Breastfeeding assessment tools for at-risk and malnourished infants aged under 6 months old: a systematic review

Background: Many small and malnourished infants under 6 months of age have problems with breastfeeding and restoring effective exclusive breastfeeding is a common treatment goal. Assessment is a critical first step of case management, but most malnutrition guidelines do not specify how best to do this. We aimed to identify breastfeeding assessment tools for use in assessing at-risk and malnourished infants in resource-poor settings. Methods: We systematically searched: Medline and Embase; Web of Knowledge; Cochrane Reviews; Eldis and Google Scholar databases. Also the World Health Organization (WHO), United Nations International Children’s Emergency Fund (UNICEF), CAse REport guidelines, Emergency Nutrition Network, and Field Exchange websites. Assessment tool content was analysed using a framework describing breastfeeding ‘domains’ (baby’s behaviour; mother’s behaviour; position; latching; effective feeding; breast health; baby’s health; mother’s view of feed; number, timing and length of feeds). Results: We identified 29 breastfeeding assessment tools and 45 validation studies. Eight tools had not been validated. Evidence underpinning most tools was low quality and mainly from high-income countries and hospital settings. The most comprehensive tools were the Breastfeeding, Evaluation and Education Tool, UNICEF Baby-Friendly Hospital Initiative tools and CARE training package. The tool with the strongest evidence was the WHO/UNICEF B-R-E-A-S-T-Feed Observation Form. Conclusions: Despite many possible tools, there is currently no one gold standard. For assessing malnourished infants in resource-poor settings, UNICEF Baby-Friendly Hospital Initiative tools, Module IFE and the WHO/UNICEF B-R-E-A-S-T-Feed Observation Form are the best available tools but could be improved by adding questions from other tools. Allowing for context, one tool for rapid community-based assessment plus a more detailed one for clinic/hospital assessment might help optimally identify breastfeeding problems and the support required. Further research is important to refine existing tools and develop new ones. Rigorous testing, especially against outcomes such as breastfeeding status and growth, is key.


Introduction
Protecting breastfeeding has been described as the single most effective child survival intervention (UNICEF, 2009;WHO, 2007). It also plays a key role in reducing the global burden of undernutrition (The Lancet Series, 2008) and is one of 13 priority interventions highlighted by the international 'Scaling Up Nutrition' movement (SUN, 2010). Despite this, suboptimal breastfeeding practices are common, accounting for significant morbidity and 804,000 deaths per year -11.6% of all deaths in children aged under 5 years worldwide (Black et al., 2013). The greatest burden of mortality and morbidity is in low income countries as defined by the World Bank (Fantom & Serajuddin, 2016). High background mortality and high rates of undernutrition and communicable disease all make the protective effects of breastfeeding critical. With collapses in infrastructure and normal societal networks, emergency affected populations are particularly vulnerable if breastfeeding is not supported and problems are not quickly identified and addressed. A group particularly higher risk of mortality and morbidity are the small and nutritionally at risk infants under six months of age compared to the infant that achieve optimal growth. At a population level, small and nutritionally at risk children are those identified as wasted, stunted and underweight and a combination of these (ENN/LSHTM, 2021).
Whilst the importance of breastfeeding is widely recognised, supporting it can be challenging. Under the overall heading of 'Promoting proper feeding for infants and young children', the World Health Organization (WHO) lists several areas of work including: the Baby-Friendly Hospital Initiative (BFHI) (WHO/UNICEF, 2009a); promotion of exclusive breastfeeding; and the International Code of Marketing of Breast-milk substitutes. These initiatives are aimed at population level breastfeeding support; there is good evidence of their effectiveness (Beake et al., 2012). More challenging is how to help those who fall through these population 'safety nets'; when an individual mother-infant pair presents with an established problem. Managing very small infants, those with growth failure and other high-risk characteristics is particularly complex. Breastfeeding problems are common in this group but there are many other potential underlying causes and contributory factors (Goh et al., 2016). Breastfeeding problems may be a primary cause or secondary to other causes. There is also a wide and complex spectrum of breastfeeding problems ranging from a simple positioning difficulty leading to insufficient milk intake, milk insufficiency perception, early complementary feeding introduction, to secondary milk insufficiency due to maternal depression, due in turn to lack of social support at home (Amir & Ingram, 2008;Moore et al., 2012;Pannu et al., 2011;WHO/UNICEF, 1994).
This review arose from a project exploring the Management of (Nutritionally) At-risk Mothers and Infants aged under 6 months (MAMI) Project (ENN/UCL/ACF, 2010b). The goal of the original MAMI Project was to investigate the management of malnourished infants under six months of age in resource-poor and humanitarian settings, and to contribute to evidence-based, better practice guidelines to improve practice. The project identified that the burden of infant less than 6 months' undernutrition is significant: worldwide, 3.8 million infants are severely wasted; 4.7 million are moderately wasted (Kerac et al., 2011). Since breastfeeding difficulties are associated with undernutrition (Gagliardi et al., 2012;Gribble et al., 2011) (Gribble et al., 2011 and exclusive breastfeeding in infants under 6 months, a common treatment goal (ENN/UCL/ACF, 2010a), the report also examined breastfeeding assessment as part of overall infant assessment. It found no 'gold-standard' breastfeeding assessment tool that catered for inpatient and community settings. This is a critical gap; correct 'diagnosis' of a breastfeeding problem is vital to inform appropriate support and treatment. Building upon and updating the work of the MAMI Project, this current review thus aims to: a) identify and profile currently available breastfeeding assessment tools; b) discuss their potential application for assessing at risk and malnourished infants under 6 months (i.e. to determine the link between breastfeeding problems and malnutrition in a particular individual; to describe the nature of that breastfeeding problem). Informed assessment is critical to targeted intervention of support.

Methods
Breastfeeding assessment tools were defined as: documented guidance for clinicians, nurses, midwives, community health workers and carers on how to observe and/or assess the breastfeeding outcomes. These could take the form of checklists, questionnaires, algorithms, indices, history taking forms or listing of the specific aspects of breastfeeding that should be assessed.

Inclusion criteria:
We included articles that: tested or used breastfeeding assessment tools; integrated at least one clinically relevant maternal or child outcome (e.g. duration of breastfeeding, infant weight gain); reported on tool performance. Articles describing complex interventions that included breastfeeding support could only be included if it was clear which tool had been used, and if breastfeeding assessment had been explicitly mentioned in the intervention description. There were no study design restrictions.

Amendments from Version 1
Following our reviewer comments and suggestions, we add two new references to clarify the concepts of small and nutritionally at-risk children and the definition of middle and low country. We described the study methodology better and improved Figure 1. We clarify that we did not include BEET tools in the recommended ones because it does not have enough validation studies, despite it covers all domains considered important. We added a further reference of the MAMI project to highlights the importance of an approach that looks at the complex spectrum of breastfeeding problems considering mother-baby died as well as a wider social contest. We proofread the article again. We hope that this makes this second version of the article clearer and more enjoyable to read.

Any further responses from the reviewers can be found at the end of the article
Exclusion criteria: Tools that focused just on artificial feeding (i.e. use of a breastmilk substitute) or that were designed for women after breast augmentation/reduction surgery were not considered in this review. Also excluded were tools that involved complex and expensive technology that are not designed for routine clinical use in resource poor settings (e.g. those using electromyographic methods; direct measurements of breastmilk composition; web-based tools; software to measure sucking strength/effectiveness; ultrasound measures of milk removal/swallowing). Tools that focused on wider breastfeeding support (e.g. employer support) rather than the actual process of breastfeeding were also excluded as were those focused solely on change in health worker knowledge, attitude or practice as an outcome. The literature search was restricted to English language articles with human subjects.

Databases and search terms:
Articles were identified by searching electronic database Medline and Embase via Ovid interface (full search strategy is free available at http: //www.doi. org/10.17037/DATA.00001881 in Extended data (Kerac et al., 2020)). Key words and MeSH terms were selected by the review on The Lancet Breastfeeding Series (The Lancet Series, 2016) and a recent similar review on feeding assessment tools (Howe et al., 2008). We also included hand search papers form grey literature, WHO and ENN websites. Searches were finalised in March 2018. This updated an earlier search done as part of the original MAMI project performed on PubMed, Web of Knowledge, Cochrane Review, Eldis and Google scholar databases which concluded in November 2013. In that original search, highly relevant journals were also searched directly: Maternal and Child Nutrition, International Breastfeeding Journal, Journal of Human Lactation, and BMC Family Practice. Reference lists and the 'related articles' were used to identify further articles. A standard two-stage search strategy was used: initial screening of titles and abstracts by 3 authors (C.B, K.L.R. and M.K.); detailed review of full articles secondly (C.B, K.L.R. and M.K.). Since tools were few but varied, risk of bias was not formally scored for each individual study but is discussed under 'limitations' for studies as a whole.

Description of the tools
To understand and characterise the tools we also examined: Tool coverage of breastfeeding 'domains' There are several aspects or 'domains' of breastfeeding. Knowing which are affected helps guide appropriate subsequent treatment. We used an established framework (Moran et al., 2000) to characterise which aspects of breastfeeding the assessment tools assessed. These included: baby's behaviour (e.g. alertness to feed), mother's behaviour (e.g. watches and listens for baby's cues), positioning (e.g. baby facing mother), attachment (e.g. lower lip turned outward on breast), effective feeding (e.g. sucking, swallowing, jaw movement and signs of milk release), health of the breast (e.g. nipple trauma), health of the baby (e.g. alert), and mother's experience (e.g. feels strong suction). We added another domain on number, timing and length of feeds. We also noted any other domains identified by individual studies.
Evidence underpinning each assessment tool Studies were grouped according to type of evidence presented. One group looked at prediction of later breastfeeding status. Another assessed test-retest, inter-rater reliability and sensitivity and specificity of tools. A final group of studies focused on assessment tools used to directly improve breastfeeding technique or experience.

Results
From a total of 15,649 titles and abstracts screened, a final count of 52 papers describing 29 distinct breastfeeding assessment tools were identified (Figure 1).

Final selection of tools
Details of the 29 tools identified are summarised in Table 1. Exclusions and reason for those are presented in web-appendix (Extended data (Kerac et al., 2020)). We were unable to get sufficient information about two tools: The LAT TM (Cadwell et al., 2004) and the Prague Newborn Behaviour Description Technique (Sulcova & Tisanska, 1994) so we could not include them in the final review.

Context
Of the 29 tools identified: 22 (76%) were developed in high-income countries and used in 31 studies carried out in high-income countries, six (21%) tools were developed in low and middle-income countries and one (3%) was developed worldwide. Sixteen tools (55%) were developed for hospital settings. Of these, 24 (83%) tools were designed and/or tested for use in infants less than 6 months with breastfeeding problems; none of these were specifically designed for or tested on at risk and malnourished infants less than 6 months. Table 2 shows that most tools covered a number of different domains but only one, the Breastfeeding Evaluation and Education Tool (Tobin, 1996), covered them all.

Coverage of breastfeeding domains
Other tools covering a wide range of domains were the Baby-Friendly Hospital Initiative (BFHI) guidelines (UNICEF, 2010;WHO/UNICEF, 2009a) and the CAse REport guidelines (CARE guidelines) (CARE, 2004). The BFHI and CARE guidelines also highlighted other items that could be useful for future testing: positions for low birth weight babies, differentiating between 'perceived' and 'real' milk insufficiency, mother's health, and the use of BMS and dummies/pacifiers. The World Health Organization/United Nations International Children's Emergency Fund (WHO/UNICEF) B-R-E-A-S-T-Feed Observation Form covered seven domains, missing out 'health of the baby' and 'mother's view of the feed' (WHO/UNICEF, 1994). Additional domains identified by other tools included: mother's comfort level, previous breastfeeding experience, other foods/liquids being given to the baby, loss of >10% of birth weight, hypertension and delivery type (Darmstadt et al., 2009;Dongre et al., 2010;Hall et al., 2002;Mannan et al., 2008;Milligan et al., 1996;Palmer et al., 1993).
Ability of tools to predict breastfeeding outcomes In total, 12 (41%) tools had been tested for their ability to predict breastfeeding outcomes (Table 3).
The present studies either tested the tools or tested the intervention or tested both. The tools with the most studies testing their ability to predict breastfeeding outcomes during an intervention study were the LATCH (n=5), the WHO/UNICEF B-R-E-A-S-T-Feed observation form (n=6) and the BAS tool (n=4). The BAS was consistently predictive in all studies, although as shown in Table 2, it covers the least number of breastfeeding domains. There were mixed findings for the LATCH tool: three studies observed positive findings, and two reported limited ability of the tool to predict breastfeeding outcomes. The WHO/UNICEF B-R-E-A-S-T-Feed Observation Form was predictive of breastfeeding outcomes in three studies, but was not predictive of exclusive breastfeeding in a fourth study.
Two further studies described the determinants of poor scores on the WHO/UNICEF B-R-E-A-S-T tool including repeated crying, colic history, shorter sleeping episodes and regurgitation (Yalcin & Kuskonmaz, 2011), and primiparity, cracked nipples, mastitis, preterm and low birth weight babies and poor suckling (Goyal et al., 2011).

LMICs
Community humanitarian settings Training materials include handouts and counselling cards on: Signs of good positioning (4 items) and attachment (5 items, 1 illustration) and effective suckling (5 points); recommendations on optimal breastfeeding practices focusing on mother's behaviour; 3 common breast conditions (including photos); perceived insufficient milk supply; 11 'special situations' including malnourished and stressed mothers, baby refusal to feed. Prevention and solutions are given.
Checklist from 'breastfeeding and the use of pacifiers'

Sweden
Hospital 16 observations to determine early breastfeeding cessation and correct vs incorrect sucking techniques: breast offering (3 items), sucking at the breast (9 items), after feeding (2 items), and conclusions (2 items).
Essential Nutrition Action Messages (Breastfeeding guidance booklet) (Guyon & Qinn, 2011;Guyon et al., 2009) LMICs Specifies multiple settings for use Illustration and recommendations to ensure optimal breastfeeding. Illustration 8 on correct positioning: 9 guidance items + 3 pictures. Illustration 9 focuses on proper attachment: 4 signs and 5 signs of efficient suckling + 1 picture. There is also illustration 10 for three different breastfeeding positions and attachment, with pictures.
History Taking Form from 'Functional assessment of infant breastfeeding patterns' (Walker, 1989) USA Not specified A sample feeding assessment with rationale that covers: general physical condition and body tone of baby; with a digital check of infant sucking ability, breast assessment (e.g. look for engorgement), nipple assessment (e.g. flat nipple), position of mother and baby whilst nursing, latch on, sucking pattern/sound, and maternal impression of the feed. It is part of a general assessment of normal and problematic situations that include a baby's feeding history and the mother's history on some physical aspects before and after pregnancy.
Hands off technique and one picture pre-latch-on 2) Six aspects of latch-on and suckling dynamic 3) three aspects of milk transfer from mother to infant 4) one aspect of mothers comfort during/between feedings 5) one aspect of infant signs of satiety Mother-Baby Assessment (MBA) (Mulford, 1992) USA Hospital 5 steps in breastfeeding are assessed for both the mother and the infant: signalling, positioning, fixing, milk transfer, ending. A score out of ten rates mother's and baby's efforts to breastfeed and the progress of both partners. Tool items based on positioning, fixing & milk transfer items from published work describing common features of effective breastfeeding. USA Not stated Breastfeed is scored 0-2 (0=absent behaviour, 1=problematic, 2=no problem): 1) latch on 2) time before latch on and suckle 3) suckling 4) degree of swallowing 5) mother's evaluation. Overall Scores 0-2 = high risk, return visit/call within 24 hours (automatic high risk if >10% birth weight lost or mother had breast surgery); 3-6 = medium risk, refer to public health nursing, visit within 3 days; 7-10 = low risk, routine calls/visits. to assess excessively sleepy baby following high dose of labour analgesia. UK Hospital 4-item tool: position, attachment, sucking and swallowing to improve targeting positioning and attachment advice. Attribution of a 0 to 2 score: 0 poor -2 good or no need advice. It can be useful on tongue-tied infant.

WHO/UNICEF B-R-E-A-S-T-Feed
Lactation history and risk assessment form (Riordan, 1989) USA Hospital 4-items form to take lactation history and evaluate breast and nipples to carry out an appropriate risk assessment: feeding choice, physical exam, history including baby weight gain, risk factors.
Positions if low birth weight, insufficient milk, mother's health, formula, dummies Breast-feeding Assessment Score  Observational within 72 hours postpartum Evaluation of internal consistency, validity, sensitivity and specificity 5-and 4-item versions of LACTH. Data were filtered: preterm deliveries were excluded because of their different suckling patterns. Only 4 or 5 outcomes. The sample were infant with body weight 3.14-0.39 Kg.
The 4-item versions can be considered as routine assessment tool to assist.
The sensitivity of the tools to correctly identify postanal woman at risk of non-exclusive breastfeeding is satisfactory (cut off point 3.5 and 5.5) the specificity is poor. Acceptable internal consistency. 6 were randomised or cluster randomised controlled trials, two reported time trends; and 1 reported intervention baseline and endline data without a control group.

LATCH
The BAS tool had four validation studies, all of which show positive results for the tool, in terms of ability to identify those at risk of breastfeeding cessation, and moderate sensitivity and specificity (Gianni et al., 2006;Hall et al., 2002;Mercer et al., 2010;Zobbi et al., 2011). The evidence to support the use of the Essential Nutrition Actions Framework tool is weak in terms of validation (i.e. no control group; not clear if the tool was routinely used) (Guyon et al., 2009). IBFAT also had a low inter-rater reliability. Furthermore, most studies were low quality (e.g. small sample size and observational designs) and were also conducted exclusively in high income settings (Furman & Minich, 2006;Matthews, 1988;Matthews, 1991b;Riordan & Koehn, 1997;Schlomer et al., 1999).
Nine tools were tested for test-retest and inter-rater reliability in eight studies -one study compared three tools. Two tools performed well: the Integrated Management of Childhood Illness (IMCI) showed good sensitivity and high specificity in highlighting breastfeeding problems judged against clinician assessments (Darmstadt et al., 2009); the Mother Infant Breastfeeding Progress Tool (MIBPT) showed high inter-rater agreement (Johnson et al., 2007). There were mixed findings for the remaining tools. Details of these studies are in Table 4.

Ability of tools to correct breastfeeding technique or improve breastfeeding experience
Few studies tested the use of tools to correct breastfeeding technique or to improve breastfeeding experience. These are shown in Table 5.

Discussion
Our review identified a number of breastfeeding assessment tools which could be used in the management of our target group of at-risk and malnourished infants aged under 6 months. Though none of the tools were developed for or tested on this group directly, characterising them and understanding the underlying evidence-base allows for better informed decisions about which might be the most helpful for future programme use.
Regarding the coverage of breastfeeding domains, only one tool (BEET) achieves full coverage of all the key assessment domains, but there were no validation study at our knowledge. The tools that achieve the widest coverage (IFE Module 2, BEET, and WHO/UNICEF B-R-E-A-S-T-Feed Observation Form and UNICEF/WHO Breastfeed Observation Aid) are generally those which have been developed with resource-poor low and middle income countries in mind. Although these tools are based on extensive clinical and field experience, they suffer from lack of validation research and miss some important domains (e.g. WHO/UNICEF B-R-E-A-S-T-Feed Observation Form misses health of the baby, IFE Module 2 misses positioning). These shortfalls could be addressed with minor modifications in the short term and with appropriately designed studies soon after to help determine which domains are the most important and relevant to patient care. Only 11 tools assess mothers' own behaviour towards the baby: this is telling about her psychosocial status and can inform management. It is important to consider and account for such gaps since an infant may be effectively breastfed but at risk and malnourished for another reason, e.g. related to child health status or maternal factors. The mother-infant dyad is at the heart of approaches to treat malnutrition, but wider family and community relationship are also important but cannot be treated extensively in this review (ENN/LSHTM, 2021b).
A challenge validating breastfeeding assessment tools is the lack of a 'gold standard' treatment option for at-risk and malnourished infants less than 6 months. This makes validation studies a challenge methodologically since it is difficult to separate out the performance of an assessment tool from the effectiveness of the subsequent management strategy in averting adverse nutrition/morbidity outcomes. It is likely that different tools and different levels of management will be appropriate to different settings, e.g.
• In primary healthcare / community settings: simple and rapid breastfeeding assessment tools, associated with easy-to-deliver interventions and to prompt referral for more specialised support. For use by community healthcare workers who may have limited training and experience.
• In secondary healthcare / outpatient clinic settings: more detailed tools could be appropriate but would need more training and staff with more background skills, expertise and time to deliver.
• In tertiary-level inpatient settings: more complex assessments would be appropriate to identify more complex problems. These could be delivered by more highly trained healthcare staff such as nurses and doctors.
No single tool meets all these needs. Which tool is more appropriate to a given setting and individual mother-infant situation is itself an important question that warrants further testing and exploration.
For immediate use, whilst refining current tools and developing new future ones, the WHO/UNICEF B-R-E-A-S-T-Feed Observation Form, the aids in Module 2 on IFE and UNICEF/WHO Breastfeed Observation Aid, offer the most promise for programmes targeting at-risk and malnourished infants aged under 6 months.
In future research testing current and new tools, there is a need to agree on the most appropriate outcomes for validation studies targeting at-risk and malnourished infants under 6 months. The fact that so many tools exist, and that they cover such a wide range of feeding outcomes and domains arguably reflects uncertainly and lack of consensus about how best to assess the effectiveness of breastfeeding. For example, must there always be sufficient infant weight gain associated with other measures of effective feeding? Most current evidence comes from high-income countries and hospital  settings. For use in tackling the significant global burden of malnutrition in infants aged less than 6 months, this is a problem. More tools for low income countries and for community settings are urgently needed (Moran et al., 2000;Mulder, 2006;Riordan, 1998;Riordan & Koehn, 1997).
Another key finding of our review was the variable -and overall low -quality of evidence underpinning existing breastfeeding assessment tools. Often the evidence-base for a particular tool is unclear, particularly their effectiveness in identifying specific breastfeeding problems and facilitating a resolution. Prospective and ideally randomised studies testing tools' ability to do this are important in the future (Da Costa et al., 2008). Simple checklists have been shown to be powerful if used consistently in clinical settings (Haynes et al., 2009;Pronovost et al., 2006). There is therefore an argument to develop checklist-based tools that can be incorporated into routine breastfeeding assessment, to maximize the chances of resolving breastfeeding problems. These should also be able to discriminate between different types of breastfeeding problems and lead clearly to specific interventions.
We found that tools varied in their level of complexity, and their scoring systems. This may make individual tools relevant only for specific contexts. For example, three tools involve two stages: IFE Module 2 includes a simple rapid assessment, followed by a full assessment (ENN et al., 2007); the BFHI guidelines may include initial use of the breastfeeding assessment form, leading on to the UNICEF/WHO breastfeed observation aid if necessary (UNICEF, 2010; WHO/UNICEF, 2009a); the IMCI algorithm includes both a brief history taking and observations of the breastfeed (Mannan et al., 2008). This is potentially a good thing. Rather than one tool trying to do everything, different tools for different levels of assessment could be helpful: e.g. a quick, basic tool for use in the community to identify and correct 'simple problems and identify referral need, complemented by a more detailed tool if problems are suspected or identified; another more detailed one for clinic/hospital use assessing more serious and complex problems flagged by the first tools. Tool developers need to consider what the key contact points with infants are, and the associated opportunities and capacities with these contact points. Coupled with this must be the capacity to respond to any problems identified. To address breastfeeding in high mortality/morbidity settings, tools need to consider not just physiological issues and techniques around breastfeeding, but also the wider social and psychological factors, which may be contributing to or perpetuating a problem (Galipeau et al., 2017).
Which tools for resource-poor, high-undernutrition settings From this review, Baby Friendly Hospital tools, the Module 2 IFE and WHO/UNICEF B-R-E-A-S-T-Feed Observation Form, have emerged as potentially useful for use in humanitarian settings with at-risk and malnourished infants under 6 months. They require a short training and they are easy-touse. Baby Friendly Hospital tools and the Module 2 IFE could benefit from adaptation by adding the missing components that we would be considered useful for humanitarian contexts.
While BFHI has become a 'gold standard' for maternity care in hospital setting, the effectiveness of the training course has been assessed but the evaluation of the breastfeeding assessment form requires more studies. Equally, these tools could be combined (e.g. by adding questions from one tool to another) in a way that might improve the quality of breastfeeding assessment, and that would take into account the specific needs and limitations of contexts with a high burden of undernutrition. It will be important to ascertain the feasibility of community health workers using these tools.
Based on coverage of domains, appropriateness to target population and setting, and underlying evidence, WHO/UNICEF B-R-E-A-S-T-Feed Observation Form appears to be the most suitable for assessing at risk and malnourished infants aged under 6 months. In two Danish RCTs, health visitors were trained to conduct home visits incorporating breastfeeding assessment and classification of technique problems (Kronborg & Vaeth, 2009;Kronborg et al., 2007). One study found a 14% lower breastfeeding cessation rate amongst intervention participants, and greater confidence of mothers that their breast milk was sufficient. However, the other found no difference in exclusive breastfeeding rate or a reduction in breastfeeding problems -this may be due to a single corrective intervention being insufficient to resolve breastfeeding problems. The authors argued for on-going breastfeeding support to ensure breastfeeding problems are truly resolved. This idea is corroborated by a third Brazilian hospital-based RCT with a low socioeconomic population, which found no impact of a single breastfeeding assessment and correction on exclusive breastfeeding rates, breastfeeding technique or breastfeeding problems 30 days post-partum (De Oliveira et al., 2006). A further RCT in Brazil also used the WHO/UNICEF B-R-E-A-S-T-Feed Observation Form but included a greater number of home visits (n=6). This observed a 39% increase in any breastfeeding, and a significant increase in exclusive breastfeeding. One limitation of this study is that it is difficult to unpick the effect of the breastfeeding observation and corrective advice from the other interventions during the home visit (Leite et al., 2005). This underlines the importance of not just having a good tool, but using it to maximum effect i.e. not just conducting a single assessment and correction, but providing on-going support through community outreach (Imdad et al., 2011). What is most encouraging about the WHO/UNICEF B-R-E-A-S-T-Feed Observation Form is its apparent usability in routine clinical settings, with relatively short training if conducted for the use of the test only. As the tool is part of a broader training on breastfeeding counselling, it is recommended to explore the whole manual, but it is also possible to adapt to the situation's needs. It would still be valuable to do further validation of this tool and possibly extend the tool components to include aspects of the baby's health, as identified in the section on coverage of breastfeeding domains.

Ways forward
As well as standard validation studies, new tools or those initially developed in/adapted from resource-rich settings should be assessed for cultural relevance and sensitivity before they are considered for use in resource-poor developing country/humanitarian settings. This formative work should ideally precede detailed validation or intervention studies. Validity is likely to vary according to target patient group and studies should therefore be sufficiently powered to explore subgroups. Tools that are designed to assess breastfeeding in healthy, well-nourished infants are not necessarily as good or adequate for assessing sick or undernourished ones. As none of those tools presented above were developed and tested in malnourished children and since these infants are at particularly high risk of morbidity and mortality, specific tools should consider the needs of infants aged less than 6 months with malnutrition -the group who inspired this review in the first place. Since there are many factors potentially underlying or contributing to malnutrition, we believe that tools for this group should be part of a wider assessment of the mother-infant dyad and take an appropriately broad perspective by considering other factors known to impact on infant nutrition e.g. maternal mental health, maternal illness, and maternal malnutrition.

Limitations
We acknowledge the limitation of our review. Firstly, it was restricted to articles written in English; there may be useful breastfeeding assessment tools published in other languages that were not captured.
Secondly, it is possible that we missed some studies, e.g. those using a broader approach to improving infant feeding may not have explicitly mentioned breastfeeding assessment tools as part of their intervention protocol; those which were using a tool in a programme but were not in the title or abstract clearly evaluating/testing the tool itself; those that may have had relevant content (e.g. maternal psychosocial status) but did not meet the inclusion criteria of one clinically relevant maternal or child outcome.
Third, we did not explicitly grade the quality of individual studies -this was felt not to add significant extra value to our review since observational studies, which comprised great majority of papers identified, are by definition low quality compared to intervention/RCT type designs. Quality grading would not have helped differentiate between more/less valuable tools, since the quality of evidence underpinning them all was generally low.
Finally, we found few tools explicitly targeted to our setting and main patient group of interest. This is not ideal since it means applicability had to be extrapolated based on our judgement rather than on hard data.
Despite these limitations, we do not believe that the overall direction or message arising from our findings are affected.

Conclusion
In this review of breastfeeding assessment tools for resource poor settings and targeting the assessment of malnourished infants less than 6 months, we have identified many possible but few stand-out 'gold standard' options. This represents an important evidence gap and highlights an urgent need for future research. The many different tools that we did find arguably show that one tool alone is unlikely to be suitable or even desirable. Tools must strike the right balance between simplicity, feasibility of use and minimal training requirements without losing the depth of information required to help healthcare workers and the women they are working with address breastfeeding difficulties. Thus, different tools for different levels of the health care system are needed: simple, quick-to-use tools for initial triage and problem identification in the community; more sophisticated tools for use in secondary and tertiary care settings where initial attempts at support have failed. Supplementary items such as pictures of good latch, and materials to help mothers and health workers understand the nature of breastfeeding problems (e.g. 'take action cards ' (Dongre et al., 2010)), may be helpful. For any tool at any level, it is important that it leads to clear corrective actions. A "diagnosis" or "problem label" by itself is not always useful. Hence, future tools might give appropriate weight to problems, which can most readily be solved, or those which have the biggest short and long term impact. Research on breastfeeding assessment tools needs to consider such impacts -again, good test inter-and intra-observer validity is necessary but not alone sufficient to make a 'good' tool. It must help improve key outcomes like breastfeeding status and infant growth. Robustly designed studies in the contexts in which they will be used are essential.
Finally, we note that time will be needed to develop and test better future breastfeeding assessment tools. Yet support for women and their infants is urgently needed now. Not having an ideal tool is not a reason to defer breastfeeding assessment of at risk and malnourished infants under 6 months. There are great opportunities at present to collect and report good quality operational data using tools that are currently available. Expanding the current literature on breastfeeding assessment will be of great benefit to future tool developers. More importantly, focus on this area will also raise the profile of and directly benefit breastfeeding as a key child nutrition, health and survival intervention.

Data availability
Underlying data All data underlying the results are available as part of the article and no additional source data are required.

4.
Regarding your suggestion for authors to proofread the paper to improve the overall writing; We have now proofread and think this looks fine in this last version. the major changes are: Table 2: corrected 'behaviour' Table 3: corrected 'postnatal'; 'paediatric'

Sandra Fucile
School of Rehabilitation Therapy, Queen's University, Kingston, ON, Canada Brugaletta et al.'s, review of breastfeeding assessment tools for assessing at-risk and malnourished infants in resource-poor settings provides a comprehensive literature search of available tools for this highly vulnerable population. The authors reveal there is no 'gold standard' tool available for at-risk and malnourished infants in resource-poor settings. However, they highlight three ready available tools, the Breastfeeding, Evaluation and Education Tool, UNICEF Baby-Friendly Hospital Initiative tools and CARE training package, that can be used with this population and emphasize the need for refining or developing new breastfeeding tools to meet the needs of infants in resource-poor settings.
Overall, the authors provide a very thorough introduction with a clear rationale for undertaking this study. The authors perform a systematic in-depth literature search, which included seven online database resources. The authors identified 29 breastfeeding assessment tools and 45 studies related to the tools' psychometric properties. They found that the evidence and psychometric properties of the tools was low quality and mainly from high-income countries. The strengths and weakness of these 29 breastfeeding tools were described in terms of the tool content of breastfeeding domains, predictive validity, reliability, and evidence underlying the content each tool. The tables provided clearly synthesize and integrate the strengths and weakness of each tool. In the discussion, the authors address the limitations of the study and bring to light the drawbacks of current available tools in achieve the defined outcome in this study.
The findings from this review are clinically significant and I have minor suggestions: I encourage the authors to thoroughly reread the manuscript to ensure there are no grammatical and editorial errors.
○ Figure 1 appears to be adopted from the PRISMA framework, ensure that this is referenced.

○
Use of terminology, the authors refer to Table 3 as validation studies, I recommend rewording to use predictive validity studies.
○ I recommend including specific definitions for high vs low income countries, define at-risk infants, define malnourished infants either in the introduction or methods section.

○
The authors note 29 tools were identified, 22 were developed in high income countries, 4 low income countries. Three tools are missing such designations.

Is the statistical analysis and its interpretation appropriate? Yes
Are the conclusions drawn adequately supported by the results presented in the review? Yes Competing Interests: No competing interests were disclosed.
Reviewer Expertise: Oral feeding in critically ill infants.
I confirm that I have read this submission and believe that I have an appropriate level of expertise to confirm that it is of an acceptable scientific standard.

Author Response 25 May 2021
Concetta Brugaletta, University College London Hospitals NHS Trust, London, UK Thank you for taking the time to read the review and write your suggestion.
Regarding your advice to proofread the manuscript to ensure there are no grammatical and editorial errors; We have proofread and we think this looks fine in the new version of the manuscript.
Regarding your advice to ensure the reference on Figure 1; we added the reference at the base of the figure in the new version of the manuscript.

Kerstin E. Hanson
Independent researcher, Paris, France Jessamyn Ressler-Maerlender Independent researcher, Ann Arbor, MI, USA This review by Brugaletta et al. addresses the important gap in tools and evidence for effective case management of at-risk and malnourished infants aged under 6 months old in low-resource and humanitarian settings. The authors focus specifically on the availability and quality of breastfeeding assessment tools for use in this population. They start with an excellent introduction highlighting the importance of this theme -describing the essential role of breastfeeding in protecting the health and lives of children and the subsequent place it has in global priority interventions, and the challenges that remain in properly addressing persistently suboptimal breastfeeding practices.
Results of the review are presented in clear tables, summarizing important features of the various tools identified. The authors provide a clear breakdown of these features into the following categories: context, coverage of breastfeeding domains, ability to predict breastfeeding outcomes, evidence underpinning the tools, and ability to correct breastfeeding technique or improve breastfeeding experience. The analysis also addresses not only technical or academic features of the tools, but also "real world" implementation issues, and ability to bring about the desired outcomes -improved breastfeeding.
The discussion highlights strengths and gaps of individual tools, as well as the overall "collection" of tools identified. The authors are clear to state that none of the tools were directly developed for or tested on at-risk and malnourished infants aged under 6 months, nor do any of the tools fully meet the various needs in terms of categories outlined above. The authors do nevertheless identify three tools that could be used "for immediate use, whilst refining current tools and developing new future ones".
We appreciate the overall approach to this review -identifying and analyzing current tools, recognizing that we do not currently have an ideal tool, explaining the key gaps and ways forward, and importantly -providing temporary best options. Minor suggestions to consider in subsequent versions: In the abstract the authors list the following as part of their search: the World Health Organization (WHO), United Nations International Children's Emergency Fund (UNICEF), CAse REport guidelines, Emergency Nutrition Network, and Field Exchange websites as parts of the search. These are not mentioned in the database and search terms of the methods section. They are perhaps listed in one of the references, but it might be helpful to include them in this later section of the manuscript as well. 1.
In figure 1 it is unclear where the "Handsearch Papers" fit. This part of the search could be expanded upon in the methods section.

2.
Under context the authors state: "Of the 29 tools identified: 22 (76%) were developed in high-income countries and used in 31 studies carried out in high-income countries and four (14%) tools were developed in low and middle-income countries." What about the 3 tools, not included in the 22 developed in high-income countries and the 4 developed in low-and 3.
middle-income countries?
The introduction does a nice job addressing the particular challenges associated with "managing very smalll infants, those with growth failure and other high-risk characteristics". It also highlights the complex spectrum of breastfeeding problems including potential underlying causes and contributory factors, including but not limited to maternal wellbeing and social support. Although these essential topics are touched upon very briefly in the results and discussion sections, and in a bit more detail in the ways forward section, the review could benefit with expansion of these critical topics.

4.
In the discussion section the authors note that only one tool, BEET, achieves full coverage of all the key assessment domains. Yet, this tool is not included amongst those listed as potentially useful for immediate use; it could be useful to note why this is the case.

5.
In the initial paragraphs of the discussion, the authors suggest: "For immediate use, whilst refining current tools and developing new future ones, the WHO/UNICEF B-R-E-A-S-T-Feed Observation Form, the aids in Module 2 on IFE and UNICEF/WHO Breastfeed Observation Aid, offer the most promise for programmes targeting at-risk and malnourished infants aged under 6 months". Later, under the heading Which tools for resource-poor, highundernutrition settings, they suggest: "From this review, Baby Friendly Hospital tools, the Module 2 IFE and WHO/UNICEF B-R-E-A-S-T-Feed Observation Form, have emerged as potentially useful for use in humanitarian settings with at-risk and malnourished infants under 6 months." Referring to the tools listed in the tables, we imagine that the Baby Friendly Hospital tools and UNICEF/WHO Breast Observation Aid noted above are referring to the same tool. If so, the same naming convention should be used in both instances for clarity.

6.
There are a number of minor typos and grammatical errors throughout the paper and tables that should be corrected.

7.
Are the rationale for, and objectives of, the Systematic Review clearly stated? Yes

Is the statistical analysis and its interpretation appropriate? Not applicable
Are the conclusions drawn adequately supported by the results presented in the review? Yes Competing Interests: No competing interests were disclosed.
Reviewer Expertise: Pediatric and nutrition programming and case management in low-resource and humanitarian settings.
We confirm that we have read this submission and believe that we have an appropriate level of expertise to confirm that it is of an acceptable scientific standard.
4. Regarding your comment on considering complex spectrum of breastfeeding problems including potential underlying causes and contributory factors, including but not limited to maternal wellbeing and social support; It is a great suggestion; the number of words required does not allow us to expand on this very interesting and crucial topic. We have however added a reference of the MAMI website, which is regularly being updated and a note saying "The mother-infant dyad is at the heart of approaches to treat malnutrition, but wider family and community relationship are also important but cannot be treated extensively in this review". (ENN/LSHTM, 2021)https://www.ennonline.net/ourwork/research/mami https://www.ennonline.net/mami/practice 5. Regarding your comment on the discussion section where only one tool, BEET, achieves full coverage of all the key assessment domains. Yet, this tool is not included amongst those listed as potentially useful for immediate use; Thank you for your comments. The justification is in the paragraph on evidence underpinning tools: 'The extent of tool testing varied substantially; 8 tools had no validation studies: Infant Feeding in Emergencies (IFE) Module 2 (ENN et al., 2007), Breastfeeding Evaluation and Education Tool (Tobin, 1996)'. This is why we do not include for immediate use. We have added a note in the text "only one tool (BEET) achieves full coverage of all the key assessment domains, but there was no validation study at our knowledge".
6. Regarding your comment on request of clarification if the Baby Friendly Hospital tools and UNICEF/WHO Breast Observation Aid are referring to the same tool. We would like to explain their difference and the rational for our subtle considerations: -The Baby Friendly Hospital tools: is a checklist (12 to 14 items) designed with the aim to identify area of problem and give advice. These tools take in consideration health professional background and day of life of the baby and one can also be self-administered. This means there are 4 slightly different tools available: for mother and midwife, for mother and health visitor, for neonatal and for mother alone. The domain covered are: baby's and mother behaviour, positioning, lactating, effective feeding, breast health, baby health, mothers view (in addition these tools look at urine and stools, formula). ( https://www.unicef.org.uk/babyfriendly/baby-friendly-resources/implementing-standardsresources/breastfeeding-assessment-tools/). -The UNICEF/WHO Breastfeeding Observation Aid is a checklist of identify dichotomous items ( 42 items/5 scales) list signs that represent that BF is going well versus possible difficulties. It cover similar breastfeeding domains of the BFH tools but is a more simple checklist and doesn't offer possible solutions. ( https://www.scribd.com/document/353627133/Breastfeed-Observation-Job-Aid) This is why we maintained 2 different names and we advise to use the BFHT for resourcepoor, high undernutrition settings where it is useful to have alongside the assessment also some initial advice.
7. Regarding your advice to proofread the paper to improve the overall writing; we have proofread and now we think this looks fine in this last of the manuscript.

Competing Interests:
No competing interests were disclosed.