From evidence to national scale: An implementation framework for micronutrient powders in Rwanda

Abstract

Micronutrient powders (MNP) are an efficacious intervention in terms of reducing anaemia among young children, yet challenges remain regarding implementation at scale. Research that can guide effective implementation of nutrition interventions and facilitate integration into existing health care platforms is needed. This paper seeks to advance the implementation science knowledge base by presenting our multiphased strategy and findings for scaling‐up MNP in Rwanda. The multiphased implementation strategy, spanning a 5‐year period (2011–2016), included (a) a feasibility study involving formative research, (b) a 30‐day trial of improved practices (n = 60 households), (c) a 12‐month pilot that included an effectiveness study (n = 1,066 caregiver/child pairs), and (d) a staggered approach to national scale‐up. At the end of Phase 4, the programme had been implemented in 19 of Rwanda's 30 districts with the scale‐up in the final 11 districts completed in the following year. The caregivers of over 270,000 eligible children 6–23 months of age received a box of 30 MNP sachets in the final 3‐month assessment period, representing a coverage rate of 87%. Initial problems with the supply chain and distribution and ongoing challenges to monitoring and reporting have been the largest obstacles. Continued success will be dependent on adequate resources for capacity development, refresher training, and responsive monitoring. Rwanda is one of the first countries to successfully scale‐up home fortification subnationally with MNP. Lessons learned have implications for other countries.

Key messages.

• A comprehensive multiphased strategy led to the successful implementation of an MNP programme for children 6–23 months of age at a subnational scale in Rwanda.

• Formative research and pilot studies prior to national scale‐up allows implementers to gain essential information on local practices and facilitates an enabling environment for implementation.

• The integration of a nutrition intervention into existing national health platforms requires careful management with formative research and adequate resources allocated at each step of the process.

• Although considered a low‐cost and easy‐to‐use nutrition intervention, MNP programmes require close monitoring and contextualized adaptation to establish community acceptance.

1. INTRODUCTION

Home fortification with micronutrient powders (MNP) is a low‐cost intervention used to improve the nutrient adequacy of complementary foods that has proven to be effective in reducing the prevalence of iron deficiency and anaemia among young children 6–23 months of age (De‐Regil, Suchdev, Vist, Walleser, & Pena‐Rosas, 2013). The efficacy of MNP was initially demonstrated in controlled studies designed primarily to evaluate biochemical indicators (Zlotkin et al., 2003; Zlotkin et al., 2005; Zlotkin, Arthur, Antwi, & Yeung, 2001). Most recently, a study of MNP containing 22 micronutrients on full‐term low‐birth‐weight children aged 6–12 months in Bangladesh showed a reduction in rates of stunting at 12 months (Shafique et al., 2016).

MNP were included in the Lancet's 2008 “Maternal and Child Undernutrition” Special Series (Black et al., 2008) and later by the World Bank (Horton, Shekar, McDonald, Mahal, & Brooks, 2010) as one of 13 efficacious and highly cost‐effective interventions to improve the nutrition of young children. Although MNP have been evaluated in the context of small‐scale projects (De‐Regil et al., 2013; Scott, Chen‐Edinboro, Caulfield, & Murray‐Kolb, 2014; Shafique et al., 2016; Zlotkin, Perumal, Ho, & Wong, 2015), and global guidelines have been developed (Home Fortification Technical Advisory Group, 2013; World Health Organization [WHO], 2016), there has been a dearth of large‐scale effectiveness studies and a lack of documentation of best practices for successful integration into existing health platforms. Although published guidelines clearly recommend the widespread use of MNP, there is a wide gap between these guidelines and knowledge on how to implement home fortification with MNP at scale and an absence of available documented programme learnings (Nyhus Dhillon et al., 2017; Reerink et al., 2017; Vossenaar et al., 2017).

The importance of implementation science has been recognized in the health community over the past decade, and attempts have been made to provide guidance and a unifying theory for implementation research (Damschroder et al., 2009; Fixsen, Naoom, Blase, Friedman, & Wallace, 2005; Peters, Tran, & Adam, 2013). In addition, there has recently been a push to separate biological efficacy from programme efficacy in nutrition interventions (Habicht & Pelto, 2014) and for increased implementation research that is nutrition‐specific (Garrett, 2008; Leroy & Menon, 2008; Tumilowicz, Neufeld, & Pelto, 2015). In 2015, the Society for Implementation Science in Nutrition was formed with the mission to “advance the science and practice of nutrition implementation world‐wide” (Society for Implementation Science in Nutrition, 2017). In the context of implementing MNP programmes, district‐level and subnational distribution programmes commonly report challenges related to utilization, adherence, acceptability, supply chain management, infrastructure and human capital capacity, and monitoring (Angdembe, Choudhury, Haque, & Ahmed, 2015; Kodish, Rah, Kraemer, de Pee, & Gittelsohn, 2011; Tumilowicz, Schnefke, Neufeld, & Pelto, 2017; Zlotkin et al., 2015), all of which create barriers to success and call for the application of implementation research to improve outcomes. Research studies on MNP mitigate these issues with heavy investment in supervision and monitoring, in some cases, employing health personnel for daily home visits to caregivers (Shafique et al., 2016). This strategy is not feasible for sustainable programmes, which must fit within existing health platforms and pose a minimal burden on health personnel.

An opportunity for implementation science frameworks to be aligned with operational methodologies presented itself in Rwanda where the government is highly supportive of evidence‐based nutrition interventions and have “District Plans for the Elimination of Malnutrition” (DPEM) integrated within the health system. Great progress had been made in Rwanda in terms of poverty reduction and social development since the 1994 genocide, yet improvements in household food security and child nutrition have been slow (Rwanda 2000, 2000; Rwanda Demographic and Health Survey [DHS] 2010, 2012; Rwanda DHS 2014‐15, 2015). The prevalence of stunting remains high at 38% of children under 5 (Rwanda DHS 2014‐15, 2015), and anaemia has only decreased by 1% since 2010, with 38% of children under 5 categorized as anaemic (Rwanda DHS 2010, 2012; Rwanda DHS 2014‐15, 2015). Recognizing this, Rwanda's National Food and Nutrition Policy states that “strategies [are] needed to solve serious and persistent problems including the high prevalence of child stunting and high levels of anaemia in children and women” (National Food and Nutrition Strategic Plan 2013‐2018, 2012).

Home fortification with MNP offers a practical strategy in the Rwandan context to address the problem of anaemia and other micronutrient deficiencies if the challenges to programme implementation can be adequately addressed. The purpose of our research was to develop and test an implementation strategy and tools for an effective MNP programme adapted to the local context and health system. With a focus on implementation research, we provide an overview of the results of our multiphase strategy including supportive evidence from biochemical indicators and morbidity data during the pilot phase and monitoring of acceptability, coverage, utilization, and adherence.

2. METHODS

The University of British Columbia (UBC) and UNICEF conducted advocacy with the Rwandan Ministry of Health (MoH) for home fortification to be introduced and received official support for the programme to be integrated within the health sector's existing DPEM in 2011. A multiphased implementation strategy was developed and tested that included (a) a feasibility study, (b) a 30‐day trial of improved practices (TIPs), (c) a 12‐month pilot including an effectiveness study, and (d) a scale‐up phase. New knowledge generated across each phase was utilized to inform the subsequent phase and build organizational capacity, with the ultimate objective of integrating an MNP programme for children 6–23 months into national health platforms. The first three phases were completed in nine sectors of Musanze (n = 4 sectors) and Nyaruguru (n = 5 sectors) districts. A team from the UBC was responsible for the design and led the trainings and research components, with field support from UNICEF, the Institute of Agriculture and Technology, the National University of Rwanda School of Public Health, and non‐governmental organization partners. The first three phases occurred consecutively (July 2011, November to December 2011, and May 2012 to June 2013) with small breaks in between; findings were then utilized to secure funding and build support for the programme prior to the scale‐up phase (November 2014 to July 2016). The specific objectives and timeline for each phase are provided in Figure 1.

Figure 1.

Implementation phases and specific objectives to scaling‐up an MNP programme in Rwanda. Details each phase of the research process along with corresponding objectives. CHW: community health worker; MNP: micronutrient powder

2.1. Phase 1: Feasibility study

Phase 1 consisted of formative work on the feasibility of implementing a home fortification programme with 18 focus groups with 10 mothers of children 6–23 months each (n = 180) and key informant interviews with mothers, grandmothers, fathers, and health care workers (n = 97). A three‐stage cluster sampling design was used. At the first stage, Musanze and Nyaruguru districts were purposively selected as representative of different agricultural and socio‐economic regions. At the second stage, four to five sectors per district were selected using systematic sampling with probability proportionate to size. At the third stage, villages were categorized as near (within 6 km) or far (beyond 6 km) from a health centre, and one of each randomly chosen per sector. Caregivers of children under 2 years of age were invited to participate in focus groups and interviews through networks of community workers. Interviews were also conducted with key personnel at the main health centre in each sector.

The primary focus was to understand and identify issues related to knowledge, attitudes, and practices pertaining to complementary feeding and health priorities for infants and young children. Interviews and focus groups were conducted over 2 months by enumerators from the research team, who then translated and transcribed recordings from Kinyarwanda to English. The cultural context, food availability, and local capacity were evaluated to guide operational components of the programme with communication messages and tools developed from the results.

2.2. Phase 2: TIPs

Phase 2 consisted of a 30‐day TIPs to determine the acceptability of MNP among 60 caregiver–child pairs invited to participate by health care workers: 30 per district (Musanze and Nyaruguru), randomly selected from village lists of children 6–23 months. The TIPs technique functions as a pretest and provides in‐depth understanding of the preferences and capabilities of programme beneficiaries, as well as the obstacles they face in adopting new behaviours, practices, and products to improve health outcomes (Manoff Group, 2005). Training materials for the TIPs were based on the results of the formative research. Local community health workers (CHWs) were trained on complementary feeding principles and MNP usage by team members, with cooking demonstrations with local foods to demonstrate age‐appropriate diversity, consistency, and quantity. Subsequently, caregivers underwent similar training in the community with cooking demonstrations led by CHWs using a participatory approach.

Following the demonstrations, caregivers received 15 sachets of MNP to add to their child's food over 30 days. Interviews were conducted with caregivers by team members at three time points: (a) at baseline, prior to training on MNP usage; (b) after 10 days of receiving MNP; and (c) after 30 days of receiving MNP. Knowledge, attitudes, and practices related to complementary feeding and child health were assessed as well as adherence, utilization, motivations, and acceptance of MNP.

2.3. Phase 3: Pilot study

Phase 3 integrated important findings from the first two phases to refine training materials in anticipation of a distribution, monitoring, and evaluation strategy that was piloted in a longitudinal study over 12 months in four districts. To determine the effectiveness of the training materials for improving complementary feeding practices and child health, a subset of children (two of four districts) was provided with educational materials but no MNP. The pilot districts were paired with adjacent comparison districts that were similar in geography, nutritional status of children, and agricultural practices in a quasi‐experimental design. Musanze (pilot) was paired with Burera (comparison) in the North Province, and Nyaruguru (pilot) was paired with Nyamagabe (comparison) in the South.

Within the selected districts, the MoH compiled a list of all eligible children 6–12 months from all the health centres. From this list, 300 children were randomly selected using Excel from within each district. The study population consisted of 1,066 children: pilot n = 580 and comparison n = 486. This sample represented approximately 1% of the children of similar age in the districts. This age group was chosen to ensure that the study children would remain in the target group for MNP (6–23 months) for the entirety of the 12‐month study period and therefore receive the maximum number of sachets for evaluation purposes. Caregivers were to receive 10 MNP sachets from a CHW who recommended that they use all sachets in 1 month but not more than one a day. This distribution and adherence schedule were developed from the TIPs study results, previous research on the efficacy of flexible adherence (Ip, Hyder, Haseen, Rahman, & Zlotkin, 2009), and programme funding. Caregivers were instructed to return empty sachets when obtaining the next MNP distribution, as an indicator of utilization.

Utilization was defined as the proportion of caregivers of eligible children who reported using MNP in accordance with a minimum of six of the nine key directions for use. Coverage was determined by the proportion of caregivers of eligible children who reported receiving a box of MNP in the previous 3 months. Adherence was defined as the proportion of caregivers of eligible children who reported using MNP in accordance with the schedule of vadministration. The same definitions were utilized during Phase 4.

Surveys were conducted by teams from UBC and Institute of Agriculture and Technology with support from the MoH and UNICEF at three time points: baseline, 6 months, and 12 months, to assess complementary feeding practices, morbidity, and MNP acceptability. Haemoglobin concentration was measured using a portable HemoCue® Hb 201+ system on finger prick blood drops, and the WHO cut‐offs were employed to identify children with mild, moderate, and severe anaemia of 100–109, 70–99, and <70 g L⁻¹, respectively, excluding biologically implausible values (WHO, 2011), adjusted for elevation.

Prior to implementation, training was provided to CHWs and subsequently caregivers, incorporating key messages from the national Infant and Young Child Feeding (IYCF) guidelines with oversight from team members. Monitoring was conducted monthly by the National University of Rwanda School of Public Health to provide feedback and response on distribution and adherence. In addition, the communication materials and the supply and distribution system were reviewed following the intervention period to further identify constraints to operationalization.

2.4. Phase 4: Scale‐up

Results from the first three phases were combined, analysed, and used to design multiple components of an integrated strategy for the scale‐up of home fortification with MNP and IYCF programming in 19/30 districts of Rwanda. Protocols were developed for (a) training of service providers, (b) monitoring and evaluation of the supply chain system and distribution model, and (c) social mobilization and behaviour change communication in the context of the national “1,000 days campaign” initiative and the District Plans for Elimination of Malnutrition.

To scale‐up the integration of MNP into national programmes with appropriate supervision, partnerships were formed with seven implementing agencies—international non‐governmental organizations already operating in the 19 districts. With support of these partners and district health centres, trainings on MNP usage as a component of IYCF, particularly complementary feeding, sanitation, and hygiene practices, were standardized and implemented using a cascade model.

Implementation in 19 districts was staggered over 15 months to allow for capacity building in regards to training and supportive supervision. Distribution occurred with cartons of MNP delivered to district hospitals by UNICEF. District hospitals and implementing partners distributed supplies to health centres, where CHWs picked up the MNP for their communities. Distribution, at no‐cost to caregivers, took place monthly during routine growth monitoring activities that included cooking demonstrations with a variety of ingredients sourced by participants. Boxes of 30 MNP sachets were provided to caregiver every 3 months with instructions to add MNP to their child's food approximately two to three times per week (Ip et al., 2009). This regimen largely followed the adherence schedule of the pilot but reduced contact with CHWs from once a month to once every 3 months so as to not overburden caregivers and CHWs. The demonstrations gave CHWs the opportunity to deliver key messages on the appropriate use of MNP and a platform for reinforcing key IYCF messages on timely introduction, quantity, variety, consistency, and frequency of complementary foods. Caregivers were provided with an opportunity to practice mixing MNP with appropriate foods and feed their child in a supportive environment with bowls and spoons provided. Distribution reports, with total eligible children and total receiving MNP, were submitted by CHWs at month‐end, compiled at health centres, and then district hospitals.

At the end of Phase 4, a programme review was completed by a small UBC‐directed team. The team completed interviews with caregivers, programme providers (CHW and health centre staff), and programme administrators (district hospital and implementation partner staff). In‐depth qualitative interviews with 120 programme providers and administrators were solicited in all implementing districts. The supply chain components were assessed for efficiency and feasibility, and the length and content of training was re‐evaluated as was monitoring and reporting.

In addition, a small‐scale quantitative survey to assess caregiver practices (including coverage, utilization, and adherence) and acceptability was completed with 186 caregivers of children who had utilized MNP for at least 12 months. Eligible caregivers were identified with the help of local CHWs from two randomly selected villages in each of the 19 implementing districts, for a total of 38 villages. Five random numbers were generated within the range of eligible children for each village and corresponding caregivers selected for an interview with the intention of sampling 10 children in each district. The sample size was determined based on available funds.

2.5. Statistical analysis and ethics

A mixed methods approach was used throughout the research components of the programme, including focus group discussions, key informant interviews, surveys, and stock reports from health facilities and pharmacies. Qualitative assessment techniques allowed for open‐ended responses that captured knowledge, attitudes, and practices, whereas quantitative assessment was used to standardize data and measure progress and make comparisons across time and by group.

For the qualitative data analysis, the framework approach was used (Ritchie & Spencer, 2002). Data were imported and managed with NVivo software, with hand‐coding for smaller data sets. The data were examined with multiple iterative processes, and visualizations were used to describe dominant themes.

All quantitative data from caregiver questionnaires, anthropometric, and haemoglobin measurements were reviewed on location by UBC's research team before being entered into Excel and converted to SPSS for Macintosh V 22 for statistical analysis. Caregivers lost to follow‐up were not replaced. Indicators of dietary diversity, meal frequency, and dietary adequacy were defined according to WHO standards (WHO, 2008). Instances of child diarrhoea, fever, and/or cough within the past 2 weeks were reported by the caregiver. Anthropometric indicators, including length‐for‐age z‐score, weight‐for‐age z‐score, and weight‐for‐length z‐score were calculated using the 2006 WHO Growth Standards (United Nations Children's Fund & World Health Organization, 2009). Continuous variables were examined for significant outliers using boxplots. Extreme outliers in the anthropometric data that were deemed biologically implausible were removed from the data set. Descriptive statistics were used to summarize characteristics at baseline. This included the calculation of means and standard deviations for continuous variables and frequencies for categorical variables. Independent sample t tests and Pearson chi‐square tests were used to compare differences between groups.

Mean haemoglobin concentrations after 6 and 12 months of follow‐up were compared between groups using analysis of covariance. The occurrence of child diarrhoea, fever, and cough was also compared between groups after 6 and 12 months of follow‐up using multivariate logistic regression modelling. Multiple logistic regressions were used to determine whether treatment with MNP influenced anaemia status (yes/no), stunting status (yes/no), and 2‐week prevalence of illness (diarrhoea, fever, and cough) at both 6 and 12 months.

The Rwandan National Ethics Committee and UBC's Behavioural Research Ethics Board granted approval for the all components of the four phases, including interviews, surveys, measurements, and participation in the 30‐day trial and pilot study. Participation was voluntary, and caregivers were allowed to withdraw from the study at any time. Prior to enrolment, informed consent was obtained from the caregivers of children participating in the TIPs and pilot.

3. RESULTS

3.1. Phase 1: Feasibility study

The focus group discussions and key informant interviews revealed poor nutrition knowledge and suboptimal IYCF practices by caregivers. Importantly, appropriate food vehicles for MNP in the local context were identified, as well as preferred communication methods, and trusted sources of information in the community, as shown in Box 1.

Box 1: Key findings of Phase 1.

• Identified poor nutrition knowledge and IYCF practices with implications for child growth and micronutrient deficiencies:

  • First, foods were typically watery porridge made from sorghum, maize, or cassava, served as a drink from a cup.
  • Knowledge and consumption of energy‐dense foods and foods containing iron was low due to lack of access, affordability, cultural practices, and understanding of their importance, for example, animal foods, fats, and oils.
  • Little knowledge of anaemia. Red‐coloured fruits (tree tomatoes) were believed to benefit blood.
• Determined appropriate food vehicles for MNP in the cultural context as

  • Thick porridge, mashed bananas, beans, and “Irish” potatoes, with vegetables (amaranth leaves), ground nut paste, eggs, small dried fish, oil, and meat identified as foods with which to enrich basic staples (where available).

• Identified signs of a healthy child for key messages including active, growing well, and good appetite

• Assessed cultural acceptability of MNP:

  • Caregivers expressed support, emphasizing the need for clear and accurate messages, knowledge of product composition, and potential side effects, delivered by well‐trained health staff.
  • Established volunteer CHWs, health centre staff, and community leaders as trusted sources of information.

• Radio was the preferred method for mass communication.

3.2. Phase 2: TIPs

The TIPs provided highly relevant information on specific complementary feeding and hygiene practices along with obstacles and motivations for changing behaviours (Box 2). Interviews at the end of the 30‐day trial showed that most caregivers had been adding the MNP to their child's food, with an average of 14/15 sachets used.

Box 2: Key findings of Phase 2.

• Inadequate IYCF knowledge and practices and high levels of morbidity at baseline included

  • Introduction of complementary foods earlier (5 months) or later (8 months) than recommended 6 months, infrequent meals, complementary foods of watery consistency, lack of active feeding, less than 33% washed hands with soap before preparing food or feeding children.
  • Most believed green vegetables and red fruit were high in iron.
  • Percentage of caregivers who reported children with the following in the previous 2 weeks: diarrhoea, 50%; fever, 42%; and cough 65%.
• High adherence, utilization, and acceptability of MNP at the household and community level after 30 days.

  • Cooking demonstrations proved useful as caregivers initially observed preparation and their child's consumption of a variety of foods with cooking oil and MNP added.
  • Caregivers were able to repeat most of the instructions and messages on MNP from counselling. More emphasis needed on handwashing prior to preparation and feeding.
  • Of the 56 caregivers followed‐up on Day 30, 55 stated they had been adding MNP to their child's food over the course of the trial.
  • Majority added MNP to midday meal of mixed soft “family foods,” for example, mashed potatoes and beans.

3.3. Phase 3: Pilot study

Selected characteristics of the pilot study population at baseline are provided in Table 1. Both intervention and comparison groups received IYCF counselling, and there were no differences at endline in complementary feeding practices. Anthropometric measures (reported elsewhere), notably stunting, did not differ between groups or across time (Michaux, 2015). A significant improvement in haemoglobin concentrations was observed among children who were anaemic at baseline and received MNP compared with those who were anaemic but did not receive MNP (Table 2). Children receiving MNP had a lower reported occurrence of cough after 6 months of treatment and fever after 12 months of treatment compared with those in the comparison group (Table 3). Problems with supply and distribution were identified as key issues in an overview of other outcomes and challenges presented in Box 3, limiting the potential impact.

Table 1.

Baseline characteristics of children enrolled in study (N = 1,066)

	Comparison group	n	MNP group	n	P value
Child's age, months	9.1 (1.9)	486	8.9 (1.8)	580	0.040
Child's sex, female	235 (48.4)	486	276 (47.6)	580	0.803
Still breastfeeding	480 (98.8)	486	571 (98.4)	580	0.661
Met minimum dietary diversitya	190 (39.6)	480	210 (36.8)	571	0.351
Met minimum meal frequencyb	302 (62.9)	480	315 (55.2)	571	0.011
Met minimum acceptable dietc	134 (27.9)	480	145 (25.4)	571	0.356
Child's Hb (g L−1)	102 (13.3)	486	103 (13.3)	580	0.125
Any anaemia (Hb < 110 g L−1)	354 (72.8)	486	393 (67.8)	580	0.081

Note. Data are expressed as mean ± SD for continuous variables and as N (%) for categorical variables. P values are for differences between comparison and MNP groups, examined using two‐tailed independent‐samples t tests and Pearson chi‐square, as appropriate. Haemoglobin (Hb) values are adjusted for an average elevation of 2,000 m above sea level (−8 g L⁻¹). Minimum meal frequency, minimum dietary diversity, and minimum acceptable diet exclude nonbreastfed children (n = 15) as the appropriate questions for formula‐fed or nonbreastfed children were not asked in the questionnaire. MNP: micronutrient powder; SD: standard deviation.

^a Proportion of breastfed children who were fed ≥4 food groups in the previous 24 hr. Based on mothers recall. Food groups include grains, roots, and tubes; legumes and nuts; dairy products; flesh foods (meat, fish, poultry, and liver/organ meats); eggs; vitamin A‐rich fruits and vegetables; and other fruits and vegetables.

^b Proportion of breastfed children who received solid, semi‐solid, or soft foods then minimum number of times or more in the last 24 hr; minimum number of times varies with age: two times if 6–8 months and three times if 9–23 months. Based on mothers recall.

^c Proportion of breastfed children who met the requirements for both the minimum dietary diversity and minimum meal frequency indicators.

Table 2.

Haemoglobin (Hb) concentration at 6 and 12 months of treatment compared between treatment groups for all children and only children who were anaemic at baseline

	Comparison group	n	MNP group	n	Mean diffa (95% CI)	P valueb
All children
Hb (g L−1) at 6 months	108 (0.6)	454	109 (0.6)	531	1.3 (−0.4, 3.0)	0.124
Hb (g L−1) at 12 months	113 (0.6)	450	114 (0.6)	531	0.5 (−1.2, 2.2)	0.579
Anaemic childrenc
Hb (g L−1) at 6 months	106 (0.7)	329	109 (0.7)	358	2.7 (0.6, 4.7)	0.010
Hb (g L−1) at 12 months	112 (0.7)	324	114 (0.7)	357	2.4 (0.4, 4.4)	0.018

Note. Children were 6 to 12 months of age at the beginning of the study, 12 to 18 months after 6 months of treatment, and 18 to 24 months after 12 months of treatment. Data are presented as adjusted means (SE) and mean difference (95% CI). Data were excluded on a pairwise basis. Group differences were compared using ANCOVA with the following baseline characteristics included as covariates: Hb concentration, sex of child, age of child (months), income generating activity (y/n), household has agricultural land (y/n), household has vegetable garden (y/n), child received prelacteal feeds (y/n), child met minimum meal frequency (two meals if 6–8 months and three meals if 9–23 months; y/n), child had diarrhoea in previous 2 weeks (y/n), child had cough in previous 2 weeks (y/n), child had fever in previous 2 weeks (y/n), LAZ, WHZ, and underweight (WAZ < −2SD). Haemoglobin (Hb) values are adjusted for an average elevation of 2,000 m above sea level (−8 g/L). ANCOVA: analysis of covariance; CI: confidence interval; LAZ: length‐for‐age z‐score; MNP: micronutrient powder; WAZ: weight‐for‐age z‐score; WLZ: weight‐for‐length z‐score.

^a Values for mean diff are for mean Hb concentration (g L⁻¹) in MNP group minus mean Hb (g L⁻¹) concentration in the comparison group (g L⁻¹).

^b Adjusted for multiple comparisons. Each test would be considered significant at P < 0.05.

^c Results are for children who were anaemic at baseline (Hb concentration < 110 g L⁻¹).

Table 3.

Multivariate logistic regression examining the effect of treatment on the prevalence of diarrhoea, fever, and cough in all children at 6 and 12 months of treatment

		95% CI for OR
	ORa	Lower	Upper	P value
Six months of treatment
Diarrhoea in last 2 weeks
MNP group	0.77	0.59	1.02	0.069
Comparison group (reference)	1.0	—	—	—
Fever in last 2 weeks
MNP group	0.81	0.62	1.06	1.29
Comparison group (reference)	1.0	—	—	—
Cough in the last 2 weeks
MNP group	0.67	0.51	0.88	0.004
Comparison group (reference)	1.0	—	—	—
Twelve months of treatment
Diarrhoea in last 2 weeks
MNP group	0.82	0.61	1.10	0.181
Comparison group (reference)	1.0	—	—	—
Fever in last 2 weeks
MNP group	0.71	0.54	0.93	0.013
Comparison group (reference)	1.0	—	—	—
Cough in the last 2 weeks
MNP group	0.77	0.58	1.01	0.062
Comparison group (reference)	1.0	—	—	—

Note. Children were 6 to 12 months of age at the beginning of the study, 12 to 18 months after 6 months of treatment, and 18 to 24 months after 12 months of treatment. Values are odds ratios and 95% CI. Adjusted for sex of child, age of child (months), income generating activity (y/n), household has agricultural land (y/n), household has vegetable garden (y/n), child received prelacteal feeds (y/n), child met minimum meal frequency (two times if 6–8 months and three times if 9–23 months; y/n), child had diarrhoea in previous 2 weeks (y/n), child had cough in previous 2 weeks (y/n), child had fever in previous 2 weeks (y/n), WHZ, LAZ, and underweight (WAZ < −2SD). CI: confidence interval; LAZ: length‐for‐age z‐score; MNP: micronutrient powder; OR: odds ratio; WAZ: weight‐for‐age z‐score; WLZ: weight‐for‐length z‐score.

^a Odds ratio is Exp β value from SPSS.

Box 3: Key findings of Phase 3.

Perceived Outcomes and Acceptability

• Perceived improvements in appetite and activity level were reported at 6 and 12 months among caregivers of children who received MNP.

  • Eighty‐five percent of caregivers with children who received MNP reported increases in their child's appetite and 88% reported their child was more active at 12 months.

• All caregivers in the intervention group stated that family and community members were supportive of MNP, and 99% stated that they would continue giving MNP to their child.

Implementation Challenges

• A suboptimal number of MNP sachets were received by caregivers at 6 months (median number of sachets reported was 30 instead of intended 60). Supply and distribution were major obstacles as many caregivers received the full 120 sachets only near the end of the pilot. MNP cartons were not reaching district hospitals and health centres in a timely manner, and changes were made to ensure a continuous supply for scale‐up.

• Stock issues were prevalent, especially in the first half of the pilot. Overstocking and poor storage conditions led to product expiring or spoiling, whereas in other areas, stock‐outs occurred limiting coverage.

• Monitoring did not occur in a timely manner leading to delays in response to programme issues; monitoring tools were too long and complicated for community use. Supportive supervision at the community level was inadequate and needed to be enhanced to provide CHWs with timely feedback.

• Anecdotal reports were made about MNP causing diarrhoea and vomiting as side effects. The effect of rumours was mitigated by prompt responses from implementing partners, through community sensitization, and assurance from health workers.

• The lack of an overall communication strategy in the districts led to inconsistencies in messaging; the need was identified for a CHW reference guide with responses to frequently asked questions.

3.4. Phase 4: Scale‐up

Distribution data submitted monthly by CHWs to health centre staff and aggregated at the district level during Phase 4 saw the number of children receiving MNP increase from 13,556 in March 2015 to 272,001 in June 2016 (Figure 2). Coverage within the 19 implementing districts was 87% with a general upward trend since scale‐up began in July 2015. A year into Phase 4, 50 hospital staff, 344 health centre staff, and 29,326 CHWs had been trained on home fortification with MNP. An estimated 575,000 caregivers had participated in MNP counselling and received a box for their child.

Figure 2.

Coverage of micronutrient powder (MNP). Shows quantity and rates of MNP coverage in graph (bar/line) format

The small‐scale survey of 186 caregivers from the 19 districts revealed high rates of acceptability and knowledge as well as adherence and utilization. In total, 95% of caregivers stated they would recommend MNP to others and the strong majority reported seeing an increase in their child's appetite (86%) and activity (73%). The recommendation to use two to three sachets per week was adhered to by 75% of caregivers surveyed, and there was evidence suggesting positive improved IYCF practices with 92% of caregivers having mixed MNP with two or more food groups at the most recent meal. Just over 25% reported their child had experienced diarrhoea during the previous 2 weeks. Implementation of the programme continued to pose challenges during scale‐up with issues pertaining to stock‐outs, delivery, and reporting among partners.

4. DISCUSSION

In this paper, we documented the experience of planning, implementing, and scaling‐up an MNP programme in Rwanda using a multiphased strategy spanning a 5‐year period (2011–2016). Strong government support and a well‐developed network of health centres within this densely populated and small country, likely minimized some of the operational challenges that would be faced elsewhere in countries with larger populations and less capacity and support.

Although implementation research in nutrition programming is still nascent, and the situation in Rwanda unique, our approach addressed several recently identified gaps in the literature related to the implementation and scale‐up of MNP interventions (Nyhus Dhillon et al., 2017; Reerink et al., 2017; Vossenaar et al., 2017). One of the key gaps in MNP implementation addressed by this work is the description of approaches, methodologies, and tools for basic formative research to inform communication strategies around MNP that focus on context‐specific messaging around complementary feeding and positive behaviour change practices that can improve adoption (Schauer et al., 2017; Tumilowicz et al., 2017). We learned that prior to full‐scale implementation, a phased introduction to address potential barriers and garner support among implementers and end‐users is a critical component of success. The concepts of “trialability” and progressive adaptation around an “efficiency core” were central to the multiphase strategy presented in this paper as a prerequisite to effective integration into health platforms. “Trialability,” a term used by the Consolidated Framework for Implementation Research, is a key construct of implementation and “promotes successful adaptation of the intervention” (Damschroder et al., 2009), whereas the “efficiency core” of MNP that includes their low cost, ease of use, and high vitamin and mineral content would be predicted to facilitate uptake by end‐users (Zlotkin et al., 2015).

The contextualized adaptation for the Rwanda MNP programme included developing country‐specific packaging and communication materials in the local language with appropriate imagery, created and tested over the initial phases. Training modules for health workers and caregivers, modified after each phase, likely resulted in improved complementary feeding practices in keeping with WHO recommendations for diversity and consistency of young children's foods (WHO, 2004).

In addition to adapting the intervention to local context, the multiphased strategy allowed for controlled interaction with the social, economic, and political forces crucial for programmatic success. Early implementation of MNP is often hindered by rumours of an association between MNP and common illnesses (Creed‐Kanashiro, Bartolini, Abad, & Arevalo, 2016; Kodish et al., 2011; Soofi et al., 2013). Access to the product alone, without a well‐informed communication strategy, is not enough to achieve continual and appropriate use. Regular ongoing support is likely necessary for successful implementation (Kodish et al., 2011; Reerink et al., 2017; Vossenaar et al., 2017).

By conducting the initial phases of implementation with supportive supervision, careful monitoring of community perceptions was possible, and the product had adequate time to achieve positive brand recognition. That said, this comes at a cost, and Rwanda had considerable support, including funded local and external staff, as well as in‐kind support from UBC.

From in‐depth interviews with health centre, district, and implementing partner staff a year into scale‐up, a recurring theme was that the programme's acceptance had improved as more caregivers were endorsing the product by word of mouth, sharing their perceptions of improved child health. Although this bodes well for long‐term success, it is noteworthy that the survey conducted during scale‐up revealed that 25% of caregivers attributed their child's diarrhoea to MNP, despite no reported differences in diarrhoea between groups in the pilot phase. Caregiver access to knowledgeable support staff is thus crucial during the first rounds of distribution but is also still required after the programme has been established. Although individual consultation may be ideal, it is cost prohibitive in most contexts and the benefits of more feasible group counselling to improve IYCF practices have been documented in a number of studies (Flax et al., 2014; Nguyen et al., 2014; White et al., 2016).

The MNP programme had consistently high acceptance across the phases, and improved caregiver practices were observed during scale‐up. In terms of the economic and political context, the multiphased strategy enabled effective advocacy by providing data and information to key decision makers and creating an enabling environment for implementation. Early formative research meant that stakeholders had relevant information before making a major financial investment. While previous research in various countries determined the “efficacy” of MNP (De‐Regil et al., 2013), sustaining support at the national level required a demonstration of the “effectiveness” in the Rwandan context. By demonstrating a significant reduction in anaemia in children during the pilot study (Michaux, 2015), we were able to successfully increase the relative priority of the programme and further engage national leadership, thereby strengthening the structural and political context—or “inner setting”—for implementation (Damschroder et al., 2009; Schauer et al., 2017).

Integration within the existing health platform began during Phase 3 by consolidating MNP activities within the “1,000 days campaign” and using the existing CHW network to counsel caregivers and provide follow‐up support. As the MNP programme and the “1,000 days campaign” have overlapping target populations of caregivers of children 6 to 23 months, there was a high degree of programme compatibility. Both programmes prioritized decentralization of health services and had stakeholders receptive to cooperation. These two characteristics have been found to be key components of successful integration (Vossenaar et al., 2017; Wallace, Dietz, & Cairns, 2009). The supply chain process for MNP for Phases 3 and 4 used a system parallel to that of the existing national supply chain for medicines and nutrition‐related products. The parallel system utilized UNICEF and implementing partners for delivery to health centres, where health care staff and CHWs then distributed and monitored coverage. Formal integration of MNP into the existing national supply chain did not occur until after the conclusion of the scale‐up phase. Integrating a new product into a national supply chain system, along with corresponding reporting mechanisms, is a lengthy process and can be difficult to modify; the parallel system allowed for greater flexibility and responsiveness during the scale‐up phase.

Although use of CHWs and health care staff made MNP delivery feasible, integration and adaptation required consistent monitoring and refinement throughout the pilot phase to offset capacity gaps in knowledge. During Phases 3 and 4, we found that CHWs could not remember all the correct steps to mixing MNP and struggled with the reporting tools they were required to complete. Health centre staff involved in delivering MNP had trouble differentiating between cumulative and monthly reporting, and often reported consumption instead of distribution, making it appear as though stock levels were higher than they were. These problems led to stock‐outs, which in turn reduced coverage. As noted by Schauer et al. (2017), weaknesses in pre‐existing systems and programmes are difficult challenges to overcome without technical support as part of long‐term financing.

In response to findings on low CHW knowledge retention and difficulties with reporting, increased supportive supervision and refresher training was provided in districts encountering challenges, and modifications to programme reporting were made to increase ease of use. Our experience confirms previous findings that adding responsibilities within existing systems requires not only initial training but also regular refresher training and follow‐up during the integration process (Schauer et al., 2017; Vossenaar et al., 2017; Wallace et al., 2009). Issues associated with reporting and supply chains affected data collection and thus estimates of coverage, especially prior to refresher training. Although Zlotkin et al. (2015) and others (Nyhus Dhillon et al., 2017; Vossenaar et al., 2017) recommend using a robust system for monitoring and evaluation, this is not always possible and is particularly challenging when the reporting system is being piloted in addition to the programme itself. Health workers, including CHWs, are generally overburdened and have varying degrees of motivation, particularly as the latter are generally low paid or unpaid (Reerink et al., 2017).

The multiphased intensive strategy presented in this paper has created a robust foundation in Rwanda for long‐term success, but consideration must be given to the challenges we identified. The process we developed is feasible and adaptable to low infrastructure settings but emphasizes the need for continuous feedback and input from end‐users, health personnel, government, and other partners. Issues that are identified early on while still manageable in scope can be addressed before the next phase and expansion of the scale‐up process. The Phase 4 scale‐up saw increasing numbers of children receiving MNP, with high rates of acceptability. An increase in coverage, adherence, and utilization would be expected to continue if integration within existing health platforms is achieved, and community acceptability stabilizes on positive perceptions. That said, long‐term costs of scale‐up should not be underestimated as the need for refresher trainings, supportive supervision, and responsive monitoring within systems that are overburdened and have high staff turnover will be ongoing for the foreseeable future.

The study period for the MNP programme in Rwanda ended in June 2016 with full‐scale implementation in 19/30 districts. Subsequently, the programme continued expansion and in 2017 was implemented within all 30 districts, making Rwanda the first country in Africa to achieve full national scale‐up of an MNP programme. In addition, the MoH and UNICEF collaborated to entrench the programme into existing health platforms. This included delivering training on home fortification with MNP as part of the Maternal Infant and Young Child Nutrition curriculum, integrating MNP with all nutrition commodities into the national supply chain for medicines and the logistics management information system and including MNP indicators in the online health information systems. The continued work towards complete integration of the MNP programme into the existing health platform in Rwanda is evidence of the value of the investment (time and money) required for effective programme implementation.

5. CONCLUSION

We have described the successes in our phased approach to scaling up MNP in Rwanda, addressing a number of gaps in the literature (Nyhus Dhillon et al., 2017; Vossenaar et al., 2017). Although funding was intermittent, it was adequate to conduct the formative research, trial, and pilot projects needed to establish the programme within this small country and develop methodologies relevant for use in different contexts.

The phased approach to implementation as described in this manuscript is not specific to Rwanda. Detailed planning is important for the successful initiation and scale‐up of most nutrition interventions, independent of geography. New or at‐scale efforts to integrate MNP into ongoing IYCF services would likely benefit from the learning outcomes of the Rwandan experience. The Government of Rwanda remains strongly committed to eliminating malnutrition in all its forms, which bodes well for the continued success of the MNP programme and ultimately a transition from donor funding to sustainable financing.

CONFLICTS OF INTEREST

The authors declare that they have no conflicts of interest.

CONTRIBUTIONS

JM and CM designed the research; MN oversaw the final phase and conducted the data collection, analysis, and reporting on the scale‐up and wrote the first draft of the paper with the support of RJR. JM, MN, and RJR had primary responsibility for the final content, whereas SZ, CS, and CM provided important intellectual content. All authors read and approved the final manuscript.

McLean J, Northrup‐Lyons M, Reid RJ, et al. From evidence to national scale: An implementation framework for micronutrient powders in Rwanda. Matern Child Nutr. 2019;15:e12752 10.1111/mcn.12752