Improving nutritional status among urban poor children in sub‐Saharan Africa: An evidence‐informed Delphi‐based consultation

Abstract In sub‐Saharan Africa (SSA), rapid urbanisation coupled with the high prevalence of infant and young child (IYC) undernutrition in low‐income settings means that interventions to support IYC nutrition are a priority. Little is known about how urbanisation influences IYC feeding (IYCF) practices, and evidence‐based interventions to improve IYC health/nutrition in the urban poor are lacking. Therefore, this research aimed to (a) systematically review evidence on interventions for improving the nutritional status of IYC aged 6–23 months living in urban poor areas (PROSPERO CRD42018091265) and (b) engage stakeholders to identify the highest ranking evidence gaps for improving IYCF programmes/policies. First, a rapid systematic review was conducted. This focused on the literature published regarding nutrition‐specific and nutrition‐sensitive complementary feeding interventions in urban poor areas, specifically low‐income informal settlements, in low‐ and middle‐income countries (LMICs). Six intervention studies met the review inclusion criteria. Intervention adherence was generally high, and indicators of maternal knowledge and IYC nutritional intake typically increased because of the interventions, but the impact on anthropometric status was small. Second, stakeholders working across SSA were engaged via a Delphi‐based approach to identify priority areas for future intervention. Stakeholders reported that a situational analysis was required to better understand IYCF in urban poor areas, particularly the causes of IYC undernutrition, and highlighted the need to involve local communities in defining how future work should proceed. Together, these findings indicate a need for more evidence regarding IYCF and the factors that drive it in urban poor areas across LMIC settings, but particularly in SSA.


| INTRODUCTION
Stunting, caused by long-term undernutrition, is a significant global health concern (UNICEF, 2020; World Health Organization [WHO], 2012). Stunted children experience impaired growth and development with lifelong consequences on morbidity and mortality (Black et al., 2008). Suboptimal complementary feeding, including inappropriate timing of introduction of semi-solid and solid foods, is associated with stunting in low-and middle-income countries (LMICs) (Marriott, White, Hadden, Davies, & Wallingford, 2012). Currently, less than one in five children in lowand lower middle-income countries receive a minimum acceptable diet (MAD; a composite indicator composed of minimum meal frequency [MMF] and minimum dietary diversity [MDD]) (UNICEF, 2019). Across sub-Saharan Africa (SSA), the proportion of children receiving the MAD during complementary feeding has been low and is declining. For instance, Demographic and Health Survey (DHS) data (2010-2013) from 10 SSA countries revealed that less than 18% of 6-23-month-olds met the MAD (Na, Jennings, Talegawkar, & Ahmed, 2015). Between 2008 and 2014, the proportion of 6-23-month-olds who received the MAD decreased from 39% to 22% in Kenya (Kenya National Bureau of Statistics [KNBS], 2015) and from 19% to 8% in Malawi (National Statistical Office, 2011;National Statistical Office, 2017), indicating a significant worsening of diet. In Tanzania, DHS data showed a decline of both MDD and MAD between 2004/2016 in 6-23-month-olds from 46% to 30% and 16.6% to 6% respectively, while MMF remained constant (Ogbo, Ogeleka, & Awosemo, 2018).
Poverty and poor living conditions in LMICs are also associated with stunting. In 2019, two out of five stunted infants and young children (IYCs) globally lived in SSA, which is significantly higher than the global average of one in five (21.3%) children under 5 years who are stunted (UNICEF, 2020). Children living in slums are more likely to experience undernutrition than other urban children (Ahsan, Arifeen, Al-Mamun, Khan, & Chakraborty, 2017;Unger, 2013). SSA is experiencing rapid urbanisation and nutrition transition with consequent changes in the social and physical environments and shifts in food habits and practices. By 2050, the African continent will have 1.5 billion urban dwellers, equal to 22% of the world's urban population (United Nations [UN], 2019). The urban population in SSA is increasing through high rates of rural-to-urban migration, natural population growth and the reclassification of towns to urban centres. The spread of urban low-income informal settlements and undernutrition have increased alongside rapid urbanisation (Awumbila, 2017;Fotso et al., 2012;Nickanor & Kazembe, 2016).
It is estimated that interventions to promote optimal IYCF practices could prevent 20% of deaths in under 5 years in countries with high child mortality (Alderman, Nguyen, & Menon, 2019;Jones et al., 2003;Kramer & Kakuma, 2004. Optimal feeding practices include breastfeeding initiated within 1 h of birth, exclusive breastfeeding for 6 months after birth, and introducing safe, nutritionally adequate complementary solid food at 6 months (alongside continued breastfeeding up to 24 months or longer) (WHO, 2020).
Current approaches to improve nutritional status include both nutrition-sensitive interventions, which address the underlying determinants of nutrition (e.g., agriculture and food security, early child development, maternal mental health, health and family planning services, and water, sanitation and hygiene [WASH]), and nutritionspecific interventions, which address proximal determinants of nutrition (e.g., adequate food and nutrient intake, feeding, caregiving and parenting practices, and low burden of infectious diseases) (Ruel et al., 2013).
A Cochrane systematic review of nutritional interventions for preventing stunting in children under 5 years in poor urban areas in LMICs showed that nutrition-specific interventions have the potential to decrease stunting in nonslum contexts, but this efficacy is inconsistent when these interventions are applied in urban slums (Goudet et al., 2019). This finding was attributed to environmental challenges associated with conducting interventions in these settings, including the lack of social services, high mobility and attrition rates (Goudet et al., 2019). Although this comprehensive review of nutrition-specific interventions exists for children under 5 years, no review has yet considered the combined evidence from nutritionspecific and nutrition-sensitive interventions during the complementary feeding period (6-23 months of age), which is of critical importance for the promotion of optimal growth, health and

Key messages
• Intervention studies to improve nutritional status in infants and young children (IYC) living in urban lowincome informal settlements or slums in low-and middleincome countries, including sub-Saharan Africa (SSA), have focused primarily on outcomes related to anthropometric status.
• Future complementary feeding interventions should include an assessment of diet quality.
• Nutrition-sensitive factors should be considered when developing and evaluating complementary feeding interventions in urban low-income informal settlements or slums; this will help to maximise the potential impact and effectiveness of nutrition-specific interventions.
• Multisectoral approaches are needed to improve the nutritional status of urban poor children of complementary feeding age in SSA. development. Furthermore, there has been no specific focus on characterising diet quality during the complementary feeding period; this early life risk factor influences childhood growth, and is a pathway to the development of chronic diseases in later life (Lynch & Davey-Smith, 2005).
It is important to develop interventions that align with the current priorities of local, national and international stakeholders in relation to IYCF programmes, alongside gathering evidence from intervention studies to address IYC undernutrition. Combining evidence-based research with multisectoral policy and planning with stakeholders to identify platforms for the delivery of community-based interventions is more likely to lead to successful and sustainable health improvements for IYC (Bhutta et al., 2013).
Indeed, engagement across different sectors and stakeholders is recognised by the WHO (2018) as an important approach to help achieve a 40% reduction in stunting in IYC under 5 years by 2025.
In summary, given the rapid urbanisation in SSA and the high prevalence of undernutrition, the nutritional status of IYC living in urban low-income informal settlements is an important area for improvement. However, such improvements require evidence regarding the likely success of nutrition-related interventions in the most vulnerable age group of IYC transitioning to complementary feeding in urban poor environments, as well as input from stakeholders to identify their priority areas for intervention. Therefore, we aimed to (a) systematically review current evidence in a rapid review of nutrition-specific and nutrition-sensitive interventions for improving the nutritional status of IYC aged 6-23 months living in urban poor areas (specifically low-income informal settlements or slums) in LMICs and (b), through stakeholder engagement, identify the highest ranking evidence gaps for improving IYCF programmes and policies (focusing on complementary feeding) in children under 24 months in SSA.

| Rapid review
A rapid review approach, with a simplified systematic review process, was used to obtain context-sensitive knowledge in a shortened timeframe (Grant & Booth, 2009;Khangura, Konnyu, Cushman, Grimshaw, & Moher, 2012). The rapid systematic review protocol was registered on PROSPERO (registration number: CRD42018091265). The search was originally conducted to review and synthesise evidence on nutrition-sensitive and nutrition-specific interventions to improve the nutritional status of IYC in urban lowincome informal settlements in SSA. After screening titles, abstracts and full texts for inclusion (completed in May 2018), this search yielded one eligible article. Following consultation with stakeholders (government, academic and NGO representatives) at a meeting held in Nairobi, Kenya, in June 2018, the protocol was modified by expanding the search to include all LMICs because of the lack of evidence in SSA. The revised protocol was updated and reregistered with PROSPERO (registration number as above) in June 2018. For this review, LMICs were defined using the World Bank Group (2019) classification scheme (based on gross national income per capita). The rapid review was designed to follow on from a scoping review from some of the authors that examined risk factors for malnutrition in slums and the impact of interventions on children's health in LMICs (Goudet, Griffiths, Bogin, & Madise, 2017). Key differences between the current protocol and the previous scoping review were to (a) include nutrition-sensitive as well as nutrition-specific interventions; (b) focus on IYC aged 6-23 months as opposed to all children under 5 years; (c) focus on diet quality outcomes, in addition to anthropometric outcomes.
The search dates were set to identify new studies published since the scoping review search was completed in December 2013 (Goudet et al., 2017).

| Inclusion/exclusion criteria
Inclusion criteria for studies were based on the Population, Intervention, Comparison(s) and Outcome (PICO) structure (Thomas, Kneale, McKenzie, Brennan, & Bhaumik, 2019): (a) IYC aged 6-23 months in urban low-income informal settlements or slums in LMICs; (b) nutrition-specific or nutrition-sensitive interventions, according to the definitions outlined by Ruel et al. (2013); and (c) quantitative outcomes of nutritional status indicators (length-for-age [LAZ], weight-for-height [WHZ] and weight-for-age [WAZ] z-scores), undernutrition (stunting, underweight, wasting and micronutrient deficiencies) and diet quality (i.e., infant and young child MDD score, MMF and MAD). Eligibility criteria for the year of publication were from January 2014 to July 2018 (date of the search completion). The search, and summary of included studies, was completed prior to the stakeholder consultation (Delphi process) to inform this aspect of the research.
We included studies that specified the location as being a slum or low-income informal settlement in accordance with the UN-Habitat (2004) definition criteria, that is, lacking one or more of the following indicators: access to improved water and sanitation facilities, security of tenure, durability of housing and sufficient living area. When the location of a study was not clearly defined in a published paper, we consulted team members/collaborators who had local knowledge and/or original authors to help us to decide if a study met our requirements for inclusion as an urban low-income informal settlement or slum.

| Search strategy
Four electronic databases were searched: Medline, Scopus, Web of Science and Embase. An example search strategy (Medline) and the list of LMICs included in the search are provided in Data S1. The reference lists of existing systematic and/or literature reviews were also screened to identify further studies for inclusion. Citation follow-up and hand searches were used to identify studies up until January 2019, which yielded one additional study (Smuts et al., 2019).

| Screening, data extraction and analysis
Screening Duplicates were removed before screening. The title, abstract andsubsequently-full-text articles were screened against the inclusion criteria (by EKR, OM, TAZ, JMNC and EK). The review team pretested the screening form with an initial pilot phase of 15 studies. This process improved clarity of the inclusion criteria and consistency among assessors. An independent reviewer (RP) checked a random 10% sample of the excluded records at the title and abstract stages. Full-text screening of articles was conducted by a second reviewer (EKR or OM) for confirmation. Any disagreements on study inclusion were discussed to reach consensus. If this was not possible, a third reviewer was consulted (PLG), and if necessary, authors were also contacted for further information about the study design to assess whether inclusion criteria were met.

Data extraction
A data extraction form was developed, piloted on two studies and then used to collate the data from all included studies. Information extracted from studies included title, author(s), year of publication, type of intervention (nutrition-sensitive or nutrition-specific), details of intervention (type of intervention/supplement, etc.), intervention duration, details of intervention compliance, outcome measures assessed (child nutritional status and nutritional quality of diet, anthropometrics), any other outcome measures related to nutritional status/diet quality/malnutrition (specifically, stunting, underweight, wasting and micronutrient deficiencies), detailed results (list all relevant outcome measures separately), study limitations, and lessons learned in delivering the intervention. Data were extracted by two independent reviewers (NP and OM) and checked by EKR.

Quality assessment process
Two reviewers (MM and EH) independently assessed the quality of the included studies using a 14-item rating tool developed by Kmet et al. (2004). Their ratings were compared, and they came to an agreed rating for each criterion. As gold-standard Cochrane guidance advises against the use of a rating score (Higgins & Green, 2006), a modified assessment classification was used. Instead of rating each criterion as either 0, 1 or 2, a qualitative, colour-coded assessment of low quality/red (high risk of bias), medium quality/yellow or high quality/green (low risk of bias) was used, as has been done successfully in previous research (e.g., Rousham et al., 2020). If all 14 criteria were rated as 'high/green' then a 'high' overall quality assessment was awarded. Papers that had a combination of green and yellow (high/medium quality) ratings received a 'medium' overall quality rating. A 'low' overall quality assessment was given if any of the 14 criteria were assessed as 'low/red'.

| Stakeholder consultation
We built on our findings from the rapid review by conducting a stakeholder consultation to determine the priority areas for improving IYCF programmes and policies in SSA. We adopted a consensus-gathering approach based on the Delphi method (Iqbal & Pipon-Young, 2009) and consulted a range of stakeholders ('panellists') who contributed to three phases of information generation and consensus gathering. Consultation methods included two face-to-face stakeholder workshops (in Nairobi, Kenya, and Lilongwe, Malawi) and a survey that was distributed either online,

| Participants and ethical approval
Participants were individuals with experience of working or conducting research into IYC nutrition in SSA. Purposive sampling was used to identify potential participants. Participants were all aged 18 years or over. Prior to recruitment, ethical clearance was obtained from Loughborough University, from the Amref Health Africa Ethics and Scientific Review Committee (AMREF-ESRC P528/2018) in Kenya and from the National Committee on Research in the Social Sciences and Humanities (No. P.12/18/336) in Malawi. Informed consent was given by all participants who took part in the workshops or completed the survey.

| Delphi consensus-gathering procedure
A modified Delphi method was implemented, comprising three rounds. The Delphi method is interested in the formation or exploration of consensus, and it is exceedingly useful for topic areas where there is limited research because input and ideas come from a range of expert stakeholders (Iqbal & Pipon-Young, 2009). Initially, evidence from the rapid review's synthesis of the recent research evidence identified gaps in knowledge in relation to IYCF, particularly complementary feeding, in LMICs. The research team drew on their knowledge and experience to summarise the research gaps in relation to SSA. These evidence-informed gaps formed the basis of the first round of the modified Delphi approach.

Round 1
Evidence gaps were shared with a range of stakeholders from Kenya and Malawi at a face-to-face meeting in Nairobi, Kenya, in June 2018 (n = 18). Stakeholders were identified by co-investigators in each country based on a list of target sectors (e.g., Ministry of Health, NGOs including practitioners and implementers, policymakers, academics, county government health officials, research institutions, professional networks [e.g., the African Nutrition Society and UNICEF]) to ensure a breadth of views would be represented. These stakeholders were invited to contribute their views about whether these evidence gaps were deemed to be (the most) important, as well as to offer their perspective-and to generate ideas-on where more evidence is needed to improve the nutritional status of IYC living in urban low-income informal settlements or slums in SSA. Stakeholders discussed the questions posed in small groups, facilitated by a discussion leader. Notetakers recorded stakeholders' views and opinions. An example question asked of the stakeholders was What is the evidence needed to enable the formulation and implementation of policies and programmes to improve child nutritional status?
The use of open-ended questions, yielding qualitative data, is recommended for Round 1 because at this stage, the Delphi method focuses on the 'future thinking' on the issue (Iqbal & Pipon-Young, 2009). Notes from the discussions were subsequently used to inform Round 2.

Round 2
Following the stakeholder meeting (Round 1), the information gained was used to create a survey, which was developed and refined by the multidisciplinary research team. An initial pool of survey items was developed, reviewed and refined by the research team. The final survey comprised 57 items, alongside some questions about participants' background to give context on their level of experience. The survey items were divided into three sections: (A) Nutritional status of urban poor families (11 items); (B) What is already known? (18 items); and, (C) What future work is needed, and how can this be done? (28 items).
At the start of each section, participants were reminded of the statements related to the nutritional status of urban poor families and programmes/interventions/policies that might exist to support IYC feeding or nutrition and to think about IYC under 2 years when responding. Participants were asked to indicate the extent to which they agreed/disagreed with each statement, using a 5-point scale. The survey was either completed online or administered as a face-to-face interview with a wide range of stakeholders who worked across SSA to gain their input into the evidence gaps and where further work/research is required. The online survey was distributed via the African Nutrition Society and by sending emails directly to relevant stakeholders. The aim of this round was to determine whether there was broad consensus in respondents regarding the evidence gaps highlighted by the rapid review.

| Round 3
The third and final round was conducted (a) at a second stakeholder meeting held in Malawi in January 2019 and (b) by emailing all of the other respondents (i.e., those not in attendance at the meeting) who had completed the Round 2 survey. Synthesised views from Round 2 were shared with stakeholders for any comments. Stakeholders were asked to select priority areas from the list of future directions.
Six items related to what future work is needed, and how this can be done, when thinking about the causes of (mal)nutrition in urban poor IYC (focusing on undernutrition), and 22 items related to what future work is needed, and how this can be done, thinking about potential solutions to (mal)nutrition in urban poor IYC (Data S2). Respondents were asked to indicate their top two priorities from the list of six items and their top seven from the list of 22 items. Responses were collated and used to determine what stakeholders working in SSA believed were the most important priority areas for future work.

| Data analysis
Descriptive statistics were calculated. Round 1 information was collated, synthesised and used to inform the generation of Round 2 questions.
Consensus in Round 2 was defined as ≥70% of participants either agreeing/strongly agreeing or disagreeing/strongly disagreeing with each statement. The level of agreement was set at ≥70%, as in previous studies (e.g., Diamond et al., 2014;Vogel et al., 2019). All 'neutral' responses were removed prior to calculating the percentage agreement/disagreement to ensure that only those who were confident with their answer were included (as per Vogel et al., 2019). For Round 3, priority areas for future work were identified, and responses were summed across participants to indicate the items with the greatest support.

| Rapid review
The search yielded 3,657 records. After removal of duplicates, 3,067 titles and abstracts were screened ( Figure 1). Of these, 71 records qualified for full-text screening, and 68 records were excluded at the full-text stage. The three remaining records, plus three additional articles identified from reviewing reference lists or journal hand (non-database) searches, met the inclusion criteria and were included in the review.
Our search also yielded three published protocol papers linked to ongoing intervention studies, one of which met our inclusion criteria.
The study of Islam et al. (2018) aims to evaluate the effect of various doses, forms and frequencies of preventative daily zinc supplementation over a 24-week period on reducing diarrhoea and improving growth, specifically LAZ, in IYC aged 9-11 months in the peri-urban low-income area of Mirpur (Dhaka, Bangladesh) (Clinical trials.gov: NCT03406793).

| Included studies
The six eligible studies included four randomised controlled trials

| Description of studies
Of the six studies, two studies were carried out in India (More et al., 2017;Strand et al., 2015), one in Haiti (Iannotti et al., 2014), two in South Africa (Smuts et al., 2019;Tomlinson et al., 2018) and one in Bolivia (Martinez et al., 2018). Three studies involved nutritional supplements: one using vitamin B12 and folic acid supplements (Strand et al., 2015) and the two others administered daily lipid-based nutrient supplements (LNSs; Iannotti et al., 2014;Smuts et al., 2019).
Two studies involved home-based visits: one included promotion of optimal IYCF practices by trained 'Mentor Mother' community health workers (Tomlinson et al., 2018) and one incorporated play-based nutritional education messages (Martinez et al., 2018). Only one study took an integrated community-based approach with the introduction of local resource centres in addition to home visits and day-care centre interventions (More et al., 2017). The primary outcomes of More and colleagues' study were broad, including uptake of family planning/contraception and uptake of child immunisation, as well as treatment of severe acute malnutrition. In Tomlinson et al.'s (2018) study, home visits were conducted as part of an intervention to provide perinatal home visits to women with antenatal depression, so child development and nutritional indicators were not the primary outcomes of the intervention. Further details of the interventions are given in Table 1.

| Quality assessment summary
All six studies achieved an overall quality rating of medium as they all contained a combination of green and yellow ratings across the 14 criteria outlined by Kmet et al. (2004). This indicates that these studies are relatively reliable sources of evidence with a fairly low risk of bias.

| Outcomes of the research studies included in the rapid review
The outcomes, process outcomes/participation rates, compliance details, study limitations and lessons learned from the intervention studies included in this review are summarised in Table 2 Open-labelled, cluster-randomised controlled parallel trial 24-month combined nutrition-specific and nutrition-sensitive intervention in children <5 years at baseline n = 20 clusters were randomly allocated to have a resource centre (intervention group) and n = 20 clusters had no centre (control group).
Each intervention group had a Society for Nutrition, Education and Health Action resource centre, which used approaches including home visits, organised group meetings, day care for malnourished children, community events, service provision and liaising with existing systems. Community intervention centres had a communication emphasis on maternal and neonatal health, child health and nutrition, sexual and reproductive health and prevention of violence against women. Integrated activities included home visits, group meetings, day care, community events, service provision and liaison.
• Double-blinded, randomised, controlled parallel trial 6-month nutrition-specific intervention in IYC aged 6-to 30-months at baseline Participants were randomly allocated to receive a placebo, folic acid and/or vitamin B-12 lipid-based paste. Supplements contained: Twice the RDA of vitamin B12 and/or twice the RDA of folic acid daily for 6 months. The vehicle for the supplementation was a lipid-based paste, which contained 54.1 kcal, 0.71 g protein, 3.31 g fat per 10 g. IYC 6-11 months and ≥12 months were supplemented with 5 g/day (one teaspoon) and 10 g/day paste (two teaspoons), respectively.
The placebo and vitamin supplements were matched in appearance and taste.
• 2) To reduce alcohol use/abuse among pregnant mothers (an adapted, onemeeting evidence-based intervention) 3) To encourage mothers to breastfeed for ≥6 months and to promote healthy food choices for optimal child growth and nutrition.
Mothers in LNS groups were instructed to offer the child one-half of the LNS sachet in the morning and one-half of the sachet in the afternoon. Instructions to mothers included washing hands and the sachet before feeding the LNS to their infant. They were also informed that the LNS should not be used as replacement for other foods.
All groups received messages about appropriate complementary feeding and hygiene practices.
(Continues) Partially blinded, randomised controlled parallel trial 6-month nutrition-specific intervention in children aged 6 months at baseline Infants were randomly allocated to one of two intervention groups (SQ-LNS/SQ-LNS-plus) or a no-supplement (control) group (1:1:1 for group assignments). SQ-LNS/SQ-LNS-plus groups (20 g/day): Parents were asked to incorporate 1 × 20 g sachet daily into their child's first meal, mixed into their usual maizebased complementary diet.
All groups were treated and monitored similarly, but the control group did not receive supplements.
(Continues) supplementation improved WAZ and LAZ in wasted, underweight and stunted children (significant baseline subgrouping variable × intervention interaction; p < .05). Iannotti et al. (2014) found that intake of an LNS for 6 months was associated with a significant increase in LAZ (0.13 z-score, 95% Haiti score and the WAZ by 0.12 ± 0.02 z-score, compared with the control group after adjustment for child age (both p = .02). Compared with the control, the effects of 6-month LNS were sustained 6 months post-intervention for LAZ: 0.10 ± 0.05 z-score (p = .04) and WAZ: 0.11 ± 0.04 z-score (p = .02). 3-month LNS supplementation had a negative effect on LAZ: −0.12 ± 0.05 z-score (p = .03), which was sustained 6 months post-intervention: −0.11 ± 0.05 z-score (p = .03), compared with the control group. There was no impact of 3-month LNS on WAZ: 0.01 ± 0.05 z-score (p = .86), compared with the control group. 74.8% and 71.3% of the participants were followed up to 6 and 12 months, respectively. The sample size had factored a 25% attrition rate. Other than the 3-month LNS group, which had an attrition rate of 35.7% at 6 months, loss to follow up for control and 6-month LNS was within the assumed rate at 24.6% and 25.7%, respectively.
reported consumption of monthly LNS.
In both groups, 6% complied with the twice-daily, split dose LNS consumption protocol.
mean age was evident in the 6-mo LNS group at baseline. This could have confounded the impact of LNS on LAZ, an outcome that is tends to be more influenced in younger IYC.
There was a high loss to follow up due to the high mobility of the population (relocation was the main reason cited for drop out). In the 3-month LNS group lost to follow-up, the mean age of mothers was younger, and children showed a significantly greater WAZ (p < .05). By visit 5 of 7, children who remained in the trial from the 3-month LNS group were younger than those in the control group how nutrition-sensitive factors impact on nutritional status in an urban poor context. Only two included studies were nutrition-sensitive or combined nutrition-specific and nutrition-sensitive approaches (Tomlinson et al., 2018 andMore et al., 2017, respectively). Our review results indicated that studies tended to focus on anthropometric outcomes rather than factors that lead to achievement of nutritional impact (e.g., complementary feeding practices). These findings informed phase two of our research and directly informed the stakeholder consultation (Delphi) process by helping to identify the evidence gaps around IYCF and complementary feeding in LMICs to inform the questions subsequently asked of the stakeholders.

| Consensus testing
Consensus was reached on 47 (82.4%) of the 57 Round 2 items (see Table 3). For the 10 questions where consensus was not achieved, these items were initially reviewed and discussed by the authors. Based on this review and following consultation with stakeholders as part of Round 3, it was agreed that these items might never achieve consensus given the divergence of views across the different sectors, disciplines, occupations and geographical locations of participants.

Section A: Nutritional status of urban poor families (11 items)
This section had the greatest number of items for which consensus was not reached. Overall, respondents agreed with six items. These items related to agreement that IYC nutrition intervention programmes exist (e.g., There are existing intervention programmes that are aimed at improving the nutritional status of children in urban poor settings). There were five items where respondents did not agree.
These related to stakeholders' perceptions surrounding a lack of monitoring/evaluation of existing intervention programmes, a lack of intervention work with urban poor families and a perceived lack of efficacy of existing programmes (e.g., There is sufficient documentation and stocktaking of the on-going IYCF work already being implemented in urban poor settings). The lack of consensus likely reflects variability across the countries in which the stakeholders work in terms of the range of interventions offered for the urban poor, as well as the monitoring, evaluation and efficacy of these.

Section B: What is already known? (18 items)
There was a good deal of agreement achieved in response to these items with all but three items receiving consensus, two of which related to exclusive breastfeeding. Where there was a lack of consensus, it reflected varied breastfeeding practices and support across the countries in which participants worked. The third item asked whether the value of micronutrient supplementation in the context of maternal, infant and young child nutrition (MIYCN) is well understood for the urban poor, and while two-thirds of participants (68%) disagreed/ strongly disagreed with this statement, the lack of consensus reflects the heterogeneous contexts across countries/areas. Respondents tended to disagree/strongly disagree with five items that asked whether complementary feeding is well understood, whether challenges of urban poor families in relation to IYCF are well understood and whether urban informal day-care centres provide safe, hygienic care for children, which indicates a need to generate evidence on these areas as a priority. Participants tended to agree/strongly agree with statements covering a variety of factors thought to impact on IYC nutrition (e.g., social support, maternal education, maternal employment, poverty, sanitation and hygiene), which suggests that these are well-established risk factors and potential intervention targets.
Section C: What future work is needed, and how can this be done?

(28 items)
Consensus (agreement) was achieved for all but one of the statements in Section C, which were a list of areas for future research and intervention. For research, this included work looking at integrated, multisectoral analysis of existing country-level data on nutrition-specific and nutrition-sensitive intervention programmes/ policies and on access to food for the urban poor population. Suggestions for future intervention included a need for both nutritionspecific and nutrition-sensitive interventions targeted at urban poor mothers feeding IYC, as well as the use of locally available foods to improve complementary feeding, and empowering urban poor women by providing education and information. There was a lack of consensus as to whether the causes of undernutrition in urban poor areas are as well understood as in rural areas; however, although not reaching the threshold for consensus, two-thirds of participants (66%) agreed that the causes were less well understood in urban poor than rural areas.

| Identification of priority areas for future research and intervention (Round 3)
Given the wide range of Section C statements which achieved consensus  Table 4). 'Efforts are needed to better understand the current practices around complementary feeding' was the next highest ranked, with 12 out of 31 stakeholders voting for this (38.7%) and 'Evidence about access to food for the urban poor population needs to be generated' was ranked third, receiving support from 11 of the 31 respondents (35.5%).
In relation to potential solutions to (mal)nutrition in urban poor IYC, two statements were joint highest ranked (see Table 5). These home-based participatory play interventions could also provide an important platform for improving IYCF practices (Martinez et al., 2018;More et al., 2017;Tomlinson et al., 2018). Martinez et al. (2018) noted that 6% of control group households reported taking part in the home-based intervention study. Therefore, it is worth considering strategies to minimise treatment 'contamination' between intervention and control groups in future studies, such as adopting a cluster randomisation study design (Torgerson, 2001). Our rapid review shows that the gap in evidence around nutrition-sensitive interventions in urban areas of low-income countries persists, confirming the earlier findings of Goudet et al. (2017 T A B L E 5 Results of Round 3 of the Delphi-based consensus-gathering approach from n = 31 stakeholders to determine priority areas for future research (listed in priority order) 54. There is a need for policies specific to urban poor settings that are aimed at supporting child nutrition 3 (10%) More than 75% of stakeholders across SSA countries reported that the recommendations around, and the value of, complementary feeding were not well understood by urban poor caregivers. Moreover, 90% believed that the reasons for poor complementary feeding in urban poor areas need to be better understood, and 100% of participating stakeholders indicated that efforts are needed to better understand current practices around complementary feeding.
Together, these findings highlight that further work around complementary feeding in urban poor areas is a priority across SSA. They also reveal that existing evidence around complementary feeding in urban poor areas (e.g., Kimani-Murage et al., 2011;Macharia, Ochola, Mutua and Kimani-Murage, 2018) is not reaching key stakeholders, so improved dissemination of research findings to relevant stakeholders is another future priority. Furthermore, much of this existing evidence tends to be limited to very large poor urban areas in mostly middle-income country contexts, which means that it is not necessarily so relevant to very different poor urban environments in low-income settings, highlighting another gap to address in the future.
In contrast to the need identified for further work around complementary feeding in urban poor areas, it was generally agreed (91%) that existing IYCF messaging and nutritional counselling interventions are effective at increasing exclusive breastfeeding, indicating that this is less of a priority area for nutrition-related interventions. Strengths of this review include the systematic search and retrieval of articles from four databases, conducting additional analyses relevant to our aims, and the synthesis of evidence from both nutrition-specific and nutrition-sensitive interventions We used a rapid, rather than a systematic, review as we needed to complete a rapid evidence synthesis that balanced time constraints with considerations for bias. For this reason, we did not include a search of the grey literature or non-English articles, nor did we aim to conduct a metaanalysis, but we did conduct a quality assessment of the six studies.
As the rapid evidence synthesis informed our stakeholder consultation (Delphi process), we limited the start of the search period based on the date of publication. Therefore, studies published before our search period (January 2014), including nutrition-sensitive study elements (e.g., WASH) and the ones with diet quality outcomes, were outside the search parameters. For example, the study by Penny et al. (2005), which was included in the earlier review by Goudet et al., observed that an intervention to improve nutrition education delivered through Peruvian health services led to some improvements in complementary feeding practices (main secondary outcome) when compared with IYC from control areas. The Goudet et al. (2017) review focused on the measures of undernutrition rather than complementary feeding specifically, so this information is not captured either here or in the earlier review. The six included studies had a broad geographical spread but had high heterogeneity in terms of nutritional supplements/interventions, outcome indicators and followup periods. Strengths of the Delphi study are the consultation of a wide range of stakeholders, the use of a three-step process to garner consensus and the contribution of stakeholders from the outset. Limitations include unequal representation from across SSA countries and a lack of representation from all SSA countries, meaning that generalisability is somewhat limited and we were unable to conduct analysis by country. A further limitation was the absence of questions relating to IYC consumption of 'junk' foods and sugar-sweetened beverages, given evidence that intake of these is increasing in many LMICs (e.g., Monteiro, Moubarac, Cannon, Ng, & Popkin, 2013;Pries, Filteau, & Ferguson, 2019), and this is something that warrants exploration in future.
In conclusion, together, these findings highlight the limited research into complementary feeding and the nutritional status of children aged 6-23 months in the urban poor context. There has been a dearth of studies regarding nutrition-sensitive programme development and a lack of research on complementary feeding practices on the pathway to anthropometric outcomes in urban poor areas. Future studies are needed to address this knowledge gap. The rapid systematic review revealed a lack of interventions to improve nutritional status of IYC living in urban low-income informal settlements in SSA and highlighted that effects were modest. When integrated with the Delphi findings, these results help to prioritise the next steps in terms of supporting the development of such interventions, namely, that more studies or evidence on the causes of undernutrition in urban poor areas are needed; a situation analysis of the urban poor needs to be undertaken; and the urban poor community should be engaged to identify gaps and potential solutions.
These priority areas identified by the stakeholders provide a clear direction for future research and intervention to improve IYC nutrition in urban poor families in SSA.