The Impact of a Household Food Garden Intervention on Food Security in Lesotho

Food insecurity is a challenge in the developing world, where many are finding healthy food inaccessible due to poverty. A pre-test, post-test design was applied to determine the impact of a vegetable gardening intervention in 25 experimental and 25 control households in Lesotho. Information about sociodemographic conditions and indicators of food security was collected by trained fieldworkers. As evidenced by the Living Poverty Index of 2.5, the sample was characterized by high levels of poverty. Although almost no households were scored very low or low using the Months of Adequate Household Food Provisioning (MAHFP) tool, less than half of households were categorized as food-secure. Household Dietary Diversity (HDD) showed infrequent intake of vegetables and fruits and regular intake of fats and sugar. After intervention, the percentage of households with a low HDD score improved significantly in the intervention group (12%) compared to the control group (40%) (95% CI (2.5%; 50.7%)). Despite this, the percentage of households that consumed vegetables during the previous day was still below 30%. Food gardens have the potential to improve availability of food and frequency of vegetable consumption, but harsh environmental conditions need to be considered.


Introduction
The small mountainous country of Lesotho is populated by about 2.2 million people, with two-thirds living in rural areas. The villages in the lowlands of the country are the most populated [1][2][3]. About 60-70% of the Lesotho population is severely food-insecure [1]. In rural areas, the majority of individuals are dependent on agriculture for survival, which proves challenging as less than 10% of the country's total area is suitable for growing crops. Land availability is further influenced by urbanization and environmental factors, such as an erratic climate, soil erosion, and climate change [4]. Thus, the number of landless households is steadily increasing [5].
Lesotho is considered to be one of the least developed and poorest countries in the world. This can be attributed to high unemployment rates, a decrease in production from agricultural activities, low life expectancy, and increased child mortality rates. Half of Lesotho's population lives below the poverty line and depends on remittances to survive [6].
In poor countries, income, food prices, and environmental factors play a larger role in what is eaten than personal choice [7]. Despite these challenges, household agricultural activities form the basis of livelihoods in the majority of households [8].
Research Ethics Committee of the University of the Free State (ECUFS NO 94/2014) and the Ministry of Health in Lesotho, a rapid appraisal census was undertaken in all households that were included in the list of beneficiaries of the intervention partner. During the census, information about the presence of a vegetable garden at each household was collected. Based on this information, households that met the inclusion criteria were identified. From the households included in the census, 50 households that met the inclusion criteria were purposively selected. From the list of 50 households, 25 households were randomly included in the intervention group and 25 in the control group.
Inclusion criteria included: • Households providing informed consent after being adequately informed; • Households situated in the beneficiary community; • Households with no physical evidence of a garden; Households with a non-functioning garden-there may have been a garden, but the garden was not being maintained and had not produced any crops during the past season.

Data Collection Procedures
Prior to the baseline household interviews, fieldworkers from the National University of Lesotho received training on the data collection process. Measures of food security included the Living Poverty Index (LPI), Months of Adequate Household Food Provisioning (MAHFP), Household Dietary Diversity (HDD), and frequency of vegetables eaten.
A pilot study was undertaken prior to the commencement of the data collection phase of the study to determine the amount of time needed to complete the questionnaires as well as the clarity of the questions. Changes that were made to questionnaires following the pilot study included the following: • Underlining of key words in order to assist the interviewer; • Instructions for the interviewer at certain questions were inserted to ensure uniformity; • Wording/phrasing of questions and answer options were changed in the questionnaire to make them more applicable to the target population.
The LPI scale was used [21] to assess the standard of living (an indirect indicator of both poverty and food insecurity) by determining the frequency that households went without basic necessities of life (namely food, water, medicine, electricity and fuel, and cash income). It was assessed for a period of 12 months prior to the interview using a set of six questions, each with six possible responses (the sixth response being 'I don't know'). The range of responses each received a score on a five-point scale. The responses were then combined to calculate the average LPI score for the household, 0 (no poverty) to 4 (complete poverty) [21].
The standardized MAFHP questionnaire developed by Billinsky and Swindale, another indirect measure of food security, was applied to determine MAHFP during the previous year [22]. For each household, MAHFP was calculated by subtracting the total number of months out of the previous 12 months that the household was unable to meet their food needs from 12 (e.g., 12-sum (A + B + C + D + E + F + G + H + I + J + K + L). Values for A through L were '0 or '1 [22]. The scores were categorized into three groups according to level of food security. A score of 12 meant that the household had year-round adequate food provisioning. Households that scored between 11 and 8 were deemed to have moderate food security, households that scored between 4 and 7 low level of food security, and households that scored between zero and 3 were considered to be food-insecure.
HDD Score (HDDS) was defined as the number of food groups consumed during the previous day. The level of diversity in the household diet was determined using the standardized questionnaire on HDD [23] after which the HDDS of a household was calculated. In this questionnaire, the number of different food groups consumed during the previous day was determined from a possible 12 food groups, which included cereals; roots and tubers; vegetables; fruits; meat, poultry, offal; eggs; fish, and seafood; pulses/legumes/nuts; milk and milk products; oil/fats; sugar/honey; and miscellaneous.
Once the data had been obtained, the HDDS was calculated by tallying the total number of food groups from 12 consumed by the members of the household. The HDDS were interpreted in the following way: 0-3 = low dietary diversity; 4-5 = medium dietary diversity, and 6-12 = high dietary diversity [23].
Once the control and intervention households were identified, the first set of baseline interviews was undertaken. A two-person team of fieldworkers, from the National University of Lesotho, conducted structured interviews with the head of the household at the household.

Implementation of the Intervention
As mentioned, the households that were included in the household food garden intervention were beneficiaries of the SWAALES. SWAALES is a nonprofit, nongovernmental organization that aims to achieve a HIV/AIDS-free world and to empower African Women and Children to claim equal rights, access to healthcare, education, economic, and sociocultural opportunities. The 25 intervention households were trained on and assisted with the implementation of their household food gardens from July to September 2014. Vegetables that were planted for the summer harvest included pumpkin, carrots, spinach, green beans, tomatoes, onion, beetroot, and potatoes. Planting for the winter harvest took place between January and June 2015. The gardeners within those households received basic training on garden layout and bed design; natural soil fertility, pest and disease control; food preservation, processing, and storing; seed harvesting and saving; and preparation for winter crops, frost, and cold damage. Training of beneficiary households of the SWAALES program was done by Lima Rural Development Foundation representatives. Maintenance and monitoring continued throughout the remainder of the project (July 2014 to September 2015). Training material(s) were made available in Sesotho. Control households were informed of the incentive they would receive at the completion of the study (the same household food garden intervention as the intervention group).
Following the implementation of the household food garden intervention, a follow-up survey took place between July and September of 2015. The same team that had conducted the baseline survey conducted the follow-up interviews using the same questionnaire.

Validity and Reliability
Validity was guaranteed by researching evidence-based literature concerning indirect measures of household food security and including these in the questionnaire. To ensure reliability, the same questionnaires was used to obtain information from participants. Structured interviews were conducted by local fieldworkers who received training on research ethics, survey methodology, and fieldwork. The use of structured interviews contributed to reliability as they helped eliminate the possibility of research bias and subjectivity, since questions were asked exactly as worded in the questionnaire. Fieldworkers were assisted by footnote instructions in the questionnaire guiding them throughout the interview. The order in which the questions were asked was also carefully considered to avoid previous questions influencing the participant's answers. The questionnaires were translated into the local language, with the original translations being back-translated by another person to improve the reliability of the translations.

Statistical Analysis
Data were captured using a double entry process in Census and Survey Processing System (CSPro) software (Washington DC, USA). An exact replica of the questionnaire was programmed into the CSpro programme and was programmed to disallow any irrelevant data entry (e.g., words instead of numbers, incorrect values), enabling the data capturers to maintain the integrity of the data. Descriptive statistics, namely frequencies and percentages for categorical data and medians and ranges for numerical data, were calculated before and after the intervention per group. The changes from before to after the intervention were calculated and described by means of 95% confidence intervals (CIs).

Household Demographics, Responsibilities, and Structure
No significant differences occurred between the control and intervention groups at baseline in terms of household demographics (Table 1). Gender distribution in the control and intervention group was 72% male and 28% female and 64% male and 36% female, respectively. In both groups, about half of participants were married (Control: 48%; Intervention: 56%), followed by individuals who were separated (Control 32%; Intervention 44%), and slightly more individuals who were unmarried and widowed in the control group (Control 12%; Intervention 0%). In terms of education, about 40% of participants had completed primary school (Control 44%; Intervention 41.7%), while a higher percentage of participants in the intervention group had completed some high school (Control: 20.0%; Intervention: 41.7%). Households in both groups (about 1 in 4) acknowledged a female as the head of the household (Control 24%; Intervention 24%), while a similar percentage acknowledged a male-headed household (Control 24%; Intervention 16%). At baseline, the main meal of the day was eaten at home by about 90% of all participants (Control 88%; Intervention 100%). In both groups, household responsibilities such as buying food were mainly the responsibility of the head of the household (Control 68%; Intervention 64%), who was also the most likely to decide who receives food and when (Control 60%; Intervention 52%). Table 2 indicates the results pertaining to the categories of LPI of the control and intervention groups. The median LPI of 2.5 (range 1.7-4.5) in the control group and 2.83 (range 1.7-4.3) in the intervention group indicated a high level of poverty. The control and intervention groups were not different, except for the variable 'enough clean water for the house.' At baseline, 92% of the participants in the control group reported that they had never or just once or twice gone without water over the past 12 months, compared to 56% in the intervention group, a difference that was statistically significant (95% CI for the percentage difference (1.5%; 44.4%). At baseline, 44% of control households and 64% of intervention households reported not having enough food to eat 'several and many times,' with households that reported going without electricity 'many times' being similar in the two groups (Control: 100%; Intervention: 96%). Households that reported the absence of enough fuel to cook their food 'several or many times' were similar in the control (24%) and intervention (20%) groups, as were the percentage of households that went without a cash income 'several or many times' (60% in both the Control and Intervention groups).

Months of Adequate Household Food Provisioning
In Table 3, the percentage of participants that experienced adequate household food provisioning during the different months is depicted.  The descriptive data in Table 3 was used to determine the categories of scores for MAHFP (Table 4). In terms of score categories for MAHFP, there were no significant differences in the percentage of respondents from scores in the different categories in the two groups at baseline and at follow-up. The households were grouped into food-secure categories according to their MAHFP score. Table 4 shows that at baseline, the control group did not have any households that scored in the very low and low category (0%), with the intervention group having only 8% of households in the low food security category. At follow-up, though not significant, a small improvement was noted in the intervention group where there were no longer any households in the low food security category.
In the control group, the median MAHFP score remained the same (11), while the median score in the intervention group improved by one point (from 10 to 11), a change that was not statistically significant (95% CI for the difference (−2; 0)). Table 5 shows the percentage of respondents that ate the indicated food groups during the previous day, while the HDD scores are depicted in Table 6.

Household Dietary Diversity
The results from the DD data of the current study showed that, as expected, almost all participants consumed starchy cereals on a daily basis. The percentage of participants who consumed dairy and flesh meat every day was also high. After the intervention, the intake of most healthy food groups remained unchanged. Although slight improvements were seen in the intake of vitamin A-rich vegetables and other vegetables in the intervention group, the percentage of households that consumed these foods on a daily basis was still low. Furthermore, a large percentage of participants reported daily intake of unhealthy foods such as fats, oils, and sweets (at follow-up, about 70% of all participants reported consuming sweets on a daily basis). These results were used to categorize households as having low, medium, or high dietary diversity (Table 6).  At baseline, about one-third of the control and intervention households had a low level of dietary diversity (Control 36%; Intervention 28%), while half of the households in both the control and intervention group had a medium level of dietary diversity (48%). At follow-up, the percentage of households with a low HDD improved from 28% to 12% in the intervention group, while 40% of control households were categorized as having a low HDD, a difference that was statistically significant (95% CI for difference (2.5%; 50.7%)).

Discussion
The sociodemographic characteristics of the intervention and control groups were similar and reflected those reported in the LDHS. High levels of poverty were identified, and although the majority of participants were educated beyond primary school level, few had completed high school.
The most basic cause of food insecurity is inadequate access to food as a result of poverty [7,17,24]. According to the United Nations (UN), poverty is defined as "a human condition characterized by the sustained or chronic deprivation of the resources, capabilities, choices, security and power necessary for the enjoyment of an adequate standard of living and other civil, cultural, economic, political and social rights" [25]. Poverty is therefore closely linked to food insecurity. In the current study, the median LPI scores confirmed a high level of poverty.
Although almost no households were scored very low or low using the MAHFP tool, less than half of households were categorized as food-secure. The MAHFP focuses on the perception of whether there is enough food to eat in the household-it does not assess the variety or quality of the food that is available. In this context, DD is universally recognized as a key component of variety and therefore a useful tool when assessing food security [26,27]). The diets of people living in developing countries often lack diversity, since they often survive on staple plant-based diets [17,28]. Almost all households in the current study reported consuming starchy cereal-based foods during the previous day. According to Akhter et al., "diet quality is an important determinant of the food and nutrition security of a population and is influenced by food availability, access, utilisation and affordability at both country and household level" [7]. Darmon and Drewnowski found that poor households chose food to satisfy hunger, selecting cheaper energy-dense foods that were high in fat and sugar rather than more nutritious foods such as fruits and vegetables [24]. In this context, household food gardens have the potential to influence food security through increasing the availability of vegetables, contributing to a more diversified diet and a higher consumption of nutritious food [16][17][18][19][20].
Although the reported frequency of vegetables eaten improved significantly in the intervention group (but not in the control group) and slight improvements were seen in the intake of vitamin A-rich vegetables and other vegetables in the intervention group, the percentage of households that consumed these foods on a daily basis was still largely inadequate. The harsh environmental conditions in Lesotho may have played a role in the limited success of the intervention [4].
Although DD is often used as an indicator of food security, care needs to be taken when interpreting this information. A higher DDS does not guarantee the consumption of a nutrient-dense, quality diet. Unhealthy foods that are high in cheap animal proteins, fats, refined cereals, and sugar are generally more energy-dense and affordable than healthier foods [29]. The results from South African studies have shown that a high DDS may be related to an increased intake of unhealthy foods, such as fast foods [30,31]. The study of Rothman et al. confirmed that women from both rural and urban households in Lesotho underwent a nutrition transition and consumed unhealthy foods such as refined starches, fatty, and sugary foods [9]. The results of the current study are consistent with these findings, with a large percentage of participants reporting daily intake of unhealthy foods such as fats, oils, and sweets. The intake of these high-energy but nutrient-poor foods is closely linked to both food insecurity and overweight and obesity [12].
In addition to the potential of vegetable gardening interventions to address food security and to serve as a source of income generation, they are also associated with social and emotional benefits [16,18]. The creation and maintenance of gardens can contribute to building resilience and a sense of community in both adults and children. Darby et al. (2020) reported that low-income gardeners are motivated by pleasure from the practice of gardening, while also reinforcing social connections and cultural traditions [16].
The small sample size, as well as the fact that the measures that were applied to measure food security focused more on the experiences of participants and on the types of foods that were eaten, are limitations of the study. No information on quantities of foods (especially quantities of vegetables) that were eaten was collected, making it difficult to accurately determine the adequacy of the diet. The inclusion of more than one measure of food security in the current study provided a holistic view of the situation in Lesotho, which is a strength of the study. Applying a variety of tools made it possible to evaluate the contribution of a number of variables to food security, since they focused on different components.

Conclusions
The households included in the current study were characterized by high levels of poverty. Despite this, some measures of food security showed that participants were not as food-insecure as expected. Although significant improvements were noted in the frequency of vegetables consumed in the intervention group that were not noted in the control group, the percentage of households that ate vegetables was still far from the ideal of 400 g per day recommended by the WHO [32]. Faber et al. noted that, for vegetable gardens to have a sustainable impact, access to high-quality natural resources is required [29]. The unfavorable agroecological conditions in Lesotho may have contributed to the limited impact of the current intervention. In view of this, it may be helpful to equip potential gardeners with knowledge on gardening practices that consider environmental challenges (e.g., drought, frost, etc.). Sharing and communal gardens may also benefit this type of community as individuals pool resources and combine knowledge. Ultimately, interventions that target the basic and underlying causes of poor food security, such as poor socioeconomic circumstances, should be prioritized. Moreover, we recommend empirical research to develop an econometric model to further elucidate the impact of household food garden interventions in a larger sample.