011). This heavy dependence on imported food has strong implications for FNS and food policy. Together with a corresponding rise in NCDs this is a problem recognised both by the local government and international organisations (FAO 2013; Government of Kiribati and MFED 2016;Kiribati Ministry of Health 2012;Kiribati Ministry of Health and Medical Services 2015). Dependence on imports also potentially makes Kiribati less climate resilient, subject to external climate shocks in the global food commodity market. While the prices of imported food items, (rice, sugar and flour) are controlled by government, (MCIC Kiribati 2018) this comes at great cost, especially when considering the volatility of food prices worldwide. (Childs and Kiawu 2009) Kiribati was subject to food crises in relation to rice in 2004 and then again in 2008, with people found unprepared for shortages in rice, which is now the main food staple (Rimon 2011;World Bank 2011).
In this study, rice, sugar and flour-based products (such as bread and pancakes) were not seen as significantly more important in rural areas (see Fig. 3). Rice, sugar and flour all received very high scores, all being in the highest quintile for importance. This data is corroborated by a recent study by a recent study by Eze Eme et al., which found that sugar and cereals were consumed in virtually all households investigated in their quantitative assessment of dietary intake in Kiribati (Eme et al. 2019). Despite access being more limited in rural islands, their presence in the communities is strong: in fact, rice is mentioned in all communities investigated (refer to supplementary Table 1) and is considered a staple food in Kiribati even though it is not grown there. One notable exception is noodles, which are seen as both more important and healthier in urban communities. Local foods however are highly valued, and consistently received high importance and health scores in the communities investigated. The question here therefore is what the Kiribati government can do to improve access to these local foods, which are healthier and provide greater nutritional diversity than imported food items. With NCD incidence gravely worsening since 1984 (Institute for Health Metrics and Evaluation (IHME) 2019; King et al. 1984) it seems that the many policies implemented seeking to address this problem haven't been successful (Kiribati Ministry of Health and Medical Services 2015;Ministry of Health and Medical Services Kiribati 2015). The government of Kiribati recognises that it currently has limited institutional and individual capacity to plan and implement actions to reduce the impacts of climate change-induced impacts on FNS (United Nations Development Program 2016).
Attempts have been made by the Government of Kiribati over the years to try to improve the infrastructure and regularity of shipments, which are an important feature of food    security in the island nation. Import levies were implemented in 1977 designed to help ensure imported goods sold in outer islands have a similar price to those on South Tarawa, with an amendment in the 1990s to include local agricultural produce sent to South Tarawa from outer islands. Sadly, this doesn't seem to be working effectively, with notable exclusions to this policy such as babai, local figs, breadfruit, coconuts and other food items such as seafood (Rimon 2011). It is important that local agricultural produce be included in any policies that encourage food trade between islands as this could have strong positive outcomes, especially when main local food items (eg. breadfruit, coconut, taro) are given high importance scores, in general. It would also be appropriate to strongly consider a reevaluation of economic instruments such as taxation or food subsidies currently existing in Kiribati. Current subsidies on flour, rice and sugar strongly incentivise the i-Kiribati to purchase items which are strongly contributing to the NCD crisis in the country. Lifting these subsidies while subsidising local foods or healthy alternatives could be an effective incentive to switch behaviour related to both food production and consumption. These options are currently being explored in Kiribati on a policy level. A draft of this policy document was provided during the field trip, but still not officially adopted.
Kiribati is already considered to be a 'canary in the coal mine for climate change' in relation to its climate vulnerability and corresponding health consequences (Hanna and McIver 2014;McIver et al. 2015McIver et al. , 2016. Being subjected to the whims of food commodity prices and external climatic shocks has grave implications on its food security. Some degree of reliance on imported food is a means of making the food security context of Kiribati more climate resilient, since external weather or climatic shocks to local food production might negatively impact the availability of food (Food and Agriculture Organisation 2019b). Infrastructure however remains a major obstacle, even on the household level. Over 82% of households in the country don't have access to refrigeration (Kiribati National Statistics Office 2016) which forces them to seek out items with a long shelf-life, often of poor nutritional quality (Hughes and Lawrence 2005).
Importantly, Kiribati could improve its current supply of local food. This can be done through agricultural tool provision, easing the bureaucratic process of setting up agricultural businesses with local produce and its processing, incenvitising local agriculture and more effective fish and reef management. Kiribati currently subsidses copra production at $2/kg, which is unsustainable as current international market prices hovered around $1.10/kg between 2014 and 2018 (Webb 2020). This is positive in some respects as it encourages Kiribati islanders to work on outer islands instead of migrating to overpopulated South Tarawa, while also serving as a means to redistribute wealth to outer islands. It is having negative impacts, however, in the form of increased expenditure on kava and alcohol on outer islands, resulting in social problems. The subsidy also comes at a significant cost to other possible products, such as crude coconut oil production (IMF 2018). This is set to double to $4 per kg, should the policy promises of the re-elected Maamau government in 2020 be implemented (Cain 2020). It could be more sustainable and cost-effective in the long-run for the Kiribati government to focus financial resources instead on production and processing of local food, boosting the domestic food market, improving public health outcomes and consequently reaping economic benefits of fewer health costs.

Food diversity scoring
The results of the diversity scoring clearly indicate that rural communities in general tend to have a greater food diversity than urban communities. Interestingly, certain urban communities (such as Tamoa) mention food items that aren't mentioned frequently on outer islands. One example is cucumber, which often fetches very high prices in South Tarawa. This might indicate that while urban communities may have less access to the high nutritional diversity of rural islands, they do have access to certain food items that are not accessible to rural communities. An interesting outcome of this scoring is the high nutritional diversity score in the 'Proximal Rural' community of Buariki in North Tarawa. While having a predominantly subsistence lifestyle, it is geographically close enough to the main shipping access ports of South Tarawa that it has a greater access to imported food items. This might indicate a 'sweet spot' for nutritional diversity, where having easier access to subsistence foods while also having relatively easy access to imported foods makes for a better food potential for FNS. As stated above, improving shipment frequencies to and from rural islands could improve FNS for both community categories, and this is commonly cited as a problem and obstacle to motivate inter-island food shipments (Rimon 2011). Conversely, this can have negative impacts if it increases the prevalence of unhealthy imported food in rural islands.

Translation and communicating effectively
As described above, the focus groups were carried out together with a staff member from KHRA/KiriCAN who was fluent in both i-Kiribati and English and who has accompanied previous researchers in the past. However, she is not a certified translator, which can be hard to come by (and afford, given limited funds for this research) in Kiribati. While she did her utmost to translate any input from the community real-time, it is possible that some important information might have been missed during the activity itself. That said, all food chart outputs were written by the community themselves, and her facilitation during the generation of these charts was crucial, especially in defining 'importance' (see below). The staff member and I regularly communicated any problems that were encountered during the focus group to work effectively as a team and ensure we were on the same page in terms of output and research goal.

Defining 'importance'
While there was a general agreement about health scoring in all communities involved, the definition of 'importance' proved more contentious, as the importance of a food item varies from person to person and has different meanings. In all communities therefore importance scores were given as an 'overall' importance score while weighing cultural values, access and availability. This score was decided by participants through consensus after discussing the item.
For example, if an item is not so commonly used (score would be 1) but has high cultural value (score would be 3) such as babai, the community would agree by consensus to give it a score of 2. This differed from one community to another, depending on its utilisation, presence and .
It would have been better to perhaps split the importance into 'utility' and 'cultural value' to get a clearer picture, however due to time constraints this was not feasible. 'Importance', therefore, here describes a community-agreed score and often led to lively discussions by participants on what constitutes an important food item. Given the generational differences in food preferences and the cultural values of foods in Kiribati, this score reflects a subjective output that varies between one community and another while offering a window into how a community would rank its food beyond mere abundance or availability.

Focus group limitations
Much attention was given to cultural norms in Kiribati in order to obtain as truthful and diverse a dataset as possible. Meetings took place, when possible, at the local maneaba, which is the heart of community life in Kiribati culture, and gender segregation ensured maximal output from participants. There were, however, some unavoidable constraints typical of focus group settings. One was a time limit of around two to two and a half hours which while adequate still limited the number of questions that could be explored. The food charts themselves were given over 30 min time if necessary, to complete. Despite having multiple flip charts and markers available, reaching the end of the page might have discouraged some communities from listing more food items. Some communities however went beyond that first page, which was encouraging.
Additionally, given the severe limitations in communication infrastructure with outer islands the research team could not know if there were special events happening on islands that might have affected focus group attendance.

Bias
There were some possible unintended selection biases in the study. Site selection and contact in South Tarawa and North Tarawa was carried out by the Kiribati Health Retreat Association (KHRA) and Kiribati Cliamte Action Network (KiriCAN), our partner organisations. Other international organisations might have already explored issues surrounding food security in Kiribati. Consequently, attending participants might have already been familiar with food security topics, possibly biasing their results. The outcomes of this study however show a consistent high scoring for traditional foods throughout all communities. The selection of communities across the outer islands however was done purposely to obtain as wide a spectrum as possible of communities in the rural category.
Recall biases in this study might have played a significant role due to seasonality. Many traditional foods in Kiribati are seasonal, with seasons for pandanus, breadfruit and other mainstay staple crops of traditional Kiribati diets. As a result, some food items might have been given more importance if they were in season, and vice versa.

Conclusion
This research presents a community perspective to the FNS issues in Kiribati, which are all-too often tackled from a topdown approach with little consideration for the subjective views of communities and, by extension, households. Qualitatively assessing their scoring of food in terms of health and importance gives us a picture of a system that gives great value to traditional food, but does not think they are more important than imported foods, which now are a key part of the Kiribati diet. One can argue indeed that while the heavy dependence on imported food that characterises Kiribati is far from ideal, it does contribute significantly to food security in the context of frequent droughts and variations in rainfall that from time to time affect the Gilbert Islands and are set to become more intense due to climate change (Australian Bureau of Meteorology and CSIRO 2011; Longépée 2016; Storey and Hunter 2010b;Werner 2017). Much can be done in determining the quality of imported food and in increasing the diversity and availability in foods present. Importantly, the reliance on rice, flour and sugar as staples is a threat to public health and is contributing to an urgent problem of NCDs in Kiribati. A recommendation of this study is that, given the high health and importance scores given to some traditional foods such as breadfruit, swamp taro (babai), coconut and pandanus, more attention should be given not only in reminding people of their health benefits but also making them more available and convenient to people. It is important to recognise that lifestyles in Kiribati, especially in urban Tarawa, have changed, and that tastes, affordability and convenience play a key role in the selection of foods on the household level. Supporting production and preparing or pre-processing local food prior to being bought might go a long way at improving peoples' choices when it comes to food. Other strategies, such as lifting subsidies on heavily processed imported foods and subsidising local produce and vegetables instead are to be encouraged.