Background

Although there has been a significant improvement in global food security, still 805 million people (one in eight people) in low and middle income countries (LMIC) remain chronically undernourished [1]. According to the key findings of the Global Food Security Index 2015 [2], the rate of under nutrition is considerably higher in low and lower middle income countries (25.4 % and 16.5 % respectively) compared to high income countries (4.9 %). It is also estimated that 29.1 % and 15.5 % of children under the age of five years in lower middle income countries are either stunted or underweight. The prevalence rate is even higher in low income countries where 39.1 % of children under the age of five years are stunted and 22.6 % are underweight [2].

In addition to the health effects of food insecurity leading to poor nutrition, household air pollution from combustion of solid cooking fuels such as firewood, charcoal, etc. is the fourth leading cause of mortality in LMIC [3]. Evidence from epidemiological studies have shown that exposure to household air pollution is associated with acute respiratory tract infection, chronic obstructive pulmonary disease (COPD), cataract and lung cancer [4–6]. Likewise diarrhoea and other common infectious diseases due to poor hygiene and sanitation are also causing significant public health problems in LMIC [3].

It is evident that health is a complex phenomenon determined by multiple risk factors. Complex environmental interactions make it difficult to determine pathways to health in many communities. Food and diet is clearly an important route for exposure to pathogens, but it should not be considered in isolation, since other environmental exposures, such as household air pollution due to burning of biomass for cooking, pesticide exposure from agricultural use and polluted water for drinking, can be equally or more important to health. Food insecurity leading to poor nutrient intake is the main cause of malnutrition, but it is also dependent on other immediate causes, such as the individual’s health status [7]. Previous studies have recognised strong linkages between agricultural interventions and nutritional health [8–10] and the development of clean fuels and improved solid fuel stoves in reducing household air pollution and adverse health effects [11]. However, the scale and effectiveness of combined household interventions to improve health, nutrition and the environment has not been investigated. It is unknown whether interventions are inter-disciplinary, crossing domains of health, nutrition, agriculture and/or environment and where these interventions are being conducted. This review determined the extent and types of community-based complex agricultural and household interventions to improve food security, health status and the household environment in LMIC.

Methods

Search strategy

A comprehensive search strategy was developed following the recommendations in the PRISMA (Preferred Reporting Items for Systematic Reviews and Meta-Analyses) statement [12]. Key search words were generated reflecting the PICOS (participants, interventions, comparators, outcomes and study design) approach [12]. A database search of Ovid EMBASE was performed using Medical subject heading (MeSH) terms, keywords and truncations covering the potential interventions, outcomes of interest and study design (Additional file 1). The search strategy was developed by combining those search terms using appropriate Boolean operators such as AND/OR/NOT. The search strategy for Ovid MEDLINE, PUBMED and SCOPUS databases were then derived from those search terms and conducted in January 2015. In addition, web and hand searches of bibliographies of identified studies were also performed manually to identify any additional potentially eligible articles.

Study selection and inclusion criteria

Community-based agricultural and household interventions such as the introduction of biogas, improved cook stoves, home gardening, animal husbandry, livestock farming and nutrition education were eligible to be included in this study if the focus of the intervention was to improve at least one of the outcome measures of interest (Table 1). Human studies employing any of these interventions, alone or in combination, and published after 1990, were included.

Table 1 Definitions of outcomes of interest measured
Full size table

The review was open to include any interventional or observational study, such as randomised control trial (RCT), cluster-randomised trial (CRT), cross-sectional study (CSS) and longitudinal studies conducted in LMIC as defined by the World Bank list of economics for 2015. As the main focus of this study was to identify community-based household interventions, clinical and occupational studies were excluded from the review. Similarly, review articles and studies from high income countries were excluded from the review.

All articles identified by electronic searching from the four databases were exported to a web-based bibliography and database manager namely, Refworks. The titles were merged in one database and duplicates removed (Fig. 1). The primary reviewer (SG) screened titles and selected potentially relevant abstracts following predefined inclusion/exclusion criteria. Then four further reviewers (DM, SS, JK and JS) independently examined 10 % of randomly selected titles and abstracts to ensure the accuracy of title and abstract screening process. Disagreements between reviewers were resolved through discussion and checking the full text articles. All articles deemed potentially eligible were retrieved in full text. Reference lists of included studies were also checked to identify other relevant studies.

Fig. 1
figure 1

PRISMA flow diagram

Full size image

Data extraction and management

A standard data extraction form (Additional file 2) was designed considering the Cochrane systematic review data collection checklist [13]. The data collection form was piloted and amended prior to starting the formal data extraction.

Data from all included studies were extracted independently by three reviewers. The extracted data from 10 % of randomly selected articles was then checked independently by a second reviewer to ensure all the correct information was recorded.

Data analysis

A narrative analysis was conducted based on interventional categorisation. Interventions were categorised according to four domains defined as follows:

  • Agricultural interventions: Interventions such as home gardening and animal husbandry that have the explicit goal of improving food productivity, nutritional status, health, dietary diversity and/or food security.

  • Air quality interventions: Interventions such as improved cook stove and biogas that have the clear aim of improving household air quality and occupant’s health.

  • Water quality interventions: Interventions such as water filters (sand and bio sand), solar disinfection technique, water treatment using chlorine tablets alone and/or combination with sanitation health and hygiene education that have the clear aim of improving drinking water quality and health.

  • Nutritional interventions: Interventions such as nutrition education, complementary food and nutritional supplements that have the clear aim of improving participants’ nutritional status, dietary diversity, and health and food security.

The studies from each interventional category were summarised in tables and narrative text provided to summarise the following aspects:

  • country where the study was conducted

  • sample size

  • setting

  • study designs followed

  • types of interventions provided

  • intervention duration

  • outcomes of interest measured

Assessment of methodological quality

An assessment of the validity of included studies was conducted alongside the data extraction using the Effective Public Health Practice Project (EPHPP) quality assessment tool for quantitative studies [14]. Studies were categorised as strong, moderate or weak based on their quality with regards to component ratings of selection bias, study design, confounders, blinding, data collection method, withdrawals and drop-outs and analysis.

Results

Identified studies

The search retrieved 10,847 unique articles (Fig. 1). After removal of 1,638 duplicates the remaining 9,209 articles were screened on the basis of title review. The first stage selection excluded 9,072 articles on the basis of predefined exclusion criteria. Studies were mainly excluded as they were conducted in high income countries, clinical or occupational settings, were not interventional studies or review articles, etc. From these 137 articles were potentially eligible for abstract screening. Finally, 112 articles met the eligibility criteria for the detailed analysis. Of the 25 articles excluded at the abstract screening stage four of them were from high income countries, five were in a clinical setting (Cl), five involved occupational settings, four were review articles, six papers were not interventional studies, and the full text of one paper was not available. Eleven additional articles were identified by hand/web searching. Finally, a total of 123 studies were included for the final review.

Study characteristics

Of the 123 included studies in the review, 27 (21.9 %) were agricultural interventions, 34 (27.6 %) were air quality interventions, 32 (26 %) were water quality interventions and 30 (24.3 %) were nutritional interventions (Fig. 2).

Fig. 2
figure 2

Overlapping intervention domains

Full size image

Characteristics of agricultural interventions (n = 27)

Of the 27 studies (Table 2) reporting agricultural interventions, 14 projects promoted and supported home gardening and household food production or the improvement of the existing garden with micronutrient-rich fruit and vegetables. Six projects promoted animal husbandry, such as pig and poultry breeding, goat farming, fisheries and dairy production. Five studies observed the effectiveness of combined home gardening and nutrition education intervention. One promoted home gardening with animal husbandry and another, a combination of home gardening, animal husbandry and nutrition education.

Table 2 Characteristics of agricultural intervention studies
Full size table

Most of the studies were either cross sectional (n = 10) or intervention studies (n = 10) with one RCT [15]. There was a wide variation of sample sizes, ranging from 58 households [16] to >10,000 participants [15]. Similarly, duration of the studies varied; from a dairy intensifying intervention in Kenya for two months [17] to a home gardening study in India for 96 months [18]. Fourteen of these studies were conducted in Asia and the other 13 in Africa. The first home gardening study was conducted in Bangladesh in 1996 [19]. Most of these studies (n = 22) were conducted in a household setting and only a few in community settings.

Nineteen of these studies examined the effect of intervention on dietary diversity and improvement in food consumption, seven on food production, seven on nutrient intake, seven on nutritional deficiencies, seven on anthropometry, three on education, two on health and two on food security.

Characteristics of air quality interventions (n = 34)

Of the 34 air quality studies (Table 3), four projects introduced biogas [13–20] as an alternative means of cooking fuel, 17 projects promoted improved cook stoves and 11 studies examined the effectiveness of improved stoves with chimney to improve the household air quality. One project evaluated the impact of improved cook stoves with solar water disinfection and hand hygiene [21], and another looked at an improved cook stove intervention with biogas fuel and solar heaters [20].

Table 3 Characteristics of air quality intervention studies
Full size table

Most of the studies provided data either on pre and post or between group comparisons with nine randomised control trial. The sample sizes of the studies ranged from 11 [22] to 4,000 households [23]. The duration of the study also varied considerably; a Peru cook stove project lasted for 3 weeks [24], while one vented stove project in the highlands of Guatemala collected data for 48 months [23]. The majority of the studies (n = 18) were conducted in South America, nine were in Asia, with the other seven in African countries. The first cook-stove intervention study was conducted in Nepal in 1990 [25]. All of these studies were conducted in household settings.

Almost all of the studies (28 out of 34) examined the improvement in household air quality parameters such as particulate matter and carbon monoxide concentrations. Twenty studies assessed the impact of the intervention on participants’ health outcomes such as incidence of pneumonia, acute respiratory infections (ARI), conjunctivitis and lung function, and three examined the impact on food production.

Characteristics of water quality interventions (n = 32)

Of the 32 water quality intervention studies (Table 4), 12 were water filter interventions; nine were chlorine tablets/solutions interventions, seven were Solar disinfection; two were hand water pumps along with hygiene education and latrine construction interventions [26]; one was a health, hand hygiene, water quality and sanitation educational intervention [27]; one involved disinfection tablets along with sanitation and hygiene education [28]; one was a water disinfection stove [29] and one a filter along with improved cook stove [30].

Table 4 Characteristics of water quality intervention studies
Full size table

Most of the studies were RCT (n = 25) or intervention studies (n = 4). The sample sizes of the studies ranged from 2 [29] to 2,193 households [31] and the interventions were delivered over periods of 2 [29] to 15 [32] months. Nine studies were conducted in South America, 10 in Asia and the remaining 13 in African countries. All of these studies were conducted in household settings.

Twenty-seven of these studies looked at the impact of intervention on health especially on the incidence/prevalence of diarrhoeal diseases; 20 on microbial contaminations and water quality; two studies examined the level of knowledge and self-compliance, two investigated air quality and one hygiene and sanitation.

Characteristics of nutrition Interventions (n = 30)

Of the 30 nutrition intervention studies included in the review (Table 5), 11 studies were supplementary food and vitamin interventions, 13 nutrition education interventions, five nutrition education together with complementary food interventions, two combined interventions of nutrition education and home gardening [33, 34] and one combined package intervention of health care, nutrition education, water and sanitation [35].

Table 5 Characteristics of nutrition intervention studies
Full size table

Most of the studies (n = 18) were intervention studies (pre and post or two group comparison), ten RCT, one randomised crossover study and one crossover trial. The sample sizes of the studies ranged from 42 [36] to 40,000 [37] participants. The duration of the study also varied; from a once-off nutrition counselling training [38] to a 48 months nutrition education intervention in Nicaragua [39]. Just over half of the studies (n = 16) were conducted in Asia, nine in Africa and the other six in South American countries. Majority of these studies (n = 17) were conducted in a household settings with some in community settings.

Eighteen of the nutrition intervention studies assessed the impact of intervention on nutritional status such as growth, prevalence of stunting (low height-for-age), underweight (low weight-for-age), and wasting (low weight-for-height), 10 studies assessed food consumption and dietary diversity, nine studies assessed the impact on nutrient deficiencies, eight studies looked at health status, six at nutrient intake, five at health and nutritional knowledge, two at feeding practice and one assessed food security.

Methodology quality

Of the 123 included studies, eight studies failed to provide sufficient detail to assess their methodological quality. Information of study selection, withdrawals, blinding and confounders were particularly under-reported in the majority of studies. Because of the nature of the intervention, it was assumed that no blinding was imposed in some studies and they were therefore categorised into moderate quality study. The most common methodological problems among the weak studies were in selection bias, confounders, reliability and validity of data collection tools and blinding.

Discussion

According to our knowledge, this systematic review is the first to explore the cross-domain overlapping of multidisciplinary research projects in agriculture, nutrition, air quality and water quality. It is obvious that there is a lot of work being done in this area but from this review it clear that there is variation in not only the type of intervention, study type, sample size, duration and setting, but also in the outcome measured.

Although a wide variety of agricultural interventions such as home gardening and animal husbandry were conducted to improve household food productivity and food consumption, this review also confirms the findings of previous reviews that only few studies were measuring the impact of those interventions on nutritional status [8–10]. Of those projects that did look at the impact of agricultural intervention on nutrition, seven examined the impact on nutrient intake, nutrient deficiencies and anthropometry. In general it is predictable that increased production and consumption of food leads to better nutrition, but due to variation in study design, duration and outcome of interest measured among the included studies, it doesn’t look likely to obtain pooled estimate for studies which look at impact of intervention on nutritional health.

While looking at the air quality interventions, it is evident that interventions to improve cook stoves are the most popular interventions (83 %) and are widely being used in all over the world. This may provide the enough roofs to perform the meta-analysis. Some biogas interventions (n = 4) [20, 40–42] have been conducted to measure the multiple benefits of intervention on indoor air quality and food production (using bio-slurry). However, as they refer to different outcome measures and are measured in different ways, the available evidence does not look strong enough to perform the comprehensive analysis.

It was identified that water purification filter interventions were the most popular (n = 12) interventions for treatment of drinking water quality in LMIC. Other interventions such as chlorine tablets or solution (n = 9) and solar disinfection (n = 7) are also common in this region. Randomised controlled trial study design was the most popular among the water quality intervention as the vast majority (78 %) of the research project applied this method. So, it is more likely that effects on the drinking water quality can be summarised across studies.

Nutrition education (n = 13) and supplementary food and vitamin (n = 11) interventions were the most popular nutritional intervention in LMIC. Some intra-domain combined interventions of nutrition education and supplementary foods (n = 5) have also been piloted in some low and middle income countries to determine the impact of intervention on dietary diversity and nutrient intake.

The main finding of this review is that the vast majority (91 %) of the academic research on agricultural, nutrition and the environmental studies are simple and discipline specific with substantially fewer (n = 11) combined interventions across domains and the result is consistent with previous domain specific reviews [7, 43]. Only six studies looked at the combined impact of agricultural and nutrition education interventions, three on air and water quality interventions, one study examined the impact of a combination of agricultural and air quality interventions and one was a combined water quality and nutritional intervention. Although poor nutrition and household air pollution are the leading cause of mortality in LMIC [3], this review did not find any studies examining the impact of a combination of air quality and nutritional interventions on health. It is also striking that none of these studies investigating the combined impact of agricultural and drinking water quality interventions on human health. The evidence reviewed here shows that silo mentality is still inherent in academic research.

Another interesting finding of this review is that certain LMIC regions seem to focus on domain-specific interventions, with most studies in Kenya and India and only a small number in other countries (Fig. 3). Asian and African countries were the most common regional target for agricultural and nutritional studies. More than half of the agricultural (52 %) and nutritional (53 %) interventions were conducted in Asian countries with the majority of them in south Asian countries. Similarly, 48 % of agricultural and 30 % of nutritional studies were conducted in Africa with the majority of them focussed in sub-Saharan African countries such as Kenya, Ethiopia and South Africa. The majority of water quality interventions were conducted in Africa (40.6 %) followed by Asia (31.3 %) and Latin America (28.1 %). However, the majority (53 %) of interventions to improve household air quality were conducted in Latin American countries particularly in Guatemala, Peru and Mexico. This restricts the generalisability of the findings to other LMIC.

Fig. 3
figure 3

Global map highlighting the regional focus of included studies

Full size image

Strengths and limitations of the study

The main strength of this review is the application of a comprehensive search strategy through four databases to capture all potentially relevant peer reviewed articles. One hundred and twenty three articles representing the four different intervention domains provide ample evidence to understand the current research gap in interdisciplinary research. The use of independent reviewers throughout the review process further strengthened the methodological quality.

The main limitation of this study is that as only peer reviewed journal articles were included in this review, there is a chance of missing those studies published in developmental organisations’ reports and bulletins (publication bias). Additionally, this review focused on household and community-based studies, so there is a chance of missing some useful studies conducted in clinical settings.

Conclusion

In conclusion, it is evident that very little interdisciplinary research has been conducted with the majority of studies on agriculture, nutrition and the environment being discipline specific. It also seems that certain LMIC regions seem to focus on domain-specific interventions. Although a wide variety of study designs have been implemented to measure the impact of agricultural, nutrition and air quality interventions on respective outcomes of interest measured, there is still not sufficient evidence which utilises robust randomised or quasi-experimental study design.

Therefore, this review emphasizes that future research needs to focus on multi-disciplinary complex interventions with standardised outcome measures. Also, rigorous research across disciplines and sharing expertise across regions is a necessity. The next phase of this review (Meta-analysis) will identify whether eliminating silos of discipline specific research can bring a significant improvement or not.