Background

No specific food group provides all the essential nutrients, and therefore, healthy dietary practices require individuals to consume food from all food groups [1]. Available studies show that dietary diversity is a good alternative measure of nutritional adequacy and food security in the household [2,3,4,5].

Nutrition or nutritional adequacy is defined as the adequate consumption of different food groups to maintain a balanced diet and meet daily nutrient needs [6]. Food security exists when all people at all times are able to access sufficient, safe and nutritious food to meet their dietary needs and food preferences, thus ensuring an active and healthy life [7]. Hence, food security is realised when every person has stable and continuous access to different foods that are of good quality and are safe, affordable and in sufficient amounts [7, 8].

Determination of household dietary diversity score (HDDS) refers to the process whereby food groups that a household has consumed are counted over a specific period of time, usually over a 24-h period [9]. The HDDS is a useful indicator of the household’s economic ability to access a variety of foods [2, 10]. The score is calculated from the dietary diversity questionnaire [10]. The HDDS can be used to assess access to food and as a result, it is commonly applied qualitatively to determine food consumption by households [11].

Low dietary diversity is most prevalent among households in developing countries [11, 12]. Although South Africa is food secure at national level [13], a large proportion of households in the country experience food insecurity [14]. Food insecurity is defined as a state whereby people and individuals lack the necessary physical and economic access to adequate, safe and nutritious food to sustain a healthy and active life [7]. Household food insecurity is the result of the application of this concept at household level [15].

Available evidence suggests that food insecurity inversely correlates with HDDS (i.e. the number of different types of foods consumed by the household) [16]. For example, households in four rural districts of Prey Veng Province, Cambodia that had high levels of food insecurity (82%), had low HDDSs (< 3) [17].

In a study carried out in the Eastern Cape province of South Africa, it was observed that the majority (61%) of the households had low dietary diversity, and that their diets consisted mainly of pulses, milk and cereals [18]. This implies that the households in the study conducted in the Eastern Cape province, consumed an average of three food groups. This is lower than the minimum dietary diversity score of six (i.e. six food groups) suggested by Cheteni et al. [18] to be a high dietary diversity status.

Available evidence suggests a relationship between poor dietary diversity and increasing under-nutrition [19]. This is because low dietary diversity scores are associated with low nutrients in the diet [20], domination of starchy foods, and lack of fruits, vegetables and animal products [11, 12]. High dietary diversity score is correlated with nutrient adequacy. Nutrient adequacy is essential for body maintenance, growth, strength, physical work, cognitive ability, immunity and good health in human beings [6]. Poor dietary diversity has proved to be associated with a high risk of chronic diseases, such as cardiovascular disease, diabetes and obesity in addition to depression and anxiety [21]. A poor diet is one of the leading causative factors of illness and mortality around the world [22].

According to previous studies in South Africa, low dietary diversity is mainly prevalent among rural households and informal urban settlements [1, 14, 15, 20]. This is attributed to the fact that the majority of households in such areas depend on government social grants and as a result are not able to afford diverse foods [23].

Participation in agricultural activities provides households with the opportunity to diversify their diets [23]. However, most rural households lack the resources to produce their own food [24]. In view of this, rural communities are not able to produce their own food [20, 21] and hence, the Mpumalanga Provincial Government introduced the Phezukomkhono Mlimi (PKM) programme in 2005 with the aim of improving availability and accessibility of food among rural households in the province [25]. The PKM programme supports rural households with production inputs (i.e. seeds, fertilisers and chemicals; mechanisation during land preparation; support with basic infrastructure for farming, such as fencing, boreholes and irrigation pipes, and extension and advisory assistance) to enable households to produce their own food. Despite the massive funds that have been invested in this programme since its inception in 2005, in several municipalities of the Mpumalanga Province, no assessment of food accessibility among the beneficiaries has been conducted. Currently little empirical evidence is available on how such programmes affect food access. A previous study that was conducted around this topic did not make use of HDDS or any of the published standardised methods [26]. Therefore, answers are needed for following questions: (i) what is the household dietary diversity score of these beneficiaries? and (ii) what factors are associated with the dietary diversity in these households?

Therefore, this paper investigates the HDDS and identifies the factors associated with dietary diversity among the agricultural households that are supported by the PKM programme in the Nkomazi Local Municipality (NKLM), South Africa.

Study area

This study was carried out in NKLM, which is situated in the Mpumalanga province of South Africa. The municipality shares its borders with Mozambique and the Kingdom of Eswatini (formerly known as Swaziland) (Fig. 1). The population in the study area is approximately 410,900 people [27]. It has a subtropical climate, with a mean annual rainfall of 755 mm and a mean temperature of 28 °C [28]. The municipality is predominantly rural, and most people depend on agriculture for a living [29]. The main agricultural activities in the study area include growing vegetables, sugar cane, bananas and citrus, with subtropical fruit farming under irrigation and maize and cotton cultivation on dry land [30]. The study area is characterised by high levels of poverty and unemployment [31].

Fig. 1
figure 1

Map of the study area showing the location of Nkomazi Local Municipality (coloured red) [29]

Full size image

Methods and materials

Study population

The target population in this study was all agricultural households in the NKLM that benefitted from the PKM programme in the 2018/19 production season (i.e. April 2018 to March 2019). According to the Department of Agriculture, Rural Development, Land and Environmental Affairs (DARDLEA), the number of agricultural households supported by the PKM programme in the study area during the 2018/19 production season was 543.

However, only household heads who were involved in food preparation or were present and ate the food prepared in the household during the reference period were included in this study. Therefore, out of the 543 households in the study area, only 355 agricultural households supported by the PKM programme during the 2018/19 production season met the inclusion criteria were thus included in this study.

Data collection

A pre-tested structured questionnaire was administered by four trained enumerators to collect data on the socio-economic characteristics and dietary diversity of the respondents through face-to-face interviews. The four enumerators were selected based on their educational qualifications in the field of agriculture. Enumerators had to have a National Diploma in Agriculture (Crop production) as the minimum qualification to be enlisted as an enumerator. In fact, one of the enumerators had graduated with a Bachelor of Agricultural Management. The enumerators were trained on how to execute their task competently and to be able to explain the nature and scope of the research. The numerators were also trained on ethical issues that needed to be observed during the data collection phase.

Each interview session lasted approximately 45 min and the data were collected from 01 February to 24 March 2020. The data collection exercise was carried out in the different villages of the NKLM. The participants were invited to a central venue in their respective villages to participate. Although the data collection was conducted mainly by the hired numerators, where the number of respondents eligible to participate in the study was high, the principal researcher also participated in the data collection. This was done to speed up the process of data collection and avoid keeping the participants waiting for a long time to be interviewed.

As suggested by Swindale and Bilinsky [10], the Household Dietary Diversity (HDD) was assessed using a 24-h recall of food groups eaten by members of the household as either a shared meal in the home or as food prepared at home to be eaten by household members outside the home.

The HDDS in the present study was based on 12 food groups, using the guidelines for measuring HDD developed by Swindale and Bilinsky [10]. The 12 food groups included cereals, tubers, vegetables, fruits, meat, eggs, fish, beans, dairy products, fats/oils, sugar/honey and condiments. One point was allocated if the food group was consumed, whereas no points were allocated if the food group was not consumed over the reference period. The sum of all the points for each household was then computed.

Data management and data analysis

Data management

The completed questionnaires were thoroughly checked by the principal researcher for any omissions at the end of each data collection session. Data were thereafter captured in Microsoft Excel by the enumerators and later imported into the Statistical Package for the Social Sciences (SPSS version 25). Thereafter, the principal researcher checked the data for missing values, duplicate observations and inconsistencies across all the variables before it was analysed.

The HDDS was measured on an ordinal scale (0 = low dietary diversity; 1 = medium dietary diversity; 2 = high dietary diversity). The marital status variable was reclassified as described by Subramanian et al. [32] into a dichotomous variable (0 = otherwise; 1 = married) by collapsing the original categories (single, divorced or widowed) into otherwise (coded 0), while the original category (married) remained (coded 1). The income of the household was recategorised into a binary variable (1 = ≤ R3000, while 2 = > R3000).

Data analysis

Descriptive statistics, computation of the HDDS and the ordered logit regression model were employed to analyse the data. The ordered logit regression model was adopted for this study because the data meet the following assumptions: the outcome or dependent variable (HDD) was measured on an ordinal level, and one or more of the independent variables were either continuous or categorical. In addition, there was no multicollinearity between the independent variables.

All categorical variables were summarised and presented as percentages in the form of tables. The HDDS was computed by adding the points allocated to the different food groups consumed by the households over the 24-h recall period. Thereafter, households were categorised based on their HDDS using the method described by Swindale and Bilinsky [10] into the following categories: low dietary diversity category (≤ 3 food groups); medium diversity category (4–5 food groups) and high diversity category (≥ 6 food groups). Adoption of this categorisation of the HDDS in this study is further justified by the fact that several studies conducted in South Africa [9, 33] and other parts of Africa [34, 35] have also adopted the same categorisation of the HDDS.

The ordered logistic regression model was fitted to the data to identify the factors that were significantly associated with the dietary diversity of the households.

The equation of the ordered logit model regression is specified as [36]

$$Y* = \sum\nolimits_{k = 1}^{k} {\beta_{k} X_{k} + \varepsilon } ,$$
(1)

where Y* is unobserved, Xk is a vector of independent variables, β denotes coefficients to be estimated and ε denotes a random error term.

From the above model, the observed or defined categorical variable Yi is determined as follows:

y = 1 if y* ≤ μ1.

y = 2 if μ1 < y* ≤ μ2.

y = 3 if μ2 < y* ≤ μ3.

y = j if μj−1 < y*.

In case y is observed in j number of ordered categories, μs are unknown threshold parameters differentiating the adjacent categories to be assessed with βs; then μ1, μ2 and μ3 represent the different levels of the HDDS that are to be estimated. The general form for the probability that the observed y falls into category j and the μs and the βs are to be estimated with an ordered logit model is

$${\text{Prob}}\left( {y = j} \right)1 - L\left( {\mu_{j - 1} - \sum\nolimits_{k = 1}^{k} {\beta_{k} X_{k} } } \right),$$
(2)

where L (·) represents cumulative logistic distribution.

Additional file 1: Table S1 presents all the variables included in the ordered logit regression model. Furthermore, the expected relationship between the outcome (HDDS) and each predictor variable is also indicated in Additional file 1: Table S1.

The model building process was carried out in two phases. The first phase involved univariate analysis to identify the variables that were significantly associated with the outcome. All variables with a p value of ≤ 0.20 in the univariate model were included in the multivariable ordinal logit model. A cut-off value of 0.20 is supported by literature [37, 38].

In the second phase, a multivariable ordinal logit regression model was fitted using the manual backward selection method. Confounding was tested in the model by checking for changes in the model coefficients and the model fit when a specific variable was removed from the model. A particular variable was considered a confounder if its removal or addition resulted in a change in the coefficient of the other variables that was greater than 10% [39]. All identified confounders were retained in the model.

Multicollinearity in the final model was tested by computation of the variance inflation factor (VIF) at a cut-off value of 3 [40]. The VIFs of all independent variables were less than 3, which confirmed that the model did not suffer from multicollinearity. The test for parallel lines was conducted, and the assumption of proportional odds was satisfied at p = 0.10 [41]. Several procedures of model suitability tests were carried out to check the model fit to the data. The likelihood ratio test that is based on the − 2 Log Likelihood was used to assess the goodness of fit of the model. The results of the likelihood ratio Chi-square test [x2(9) = 35.13; p = 0.00] proved that there was a significant improvement in the fit of the model with predictor variables over the model without predictor variables [42]. Furthermore, the deviance and Pearson’s Chi-square test were used to check the goodness of fit of the model. The results of the Pearson’s Chi-square test [x2(523) = 518.56; p = 0.55] and the deviance test [x2(523) = 592.73; p = 0.83] were not significant, which suggests that the model fit the data well [42].

Results

Socio-economic characteristics of respondents

The socio-economic details of the respondents are summarised in Table 1. The respondents were household heads supported by the PKM programme during the 2018/19 production season. A total of 355 households agreed to participate in this study and of this, more than half (56.10%; n = 199) were over 60 years old. Females made up 59.44% (n = 211) of the study population.

Table 1 The socio-economic profile of respondents (n = 355)
Full size table

Half (49.86%; n = 177) of the respondents were married and 5.63% (n = 20) were divorced. The largest proportion (43.66%; n = 155) of the respondents had attained primary education, followed by 41.97% (n = 149) who were uneducated. Very few respondents (4.51%; n = 16) had tertiary education.

Most respondents (52.39%; n = 186) were members of a large household that consisted of between six and ten members. The majority (60.28%; n = 214) of households farmed on less than three hectares of land and only 3.38% (n = 12) of the households had more than 10 hectares of land.

Food groups consumed by the respondents

The results of the present study show that all the households (100%; n = 355) consumed cereals in the 24-h recall period (Table 2). Over two-thirds (69.01%; n = 245) of the households in the present study reported having consumed condiments, beverages, such as tea and coffee, and soft drinks, such as cool drink and juices. This was followed by 42.25% (n = 150) who indicated that they had consumed sugar and honey. A small proportion of respondents indicated that they had consumed food from the following groups: beans (16. 62%; n = 59); fruits (10.14%; n = 36); fats and oils (10.14%; n = 36); dairy products (8.17%; n = 29); tubers (5.35%; n = 19); fish (3.66%; n = 13) and eggs (3.10%; n = 11).

Table 2 The distribution of households by food groups consumed in the preceding 24 h (n = 355)
Full size table

Dietary diversity status of the households

Table 3 shows the distribution of households by HDD. Forty percent (40.28%; n = 143) of the households had a low dietary diversity level (i.e. consumed 1–3 food groups), half of the households (49.86%; n = 177) had a medium dietary diversity score (i.e. consumed 4–6 food groups) and only 9.86% (n = 35) of households had a high dietary diversity (i.e. consumed 7–8 food groups).

Table 3 Distribution of households by level of household dietary diversity (n = 355)
Full size table

Overall, the dietary diversity scores of the households ranged between one and eight food groups with a mean of 3.90 (SD = 1.34) (Table 3). These results show that on average, households in the study area consume on average four food groups. Based on a mean HDDS of 3.90, it implies that the households in the study area have a medium dietary diversity level.

Factors associated with the dietary diversity of the households

The results of multivariate ordered logit model (Table 4) revealed that households headed by respondents with no formal education had lower odds (OR = 0.20; 95% CI 0.06–0.61) of having a higher dietary diversity compared to households headed by heads who had tertiary education. Similarly, households with income ≤ R3000.00 had lower odds (OR = 0.51; 95% CI 0.31–0.85) of having a higher dietary diversity as compared to households that had income > R3000.00. On the other hand, households with 1–5 members had higher odds (OR = 10.41; 95% CI 1.05–103.20) of having a higher dietary diversity as compared to larger households (i.e. had 16–20 members). With every unit increase in the age (expressed in years) of the household head, the odds of the household of having a higher dietary diversity increased by 1.03 (OR = 1.03; 95% CI 1.01–1.05).

Table 4 Results of the Ordered logit regression model showing factors that were correlated with dietary diversity among the households in the Phezukomkhono Mlimi programme (n = 355)
Full size table

Discussion

The results show that all the households consumed cereals in the 24-h recall period. The dominance of cereals in the diets of the respondents in the present study was anticipated given that cereals, such as maize, are a common staple food throughout South Africa [43] and more so among emerging farmers [44]. Similar results were observed by Jebessa et al. [45] who noted that the diet of the households in Yayu Biosphere Reserve, Ethiopia consisted mainly of cereals.

While previous studies carried out in Tanzania and South Africa observed that only a few participants had consumed vegetables [9, 43], over three-quarters of respondents in the present study indicated that they had consumed vegetables over the recall period. The results reported in the present study suggest that participation in the PKM programme by households encourages food production throughout the year (from April to March of the following year). This could explain the differences observed in this study and that conducted in Tanzania. It is possible that participation in the PKM programme led to increased availability and accessibility of vegetables in the study area, with the resultant high proportion of respondents in the study reporting having consumed vegetables. This view is supported by studies that observed that home food production leads to improved access to vegetables and subsequent increased dietary diversity [46, 47]. Since vegetables are a good source of vitamins and minerals [48], an increased intake of vegetables has the potential to translate into improved diet quality. This is confirmed by Ochieng et al. [49] who are of the view that vegetable production is essential in increasing dietary diversity.

Over two-thirds of the households in the present study reported having consumed condiments, beverages, such as tea and coffee, and soft drinks, such as cool drink and juices. These findings are consistent with those reported in a study by Udoh and Udoh [50] on the dietary diversity of households in Akwa Ibom State, Nigeria that also observed that a high number (80.7%) of households consumed condiments.

However, with respect to consumption of honey and sugars, the results of the present study contrast with the findings of the same study by Udoh and Udoh [50] that indicated that the number (82.7%) of households that consumed honey and sugar was double the number (42.25%) observed in this study. The low level of consumption of condiments and sugar and honey observed in the present study could be attributed to the fact that the majority of farmers in this study were unemployed and had low farm income. This has the potential to limit the ability of the respondents to afford food groups, such as condiments, honey and sugar [23].

Consistent with the findings by Mekuria et al. [37] and Sinyolo et al. [51], food from groups such beans, fruits, fats and oils, dairy products, tubers, fish and eggs were consumed by very few respondents in this study. This could be due to the low education level of the respondents. It has been reported that households headed by persons with no formal education lack understanding of nutrition [43] and the health benefits of a well-balanced diet. As a result, they are less likely to diversify their diet as compared to farmers with formal education [9, 45]. Contrary to the findings of this study, Udoh and Udoh [50] reported that more than 80% of the households in their study consumed food from these groups. The study by Udoh and Udoh [50] has a higher proportion (91.25%) of married respondents as compared to this study (49.86%). Other authors have shown that access to different food groups increases when the head of the household is married [52]. The findings of this study suggest that the diet of the majority of households in the study area was deficient in proteins that could be sourced from fish, eggs and beans [53]. The low consumption of proteins by respondents in the study area was concerning, especially considering that beans and other legumes are excellent sources of protein that can easily be grown [54]. This further suggests that beneficiaries and programme planners require assistance to be able to select appropriate crop varieties with potential to make a significant contribution towards food security. Previous studies have shown that collaboration between different stakeholders is important for knowledge sharing and the subsequent design of innovative approaches [55]. This view is supported by Falcone [56] who noted that stakeholder collaboration is critical in resolving differing and sometimes conflicting views. Because of this stakeholder collaboration has the potential to create an enabling environment to design well-suited policy interventions. In addition, collaboration between different stakeholders has also been proven to be beneficial in formulating effective interventions especially for farmers [57].

The low consumption of fats and oils in this study is worth noting, and this practice should be encouraged to reduce the risk of obesity and diseases of the lifestyle [58]. This finding was, however, unexpected since previous research on dietary diversity showed dominance of fats and oils in the diets of most households in South Africa [18, 35, 43].

The findings of the present study revealed that most of the households had a medium dietary diversity score which contrasts with the findings by Megbowon and Mushunje [43] in the study conducted in the Eastern Cape province of South Africa in which they reported a higher proportion (61.75%) of respondents with high dietary diversity. The HDD observed in this study could be attributed to a high proportion (42.97%; n = 149) of respondents in the present study not having formal education. Past studies have observed that education has a positive influence on dietary diversity, with higher education levels associated with higher HDD [59, 60]. Secondly, the HHD observed in the present study could be attributed to the predominantly large households in the study area. Previous research has observed that household size and dietary diversity are negatively correlated [61]. In addition to this, households in this study could be constrained to improve their HDDS due to the small farming plots. It has been demonstrated that households farming on larger plots tend to have a higher HDDS [49]. In view of this, a land reform programme aimed at increasing the farm size could be instrumental in improving the HDDS in the study area. For example, available evidence indicates that the HDDS of farmers who benefitted from the land reform programme in the Waterberg District Municipality, South Africa is higher than the HHDS of those who did not benefit from the land reform programme [62]. Furthermore, due to the increase in the number of people who were able to access land through the land reform programme, the number of households in the area with adequate food quantity increased together with their HDDS [63]. It is also possible that the medium dietary diversity score observed in the study could be attributed to post-harvest losses. In Sub-Saharan Africa, losses due to lack of appropriate post-harvest techniques among small-scale farming is estimated to be around 40–50% [64]. This undermines the food production efforts by leading to food losses and subsequent food insecurity in this sector [64, 65]. Literature suggests that introduction of appropriate low-cost post-harvest techniques among small-scale farmers could contribute to economic accessibility of fresh produce, food security and graduation of subsistence farmers to commercial farming [64].

The medium HDDS observed in this study is consistent with the findings reported by Shisana et al. [66] who observed that generally households in the Mpumalanga province of South Africa (where the present study was also conducted) tend to have a medium HDDS. However, the results reported in this study suggest that the HDDS in the study area was better than that reported in the Eastern Cape province of South Africa by Cheteni et al. [18], who observed a low dietary diversity among households (with a mean HDDS of 3). Therefore, the difference between these two studies could be attributed to the fact that some households in the study by Cheteni et al. [18] were not beneficiaries of government supported food security programmes. It has been shown in literature that government programmes do assist in improving the food security status of vulnerable households [67].

Findings reported here are consistent with what other authors have observed. For example, some authors have previously reported that the age of the household head is a significant predictor of HDD [43]. It has been suggested that as the head of household becomes older, he or she is likely to become more aware of their diets due to old-age-related health challenges that older people tend to experience [51]. In contrast with the findings of this study, other studies [59, 60] have showed that age of household head is negatively correlated with HDD. This view is also supported by Oduniyi and Tekana [68] who noted that the adoption of improved farming technology tends to very low among older heads of household. Therefore, this has a negative effect on food production and, as a result, little income to spend on food items.

It was observed that households headed by household heads with no formal education were less likely to have higher HDD. This is supported by previous studies [35, 43] that demonstrated a positive correlation between the level of education and a higher HDD. This is because the more educated the head of the household is, the better their understanding of nutrition [43] and the health benefits of a well-balanced diet. As a result, they are more likely to diversify their diet as compared to farmers with less formal education or farmers who totally lack formal education [9, 45]. Moreover, it has been shown that the higher the formal education status attained by members of a household, the higher the likelihood of such members gaining employment [61]. Being employed has been shown to positively influence the HDD [69].

Consistent with previous studies [59, 70, 71], this study also observed that household income was positively and significantly associated with HDD. This was expected because according to Jebessa et al. [45] households with resources and of a higher income status can afford to diversify their diets. Several other authors also hold a similar view that the economic status of a household positively influences dietary diversity and food security [35, 72, 73].

The household size is among various socio-demographic factors that showed association with HDD in this study. The odds of belonging to a higher dietary diversity was 10.40 times higher for households with 1–5 members than that for larger households (16–20 members). This is consistent with previous studies [61] that observed that household size was negatively and significantly correlated with HDD.

Limitations of the study

Seasonal variation influences food availability, and this could lead to variation of HDDS of the households, with the pre-harvest season having the lowest HDD. In view of this, since this study was conducted during the pre-harvest season, a period when households normally face food shortages, the results reported here should be interpreted with caution given that they do not cover both pre- and post-harvest seasons. Furthermore, it is possible that there was under-reporting because the data were collected at one recall period. Nonetheless, this study presents baseline information on HDD of agricultural households participating in the PKM programme, which has previously not been available in the study area. It also provides baseline information upon which future research can be developed.

Conclusion and recommendations

The diets of the residents of the study area are characterised by very low protein intake (i.e. are unbalanced) and low dietary diversity. In view of this, programmes aimed at creating awareness about cheaper sources of protein are recommended. Awareness about these sources of protein could trigger interest in growing such crops and their subsequent inclusion in the diets of the beneficiaries. In addition, studies that assess the barriers to utilisation of these cheap sources of proteins are needed to help improve understanding of of the low utilisation in this study. Since the PKM programme offers only assistance in form of production inputs, mechanisation and infrastructure, it is recommended that the programme should include education of the beneficiaries about the benefits of consuming a diet with a higher HDD as one of its objectives. This will foster adoption of diets with a higher HDD. Furthermore, to enhance inclusion of the new foods in the diets of the recipients, the programme should consider introducing training of the beneficiaries on how to prepare the new foods types that the programme introduces in the study area. Policy measures aimed at improving the dietary diversity status of the households in the study area should prioritise the following: (i) strategies that enhance farm income and level of education, and (ii) agricultural training and nutrition education awareness programmes that provide knowledge on the importance of food and nutrition, as well as practical guidance on how to grow and prepare nutritious foods. Furthermore, collaborations between different stakeholders, such as nutritionists, extension workers and researchers, should be encouraged in order to develop a holistic approach to address HDD. Integrating legumes in the existing gardens (e.g. chickpeas, beans and lentils) as part of the programme could help address the low intake of protein identified in the study area. Secondly, nutrition education which includes meal plans, and meal preparation of the suggested protein sources could be initiated by nutritionists for this community to help improve on protein intake among the respondents. Lastly, post-harvest losses could be investigated in future studies so that the introduction of low-cost harvest techniques can be explored.