STUDY PROTOCOL Open Access

Improving child nutrition and developmentthrough community-based childcarecentres in Malawi – The NEEP-IE study:study protocol for a randomised controlledtrialAulo Gelli1* , Amy Margolies2, Marco Santacroce1, Katie Sproule1, Sophie Theis1, Natalie Roschnik3, Aisha Twalibu3,George Chidalengwa3, Amrik Cooper4, Tyler Moorhead4, Melissa Gladstone5, Patricia Kariger6

and Mangani Kutundu7

Abstract

Background: The Nutrition Embedded Evaluation Programme Impact Evaluation (NEEP-IE) study is a clusterrandomised controlled trial designed to evaluate the impact of a childcare centre-based integrated nutritionaland agricultural intervention on the diets, nutrition and development of young children in Malawi. The interventionincludes activities to improve nutritious food production and training/behaviour-change communication to improve foodintake, care and hygiene practices. This paper presents the rationale and study design for this randomised control trial.

Methods: Sixty community-based childcare centres (CBCCs) in rural communities around Zomba district, Malawi, wererandomised to either (1) a control group where children were attending CBCCs supported by Save the Children’s EarlyChildhood Health and Development (ECD) programme, or (2) an intervention group where nutritional and agriculturalsupport activities were provided alongside the routine provision of the Save the Children’s ECD programme. Primaryoutcomes at child level include dietary intake (measured through 24-h recall), whilst secondary outcomes include childdevelopment (Malawi Development Assessment Tool (MDAT)) and nutritional status (anthropometric measurements). Athousehold level, primary outcomes include smallholder farmer production output and crop-mix (recall of last productionseason). Intermediate outcomes along theorised agricultural and nutritional pathways were measured. During this trial, wewill follow a mixed-methods approach and undertake child-, household-, CBCC- and market-level surveys andassessments as well as in-depth interviews and focus group discussions with project stakeholders.

Discussion: Assessing the simultaneous impact of preschool meals on diets, nutrition, child development and agricultureis a complex undertaking. This study is the first to explicitly examine, from a food systems perspective, the impact of apreschool meals programme on dietary choices, alongside outcomes in the nutritional, child developmentand agricultural domains. The findings of this evaluation will provide evidence to support policymakers in thescale-up of national programmes.

Trial registration: ISRCTN registry, ID: ISRCTN96497560. Registered on 21 September 2016.

Keywords: Preschool feeding, Impact evaluation, Nutrition, Agriculture, Child development

* Correspondence: a.gelli@cgiar.org1International Food Policy Research Institute (IFPRI), 2033 K Street NW,Washington, DC 20006, USAFull list of author information is available at the end of the article

© The Author(s). 2017 Open Access This article is distributed under the terms of the Creative Commons Attribution 4.0International License (http://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution, andreproduction in any medium, provided you give appropriate credit to the original author(s) and the source, provide a link tothe Creative Commons license, and indicate if changes were made. The Creative Commons Public Domain Dedication waiver(http://creativecommons.org/publicdomain/zero/1.0/) applies to the data made available in this article, unless otherwise stated.

Gelli et al. Trials (2017) 18:284 DOI 10.1186/s13063-017-2003-7

BackgroundThe 2013 Lancet series on Maternal and Child Nutritionestimated that the aggregate global burden of undernutri-tion causes over three million child deaths per year. Stunt-ing prevalence in children under 5 years also affects atleast 165 million children [1]. Recent reviews of the contri-bution of agriculture in improving nutrition [2, 3] concludethat although agricultural programmes have immense po-tential to improve nutrition, this potential is yet to beunleashed. Current evidence suggests that limitations inthe design, targeting and implementation of agricultural in-terventions, as well as their lack of clarity in nutrition goalsand the exact interventions which are being provided, arepartly responsible for this weak evidence base [3].Early Childhood Health and Development (ECD) pro-

grammes are designed to improve young children’s sur-vival, growth and development. They are considered themost cost-effective form of human capital investmentwhen compared to any subsequent schooling interven-tions [4]. ECD investments can increase the efficiency ofongoing public spending in education, in turn improvingthe overall allocation of public resources [5]. A recent re-view of ECD programmes, which included both efficacytrials and programme evaluations, has demonstrated howimproving diets of pregnant women and young childrencan prevent stunting and result in enhanced cognitive andmotor development of children [6, 7]. The most effectiveprogrammes provided activities targeted to younger anddisadvantaged children that were integrated with healthand nutritional services. Providing services to children dir-ectly and involving parents in practice and skill-buildingsessions were more effective strategies than providing in-formation alone. Further evidence shows that the qualityof the child’s learning environments, including in thehome and in preschools, has a strong impact on child de-velopment. The potential for additive and synergistic ef-fects of combining ECD, and nutrition, and agricultureand nutrition is clear [1]. Combining these three sectorsmay lead to substantial gains in cost, efficiency and effect-iveness but these need to be rigorously evaluated.The Nutrition Embedded Evaluation Programme Im-

pact Evaluation (NEEP-IE) study is designed to evaluatethe impact of an integrated package of nutritional andagricultural interventions (including improved nutritiousfood production and behaviour-change communicationrelated to food intake, care and hygiene practices) on thediets, nutrition and development of young children andtheir households in rural areas of Malawi.This paper presents the rationale and study design for

this randomised trial.

Country contextMalawi ranks 73rd out of 104 countries on the GlobalHunger Index [8] and has one of the highest rates of

chronic malnutrition in the world with 37% of childrenaged 6–59 months being moderately or severely stunted[9]. It is also one of the most committed countries toimproving nutrition, ranking 3rd on the Hunger andNutrition Commitment Index (HANCI). In March 2011,the Malawi Government joined the Scaling Up Nutrition(SUN) Movement promoting a multisectoral approachto tackling malnutrition. Malawi is one of the few coun-tries that meet the African Union’s Maputo Declarationto spend over 10% of public expenditure on agriculture.An intersectoral coordinating body (National NutritionCommittee) was set up to support cross-sectoral integra-tion and the Department of Nutrition has been placed inthe Office of the President and Cabinet to highlight thegovernment’s commitment to nutrition.The national ECD programme, led by the Ministry of

Gender, Children and Social Welfare (MoGCSW) targetsall children aged 0–8 years. Preschools (known ascommunity-based childcare centres (CBCCs)) and parent-ing groups are the two main components of the govern-ment ECD programme. CBCCs are community-initiatedand community-owned childcare centres which aim topromote holistic child development by providing safe,stimulating environments, access to health and nutritionalservices, and training for parents and caregivers. Today,there are an estimated 9000 CBCCs across Malawi, serv-ing 32%percent of all 3–5 year olds in the country. Parent-ing groups are typically linked to the CBCC and aim toreach parents of children aged 0–8 years, as well as par-ents to be (newly married and pregnant women).1

One of the main challenges facing CBCCs is lack offood (mid-morning porridge) [10]. The absence of foodat the CBCCs is one of the main causes of child absen-teeism and CBCC closure and the MoGCSW is lookingfor cost-effective ways to address the food gap inCBCCs. In the past 2 years, Save the Children, driven bydemand from the CBCC communities it supports, beganintegrating nutritional, agricultural and livelihood com-ponents into its ECD programme. The goal is to helpcommunities provide nutritious food to CBCCs all yearround and improve food security and child nutrition athousehold level at the same time. The programme drawslearnings from the USAID-funded Wellness and Agricul-ture for Life Advancement (WALA) project imple-mented across two districts for 5 years and the ConradN Hilton Foundation-funded “Improving CBCC Meals”project implemented in over 200 CBCCs across four dis-tricts. Anecdotal evidence from the 68 communities cur-rently benefiting from the integrated nutritional,agricultural/livelihood and ECD package suggests thatCBCCs, CBCC gardens and parenting groups provide avery effective, and yet untested, platform for improvingagricultural production, food security, child nutritionand caregiving practices at household level. As Save the

Gelli et al. Trials (2017) 18:284 Page 2 of 12

Children scales up this integrated package to CBCCscurrently supported in Zomba district, the impact evalu-ation will assess the feasibility, cost-effectiveness and im-pact on indicators relevant to each sector (nutrition,agriculture and ECD).The Government of Malawi also recognises the

need for multisectoral programming, particularly re-lating to nutrition [11] and is looking for evidence-based models to guide their national policies andstrategies. The proposed intervention supports a num-ber of Malawi’s national targets, most importantlythose for nutrition, food security, education and childdevelopment and brings together three sectors atcommunity level to implement a multisectoral, inte-grated, synergistic package adapted to the Malawiancontext.

The standard ECD packageThe standard package provided to Save the Children-supported communities includes the following activities.For the CBCC activities:

� CBCC caregivers receive a 2-week training, providedby government-approved trainers, using a standardgovernment-approved training manual when theystart. They receive 5-day refresher trainings annuallyand they are mentored by one visit/month by ECDfacilitators

� CBCCs target all 3–5 year olds in the community.They are managed by the community and are openfrom 8 a.m. to 11 a.m., 5 days per week. Porridge(mid-morning meal) is provided at about 10 a.m. byparents/facilitators with food contributions from thecommunity

For parenting education:

� CBCC caregivers and mentors received a 3-daytraining to organise parenting sessions using a stand-ard government parenting education manual

� The parenting sessions are organised once or twice amonth on average. Topics covered by parentingsessions include: child nutrition, stimulation, andparental role in school readiness, amongst others.Parenting sessions target all parents of children aged0–8 years old with particular focus on children from0–2 years

� Agricultural extension workers received a 5-daytraining to organise parenting sessions using a stand-ard government parenting education manual

� Trainings were conducted between April and May2016. The total number of people trained was 437with 382 women and 55 men

� Other ECD interventions include children’s corners,mobile phones and interactive radio instructions(IRI)

The integrated agricultural and nutritional interventionThe NEEP-IE integrated nutritional and agriculturalintervention is aimed at improving the diets, feeding,health and hygiene knowledge and practices in house-holds with infants and young children. This includespromoting optimal feeding and caring practices throughparenting groups; engaging parents, adolescents andother adults in the community in the planning and prep-aration of meals for children within CBCCs; improvingagricultural production of nutritious foods and food di-versification by using CBCC gardens as a learning sitefor communities; forming Village Savings and Loans(VSL) groups to help communities save and access fundsto purchase supplies for the CBCC garden and CBCCmeals or to start new home-gardens; and by organisingfarmers into collectives to increase their purchasing andselling power.More specifically, the agricultural training includes:

� Training on nutritious food production, includingthe following topics: production of foods to use inCBCCs, i.e. production of selected cereals andstaples (orange maize, orange-fleshed sweet potatoand cassava); selected legumes and nuts (soya beans,pigeon peas, cowpeas and groundnuts); and greenleafy vegetables (Amaranthus). In all this the sub-topics included: land preparation and planting,weeding and fertiliser application, pests and diseasecontrol, harvesting, storage and processing and util-isation in CBCCs. The use of manure as fertiliserwas also promoted

� The CBCC garden was used as a demonstration site.At a demonstration plot, the community preparedthe gardens with guidance from the AgricultureExtension District Officers (AEDOs). An AEDOwould start a ridge with the correct spacing and thecommunity would take over after that with theAEDOs making sure that the correct spacing is usedand the correct type of ridge is made. This wouldcontinue until the rest of the garden has beenprepared. The same process was adopted for all thegardens. Where the demonstration coincided withthe first rains, the planting was also done, again withguidance from the AEDOs. AEDOs and communityagents (CAs) monitor the gardens two to threetimes per month on average

� The first agricultural production training wasfacilitated by AEDOs and involved 900 participantsin total. Each training session lasted 3 days andincluded 45 participants. The training took place in

Gelli et al. Trials (2017) 18:284 Page 3 of 12

December 2015 and farmers (households) beganplanting soon after the training. The training wastwo-fold: theory and practice. A second agriculturalproduction training was also facilitated by AEDOs to120 participants. Each training session lasted 3 daysand each session had 20 participants. These trainingstook place in August 2016 and covered the followingtopics: land preparation and planting, weeding andfertiliser application, pests and disease control, har-vesting, storage and processing and utilisation inCBCCs for carrots, spinach and tomatoes

In addition, participating households received a rangeof agricultural inputs as summarised in Table 1.The nutritional training activities included:

� Training CBCC management committee members,caregivers, lead farmers and parents on thenutritional needs of children, healthy meals all yearround, food selection, storage and preservation, foodhygiene, safety and preparation, waste disposal,hand-washing, meal planning and monitoring ofmeal provision and recipes for CBCCs andhouseholds

� Each training session lasted 3 days and was providedby AEDOs and nutrition assistants The trainingcombined both theory and practice, includingcooking demonstrations. The participants weredivided into groups of three to five people. Trainerswould introduce selected recipes for nutritiousmeals, with instructions written on a card that wasleft with the group. The trainer would facilitate eachstep of the process. Thereafter, the groups wouldcome together and display what they had cooked. Atthe display table, a member from each group wasselected to explain how they made the meal, howmany types of food groups the recipe contained, anyalternatives and, lastly, answer any questions from

the group. After the display, the participants tastedthe food that they had made. The recipes included:1. Porridges: maize flour porridge with groundnut

powder, maize flour porridge with dry fishpowder, maize flour porridge with dry vegetables,mango porridge, porridge made from a mixtureof maize, soya, millet, beans and groundnutsflour, rice porridge with carrot and oil, and riceporridge with groundnut flour

2. Milk and soya milk production3. Legume products: pigeon peas sausage, cassava

kidos (boiled cassava dipped in eggs andvegetables and fried), soya coffee, sweet potatodoughnuts, peanut butter and soya snacks

4. Vegetable products: pumpkin leaf meatballs(pumpkin leaves, salt and eggs as ingredients);orange-fleshed sweet potato (OFSP) juice, sweetpotato leaf juice, dried vegetables (for preserva-tion); pumpkin leaves and Amaranthus ingroundnut powder and sweet potato leaf snack

5. Fish products: dry fish with groundnut powder,dry fish with tomato and onion

6. Fruit products: pawpaw, guava and lemon juice;pawpaw relish (unripe pawpaw cooked withgroundnut powder, tomato and onion) andbanana bread

� Trainings were conducted in February 2016 andCBCC caregivers/communities began putting intopractice what they had learned from March 2016.CBCC caregivers and parents were then mentoredand supervised by community-based organisations(CBOs) with four visits per month

Study aim and objectivesThe purpose of the NEEP-IE study is to provide evidenceon the effectiveness and costs of delivering an integratedagricultural and nutritional intervention through CBCCsand parenting groups on the diets, nutrition and develop-ment of young children in rural areas of Malawi. The find-ings of the evaluation will be used to inform theGovernment of Malawi and development partners on theeffectiveness and feasibility of scaling-up the intervention.The objectives of the evaluation include:

1. Evaluating the impact of integrating nutrition andagriculture with ECD on the diets, nutrition anddevelopment of children aged 36 to 72 months

2. Evaluating the impact of integrating nutrition andagriculture with ECD on CBCC meal provision,CBCC attendance and enrolment

3. Identifying the main factors that influence theimpact of the intervention on child- and household-level outcomes

Table 1 Agricultural inputs provided to households as part ofthe agricultural and nutritional intervention

Seedsprovided

Plantingperiod

Harvest period Consumptionperiod

Orange maize December March-April May onwards

Pigeon peas December April April onwards

Cow peas December March March onwards

Beans December

Groundnuts December

Soya December April April onwards

OFSP January April-March April onwards

Carrots July October October onwards

Amaranthus August August/September August onwards

OFSP orange-fleshed sweet potato

Gelli et al. Trials (2017) 18:284 Page 4 of 12

4. Evaluating the effectiveness of CBCCs and parentinggroups as entry points to improve nutrition-relatedoutcomes to other critical age groups, includingyounger siblings of preschoolers

5. Evaluating the cost, feasibility and sustainability ofscaling-up the integrated nutritional and agriculturalpackage through CBCCs and parenting groups

MethodsProgramme theoryThe overall programme theory for the package of nu-tritional and agricultural interventions is guided bythe Lancet series framework on Maternal and ChildNutrition [1] and broadly summarised in Fig. 1. For amore detailed analysis of the complex pathways link-ing agriculture and nutrition, including the differentprocesses, actors, effects and lags, see [2, 12–14]. Thepackage of interventions affects health and nutritiondirectly by improving diets and feeding practicesthrough the behaviour-change communication andnutritional education. This, in turn, has an indirectimpact on preschooling, as improving health and nu-trition has a positive impact on attention, cognitionand learning. By increasing the regularity and qualityof the CBCC meals the interventions will also directlyinfluence children’s participation in the CBCC. Theinterventions can also affect agriculture by increasingproduction, sales and profits, and changing the cropproduction mix.

Impact on children’s nutrition and healthThe three determinants of undernutrition in children in-clude food, health and care practices [1]. The mainchannels through which the intervention has an impacton nutrition and health is through improved diets via in-creased consumption of nutrient-rich foods, and throughimproved nutrition, health and hygiene practices.The proposed intervention package can potentially

have an impact on the nutrition and health of childrenenrolled in CBCCs and their younger siblings, as sum-marised in Fig. 3 in the Appendix. This involves a com-bination of direct transfers to the CBCC children (e.g.transfer of nutritious food through preschool meals) andindirect channels involving the behaviour-change cam-paigns promoting the consumption of nutritious foodsand improved nutrition practices at household level. Im-proved diets, when accompanied by adequate feeding,health and hygiene practices can then contribute to im-proved health and nutrition. In particular:

� The nutritional impact is mediated by the extent offood substitution effects within the household, andthe use of the nutritional intake by the child andtheir siblings

� Diversifying diets and increasing the intake ofmicronutrient-rich foods can have direct effects onattention and cognition

� Healthier and better-nourished children are thenbetter positioned to learn new skills both at theCBCC and elsewhere

Fig. 1 Overall programme theory for the agricultural and nutritional intervention

Gelli et al. Trials (2017) 18:284 Page 5 of 12

� All these effects are mediated to some degree bywomen’s roles in the household, time allocation anddecision-making

Substitution, or intrahousehold reallocation, mayoccur when households readjust consumption patternsin response to the CBCC meals, or to a change indiet, by substituting foods normally consumed athome, or with other foods with similar properties.Substitution is a complex issue involving changinghousehold dynamics where gender plays a fundamen-tal role. Influencing possible substitution effects willbe critical in determining the potential impacts onchildren’s nutrition and health.Changes in individual-level dietary diversity have

been found to be strongly associated with micronu-trient adequacy of diets for women [15, 16] andmicronutrient density adequacy of diets in children[15]. Addressing micronutrient deficiencies can im-prove a range of health, nutritional and developmen-tal outcomes in infants and young children,particularly if implemented alongside behaviourchange on health and nutritional practices [17]. Con-ceptually, this framework suggests that the emphasisof the interventions in the short term should focuson integrating nutritional education and messaging,alongside the CBCC meals, to deliver improved intakeof micronutrient-rich foods, with the potential ofleading to improved household diets.Restoring micronutrients and enhancing energy intake

can also have an impact on attention and motivation.Energy [18] and iron intake [19] can have an impact onhyperactivity, withdrawal, nervousness, hostile behaviourand happiness. The emotional status of children mayalso affect attention span and have other positive spill-overs. Caregivers and peers are also likely to be affectedby the increase in attention and concentration.

Impact on agricultureFrom an agricultural perspective, the intervention fo-cusses on increasing food production in the CBCC gar-den and home-gardens by increasing yields orefficiencies through input provisions or training onfarming practices. The intervention can also influencethe basket of products that are being produced, support-ing the production of higher-value crops and/or morenutritious crops through the provision of seeds, educa-tional campaigns, or the opening of new market chan-nels. The selection of particular crops involves balancingthe pros and cons of substitution between crops for pro-duction, sale and consumption and the long-term im-pacts for both incomes and nutrition (see [20] for moredetails).

Main hypotheses and outcome indicatorsThe expected impact of the intervention discussed inthe analysis of the programme theory is summarisedbelow. The intervention is expected to have a positiveimpact on:

� Preschool enrolment and attendance� Children’s diets and household diet diversity� Infant and young children’s nutritional status and

childcare practices� Agricultural production

A limited impact is expected on:

� Micronutrient status, child development, physicalgrowth

� Agricultural income

The main indicators for the evaluation are summarisedin Table 2.Note that in addition to outcome indicators we will

also observe the programme impact on intermediate in-dicators, particularly for those outcomes that are moredifficult to observe directly. In the agricultural domain,we will look at intermediate outcomes such as input use(labour, land, seeds and fertiliser), investments (farmcapital, such as tools and machinery) and market access(marketed surplus, prices and markets). The quantity,quality and timely preparation and delivery of food inthe CBCCs will also be explored.

Design of the randomised evaluationA cluster randomised trial (CRT) is being implementedin 60 rural communities with CBCCs supported by Save

Table 2 Main outcome indicators of the intervention

Type Domain Indicators

Primary Diets Individual intake and diet diversity score(children 36–72 m)

Primary Childcarepractices

WHO IYCF practices

Primary CBCCparticipation

CBCC enrolment and attendance(children 36–72 m)

Primary Agriculture Production output, crop-mix

Secondary Health andnutritionalstatus

Anthropometry (weight-for-age, height-for-age, weight-for- age z-scores and MUAC)(children 6–72 m)

Secondary Childdevelopment

Malawi Development Assessment Toolz-scores (fine motor, gross motor, languageand social domains) (children 36–72 m)

Secondary Gender Women’s asset ownership, time useand productivity

Process Meal service Quality of CBCC meals, portion sizes, frequency

CBCC community-based childcare centres, IYCF Infant and Young Child Feeding,MUACmid-upper arm circumference, WHO World Health Organization

Gelli et al. Trials (2017) 18:284 Page 6 of 12

the Children’s ECD programme in Zomba district,Malawi. The evaluation follows a mixed-methods ap-proach (combining quantitative and qualitative methods)with two rounds of surveys and assessments timed 1 yearapart, including child, caregiver, household, CBCC andmarket-level data collection.

Study siteThe intervention is targeted to disadvantaged communi-ties within Zomba district in Malawi. The proposedstudy population includes CBCCs currently supportedby Save the Children with an ECD package, includingparenting and CBCC quality improvement. The geo-graphical area for intervention was targeted by Save theChildren on the basis of a set of education variables thatimpact pupil attendance and achievement in school.Within Zomba, two traditional authorities (TAs) andtwo sub-TAs were selected based on need (educationand health) and the presence of other NGOs to imple-ment the ECD programme. Save the Children’s ECDprogramme currently reaches 228 communities, 109 inTA Chikowi, 42 in TA Mbiza, 37 in sub-TA (STA) Ntho-lowa and 40 in STA Ngwelero. Sixty-eight of these (27in TA Chikowi and 41 in TA Mbiza) had benefited fromthe agricultural and nutritional components already andwere, therefore, excluded from the evaluation.

Study populationThe evaluation targets all children aged 0–6 years in the60 selected communities and their caregivers. The pri-mary reference group for this study is children aged 3–6years old living in the service area of a Save the Chil-dren‐supported CBCC. Secondary reference groups in-clude their siblings and caregivers.

Random assignment and manipulation of treatmentsThe 60 communities were randomly assigned to one oftwo treatment arms:

1. Control group: communities with CBCCs supportedby Save the Children’s ECD programme with noadditional nutritional or agricultural support

2. Intervention group: communities with CBCCssupported by Save the Children’s ECD programmewith additional nutritional and agricultural support

The integrated intervention package will be imple-mented in 30 of the 60 rural communities after the base-line survey and extended to the control communitiesafter the end-line survey. There are several reasons whythe control group in this case is not a control withoutintervention. The Government of Malawi is committedto scaling-up the ECD quality improvement across allCBCCs and an impact evaluation on the cost-

effectiveness of different ECD quality improvementstrategies is underway.2 The proposed evaluation com-plements the ongoing work by examining the relativeimpact and costs of alternative implementation models,focussing on how to enhance participation in the CBCCsand, at the same time, supporting the nutrition of chil-dren at a critical age in their development.The 60 CBCCs were randomly selected in two stages

from a pool of 235 CBCCs in 47 clusters currently assistedby Save the Children in Zomba. Due to the clustering ofthe CBCCs around primary schools, the list of 235 CBCCswas screened to flag clusters where more than one CBCCwas being assisted. Twenty-six clusters were excludedfrom the first stage of randomisation to minimise possiblecontamination. An additional 10 CBCCs were dropped asthey had ongoing activities. The 20 clusters were then ran-domly assigned to two groups of 10 clusters, where therandomisation was stratified geographically by traditionalauthority areas. In the second stage of randomisation,within each cluster, three CBCCs were then selected atrandom for the study. As six clusters had fewer than threeCBCCs available for the study, in order to select a fullsample of 30 CBCCs per treatment arm, additionalCBCCs were randomly selected from three clusters(Gologota, St. Pius and Machereni).

Sample sizesInitial power calculations and resource availability hadsuggested the adoption of a sample with 30 clusters (com-munities) per treatment arm with 20 households in eachcluster to identify reasonable treatment impacts of theintervention on the primary study outcomes. Data forpower calculations was obtained from the 2010 Demo-graphic and Health Survey (DHS) survey. We calculatedmeans, standard deviations and intracluster correlationcoefficients (ICCs) for rural children in Malawi. For Diet-ary Diversity Score (DDS), the mean and standard devi-ation for rural children aged 0–5 years were estimated tobe 2.5 and 1.03, respectively. The ICC was 0.01. Plots ofthe standardised minimum detectable effect size (MDES)against the number of clusters assuming a sample of 25children measured in each community (cluster size), con-sistent with 20 household interviews per community andconsidering that several children may end up not beingtested, showed that only a marginal gain can be obtainedby expanding the sample beyond 20 as power is mainlydriven by the number of clusters [21] (see Fig. 2 for ex-ample simulations of MDES versus number of clusterswith high and low ICCs).After preliminary design visits to the targeted commu-

nities, the sampling strategy was modified to account forthe implementation approach adopted by Save the Chil-dren, involving the clustering of CBCCs around sur-rounding primary schools. As a result, the cluster, or

Gelli et al. Trials (2017) 18:284 Page 7 of 12

unit of randomisation was the primary school clusterthat included a number of different CBCCs, rather thanthe CBCC itself. Adjusting for ICCs at the primaryschool cluster level, where 60 CBCCs were clusteredinto two groups of 15 primary school clusters, wouldprovide 80% power to detect a 0.24-SD difference be-tween treatment groups at the 5% level of significance(Table 3).The sampling of households was conducted through a

census within a certain catchment area for each CBCCincluding information on the target age groups livingwithin each household. Households with children aged3–5 years were then randomly selected for participationin the survey.

Data collection toolsThe impact evaluation includes child-, caregiver-, house-hold-, CBCC- and village-level data collection (Table 4).The household questionnaire collected data at thehousehold level as well as for each relevant householdmember separately (main caregiver and all children inreference age groups).

Methods of analysisThe randomised design allows for the identification ofcausal impacts of interventions using comparisons ofmean outcomes between the randomised treatment armsat end line. The analysis will follow the intention-to-treat approach as protocol and as treated, using econo-metric analysis for all the relevant outcomes of the inter-vention. Following [22], impact will be assessed usingboth a ‘difference-in-difference’ (DID) estimator and asingle difference analysis of covariance (ANCOVA)model. Depending on the level of clustering of the out-come, we will employ multilevel regression models thataccount for the hierarchical nature of the data [23].Multilevel models, also known as mixed-effects models,use both fixed effects (covariates) and random effects inat-school and household levels.The DID estimate is calculated as the average change

in the outcome of interest in the treatment arm minusthe change in outcome in the control group. A difficultyof DID analysis involves serial correlation [15] resultingfrom unobserved factors affecting the outcomes that arethemselves correlated over time. Serial correlation affectsestimated standard errors and can lead to erroneous ac-ceptance or rejection of null hypotheses but not the esti-mation of the effect size of the intervention. It may,therefore, lead to erroneously finding or not finding astatistically significant impact of the intervention. Thisproblem can be addressed by calculating clustered stand-ard errors [24]. Clustered standard errors will also beemployed in all cases in which correlated outcomes areobserved within the same unit of analysis. The analysisof covariance (ANCOVA) estimator has been shown toprovide a more efficient estimate of programme impactwhen autocorrelation of outcomes is low [22].

Heterogeneity of impactThe large dataset will allow for extensive subgroup ana-lyses, including gender, age and geographic characteris-tics. The impacts of preschool feeding may be quiteheterogeneous and context specific [24, 25]. Schoolmeals, for instance, have been associated with markedimprovements in school participation of girls in ruralareas where there are large gender disparities in accessto education [26]. Furthermore, smallholder farmers tar-geted by the programme are mostly female.

Cost-effectivenessCost data will be collected retrospectively following aningredients approach using a semistructured question-naire. The survey will be based on a standardised costingframework capturing capital (fixed) and recurrent costsincurred at the school level. The questionnaire will alsocover both cash and in-kind contributions and will beused to estimate both financial and economic costs.

Table 3 Sample sizes

Primaryclusters

Communities/CBBCs

Householdswith childrenof target age

Children(0–5 y)

Children(3–5 y)

Control 15 30 600 936 648

Intervention 15 30 600 936 648

Total 30 60 1200 1872 1296

Note: Number of children estimated based on demographic data from theDemographic and Health Survey (DHS), 2010CBBCs community-based childcare centres

Fig. 2 Diet diversity: minimum detectable effect size versus numberof clusters, simulations with high and low intracluster correlationcoefficients (ICCs)

Gelli et al. Trials (2017) 18:284 Page 8 of 12

Financial costs capture actual expenditures in terms ofprogramme implementation on an annual basis. Economiccosts included the opportunity costs of community mem-bers, teaching staff and other stakeholders involved in theintervention provision. Opportunity costs of preschool staffand community members will be calculated using local payscales. Capital costs will be annuitised over the useful lifeof all relevant school-level assets using a discount rate of3% as per World Bank recommendations. Annuitisationenables an equivalent annual cost to be estimated and re-flects the value in-use of capital items, rather than reflect-ing when the item was purchased. Process and output datacovering the adequacy of the service delivery will be col-lected from monitoring visits on a quarterly basis usingstandardised data collection forms. Output data will becombined with the costs to provide estimates of cost-efficiency metrics, including costs per beneficiary, kilocalo-ries, iron, and vitamin A delivered. Sensitivity analysis willbe undertaken to account for uncertainties in the economicevaluation. The figures obtained in this way will then becompared to figures calculated for other interventions.

Table 4 Survey questionnaire modules

Questionnaire Module Description

CBCC Location andaccess

Identification, location

Infrastructure Physical infrastructure, includinglearning space, water andsanitation, cooking and storagefacilities

Staff Staff roster, education and training

Curriculum andservices

Quality of CBCC activities andrelated services

Caregivers healthand nutritionknowledge

Knowledge related to optimalinfant and young childcare andfeeding practices

Health and hygienepractices

Health and hygiene practices ofCBCC staff

Meal provision Meal quality, portion sizes, mealplanning, management anddistribution

Food procurement List of food procured/sourcedby the CBCC

Garden land Land used by CBCCs, includingownership and use, size of theplot, crops planted, input andlabour for each plot

Garden production Crop production and use

Garden sales Crop sales, volumes and prices

Food storage Food storage infrastructure andpractices

Household Roster Listing of demographiccharacteristics of householdmembers

Dwellingcharacteristicsand assets

Basic features of the household’sprimary dwelling place, includinginfrastructure, access to water andelectricity

Assets Assets owned (by men andwomen separately)

Land Land owned and used byhousehold’s women and men,including ownership and use, sizeof plot, crops planted, labour foreach plot

Agriculturalproduction

Crop production and use

Agriculturalmarketing

Crop sales, volumes, prices,

Agricultural storage Storage volumes and management

Farm investments On-farm investments and labour

Farming practices Pre- and post-harvest practices

Livestock Livestock holdings, revenue andcosts, women ownership

Employment andbusiness enterprise

Non-farm sources of income(including employment), costs,male and female members

Table 4 Survey questionnaire modules (Continued)

Shocks Unexpected events that may haveinfluenced household’s wellbeingand responses taken by household

Food security Household vulnerability withrespect to food frequency

Food expenditures Food expenditures and quantitiesconsumed at household level

Non-foodexpenditures

Expenditures on household items,clothing and personal expendituresover the last month

Caregiver Caregiver’s healthand hygiene

Health- and hygiene-relatedquestions

Caregiver’s healthand nutritionknowledge

Knowledge related to optimalinfant and young childcare andfeeding practices

Caregivers IYCFpractices

Infant and child feeding practices

Childcare practices Childcare practices, includingsupport for learning andstimulation

Women’s timeallocation

Women’s use of time, perceptionson women’s time use

Child Child health Child immunisation history andhealth-related questions

Dietary assessment Interactive 24-h recall on foodintake for children aged 3–5 years

Anthropometry Physical measurements of allchildren and their parents

Development MDAT scores (fine and grossmotor, language and socialdomains)

CBCC community-based childcare centre, IYCF Infant and Young Child Feeding,MDAT Malawi Development Assessment Tool

Gelli et al. Trials (2017) 18:284 Page 9 of 12

Data collectionThe enumerators will be recruited from Chancellor Col-lege, University of Malawi and trained for the baseline sur-vey. Each team, led by a supervisor and assisted bycommunity leaders, will conduct household listings andsampling in each enumeration area. The data collectionwill be undertaken using electronic tablets. Data collectionwill be reviewed daily by a team supervisor and inconsist-encies clarified. Dietary assessment will be undertaken bytrained and supervised enumerators using the interactive24-h recall method. Prior to the recall interview, care-givers will be briefed on the purpose and methods ofinterview. The interview will be conducted using visualaids to assist in estimating portion sizes of the foods con-sumed. The 24-h dietary assessment will be repeated onnonconsecutive days for a subset of households (approxi-mately 20%) to obtain estimates of usual intake [27]. An-thropometry collection will include measurements ofchildren’s height and weight. Height or length will be mea-sured to the nearest 0.1 cm using portable fixed base sta-diometers and weight will be measured to the nearest0.1 kg using electronic scales. The height and weight mea-sures will be assembled and placed on a level surface. Inthe absence of a level ground in the household, a suitableplace will be identified for the measurement in the

community. Training on the MDAT will be provided toall supervisors and enumerators by trained staff from theCollege of Medicine, Malawi. During the MDAT trainingall enumerators will be reviewed for consistency andreliability.

DiscussionEarly childhood development programmes are consid-ered as the most cost-effective form of human capital in-vestment compared to any subsequent schoolinginterventions [4]. Emerging evidence highlights the po-tential additive and synergistic effects combining ECD,and nutrition [6, 7], and agriculture and nutrition [3].Combining these three sectors may lead to substantialgains in cost, efficiency and effectiveness; however, thereis a need to rigorously assess these potential synergies.This paper describes the design for a CRT of a

preschool-feeding-based intervention linked to smallholderagricultural and community engagement. As far as we areaware, it is the first to explicitly examine (from a food sys-tems perspective) the simultaneous impact of a preschoolmeals programme on dietary choices, alongside outcomesin the nutritional, child development and agricultural do-mains. As the intervention is complex, the scope of this

Fig. 3 Impact pathways for the intervention on child nutrition and development (Fig. 3). (Source: adapted from [28])

AppendixImpact pathways for preschool meals

Gelli et al. Trials (2017) 18:284 Page 10 of 12

theory-based evaluation includes measurement of a rangeof outcome and process metrics across multiple disciplines.The data collection requires multisectoral expertise in thesurvey teams, including measurements of dietary intake,anthropometry and child development, alongside expend-iture and other socioeconomic-related dimensions.

Trial statusThe first participant was enrolled on 7 September 2015.Data collection for the last of 120 participants will becompleted by December 2016 (see the SPIRIT Checklistfigure (Additional files 1 and 2)) (Fig. 4).

Endnotes1Save the Children, unpublished.2The Protecting Early Childhood Development (PECD)

trial.

Additional files

Additional file 1: The schedule of enrolment, interventions andassessments for the NEEP-IE study. (DOCX 14 kb)

Additional file 2: SPIRIT 2013 Checklist. (DOC 120 kb)

AbbreviationsANCOVA: Analysis of covariance; CBCCs: Community-based childcarecentres; CBO: Community-based organisation; CRT: Cluster randomisedtrial; DDS: Dietary Diversity Score; DID: Difference-in-difference; ECD: EarlyChildhood Health and Development; FVS: Food Variety Score; HANCI:Hunger and Nutrition Commitment Index; HAZ: Height for age z-score;HH: Household; IRI: Interactive radio instructions; IYCF: Infant and YoungChild Feeding; MDAT: Malawi Development Assessment Tool; MDES: Minimumdetectable effect size; NEEP-IE: Nutrition Embedded Evaluation ProgrammeImpact Evaluation; OFSP: Orange-fleshed sweet potato; SUN: Scaling UpNutrition; TA: Traditional Authority; WALA: Wellness and Agriculture for LifeAdvancement

AcknowledgementsWe are grateful to the Government of Malawi and acknowledge the inputsand feedback from the Ministry of Gender, Children, Disability, and SocialWelfare, including Mr. McKnight Kalanda (Director of Child Affairs) and Mr.Francis Chalamanda (National Coordinator for Early Childhood

Development). We would also like to thank the following experts for inputsand feedback on the study design: Christin McConnell (World Bank), MelissaHidrobo (IFPRI), Harold Alderman (IFPRI) and Marie Ruel (IFPRI); Peter Phiri forinputs in the study design and linkages to the ECD programme; LexonNdalama for links with policy and programme implementation; and HelenMoestue for feedback on the baseline report.

FundingWe acknowledge the support from the CGIAR Research Programme onAgriculture for Nutrition and Health (A4NH), led by IFPRI. This work wassupported by a Nutrition Embedding Evaluation Programme (NEEP) grantfrom PATH, funded by the UK Government’s Department for InternationalDevelopment. The views expressed do not necessarily reflect the UKGovernment’s official policies.

Availability of data and materialsNot applicable.

Authors’ contributionsThis study was implemented in a partnership between IFPRI, Save theChildren and Chancellor College, University of Malawi. We would like toacknowledge the different author contributions: at IFPRI: AG led the researchdesign and development. AM provided inputs into the study design,questionnaire development and enumerator training. KS for the qualitativeresearch and inputs into the gender elements of the study design,questionnaire development and for supporting the enumerator training. MSfor the data cleaning, processing and analysis and ST for the qualitativeresearch and inputs into the gender elements of the study design andquestionnaire development. At Save the Children: AT for supporting theenumerator training, translation, and survey management and coordinationwithout which the study would not have been possible. GC for inputs intothe design and supporting the survey implementation. NR for inputs intothe study design and linkages with policy and programme implementation;at Ikapadata: AC and TM for the CAPI programming and supporting theenumerator training and data collection. MG (University of Liverpool) forinputs into the study design and use of MDAT. PK (University of California atBerkeley) for inputs into the study design; at Chancellor College: MK forinputs into the study design and the baseline survey. All authors read andapproved the final manuscript.

Competing interestsThe authors declare that they have no competing interests.

Consent for publicationNot applicable.

Ethics approval and consent to participateEthical clearance was obtained from the Malawi National Commission forScience and Technology, (approved 12/11/2015, ref: NCST/RTT/2/6) and atthe International Food Policy Research Institute IRB in Washington, DC,

Fig. 4 Study timeline

Gelli et al. Trials (2017) 18:284 Page 11 of 12

(approved 26 March 2015, ref: IRB00007490). Meetings were held from withGovernment Ministries (including Gender, Children, Disability, and SocialWelfare) to discuss the purpose, procedures and risks involved. Informedconsent was obtained from the parents/guardians of children throughwritten and verbal information provided before interviews.

Publisher’s NoteSpringer Nature remains neutral with regard to jurisdictional claims inpublished maps and institutional affiliations.

Author details1International Food Policy Research Institute (IFPRI), 2033 K Street NW,Washington, DC 20006, USA. 2Johns Hopkins University, Baltimore, MD, USA.3Save the Children, USA/Malawi, Lilongwe, Malawi. 4Ikapadata, RSA, CapeTown, South Africa. 5University of Liverpool, Liverpool, UK. 6University ofCalifornia at Berkeley, Berkeley, CA, USA. 7Chancellor College, University ofMalawi, Zomba, Malawi.

Received: 30 September 2016 Accepted: 20 May 2017

References1. Black RE, Victora C, Walker SP, Bhutta ZA, Christian P, De Onis M, et al.

Maternal and child undernutrition and overweight in low- and middle-income countries. Lancet. 2013;382(9890):427–51.

2. Masset E, Haddad L, Cornelius A, Isaza-Castro J. A systematic review ofagricultural interventions that aim to improve nutritional status of children.London: EPPI-Centre, Social Science Research Unit, Institute of Education.University of London; 2011.

3. Ruel MT, Alderman H, and the Maternal and Child Nutrition Study Group.Nutrition-sensitive interventions and programs: how can they help toaccelerate progress in improving maternal and child nutrition? Lancet. 2013;382(9891):536–51.

4. Heckman JJ. Skill formation and the economics of investing indisadvantaged children. Science. 2006;312:1900–2.

5. Alderman H. The economic cost of a poor start to life. J Dev Orig HealthDis. 2010;1(1):19–25.

6. Engle PL, Black MM, Behrman JR, Cabral de Mello M, Gertler PJ, Kapiriri L, et al.Strategies to avoid the loss of developmental potential in more than 200million children in the developing world. Lancet. 2007;369(9557):229–42.

7. Grantham-McGregor SM, Fernald LC, Kagawa RM, Walker S. Effects of integratedchild development and nutrition interventions on child development andnutritional status. Ann NY Acad Sci. 2014;1308:11–32. doi:10.1111/nyas.12284.

8. von Grebmer K, Bernstein J, de Waal A, Prasai N, Yin S, Yohannes Y. 2015Global hunger index: Armed conflict and the challenge of hunger. Bonn,Washington, and Dublin: Welthungerhilfe, International Food PolicyResearch Institute (IFPRI) and Concern Worldwide; 2015.

9. National Statistical Office/Malawi and ICF. Malawi Demographic and HealthSurvey 2015-16. Zomba: National Statistical Office and ICF; 2017.

10. Neuman MJ, McConnell C, Kholowa F. From early childhood developmentpolicy to sustainability: the fragility of community-based childcare servicesin Malawi. Int J Early Child. 2014;46(1):81–99. doi:10.1007/s13158-014-0101-1.

11. Scaling-Up Nutrition (SUN) Country Profile Malawi. http://scalingupnutrition.org/sun-countries/malawi. Accessed 06 June 2017.

12. Gillespie S, Harris L, Kadiyala S. The agriculture-nutrition disconnect in India:what do we know? Discussion Paper 01187. Washington: International FoodPolicy Research Institute; 2012.

13. Webb P. Impact pathways from agricultural research to improved nutritionand health: literature analysis and research priorities. Rome: Food andAgriculture Organization; 2013.

14. Turner R, Hawkes C, Jeff W, Ferguson E, Haseen F, Homans H, et al.Agriculture for improved nutrition: the current research landscape. FoodNutr Bull. 2013;34(4):369–77.

15. Arimond M, Wiesmann D, Becquey E, Carriquiry A, Daniels MC, Deitchler M,et al. Simple food group diversity indicators predict micronutrient adequacyof women’s diets in 5 diverse, resource-poor settings. J Nutr. 2010;140(11):2059S–69. Epub 2010 Sep 29.

16. Working Group on Infant and Young Child Feeding Indicators. Developingand Validating Simple Indicators of Dietary Quality of Infants and YoungChildren in Developing Countries: Additional analysis of 10 data sets. Report

submitted to the Food and Nutrition Technical Assistance Project )17/FHI360, September 2007.

17. Bhutta ZA, Das JK, Arjumand R, Gaffey MF, Walker N, Horton S, et al.Evidence-based interventions for improvement of maternal and childnutrition: what can be done and at what cost? Lancet. 2013;382(9890):452–77.

18. Polliltt E, Gersowitz M, Gargiulo M. Educational benefits of the United Statesschool feeding program: a critical review of the literature. Am J PublicHealth. 1978;68:477–81.

19. Grantham-McGregor SG, Ani G. Iron-deficiency anemia: reexamining thenature and magnitude of the public health problem. J Nutr. 2001;131:649S–68.

20. Masset E, Gelli A. Community participation and the links betweenagriculture, nutrition and education: design of a randomised fieldexperiment of ‘home-grown’ school feeding in Mali. Trials. 2013;14:55.

21. Hayes R, Moulton LH. Cluster randomised trials. Oxford: Chapman & Hall/CRC Press; 2009.

22. Bruhn M, McKenzie D. In pursuit of balance: randomization in practice indevelopment field experiments. World Bank Policy Research Working Paper4572. Washington: World Bank; 2009.

23. Goldstein H. Multilevel statistical models. 3rd ed. London: Edward Arnold;2003.

24. Angrist JD, Pischke JS. Mostly harmless econometrics. Princeton: PrincetonUniversity Press; 2009.

25. Kristjansson EA, Robinson V, Petticrew M, MacDonald B, Krasevec J, Janzen L,et al. School feeding for improving the physical and psychosocial health ofdisadvantaged elementary school children. Cochrane Database Syst Rev.2007;24(1):CD004676.

26. Gelli A. School feeding and girls’ enrollment: the effects of alternativeimplementation modalities in low-income settings in sub-Saharan Africa.Front Public Health. 2015;3:76. doi:10.3389/fpubh.2015.00076.

27. Willett W. Nutritional epidemiology. Oxford: Oxford University Press; 2013.28. Gelli A, Espejo F, Shen J, Kristjansson E. Putting it all together: aggregating

impacts of school-feeding programmes on education, health and nutrition:two proposed methodologies’. WIDER Working Paper 2014/036. Helsinki:UNU-WIDER; 2014.

• We accept pre-submission inquiries

• Our selector tool helps you to find the most relevant journal

• We provide round the clock customer support

• Convenient online submission

• Thorough peer review

• Inclusion in PubMed and all major indexing services

• Maximum visibility for your research

Submit your manuscript atwww.biomedcentral.com/submit

Submit your next manuscript to BioMed Central and we will help you at every step:

Gelli et al. Trials (2017) 18:284 Page 12 of 12