Educational interventions for improving primary caregiver complementary feeding practices for children aged 24 months and under
Abstract
Background
Although complementary feeding is a universal practice, the methods and manner in which it is practiced vary between cultures, individuals and socioeconomic classes. The period of complementary feeding is a critical time of transition in the life of an infant, and inappropriate complementary feeding practices, with their associated adverse health consequences, remain a significant global public health problem. Educational interventions are widely acknowledged as effective in promoting public health strategy, and those aimed at improving complementary feeding practices provide information about proper complementary feeding practices to caregivers of infants/children. It is therefore important to summarise evidence on the effectiveness of educational interventions to improve the complementary feeding practices of caregivers of infants.
Objectives
To assess the effectiveness of educational interventions for improving the complementary feeding (weaning) practices of primary caregivers of children of complementary feeding age, and related health and growth outcomes in infants.
Search methods
In November 2017, we searched CENTRAL, MEDLINE, Embase, 10 other databases and two trials registers. We also searched the reference lists of relevant studies and reviews to identify any additional studies. We did not limit the searches by date, language or publication status.
Selection criteria
Randomised controlled trials (RCTs), comparing educational interventions to no intervention, usual practice, or educational interventions provided in conjunction with another intervention, so long as the educational intervention was only available in the experimental group and the adjunctive intervention was available to the control group. Study participants included caregivers of infants aged 4 to 24 months undergoing complementary feeding. Pregnant women who were expected to give birth and commence complementary feeding during the period of the study were also included.
Data collection and analysis
Two review authors independently extracted data on participants, settings, interventions, methodology and outcomes using a specifically‐developed and piloted data extraction form. We calculated risk ratios (RR) and 95% confidence intervals (CIs) for dichotomous data, and mean differences (MD) and 95% CIs for continuous data. Where data permitted, we conducted a meta‐analysis using a random‐effects model. We assessed the included studies for risk of bias and also assessed the quality of evidence using the GRADE approach.
Main results
We included 23 studies (from 35 reports) with a total of 11,170 caregiver‐infant pairs who were randomly assigned to receive an educational intervention delivered to the caregiver or usual care. Nineteen of the included studies were community‐based studies while four were facility‐based studies. In addition, 13 of the included studies were cluster‐randomised while the others were individually randomised. Generally, the interventions were focused on the introduction of complementary feeding at the appropriate time, the types and amount of complementary foods to be fed to infants, and hygiene. Using the GRADE criteria, we assessed the quality of the evidence as moderate, mostly due to inadequate allocation concealment and insufficient blinding.
Educational interventions led to improvements in complementary feeding practices for age at introduction of complementary foods (average RR 0.88, 95% CI 0.83 to 0.94; 4 studies, 1738 children; moderate‐quality evidence) and hygiene practices (average RR 1.38, 95% CI 1.23 to 1.55; 4 studies, 2029 participants; moderate‐quality evidence). For duration of exclusive breastfeeding, pooled results were compatible with both a reduction and an increase in the outcome (average RR 1.58, 95% CI 0.77 to 3.22; 3 studies, 1544 children; very low‐quality evidence). There was limited (low to very low‐quality) evidence of an effect for all growth outcomes.
Quality of evidence
There is moderate to very low‐quality evidence that educational interventions can improve complementary feeding practices but insufficient evidence to conclude that it impacts growth outcomes.
Authors' conclusions
Overall, we found evidence that education improves complementary feeding practices.
PICOs
Plain language summary
Educational interventions for improving complementary feeding practices
Background
Complementary feeding is the period when an infant moves from taking only breast milk or breast‐milk substitutes (such as infant formula) to family food. It is a critical period in the life of an infant. Inappropriate complementary feeding practices, with their associated adverse health consequences, remain a significant global public health problem. This is because inappropriate complementary feeding practices, such as introduction of semi‐solid foods too early (before six months of age), poor hygiene or giving foods that do not contain adequate nutrients, are all major causes of illness. Such illnesses include malnutrition, diarrhoea, poor growth, infections and poor mental development of children. Education has been proposed as an effective means of improving complementary feeding practices.
Review question
Does education improve complementary feeding practices of caregivers of infants as well as the health and growth of the infants?
Study characteristics
We searched for randomised controlled trials (a type of experiment in which people are randomly allocated to one or more treatment groups) up until November 2017. The search identified 23 studies involving a total of 11,170 caregivers and their children. The ages of the children ranged from birth to 24 months. The caregivers received educational interventions alone while the control group received no intervention, usual care or any other non‐educational intervention. The educational methods included printed materials such as leaflets, counselling, teaching sessions, peer support, videos and practical demonstrations. Generally, the education messages were focused on the introduction of semi‐solid foods at the appropriate age, the types and amount of complementary foods to be fed to infants, and hygiene.
Key results
Education reduced the number of caregivers that introduced semi‐solid foods to their infants before six months of age by up to 12% (moderate‐quality evidence). Hygiene practices of caregivers who received education also showed some improvement compared to those that did not (moderate‐quality evidence). In studies conducted in the community, education increased the duration of exclusive breastfeeding, but not in studies conducted in health facilities. There was no convincing evidence of an effect of education on the growth of children (low to very low‐quality evidence). We could not combine the results from different studies for diarrhoea, knowledge of caregivers and adequacy of complementary food. However, from the individual reports of the study authors, education led to a reduction in diarrhoea and an improvement in the knowledge of caregivers. It also led to improvement in the quality and quantity of complementary foods fed to infants.
Overall, we found evidence that education improves complementary feeding practices.
Authors' conclusions
Summary of findings
| Educational intervention versus no educational intervention for improving complementary feeding practices | ||||||
| Patient or population: children of complementary feeding age Comparison: no educational intervention | ||||||
| Outcomes | Illustrative comparative risks* (95% CI) | Relative effect | Number of participants | Quality of the evidence | Comments | |
| Assumed risk | Corresponding risk | |||||
| No educational intervention | Educational intervention (ICC = 0.02) | |||||
| Age at introduction of complementary foods | Study population | RR 0.88 | 1738 | ⊕⊕⊕⊝ | ‐ | |
| 661 per 1000 | 581 per 1000 | |||||
| Moderate | ||||||
| 746 per 1000 | 656 per 1000 | |||||
| Duration of exclusive breastfeeding (≥ 4 months of age) | Study population | RR 1.58 | 1544 | ⊕⊝⊝⊝ | ‐ | |
| 129 per 1000 | 204 per 1000 | |||||
| Moderate | ||||||
| 0 per 1000 | 0 per 1000 | |||||
| Duration of exclusive breastfeeding (≥ 4 months of age): community‐based intervention | Study population | RR 2.32 | 1167 | ⊕⊕⊝⊝ | ‐ | |
| 40 per 1000 | 92 per 1000 | |||||
| Moderate | ||||||
| 0 per 1000 | 0 per 1000 | |||||
| Duration of exclusive breastfeeding (≥ 4 months of age): facility‐based intervention | Study population | RR 0.95 | 377 | ⊕⊕⊝⊝ | ‐ | |
| 426 per 1000 | 405 per 1000 | |||||
| Moderate | ||||||
| 0 per 1000 | 0 per 1000 | |||||
| Hygiene practices: community‐based intervention | Study population | RR 1.38 | 2029 | ⊕⊕⊕⊝ | ‐ | |
| 546 per 1000 | 754 per 1000 | |||||
| Moderate | ||||||
| 0 per 1000 | 0 per 1000 | |||||
| *The basis for the assumed risk (e.g. the median control group risk across studies) is provided in footnotes. The corresponding risk (and its 95% CI) is based on the assumed risk in the comparison group and the relative effect of the intervention (and its 95% CI) | ||||||
| GRADE Working Group grades of evidence | ||||||
| aWe downgraded the quality of the evidence by one level due to serious risks of bias; the method of sequence generation, allocation concealment and blinding of outcome assessors was unclear or not undertaken in some of the studies | ||||||
| Educational intervention versus no educational intervention for improving complementary feeding practices: growth outcomes | ||||||
| Patient or population: children of complementary feeding age Comparison: no educational intervention | ||||||
| Outcomes | Illustrative comparative risks* (95% CI) | Relative effect | Number of participants | Quality of the evidence | Comments | |
| Assumed risk | Corresponding risk | |||||
| No educational intervention | Educational intervention (ICC = 0.05) | |||||
| Weight (at 6 months of age) | ‐ | The mean weight at 6 months of age in the intervention groups was 0.03 kg higher (0.10 lower to 0.17 higher) | ‐ | 1221 | ⊕⊝⊝⊝ | ‐ |
| Weight (at 12 months of age) | ‐ | The mean weight at 12 months of age in the intervention groups was 0.06 kg higher (0.04 lower to 0.15 higher) | ‐ | 2464 | ⊕⊝⊝⊝ | ‐ |
| Height/length (at 6 months of age) | ‐ | The mean height/length at 6 months of age in the intervention groups was 0.16 cm higher (0.21 lower to 0.52 higher) | ‐ | 1221 | ⊕⊝⊝⊝ | ‐ |
| Height/length (at 12 months of age) | ‐ | The mean height/length at 12 months of age in the intervention groups was 0.32 cm higher (0.11 to 0.52 higher) | ‐ | 2464 | ⊕⊕⊝⊝ | ‐ |
| Nutritional status: stunting (H/LAZ ≤ −2 SD) | 199 per 1000 | 177 per 1000 | RR 0.89 | 3487 | ⊕⊕⊝⊝ | ‐ |
| Nutritional status: wasting (WH/LZ ≤ −2 SD) | 400 per 1000 | 316 per 1000 | RR 0.79 | 2000 | ⊕⊕⊝⊝ | ‐ |
| Nutritional status: underweight (WAZ ≤ −2 SD) | 138 per 1000 | 136 per 1000 | RR 0.99 | 2900 | ⊕⊕⊝⊝ | ‐ |
| *The basis for the assumed risk (e.g. the median control group risk across studies) is provided in footnotes. The corresponding risk (and its 95% CI) is based on the assumed risk in the comparison group and the relative effect of the intervention (and its 95% CI) | ||||||
| GRADE Working Group grades of evidence | ||||||
| aWe downgraded the quality of the evidence by two levels due to very serious risks of bias; the method of sequence generation, allocation concealment and blinding of outcome assessors was unclear or not undertaken in most of the studies | ||||||
Background
Description of the condition
Complementary feeding is defined as, "the process starting when breast milk alone or infant formula alone is no longer sufficient to meet the nutritional requirements of infants, and therefore, other foods and liquids are needed, along with breast milk or a breast‐milk substitute" (WHO 2008, p v). It is the period of transition from breast milk or breast‐milk substitute to family foods, and entails, "introducing a range of foods gradually until the baby is eating the same foods as the rest of the family" (UNICEF 2008, p 3; WHO 2015a).
Although complementary feeding is a universal practice, the methods and manners in which it is practiced vary between cultures, individuals, and socioeconomic classes. For example, although the recommended time for initiation of complementary foods is six months of age (World Health Assembly 2001), when breast milk alone is insufficient for the infant, some caregivers may initiate complementary feeding before this time for personal or cultural reasons. Alternatively, some caregivers may give teas or sugary drinks to infants based on personal reasons or the influence of family members or peers (Black 2001). Therefore, although complementary feeding may be defined in different ways based on these variances, for the purpose of this review we will adopt the WHO 2008 definition of complementary feeding stated above.
Most babies at the age of six months are developmentally prepared for the consumption of other foods. As this period is usually characterised by increases in the nutritional needs of the infants for growth and physiological development, and as breast milk alone or breast‐milk substitute alone are insufficient for meeting these requirements, complementary feeding is needed (World Health Assembly 2001).
Complementary foods are, ''any food or liquids, whether manufactured or locally prepared, suitable as a complement to breast milk or to a breast‐milk substitute, fed to infants during the complementary feeding period" (WHO 2008, p v). This should not include drinks and beverages that are low in nutrient content, like coffee, teas, and sugary drinks like soda. Coffee and teas also contain compounds that can inhibit the absorption of iron (PAHO/WHO 2003). Proper complementary feeding is essential for healthy growth, survival and the attainment of a child's human potential (PAHO/WHO 2003). The introduction of complementary foods should be timely and adequate in nutritional content, tailored to meet the age‐specific needs of the infant, and should provide all the micronutrients and vitamins needed by infants for adequate growth and cognitive development. In settings where complementary foods lack basic micronutrients, there may be a need for food fortification and micronutrient supplementation to boost the dietary content of these foods (Lutter 2003; PAHO/WHO 2003). Vitamin supplements given to babies as part of recommended public health interventions are not considered part of complementary feeding.
The period starting from birth to two years of age has been identified as a critical period in the life of infants for the promotion of optimal growth, health and development (Shrimpton 2001; Victora 2008), and poor nutrition at this stage will result in malnutrition in many infants (WHO 2008). Most incidents of stunting occur in the first two years of life when there is increased demand for adequate nutrition to fuel infant growth and physiological development (Shrimpton 2001). Inappropriate complementary feeding practices during this period, such as early onset of complementary foods, inadequate nutritional content of complementary foods and poor hygiene behaviours, have been identified as the leading causes of undernutrition, growth faltering, diarrhoea, increased rate of infections, vitamin‐mineral deficiency, poor cognitive development and increased mortality among children (Motarjemi 1993; WHO 2012a; WHO 2015a). Undernutrition results from poor dietary intake and repeated infections and, “occurs when infants do not eat (or absorb) enough nutrients to cover their needs for energy and growth, or to maintain a healthy immune system” (Burgess 2012, p 1). An undernourished infant, "can no longer maintain natural bodily capacities, such as growth, resisting infections and recovering from disease" (UNICEF 2006, p 1). Undernutrition can have far‐reaching implications for the infant that can persist throughout his or her lifespan. Stunting that occurs during the first two to three years of a child's life is irreversible (Martorell 1994; Shrimpton 2001), and chances are high that a malnourished girl child would give birth to a malnourished and low‐birth‐weight infant (PAHO/WHO 2003). Malnutrition is responsible directly or indirectly for over half of all childhood deaths globally (WHO 2012a), with 45% of childhood deaths associated with undernutrition. More than two‐thirds of undernutrition‐associated deaths happen in the first year of life, and are usually correlated with poor complementary feeding practices (WHO 2003). A number of epidemiological studies have traced a nexus between poor complementary feeding practices, malnutrition and stunting in young children (Arimond 2004; Black 2008; Philips 2000; Shrimpton 2001). Black 2008 identifies suboptimum complementary feeding to be a causal factor of stunting and states categorically that, "even with optimum breastfeeding children will become stunted if they do not receive an adequate quantity and quality of complementary foods after 6 months of age" (p 251). Also, many studies have reported that the incidence of diarrhoeal disease is especially high after complementary feeding is initiated due to bacterial contamination (Black 1982; Henry 1990; Motarjemi 1993; Sheth 2006). Bacterial contamination can result from complementary foods of poor quality and improper food handling practices, which include unhygienic preparation, storage and preservation of complementary foods Motarjemi 1993.
In 2016, about 155 million children under five years of age were estimated to be stunted while 52 million children were estimated to be wasted (WHO 2018). It is reported that two out of five children in low‐income countries are stunted, "while 50‐70% of the burden of diarrhoeal diseases, measles, malaria and lower respiratory tract infections in childhood is attributable to undernutrition" (WHO 2003, p v). Diarrhoeal disease, which is the second‐leading cause of death in children aged from birth to 59 months, accounts for about 760,000 deaths in children under five years of age annually (Fischer Walker 2012; Fischer Walker 2013; Kosek 2003; WHO 2013a).
A number of factors have been identified to influence complementary feeding practices. Studies conducted in Bangladesh (Kabir 2012), Ireland (Tarrant 2010), and Tanzania (Victor 2014), found that the socioeconomic status of caregivers, maternal education level and age, opinions of family and friends, traditional feeding practices, influence of social network, father's occupation, postnatal care, and lack of professional advice influence complementary feeding practices. Some of the problems commonly associated with complementary feeding include starting complementary feeding too early, poor nutrient content of complementary foods, inadequate feed rations, insufficient breastfeeding, poor feeding practices, poor hygiene, and bacterial contamination of complementary foods and feeding utensils. Studies show that about 20% of mothers in the USA and Ireland introduce solid foods to their infants before four months of age (Fein 2008; Tarrant 2010). Recent studies from Nepal (Khanal 2013) and Tanzania (Victor 2014) report that an average of about 35% of complementary foods fed to infants in both countries met the minimum requirement for dietary diversity.
These variations or problems associated with complementary feeding, and the need to make safe the period of complementary feeding for the infant, necessitated the development of evidence‐informed guidelines for complementary feeding by the World Health Organization (WHO) and appropriate indicators to evaluate the process of complementary feeding (PAHO/WHO 2003). Caregivers need skilled support to provide adequate nutrition for their infants (WHO 2015a), and educational interventions to improve the timing and process of complementary feeding may be believed to be helpful in ensuring safe complementary feeding for infants. It is therefore necessary to evaluate the effects of educational interventions on the complementary feeding practices of caregivers of children of complementary feeding age.
Description of the intervention
In this review, educational interventions refer to health education interventions. Health education is defined by the WHO as, "consciously constructed opportunities for learning involving some form of communication designed to improve health literacy, including improving knowledge, and developing life skills, which are conducive to individual and community health" (WHO 1998, p 4). The Committee on Health Education and Promotion defines health education as, "any combination of planned learning experiences based on sound theories that provide individuals, groups and communities the opportunity to acquire information and the skills needed to make quality health decisions" (Gold 2002, p 3).
Health education interventions can be delivered to individuals or groups, face to face or by telephone in communities, hospitals, homes, schools, or organisations. They may be delivered by verbal, written or audiovisual means such as printed materials, multimedia (video messages, PowerPoint presentations), counselling sessions, practical demonstrations, lectures, and role plays (Ciciriello 2013; ILEP 1998; Nkhoma 2013). Within this review, we define educational interventions as consciously planned interventions that seek to communicate information (verbal, written or audiovisual) to individuals, groups or communities, with the aim of improving their knowledge and life skills to enable them to make quality health decisions. These interventions are usually consciously planned and constructed based on sound theories.
Educational interventions are widely acknowledged as effective in promoting public health strategy (Brunello 2012; Higgins 2008; Shah 2009). They have been used to prevent diseases; help patients or their caregivers to effectively manage health conditions; and improve or encourage adoption of healthy lifestyles, practices, and behaviours in individuals and the community (Darity 1997; Fredericks 2013; Hunter 2010; Ofotokun 2010; Saunders 1986). Educational interventions for improving weaning practices provide information about proper weaning practices (proper timing for initiation of complementary feeding; continuation of breastfeeding after introduction of semisolid foods; hygiene; composition, amount, consistency, and frequency of complementary food; and feeding of the infant during or after illness; to caregivers of infants/children (PAHO/WHO 2003). (We define caregivers as mothers, guardians or other family members responsible for caring for and feeding the infant, and personnel charged with the responsibility of looking after infants in childcare centres).
A number of studies suggest that educational interventions can be used to improve complementary feeding practices (Monte 1997; Roy 2007). Guldan 2000 and Kilaru 2005 reported that counselling sessions on appropriate complementary feeding practices improved outcomes such as growth of infants, infant feeding practices, and knowledge of mothers. Studies by Hotz 2005 and Saleem 2014 found that lectures or nutritional messages delivered to caregivers of infants were effective in improving energy intake and growth of infants. In Black 2001, an educational videotape intervention integrated into home visits improved time of initiating complementary feeding among adolescent mothers, while in Guldan 2000 and Yin 2009, lectures and counselling improved nutritional knowledge of caregivers. Nutrition education through focus group discussions have also been reported to be effective in preventing malnutrition and growth faltering in children under two years of age (Roy 2007).
How the intervention might work
Educational interventions essentially seek to achieve change in knowledge, attitudes, and behaviours by providing information, opportunities, or both, for participants to acquire or improve the skills required for the desired change. The scientific rigour and potential effectiveness of health promotion interventions depend on the availability of an evidence‐informed theoretical framework that can inform their design and implementation. Research suggests that health promotion and public health interventions built on social behavioural theories, such as the theory of planned behaviour, the health belief model, social cognitive theory, social ecological model, amongst others, are likely to be more effective than those that do not have strong theoretical foundations (Bluethmann 2017; Davis 2015; Glanz 2010; NCI 2005). This is more so if the theoretical models used include appropriate explanatory as well as action models, and provide a broad framework that addresses interpersonal, organisational, and environmental factors that influence health behaviour and not just the individual (Glanz 2014).
According to McLeroy's ecological model for health promotion, health behaviour is said to be influenced by five major factors or processes, namely intrapersonal, interpersonal, institutional (or organisational), community, and public policy factors (McLeroy 1988). Institutional, community, and public policy factors together constitute environmental factors (WHO 2012b). Intrapersonal factors include the attitudes, beliefs, skills, self‐efficacy and self‐concept of the individual. Interpersonal factors that influence health behaviour comprise the formal and informal social networks and support systems of an individual such as family members, peers or friends, or work group. Organisational or institutional factors include social institutions or organisations that provide formal (and informal) rules and regulations for operation, while community factors include social networks or norms (formal and informal) among individuals, groups or organisations. Public policy factors are local, state and federal laws and policies that promote healthy behaviours.
Educational interventions, which are expected to be effective in promoting health behaviours, must therefore seek to address not only intrapersonal factors, such as knowledge, attitudes and beliefs of individuals, but must also take cognisance of interpersonal and environmental factors. The way the intervention works can be explained using the theory of planned behaviour, which states that the likelihood that an individual will adopt a new behaviour is determined by his or her 'intention' to perform that behaviour, which in turn is influenced by his or her attitude, subjective norms and perceived behavioural controls (Ajzen 1991). Attitudes refer to an individual's positive or negative attitudes towards the desired behaviour. Subjective norms are the social pressures the individual experiences to adopt or avoid the desired behaviour (that is, how others view the behaviour). Perceived behavioural controls are a person’s perception of their ability to perform a given behaviour. Interventions that seek to improve complementary feeding practices are likely to focus on inducing and sustaining behaviour change that will minimise the risk of undernutrition and diarrhoea, which have been identified as the key morbidity consequences of poor complementary feeding practice. As a first step, these interventions may involve interfacing with communities to identify the common challenges associated with complementary feeding, which may include understanding their perceptions and constraints in adopting adequate complementary feeding practices (USAID 2011). The outcome of this often reveals knowledge gaps and deficiencies in practice, which are usually amenable to educational interventions specifically tailored to address the knowledge gaps and complementary feeding problems that have been identified (Gibbons 1984). The explanatory model would therefore be expected to explain the mechanisms and steps through which known undesirable behaviours (inappropriate complementary feeding practices) cause undernutrition, diarrhoea and other childhood problems, and also provide unambiguous information on the benefits of appropriate complementary feeding practices, which is expected to stimulate the adoption of appropriate complementary feeding practices. On the other hand, the action model would show how the proposed interventions would eliminate barriers or induce positive actions that would reverse or prevent the mechanisms that lead to diarrhoea or undernutrition during complementary feeding. Critical appraisal of studies included in this review will extract and report information on the use and appropriateness of theoretical models based on these basic constructs.
Educational interventions to improve complementary feeding practices that provide knowledge alone, without addressing barriers as a result of social norms and perceived behavioural controls, may not be effective in improving complementary feeding practices. Interventions may therefore seek to address social norms, such as cultural practices, which may pose as barriers to adopting recommended complementary feeding practices, and to improve self‐efficacy of caregivers by boosting their confidence and improving their skills to take action and, if need be, change their physical and social environments to aid behaviour change (USAID 2011).
In line with the theory of planned behaviour, a number of empirical studies have shown that attitudes, normative influences, and perceived behavioural controls influence breastfeeding and complementary feeding practices of caregivers (Hamilton 2011; Swanson 2005; Walingo 2014; Zhang 2009). The theory of planned behaviour agrees with McLeroy's ecological model for health promotion in that it proposes that the individual's intention to perform a health behaviour is determined by attitudes of the individual (intrapersonal factors), social norms, and perceived behavioural controls (interpersonal and environmental factors).
We have presented an example of a logic model or theory of change in Figure 1, which illustrates educational interventions to improve complementary feeding practice based on the health belief model. The health belief model hypothesises that a person’s decision to take a recommended health action is determined by their perceived susceptibility to the health problem, perceived severity of problem, perceived benefits of the health action, and perceived barriers to adopting the recommended action, as well as cues to action and self‐efficacy (Janz 1984; Rosenstock 1974). According to this model, knowledge about dangers or benefits (or both) of a health action (in this case proper complementary feeding practices), as well as self‐efficacy, determine a person’s decision to take the recommended action.
Theoretical model: educational interventions for improving complementary feeding practices
Footnotes
aGSM: global system for mobile communication.
Caregivers with improved knowledge, skills, and self‐efficacy are more likely to practice better hygiene in food preparation, as well as ensure proper composition of complementary diets. Improved complementary foods will lead to reduced incidence of undernutrition, diarrhoea, and growth faltering (Monte 1997; Shi 2011).
Why it is important to do this review
The period of complementary feeding is a critical time of transition in the life of an infant, and inappropriate complementary feeding practices, with their associated adverse health consequences, remain a significant, global public health problem. A recent review of the epidemiology of global nutrition identified poor complementary feeding practices as major contributors to undernutrition and increased rates of infections in children under five years of age, and has proposed improvement in complementary feeding practices along with promotion of breastfeeding and micronutrient supplementation as strategies for combating undernutrition (Bhutta 2012). We can therefore expect that educational interventions aimed at improving complementary feeding practices would reduce the risk of malnutrition and food‐borne infections, especially diarrhoeal diseases.
A number of reviews have been conducted to evaluate the effectiveness of complementary feeding interventions, but none have been conducted to evaluate the effectiveness of educational interventions in promoting appropriate or recommended complementary feeding practices. Dewey 2008 conducted a non‐Cochrane systematic review on ‘The efficacy and effectiveness of complementary feeding interventions in developing countries'. This study did not focus on educational interventions, but looked broadly at different types of complementary feeding strategies. In addition, the authors only included studies conducted between 1996 and 2006 in the review, and they have not updated it to include studies from 2007 to date. Imdad 2011 and Lassi 2013 conducted two other non‐Cochrane systematic reviews assessing the impact of education and the provision of complementary feeding on growth and morbidity in children. Although the studies included children under two years of age, they were limited to low‐ and middle‐income countries and were not based strictly on randomised studies. Shi 2011 conducted a literature review on ‘Recent evidence of the effectiveness of educational interventions for improving complementary feeding practices in developing countries’ from 1998 onward. The systematic reviews listed above focused on growth and morbidity (stunting), but did not assess the effects of these interventions on behavioural outcomes and changes in knowledge of infant caregivers.
This Cochrane Review aims to summarise evidence on the effectiveness of educational interventions to improve complementary feeding practices of caregivers of infants. We will not limit the review to studies from low‐ and middle‐income countries alone, but will also include studies from high‐income countries. In addition to growth and morbidity outcomes, we will assess a number of other key outcomes, including changes in complementary feeding behaviour and knowledge of caregivers. This review will provide useful information on which educational intervention approaches are effective for promoting recommended complementary feeding practices.
Objectives
To assess the effectiveness of educational interventions for improving the complementary feeding (weaning) practices of primary caregivers of children of complementary feeding age, and related health and growth outcomes in infants.
Methods
Criteria for considering studies for this review
Types of studies
We included randomised controlled trials (RCTs), including cluster‐RCTs.
Types of participants
Study participants comprised caregivers of infants aged 4 to 24 months undergoing complementary feeding. Pregnant women who were expected to give birth and commence complementary feeding during the period of the study were also included.
Caregivers were defined as mothers, guardians, or other family members responsible for caring for and feeding the infant.
Types of interventions
We included studies that compared:
-
educational intervention to no intervention or usual practice (e.g. usual weaning or child care practice); and
-
educational interventions provided in conjunction with another intervention (e.g. provision of complementary food), so long as the educational intervention was only available in the experimental group and the adjunctive intervention was available to the control group.
We defined educational interventions as comprising one or more of the following, delivered in any setting: multimedia, lectures, workshops, practical demonstrations, printed materials, skills training, counselling, campaigns, or other instructional methods (written, verbal, or audiovisual).
Types of outcome measures
Primary outcomes
-
Improved complementary feeding practices (measured as a continuous outcome or dichotomous outcome), of the following:
-
age at introduction of complementary foods;
-
duration of exclusive breastfeeding;
-
adequacy of complementary foods (measured by number of children fed with adequate amount and consistency of complementary foods, children fed with at least five different classes of food, consisting mainly of protein, carbohydrate, vegetable, fats and oils, fruits; vitamin supplementation (for infant and mother); energy density of complementary foods; and meal frequency (number of times children are fed in a day); or based on the WHO minimum acceptable diet, minimum dietary diversity, minimum meal frequency or as assessed by study authors); and
-
hygiene practices: safe preparation and storage of complementary foods (measured by handwashing practices (washing of caregiver's and child's hands with soap before cooking, feeding, or eating); water sanitation practices; food preparation and storage practices; serving foods immediately after preparation; using clean utensils, plates, pots, etc. for preparing or serving food and for feeding the child; and avoiding the use of feeding bottles).
-
-
Adverse events (as defined by study authors). For example, overburdening of personnel delivering the intervention who were also responsible for other tasks in the health facility, stress on caregivers.
Secondary outcomes
-
Growth (measured by weight, height/length, head circumference, mid upper‐arm circumference (MUAC), weight‐for‐age (WAZ), height/length‐for‐age (H/LAZ), weight‐for‐height/length (WH/LZ) z scores, etc.)
-
Incidence of malnutrition among participants (as defined by WHO guidelines: WHO 2013b)
-
Morbidity (measured by episodes of diarrhoea)
-
Mortality (indicated by all‐cause mortality, diarrhoea‐specific mortality, malnutrition‐associated mortality)
-
Hospitalisation (indicated by the number hospitalised, length or duration of hospital stay)
-
Change in knowledge (measured by a difference in the pre‐test (baseline) and post‐test (postintervention) results in the intervention and control arms)
We presented our primary outcomes in summary of findings Table for the main comparison, and our secondary outcomes in summary of findings Table 2.
Search methods for identification of studies
Electronic searches
In November 2017, we searched the following electronic databases and trials registers from inception onwards. We did not limit our searches by date, language or publication status. All of the search strategies are reported in Appendix 1.
-
Cochrane Central Register of Controlled Trials (CENTRAL; 2017, Issue 10) in the Cochrane Library, and which includes the Cochrane Developmental, Psychosocial and Learning Problems Specialised Register (searched 6 November 2017)
-
MEDLINE Ovid (1946 to October week 4 2017)
-
MEDLINE In‐process and Other Non‐indexed Citations Ovid (3 November 2017)
-
MEDLINE Epub Ahead of Print Ovid (3 November 2017)
-
Embase Ovid (1974 to 2017 week 45)
-
CINAHL EBSCO (Cumulative Index to Nursing and Allied Health Literature; 1937 to 6 November 2017)
-
Science Citation Index Web of Science: Clarivate Analytics (SCI; 1970 to 6 November 2017)
-
Social Sciences Citation Index Web of Science: Clarivate Analytics (SSCI; 1970 to 6 November 2017)
-
Conference Proceedings Citation Index ‐ Science Web of Science: Clarivate Analytics (CPCI‐S; 1990 to 6 November 2017)
-
Conference Proceedings Citation Index ‐ Social Science & Humanities Web of Science: Clarivate Analytics (CPCI‐SS&H; 1990 to 6 November 2017)
-
Cochrane Database of Systematic Reviews (CDSR; 2017, Issue 11) part of the Cochrane Library (searched 6 November 2017)
-
Database of Abstracts of Reviews of Effects (DARE; 2015, Issue 2 of 4; final issue) part of the Cochrane Library (last searched 1 July 2015)
-
LILACS (Latin American and Caribbean Health Science Information database; lilacs.bvsalud.org/en; searched 7 November 2017)
-
ClinicalTrials.gov (clinicaltrials.gov; searched 7 November 2017)
-
WHO International Clinical Trials Registry Platform (ICTRP; apps.who.int/trialsearch; searched 7 November 2017)
Searching other resources
We checked the reference lists of relevant studies and reviews identified by the electronic searches to identify any additional studies. In addition, we contacted relevant individuals and organisations for information about any ongoing or unpublished studies.
Data collection and analysis
Selection of studies
Two review authors (DA, MTC) independently screened titles and abstracts for eligibility, and obtained the full reports of any potentially relevant studies. The same review authors independently applied the inclusion criteria to the full reports using an eligibility form and scrutinised publications to ensure that we included each study in the review only once. We also contacted study authors for clarification if eligibility was unclear, and resolved disagreements through discussion with a third review author (EE or FO).
We listed studies that were excluded after their full‐texts were assessed and the reasons for their exclusion in Characteristics of excluded studies tables.
We recorded our decisions in a PRISMA study flow diagram (Moher 2009).
Data extraction and management
Two review authors (DA, MTC) independently extracted data on the following, using a specifically developed and piloted data extraction form.
-
General information about the study
-
Study characteristics, including study settings and characteristics of the participants
-
Methods and quality of the study, including duration of the study, study design, type of randomisation employed, inclusion and exclusion criteria, details of the control and comparison groups, description and number of participants, duration of follow‐up
-
Details of the intervention
-
How information was collected and outcome measures assessed
-
Results
Both review authors (DA, MTC) compared the extracted data for discrepancies and resolved any disagreements through discussion with all review authors. Where information was unclear or data were missing, we contacted the corresponding authors of identified publications (see section on Dealing with missing data).
DA entered relevant data into Cochrane's statistical software: Review Manager 5 (RevMan 5) (Review Manager 2014).
Assessment of risk of bias in included studies
Using the Cochrane 'Risk of bias' tool (Higgins 2017, Section 8.5, Table 8.5a), two review authors (DA, MC) independently assessed the risks of bias of each included study across the domains described below.
Sequence generation
Description: we examined the method used to generate the allocation sequence in sufficient detail to assess whether it would produce comparable groups.
Review authors' judgement: what is the risk of selection bias due to inadequate generation of a randomised sequence?
Allocation concealment
Description: we described the method used to conceal the allocation sequence in sufficient detail in order to assess whether intervention allocation schedules could have been foreseen in advance of, or during, recruitment.
Review authors' judgement: what is the risk of selection bias due to inadequate concealment of allocations prior to assignments?
Blinding of participants and personnel
Description: we examined the measures used, if any, to blind study participants and personnel from knowledge of which intervention a participant received and any information as to whether the intended blinding was effective.
Review authors' judgement: what is the risk of performance bias due to knowledge of the allocated interventions by participants and personnel during the study?
Blinding of outcome assessment
Description: we examined the measures used, if any, to blind outcome assessors from knowledge of which intervention a participant received and any information as to whether the intended blinding was effective.
Review authors' judgement: what is the risk of detection bias due to knowledge of the allocated interventions by outcome assessors?
Incomplete outcome data
Description: we examined the completeness of outcome data for each main outcome, including attrition and exclusions from the analysis, and if attrition and exclusions were reported. We also examined if the reasons for the attrition and exclusion, numbers in each intervention and control group, and any re‐inclusions in the analyses performed by the review authors were reported.
Review authors' judgement: what is the risk of attrition bias due to the amount, nature and handling of incomplete outcome data?
Selective outcome reporting
Description: we assessed how the study authors examined the possibility of selective outcome reporting and their findings.
Review authors' judgement: what is the risk of reporting bias due to selective outcome reporting?
Other bias
Description: we examined other sources of bias not covered by the 'Risk of bias' tool.
Review authors' judgement: what is the risk of bias due to issues not addressed in the other domains of the 'Risk of bias' tool?
We assigned ratings of low, high, or unclear risk of bias to each of the domains for each included study and recorded these ratings in the 'Risk of bias' tables (beneath the Characteristics of included studies tables). We assigned a low risk of bias to studies that provided adequate information to ascertain that the investigators used the appropriate methods to successfully reduce bias. We assigned a high risk of bias to studies that provided adequate information to ascertain that investigators did not use appropriate methods to reduce bias, and we assigned an unclear risk of bias to studies that did not provide adequate information to ascertain whether or not investigators used the appropriate methods to reduce bias (Higgins 2017, Section 8.5, Table 8.5d). We resolved any differences by discussion with all review authors.
We presented our judgements about each 'Risk of bias' item as percentages across all included studies (Figure 2), and summarised our assessment in a 'Risk of bias' summary graph (Figure 3).
Risk of bias graph: review authors' judgements about each risk of bias item presented as percentages across all included studies
Risk of bias summary: review authors' judgements about each risk of bias item for each included study
Measures of treatment effect
Dichotomous outcomes
We calculated risk ratios (RR) and 95% confidence intervals (CIs) for dichotomous outcomes, such as adequate hygiene (handwashing).
Continuous outcomes
We calculated mean differences (MD) and 95% CIs for continuous data measured using the same scale (e.g. kilograms (kg)). We did not calculate a standardised mean difference since outcomes were reported using the same scale.
See Arikpo 2015 and Table 1.
| Measures of treatment effect | Event rate outcomes In this review, it is possible that some outcomes (e.g. diarrhoea, hospitalisation, malnutrition) may have been recorded as counts where the event can occur multiple times to the same participant. Where study data allow (i.e. data are available on both events and person‐years at risk), we will calculate rate ratios for count outcomes. However, study authors can report count data in a number of ways. As such, our strategy will be to extract count data in the form as reported by the original authors. For example, if study authors have reported the outcome using a rate ratio, we will extract it as such. If study authors have reported the outcome as dichotomous, we will extract it as a dichotomous outcome, noting the potential disadvantages of doing so. Multiple outcome data It is possible that studies will summarise outcomes in several ways, for example, both as a continuous and dichotomous measure. For the primary outcomes, if person‐years at risk are available, our preference will be to analyse count data as a rate ratio. However, if sufficient information is not available, and the event is common, we will analyse count data as if it were continuous. We consider the continuous measure to be clinically reasonable and preferable to dichotomising the primary outcomes. If neither of these approaches is suitable, we will extract the data as if it were dichotomous, ensuring that we classify all participants into one of two possible groups only. |
| Unit of analysis issues | Multiple intervention groups Studies with more than two intervention arms can pose analytical problems in a meta‐analysis. For example, it is important to avoid 'double‐counting' of participants. Where studies may have two or more active arms to be compared against a control, or two control conditions versus an experimental condition, we will combine similar interventions to generate a single pair‐wise comparison for the meta‐analysis. If interventions are not similar, we will split the 'shared' comparator into two groups and include as two comparisons. |
| Dealing with missing data | If we are unable to retrieve missing dichotomous data, we will conduct an available‐case analysis. We plan to undertake a sensitivity analysis assuming that participants who withdrew from either arm after randomisation experienced a negative event. In common with many public health educational interventions, dropouts are often due to perceived difficulties with the intervention or information contradictory to existing beliefs or community norms (among other reasons). As such, it is not realistic to consider a 'best case' sensitivity analysis where all dropouts successfully adhered to the intervention, for weaning practice. We will analyse missing continuous data on a completers basis, including only those participants with a final assessment. Where we are unable to obtain the missing SDs from the study authors, we will calculate them from P values, t values, confidence intervals, or standard errors, where these have been reported. If this is not possible, and only a minority of studies are missing SDs, we will impute the SD using other studies in the meta‐analysis. We will also report the extent of the missing data, describe the attrition for each study in the 'Risk of bias' tables, and discuss the possible impact of this missing data on the results of the review. We will perform a sensitivity analysis to assess the impact of the inclusion of studies with missing data on the findings of the review (Deeks 2017, Section 9.7). |
| Assessment of reporting biases | We will try to minimise publication bias by doing a comprehensive search of multiple sources and databases, and by including studies of good methodological quality and data from unpublished and ongoing studies (Sterne 2017, Section 10.3). If we have a sufficient number of included studies (at least 10), we will use outcome data to run a funnel plot regression to investigate the possibility of publication bias (Sterne 2017, Section 10.4). Funnel plot asymmetry could be due to publication bias, poor methodological quality, true heterogeneity, or a real relationship between study size and effect size or chance. We will further investigate publication bias by comparing the data extracted from published and unpublished studies in a sensitivity analysis (Sterne 2017, Section 10.4.4) |
| Subgroup analysis and investigation of heterogeneity |
|
| Sensitivity analysis | We will conduct a sensitivity analysis in order to detect the effect of excluding studies with missing data, unpublished studies, and studies with high risk of bias (judged using Cochrane’s tool for assessing risk of bias (Higgins 2017)) on the overall results of the meta‐analysis. In this analysis, we will explore the possible effects of marked differences between included studies. We will also undertake a fixed‐effect meta‐analysis to determine the robustness of the results from the random‐effects meta‐analysis |
SD: standard deviation.
Unit of analysis issues
Multiple intervention groups
For studies with two or more intervention arms, we included only the intervention arm of interest (the arm that received educational interventions alone or educational interventions provided in conjunction with another intervention, so long as the educational intervention was only available in the experimental group and the adjunctive intervention was available to the control group) and the control arm.
Cluster‐randomised studies
For appropriately analysed studies, where the analysis was adjusted for clustering, we extracted data for the estimates of treatment effect, as reported by the study authors, to use directly in the meta‐analysis. However, for the majority of studies that reported results at the individual level without explicitly accounting for clustering, we followed the guidance on inflating the standard error for incorporating cluster‐randomised studies in meta‐analyses, as reported in the Cochrane Handbook for Systematic Reviews of Interventions (Higgins 2011, Section 16.3). In order to calculate the design effect, we used the original randomised sample at baseline for both dichotomous and continuous outcomes. Where the study investigators did not report the intra‐cluster correlation coefficient (ICC), number of clusters, or mean cluster size, we contacted them in the first instance to request the additional information. If the ICC was not available, we used estimates from similar studies included in the review or appropriate external studies. We considered sensitivity analyses for a range of ICCs (see Sensitivity analysis section). If information on cluster size was unavailable, we excluded the study from the meta‐analysis.
Dealing with missing data
We contacted the contact authors of included studies to retrieve missing data needed for analysis up to three times, and included the data in the analyses. We describe the attrition for each study in the Characteristics of included studies and 'Risk of bias' tables.
We included dichotomous outcomes in the main analysis on an intent‐to‐treat basis, where we assumed missing participants did not experience the event. However, we examine this assumption in a best‐worse case sensitivity analysis (see Sensitivity analysis section). For continuous outcomes we analysed data for completers only.
Assessment of heterogeneity
We assessed clinical and methodological heterogeneity by examining study characteristics such as design; setting; participant; intervention; follow‐up; outcome measures; method of randomisation; sequence generation; allocation concealment; and blinding of outcome assessors, interventions, or outcome measures. The similarities and differences between included studies in terms of these study characteristics are discussed in the Results section. Due to concerns regarding the low power of the Chi2 test, we also report the Tau2 and I2 statistics in the main text. Tau2 provides an estimate of the between‐study variance in a random effects meta‐analysis. I2 describes the proportion of variation in the estimates of intervention effect that is attributable to heterogeneity, rather than sampling error (Higgins 2003). We had planned to use the guideline ranges reported in the Cochrane Handbook for Systematic Reviews of Interventions for the interpretation of the I2 (Deeks 2017), where a I2 value of 0% to 40% may indicate non‐important heterogeneity, 30% to 60% may indicate moderate heterogeneity, 50% to 90% may indicate substantial heterogeneity, and 75% to 100% may indicate considerable heterogeneity (Section 9.5.2). However, having too few studies in a meta‐analysis can present challenges for the estimation of heterogeneity, which may not be reliable when only two or three studies are available. As such, we did not apply the I2 ranges as specified in the protocol (Arikpo 2015). Where heterogeneity was observed (e.g. I2 greater than 50%, with consideration of the direction of effects and strength of evidence for heterogeneity (P value)), we had also planned to conduct a subgroup analysis to investigate possible explanations (see Subgroup analysis and investigation of heterogeneity section). However, as few studies were available for meta‐analysis, we report subgroup analyses for illustrative purposes only.
Assessment of reporting biases
We were unable to assess reporting bias using a funnel plot analysis as planned, due to the insufficient numbers of studies included in each category of the meta‐analyses. Our strategy for assessing reporting biases in future updates of this review is documented in our protocol (Arikpo 2015) and also presented in Table 1.
Data synthesis
We performed a meta‐analysis to obtain the overall estimate of the effect of educational interventions when more than one study was sufficiently comparable in terms of methodology, population and outcomes. We compared the information extracted for each study in the Characteristics of included studies tables to determine whether the quantitative combination of studies was appropriate. Where data were from individually‐randomised, parallel‐group studies, we conducted the meta‐analysis using RevMan 5 (Review Manager 2014), employing the random‐effects model, since we had anticipated the possibility of substantial clinical heterogeneity, given the nature of educational interventions. We used the Mantel‐Haenszel method for dichotomous outcomes, and the inverse variance method for continuous outcomes. However, where we needed to account for clustering in studies (Unit of analysis issues), we followed the guidance in the Cochrane Handbook for Systematic Reviews of Interventions (Higgins 2011, section 16.3.6), and combined studies using the generic inverse variance approach in RevMan 5 (Review Manager 2014).
We provided a narrative summary for outcomes where a meta‐analysis was not feasible. This was for two reasons:
-
either insufficient statistics were reported/provided by an individual study to enable a calculation of an effect estimate; or
-
the study‐reported outcome was incompatible with the others in the meta‐analysis.
In both cases, we report the fullest information possible as extracted from the individual study report, that is, where an effect estimate was not provided or was possible to calculate, we state this in the text. We also clearly annotate extracted metrics as 'study author‐reported'.
Asessment of the quality of evidence
Using the GRADE approach, we assessed the quality of evidence for each outcome pooled in the meta‐analysis, according to the presence of the following five factors: risk of bias, consistency, directness, precision, and publication bias (Guyatt 2008). We exported data from RevMan 5 (Review Manager 2014) to GRADEprofiler GDT (GRADEpro GDT 2015) to produce 'Summary of findings' tables for the comparisons: educational intervention versus no educational intervention for improving complementary feeding practices and educational intervention versus no educational intervention for improving complementary feeding practices: growth outcomes. We included the following outcomes in these tables.
summary of findings Table for the main comparison: Improved complementary feeding practices
-
Age at introduction of complementary food
-
Duration of exclusive breastfeeding
-
Hygiene practices
summary of findings Table 2: growth outcomes
-
Weight at 6 month
-
Weight at 12 months
-
Height/length at 6 months
-
Height/length at 12 months
-
Nutritional status: stunting
-
Nutritional status: wasting
-
Nutritional status: underweight
Subgroup analysis and investigation of heterogeneity
We conducted the following subgroup analysis for the study setting.
-
Setting: community‐based studies and facility‐based studies
There were insufficient studies to perform a subgroup analysis for educational intervention delivery strategy. We were also unable to conduct subgroup analyses for educational intervention focus/message because the intervention focus/messages of the studies overlapped with the different aspects of complementary feeding. These analyses have been archived for use in future updates of this review (see Arikpo 2015; Table 1).
Sensitivity analysis
Due to the limited number of studies we were able to include in our meta‐analyses, we did not conduct the planned sensitivity analyses to detect the effect of excluding studies with missing data, unpublished studies, and studies with high risk of bias (judged using Cochrane’s tool for assessing risk of bias; Higgins 2017) on the overall results of the meta‐analysis. These have been archived for use in future updates of this review (see Arikpo 2015; Table 1).
We conducted sensitivity analyses for the primary outcomes only, to investigate the impact of assuming an alternative ICC on the summary effect estimates.
Results
Description of studies
See: Characteristics of included studies; Characteristics of excluded studies; Characteristics of studies awaiting classification; and Characteristics of ongoing studies.
Results of the search
The search strategy identified 11,079 records while our search of other sources yielded 11 records for possible inclusion. We identified 10,880 records for further consideration after removing 210 duplicates. After screening titles and available abstracts, we excluded 10,766 records and assessed 114 full‐text reports for eligibility. Three of these full‐text reports were published in other languages (Koehler 2007; Vitolo 2005; Yin 2009), and were translated to English for data extraction. We included 23 studies from 35 reports, excluded 51 studies from 57 full‐text reports with reasons (Excluded studies), categorised 10 other studies (from 12 reports) as awaiting classification because we were unable to obtain their full‐text reports, and identified 10 ongoing studies. See Figure 4.
Study flow diagram
Included studies
Details of the 23 included studies are summarised in the Characteristics of included studies tables.
Design
Of the 23 studies that met the inclusion criteria, 13 were cluster‐RCTs (Aboud 2008; Aboud 2009; Aboud 2011; Bhandari 2004; Campbell 2013; Kang 2017; Penny 2005; Reinbott 2016; Saleem 2014; Schroeder 2015; Shi 2010; Tariku 2015; Vazir 2013), while 10 were individually randomised (Bhandari 2001; Daniels 2012; de Oliveira 2012; Edward 2013; Koehler 2007; Negash 2014; Olaya 2013; Vitolo 2005; Wen 2011; Yin 2009).
Ten of the cluster‐RCTs reported using appropriate statistical approaches to allow for clustering in the analysis (Aboud 2008; Aboud 2009; Aboud 2011; Bhandari 2004; Campbell 2013; Kang 2017; Reinbott 2016; Saleem 2014; Tariku 2015; Vazir 2013). However, not all outcomes from these studies were reported as having allowed for the effect of clustering. One study did not appear to have adjusted for clustering (Schroeder 2015). One study reported that they omitted the ICC in the final analyses as it did not impact on results (Shi 2010), while another study stated that the outcomes were reported at an individual level and not at the cluster level (Penny 2005). In order to include these three studies in our analyses (Schroeder 2015; Shi 2010; Penny 2005), we calculated effective sample sizes and inflated the standard errors in accordance with the approximate approach outlined in section 16.3.4 of the Cochrane Handbook for Systematic Reviews of Interventions (Higgins 2011).
Settings
Five studies were conducted in high‐income countries: Australia (Campbell 2013; Daniels 2012; Wen 2011), Germany (Koehler 2007) and the USA (Schroeder 2015). Six studies were conducted in upper‐middle‐income countries: Brazil (de Oliveira 2012; Vitolo 2005), China (Shi 2010; Yin 2009), Colombia (Olaya 2013), and Peru (Penny 2005). Eight studies were conducted in lower‐middle‐income countries: Bangladesh (Aboud 2008; Aboud 2009; Aboud 2011), Cambodia (Reinbott 2016), India (Bhandari 2001; Bhandari 2004; Vazir 2013), and Pakistan (Saleem 2014). Three studies were conducted in a low‐income country: Ethiopia (Kang 2017; Negash 2014; Tariku 2015). The location of one study was not explicitly stated in the study report (Edward 2013).
Of these studies, 19 were community‐based (Aboud 2008; Aboud 2009; Aboud 2011; Bhandari 2001; Bhandari 2004; Campbell 2013; Daniels 2012; de Oliveira 2012; Edward 2013; Kang 2017; Negash 2014; Reinbott 2016; Saleem 2014; Shi 2010; Tariku 2015; Vazir 2013; Vitolo 2005; Wen 2011; Yin 2009), while four studies were facility‐based (Koehler 2007; Olaya 2013; Penny 2005; Schroeder 2015).
Eight studies were conducted in urban settings (Daniels 2012; de Oliveira 2012; Edward 2013; Koehler 2007; Olaya 2013; Schroeder 2015; Vitolo 2005; Wen 2011), two in peri‐urban settings (Penny 2005; Saleem 2014), one in an urban slum (Bhandari 2001), one in local government areas (Campbell 2013), and 11 in rural settings (Aboud 2008; Aboud 2009; Aboud 2011; Bhandari 2004; Kang 2017; Negash 2014; Reinbott 2016; Shi 2010; Tariku 2015; Vazir 2013; Yin 2009).
Participants
Twenty‐three studies, including 11,170 caregiver‐child pairs met the inclusion criteria (Criteria for considering studies for this review). Nineteen studies included mother/caregiver‐child pairs, three studies enrolled pregnant women (Edward 2013; Penny 2005; Vazir 2013), and one study enrolled first‐time mothers (Wen 2011). The range of the sample size was 85 to 2064 caregivers, while that of the cluster size was 4 to 60 clusters.
All outcomes were assessed in children except for adverse events, which were assessed in both children and caregivers, and knowledge outcomes, which were assessed in caregivers. The ages of the children ranged from birth to 24 months with 10 studies including newborn infants.
Interventions
See Table 2 and Table 3 for details of the educational interventions.
| Study | Promotional activity | Message content | Ways information was collected/outcome measure assessed | Intervention providers | Delivery (e.g. mechanism, medium, intensity, fidelity) |
|
|
|
| Peer educators | During weekly group sessions | |
|
|
|
| Peer educators | Group training sessions held weekly | |
|
|
|
| Peer educators | Group training sessions held weekly | |
| Counselling sessions using a nutritional counselling guide book | Not described |
| Trained nutritionists | Monthly counselling sessions | |
|
|
|
|
| Counselling on complementary feeding conducted as follows:
| |
|
| Intervention materials incorporated 6 purpose‐designed key messages (for example, “Color Every Meal With Fruit and Veg,” “Eat Together, Play Together,” “Off and Running”) within a purpose‐designed DVD and written materials |
| Dietician | 6 x 2‐h sessions delivered quarterly at first‐time parents’ group regular meeting | |
|
| Messages in:
|
|
| Interactive group sessions at a choice of days and times, and at the same child health centres as those used for measurements | |
|
|
|
|
| The counselling sessions occurred in the maternity ward close to the time for hospital discharge and at 7, 15, 30, 60, and 120 days after the birth at the mother's home | |
|
| Doulas discouraged the introduction of solid food during the early months of life for both breast‐fed and formula‐fed infant |
| Doulas |
| |
|
| Mothers discussed messages around:
with the operators |
| Female operators | During group nutrition education sessions | |
|
|
|
|
| The counselling was carried out during education sessions in the community | |
|
|
|
| Trained community nutrition promoter (CNP) together with local NGO conducted the nutrition education sessions | The 7 nutrition education sessions were held 2–4 hours weekly or biweekly depending on the availability of the participants | |
|
|
| Unclear | 2 female research assistants (with at least 14 years of schooling) and 2 female community health workers (with at least 10 years of schooling) | Interventions were offered in participants' homes | |
|
| Not described |
| Healthcare providers in the intervention areas |
| |
|
|
| Interviews using questionnaires |
|
| |
|
|
|
| High‐school‐educated village women who were themselves mothers | Home visits | |
|
|
Intervention messages were based on the “Ten steps for healthy feeding for Brazilian children from birth to 2 years of age” |
| Trained field workers who were undergraduate students in nutritional sciences | Home visits | |
| Counselling sessions on infant feeding practices, infant nutrition and active play, family physical activity and nutrition, as well as social support |
|
| Trained research nurses | Home visits | |
|
| Mothers were educated with feeding guideline on infants and young children
| ‐ | Experts in maternal and child nutrition | ‐ |
NGO: non‐governmental organisation; TN: study number
| Study | Promotional activity | Message content | Ways information was collected/outcome measure assessed | Intervention providers | Delivery (e.g. mechanism, medium, intensity, fidelity) |
|
| Nutrition counselling was based on the Dietary Schedule for the First Year of Life (Dietary Schedule) recommended by the Nutrition Committee of the German Pediatric Society. Recommendations of the schedule include:
|
| Counsellors | Telephone calls and printed materials | |
|
| Guidelines focused on the following 3 main messages that were emphasised at all study visits:
Mothers were offered specific advice on the number of portions of meat that should be given; mothers were also advised to include chicken liver and heart as affordable forms of meat, and suggestions were given for the preparation of recommended foods. Mothers were also advised to give fruit and vegetables daily |
| Researchers | Clinic visits | |
|
|
|
| Health workers | Health facility | |
|
| The intervention was based on the modules of Growing Leaps and Bounds, a set of educational materials developed by a group of experts and funded by the Dannon Institute. These materials aim at:
While the brochures emphasise a few key points, they also provide detailed advice on infant feeding practices, physical activity, and developmental milestones related to eating patterns |
|
| Paediatric visits at 1, 2, 4, 6, 9, 12, 15, 18, and 24 months of age and at annual visits thereafter up to 5 years of age |
S/N: study number
Five studies had multiple intervention arms. Aboud 2011 was a three‐arm study in which intervention group one received six‐weekly sessions on responsive parenting (feeding and stimulation) in addition to the regular programme, intervention group two received six‐weekly sessions on responsive parenting (feeding and stimulation) in addition to the regular programme and six months of a food powder fortified with minerals and vitamins, and the control group continued with the regular programme (standard care). We considered group one (weekly sessions on responsive feeding and parenting) versus standard care in this review. Vazir 2013 was also a three‐arm study where intervention group one (complementary feeding group) received the WHO recommendations on breastfeeding and complementary foods in addition to routine integrated child development services, intervention group two (responsive complementary feeding and play group) received the same intervention as the complementary feeding group plus skills for responsive feeding and psychosocial stimulation, and the control group received the routine Integrated Child Development Services (ICDS) ‐ standard care. In this review we considered group one versus standard care only. Bhandari 2001 was a four‐arm study where intervention group one received a milk‐based cereal and nutritional counselling, intervention group two received monthly nutritional counselling alone, intervention group three was the visitation group which received home visits for morbidity assessment only (used as the control group in the study), while the no‐intervention group were contacted at three time points for anthropometric measurements and dietary assessment. We considered intervention group two versus intervention group three for morbidity outcomes and intervention group two versus the control group for growth and dietary outcomes. Koehler 2007 was also a four‐arm study and had three intervention arms and one control arm. All of the intervention arms received nutritional counselling via telephone but the interventions were slightly varied among the intervention groups. Intervention group one received the intervention by means of a telephone hotline, which was accessible for two hours each, three times per week. Intervention group two received, "additional written information on the dietary schedule distributed in 3 parts, each dealing with the diet in the coming period" (p 107). Intervention group three received additional personal telephone counselling while the control group received no intervention. Tariku 2015 also had two intervention arms with one of the arms receiving educational interventions in line with the constructs of the health belief model, while the other group received educational intervention via the traditional (didactic) method. The control group was without intervention. We discussed the results of Koehler 2007 and Tariku 2015 narratively since all of the intervention arms received educational interventions. Details of these interventions are reported in Table 4.
| Study | Interventions |
| Intervention group 1 (RFS): 6 weekly sessions on responsive parenting (feeding and stimulation) in addition to the regular programme Intervention group 2 (RFS plus Sprinkles): 6 weekly sessions on responsive parenting (feeding and stimulation) in addition to the regular programme and 6 months of a food powder fortified with minerals and vitamins Control: regular programme | |
| Intervention group 1: received a milk‐based cereal and nutritional counselling Intervention group 2: monthly nutritional counselling alone Intervention group 3: visitation group (used as the control group in the study) | |
| Intervention group 1: were offered a telephone hotline 3 times per week, open for 2 hours each time Intervention group 2: received additional written information on the Dietary Schedule distributed in 3 parts, each dealing with the diet in the coming period Intervention group 3: were offered additional personal telephone counselling | |
| Intervention group 1: the complementary feeding group (CFG) received the integrated child development services plus the World Health Organization recommendations on breastfeeding and complementary foods Intervention group 2: the responsive complementary feeding and play group received the same intervention as the CFG plus skills for responsive feeding and psychosocial stimulation Control: routine Integrated Child Development Services ‐ standard of care |
RFS: responsive feeding and stimulation
All other studies were two‐arm studies with the intervention arms receiving educational interventions or nutritional counselling and the control groups receiving routine services (usual care) or no intervention or an agriculture intervention. The control group intervention was not described in detail in two studies (de Oliveira 2012; Penny 2005).
In the study by Reinbott 2016, the intervention group received nutrition education plus agriculture intervention, while the control group received agriculture intervention alone.
One study stratified the participating mothers into two groups, namely co‐habiting with the grandmother and not co‐habiting with the grandmother, before randomising into intervention or control arms (de Oliveira 2012).
The educational interventions' promotion activities included: group education or counselling sessions, demonstration or practical sessions and role plays (Aboud 2008; Aboud 2009); stories (Aboud 2008; Aboud 2009); use of posters (Aboud 2009; Bhandari 2004; Negash 2014; Reinbott 2016); flip charts (de Oliveira 2012; Penny 2005; Vazir 2013); work books (Daniels 2012); booklets and picture books (Aboud 2008; Aboud 2009; Bhandari 2004; de Oliveira 2012; Saleem 2014; Shi 2010; Vazir 2013; Vitolo 2005); flyers and leaflets (Olaya 2013; Penny 2005); brochures and post cards (Schroeder 2015); peer support (Aboud 2011; Campbell 2013); women's group meetings (Bhandari 2004); sharing meetings (Reinbott 2016); village rallies (Bhandari 2004); feeding recommendation cards (Bhandari 2004); video tapes (Campbell 2013; Edward 2013); telephone counselling (Edward 2013; Koehler 2007; Schroeder 2015); text messaging and mail outs (Campbell 2013). With the exception of two studies (Wen 2011; Yin 2009), all of the included studies used multiple promotion activities.
Intervention messages were centred on the appropriate time to introduce complementary foods; specific foods to be offered or avoided and how to offer them; meal frequencies; amounts of complementary foods to be fed to infants at different ages while continuing breastfeeding; offering a variety of foods from different food groups; family nutrition; health seeking; child nutrition during illness; hand washing at critical points; reading infant's signals by watching, listening and interpreting them, and being responsive to infant cues; and using or enriching locally available foods for complementary feeding.
The common sources of intervention information included messages developed by the implementing organization or researchers (Aboud 2008; Aboud 2009; Aboud 2011; Olaya 2013; Penny 2005), WHO/UNICEF (Saleem 2014), Dietary Schedule for the First Year of Life recommended by the Nutrition Committee of the German Pediatric Society (Koehler 2007), the Alive and Thrive programme (Negash 2014), Modules of Growing Leaps and Bounds (Schroeder 2015), Ten Steps to Healthy Feeding (Vitolo 2005), National Nutrition Programme and UNICEF in Cambodia (Reinbott 2016), and the Integrated Management of Childhood Illnesses training manual on nutrition counselling (Bhandari 2004).
Seven studies' reports stated explicitly that their respective studies were theory based. The theories deployed in these studies included social cognitive learning theory (Aboud 2008; Aboud 2009; Aboud 2011; Campbell 2013), the health belief model (Tariku 2015), the positive deviance approach (Kang 2017), and the cognitive behavioural approach (Daniels 2012). Other study reports did not specify whether or not they were theory‐based.
Comparators
The control arms in all of the included studies did not receive the educational intervention but rather continued with the routine care or regular programme or an agriculture intervention (one study, Reinbott 2016). This was also applicable to studies with more than one intervention arm.
Duration of the intervention
The duration of the interventions ranged from four to nine months (Aboud 2008; Aboud 2009; Bhandari 2001; Edward 2013; Negash 2014; Saleem 2014; Tariku 2015; Yin 2009), 10 to 20 months (Bhandari 2004; Campbell 2013; Kang 2017; Koehler 2007; Vazir 2013), two years (Penny 2005; Reinbott 2016), and eight years and four years respectively (Vitolo 2005; Wen 2011). It was unclear in six studies (Aboud 2011; Daniels 2012; de Oliveira 2012; Olaya 2013; Schroeder 2015; Shi 2010).
Outcomes and method of assessment
Outcomes commonly reported across the studies include the following.
Primary outcomes
-
Age at introduction of complementary foods: seven studies reported this outcome (Daniels 2012; de Oliveira 2012; Edward 2013; Reinbott 2016; Schroeder 2015; Vitolo 2005; Wen 2011). This outcome was assessed by information provided by the mothers/caregivers (self‐report) during home or hospital visits.
-
Duration of exclusive breastfeeding: four studies reported this outcome (de Oliveira 2012; Penny 2005; Vitolo 2005; Wen 2011). This outcome was assessed by information provided by caregivers (self‐report) during home or hospital visits.
-
Adequacy of complementary foods: 17 studies reported this outcome (Aboud 2008; Aboud 2009; Aboud 2011; Bhandari 2001; Bhandari 2004; Campbell 2013; Daniels 2012; Koehler 2007; Negash 2014; Olaya 2013; Penny 2005; Reinbott 2016; Schroeder 2015; Shi 2010; Tariku 2015; Vazir 2013; Vitolo 2005). This outcome was assessed by information on the types of foods fed to infants, mouthfuls consumed, energy intakes, diet scores, consistency of foods fed to infants, and dietary diversity. This information was usually provided by caregivers (self‐report) during home or hospital visits, dietary recalls, or records based on the observations of research assistants or field workers.
-
Hygiene practices: six studies reported this outcome (Aboud 2009; Aboud 2011; Bhandari 2004; Negash 2014; Shi 2010, Tariku 2015). This outcome was assessed by information provided by caregivers (self‐report) during home or hospital visits, and observations by research assistants or field workers during home visits.
Secondary outcomes
-
Growth: 12 studies reported this outcome (Aboud 2008; Aboud 2009; Bhandari 2001; Bhandari 2004; Campbell 2013; Daniels 2012; Olaya 2013; Penny 2005; Saleem 2014; Schroeder 2015; Vazir 2013; Vitolo 2005). This outcome was commonly assessed by anthropometric measurements during home or clinic visits.
-
Diarrhoea: four studies reported this outcome (Bhandari 2001; Bhandari 2004; Reinbott 2016; Vitolo 2005). This outcome was assessed by information provided by mothers/caregivers (self‐report) during home or hospital visits.
-
Hospitalisation: one study reported this outcome (Vitolo 2005). This outcome was assessed by information by provided by mothers/caregivers (self‐report) during home visits and medical/hospital records.
-
Knowledge: seven studies reported this outcome (Aboud 2008; Aboud 2009; Aboud 2011; Negash 2014; Penny 2005; Vazir 2013; Yin 2009). This outcome was assessed by messages recalled by caregivers, change in knowledge scores, and change in knowledge, attitude and practice scores.
In general, outcomes were commonly assessed across the studies via information provided by caregivers (self‐report) and observations by research assistants or field workers during home or hospital visits. Data collection methods included: records taken during home visits; use of questionnaires; structured face‐to‐face interviews during home or hospital visits; data retrieval from medical or hospital records; dietary recalls; anthropometric measurements during home or clinic visits; and observations by research assistants or field workers.
Anthropometric measurements were usually carried out by trained data collectors or by clinic or hospital staff. In addition to these methods, some studies also used telephone calls and standardised telephone interviews to collect data on outcomes of interest (Campbell 2013; de Oliveira 2012; Koehler 2007; Wen 2011).
Excluded studies
We excluded 51 studies (from 57 reports) after assessing the full‐text reports. These studies were mainly excluded on the basis of having an ineligible population, intervention or design. The excluded studies and the reasons for their exclusion are found in the Characteristics of excluded studies.
Studies awaiting classification
We grouped 10 studies as awaiting classification because we were unable to obtain their full‐text reports (Dunlevy 2010; Dunlvey 2012; Guan 2016; Jordan 2015; Palacios 2017; Paul 2011; Rabadi 2013; Savage 2010; Shafique 2013; Toure 2016). From their abstracts, the studies included mothers of infants from birth to two months of age; mother‐infant dyads; full‐term, low birth‐weight infants; and rural women who were pregnant or had a child under two years of age. Common interventions included nutrition education, nutrition, health and hygiene education, soothe and sleep interventions, messages for improving feeding practices delivered via short mobile messages (SMS), and infant weaning talks. Some of these interventions were delivered with additional interventions such as agricultural interventions, home gardening, provision of hand sanitisers, provision of micronutrient powders and gender sensitisation. Details of these studies can be found in the Characteristics of studies awaiting classification tables.
Ongoing studies
We identified 10 ongoing studies that are likely to meet our inclusion criteria (Campbell 2016; Cloutier 2015; Helle 2017; Hernes 2013; Horodynski 2011; Horodynski 2015; Kimani‐Murage 2013; Kulwa 2014; SHINE Team 2015; Wasser 2015). These studies were either cluster‐RCTs or RCTs. Some of these studies included first‐time parents of infants less than two years of age or infants less than two years of age and their mothers or caregivers, while others included pregnant women in their last trimester. The common interventions in these studies included educational interventions delivered via web‐based materials, written sources, telephone contacts, face‐to‐face sessions, home visits, skill‐set training, personalised home‐based counselling, cooking courses, etc. The interventions were delivered by dieticians or community health workers. All the control arms received usual care except one study that had an attention control that received safety education. Details of these studies can be found in the Characteristics of ongoing studies tables.
Risk of bias in included studies
See Figure 2 and Figure 3 for a summary of the 'Risk of bias' assessment of all included studies.
Allocation
Random sequence generation
Community‐based studies
Twelve of the 19 community‐based studies used appropriate methods to generate the random sequence (Aboud 2008; Aboud 2009; Bhandari 2004; Daniels 2012; Edward 2013; Kang 2017; Reinbott 2016; Saleem 2014; Shi 2010; Tariku 2015; Vazir 2013; Wen 2011). The method of random sequence generation was unclear in seven studies (Aboud 2011; Bhandari 2001; Campbell 2013; de Oliveira 2012; Negash 2014; Vitolo 2005; Yin 2009).
Facility‐based studies
Three of the four facility‐based studies used appropriate methods to generate the random sequence (Koehler 2007; Olaya 2013; Penny 2005), while the remaining study was unclear (Schroeder 2015).
Allocation concealment
Community‐based studies
The allocation sequence was adequately concealed in five studies (Bhandari 2004; Campbell 2013; Daniels 2012; Edward 2013; Wen 2011), but was unclear in 14 studies (Aboud 2008; Aboud 2009; Aboud 2011; Bhandari 2001; de Oliveira 2012; Kang 2017; Negash 2014; Reinbott 2016; Saleem 2014; Shi 2010; Tariku 2015; Vazir 2013; Vitolo 2005; Yin 2009).
Facility‐based studies
The allocation sequence was adequately concealed in one study (Olaya 2013) but unclear in three studies (Koehler 2007; Penny 2005; Schroeder 2015).
Blinding
Blinding of participants and personnel (performance bias)
Community‐based studies
Blinding of participants and personnel was unclear in 13 of the 19 community‐based studies (Aboud 2011; Bhandari 2001; Bhandari 2004; Daniels 2012; de Oliveira 2012; Edward 2013; Negash 2014; Reinbott 2016; Saleem 2014; Tariku 2015; Vazir 2013; Wen 2011; Yin 2009), and judged to be at high risk of bias in six studies (Aboud 2008; Aboud 2009; Campbell 2013; Kang 2017; Shi 2010; Vitolo 2005).
Facility‐based studies
All of the four included facility‐based studies were unclear on blinding of participants and personnel (Koehler 2007; Olaya 2013; Penny 2005; Schroeder 2015).
Blinding of outcome assessment (detection bias)
Community‐based studies
We assessed eight of the 19 community‐based studies as having low risk of detection bias (Aboud 2008; Aboud 2009; Campbell 2013; Daniels 2012; de Oliveira 2012; Reinbott 2016; Vazir 2013; Wen 2011), while blinding of outcome assessment was unclear in eight studies (Aboud 2011; Bhandari 2001; Bhandari 2004; Edward 2013; Negash 2014; Saleem 2014; Tariku 2015; Yin 2009). We considered three studies to be at high risk of detection bias (Kang 2017; Shi 2010; Vitolo 2005).
Facility‐based studies
Blinding of outcome assessors was unclear in two of the four facility‐based studies (Koehler 2007; Schroeder 2015) and judged to be at high risk of bias in the other two studies (Olaya 2013; Penny 2005).
Incomplete outcome data
Community‐based studies
We assessed 11 studies as having low attrition bias because they met at least one of the following criteria: the losses were similar across intervention and control groups; study authors accounted for losses to follow‐up and also used appropriate statistical analysis methods to make up for the losses to follow‐up (Aboud 2008; Aboud 2009; Aboud 2011; Bhandari 2004; Campbell 2013; Daniels 2012; de Oliveira 2012; Edward 2013; Shi 2010; Tariku 2015; Wen 2011). In four studies the risk of attrition bias was unclear (Reinbott 2016; Saleem 2014; Vitolo 2005; Yin 2009). We assessed four studies at high risk of attrition bias (Bhandari 2001; Kang 2017; Negash 2014; Vazir 2013). In Bhandari 2001 and Vazir 2013 the attrition rates were reported to be 12% and 15% respectively, although the reasons for attrition were provided for all participants in both studies, while the attrition rates in Kang 2017 and Negash 2014 were about 18% and 20%.
Facility‐based studies
We assessed two studies at low risk of bias as the losses were balanced across groups, study authors accounted for losses to follow‐up and also used appropriate statistical analysis methods to make up for the losses to follow‐up (Olaya 2013; Penny 2005). We rated one study at unclear risk of bias as there was no information on total number of participants lost to follow‐up (Koehler 2007). We rated one study at high risk of bias as the attrition rate was high (21%) and no reason was given for the losses to follow‐up (Schroeder 2015).
Selective reporting
Community‐based studies
We assessed one study as having low risk of reporting bias (Campbell 2013). The study protocol was available for assessment and study authors reported on all outcomes listed in the Methods section of the study reports. We assessed 16 other studies as having unclear risk of reporting bias in this domain because the study protocols were not available for assessment (Aboud 2008; Aboud 2009; Aboud 2011; Bhandari 2001; Bhandari 2004; Daniels 2012; de Oliveira 2012; Edward 2013; Kang 2017; Reinbott 2016; Saleem 2014; Shi 2010; Vazir 2013; Vitolo 2005; Wen 2011; Yin 2009), and one of which, Yin 2009, was originally published in Chinese and we were limited by the translated study. We assessed two studies as having high risk of reporting bias (Negash 2014; Tariku 2015). Negash 2014 did not report the results of the anthropometric measurements although the authors reported that the measurements were taken, while Tariku 2015 did not clearly present data for some outcomes.
Facility‐based studies
We assessed the four facility‐based studies as having unclear risk of reporting bias (Koehler 2007; Olaya 2013; Penny 2005; Schroeder 2015). The studies reported on all outcomes listed in the Methods section of the study reports but study protocols were unavailable for assessment.
Other potential sources of bias
Community‐based studies
We assessed 12 studies at low risk of other biases (Aboud 2008; Aboud 2009; Aboud 2011; Bhandari 2001; Bhandari 2004; Daniels 2012; de Oliveira 2012; Edward 2013; Negash 2014; Saleem 2014; Tariku 2015; Vazir 2013). We assessed four studies at unclear risk of other biases (Campbell 2013; Shi 2010; Wen 2011; Yin 2009), two of which reported baseline imbalances (Shi 2010; Wen 2011). We were unable to assess Yin 2009 since it was originally published in Chinese and we were limited by the translated study report. We considered three studies as having high risk of other biases (Kang 2017; Reinbott 2016; Vitolo 2005).
Facility‐based studies
We judged two studies, which reported adequate comparability between study arms at baseline, at low risk of bias (Koehler 2007; Penny 2005). We judged Schroeder 2015 at unclear risk of bias. Although it reported baseline imbalances in the ethnicity, employment, household income, education, home ownership, usage of food stamps, usage of WIC (women, infants and children) program services and breastfeeding rates, we were not sure how this affected the results following the intervention, since it was a facility‐based study and participants would have been exposed to the same conditions. We judged Olaya 2013 at high risk of bias due to baseline differences in child growth indices.
Effects of interventions
See: Summary of findings for the main comparison Educational intervention versus no educational intervention for improving complementary feeding practices; Summary of findings 2 Educational intervention versus no educational intervention for improving complementary feeding practices: growth outcomes
Primary outcomes
1a. Age at introduction of complementary foods
Community‐based studies
Pooled results
Six, individually‐randomised, community‐based studies reported the effect of educational intervention on age at introduction of complementary foods. Four studies reported data suitable for quantitative analysis (de Oliveira 2012; Edward 2013; Vitolo 2005; Wen 2011). The pooled effect estimate suggests that, compared to standard care, educational intervention reduces the risk of early introduction of complementary food (before four to six months of age) by 12% (average RR 0.88, 95% CI 0.83 to 0.94; 4 studies, 1738 children; Tau2 = 0.00, I2 = 0%; moderate‐quality evidence; Analysis 1.1; summary of findings Table for the main comparison). Studies used intervention delivery strategies that ranged from counselling sessions to the use of printed materials (booklets, brochures, leaflets), flip charts and videos, with some studies using a combination of at least two of the listed delivery strategies.
Single study results
Two community‐based studies were not included in the meta‐analysis. Daniels 2012 reported a difference in mean age of complementary food introduction (intervention mean age 22.8 (± 4.4) weeks versus control mean age 22.7 (± 4.9) weeks; study author‐reported P = 0.85). Reinbott 2016 reported the proportion of children introduced to semi‐solid/soft foods between the WHO recommended ages of six to eight months (intervention 88.1% versus control 92.6%; study author‐reported P = 0.349). Insufficient information was reported by Reinbott 2016 to estimate an intervention effect and the study could not be included in the meta‐analysis. (See Table 5).
| Study | Result |
| The incidence and prevalence of diarrhoea and ALRI were not significantly affected by either intervention Nutritional counselling group: episodes per child 6.9 (± 3.2), prevalence per 100: d 14.6 (± 12.0) Visitation group: episodes per child 6.7 (± 3.4), prevalence per 100: d 13.2 (± 9.8) | |
| The reported prevalences of common illnesses in the previous 7 days did not differ in the 2 groups | |
| Diarrhoeal illness in the past 2 weeks (%) Baseline: intervention = 36.9%, control = 41.6% Impact: intervention = 27.9%, control = 26.2% | |
| Number with event: intervention = 46, control = 98 |
ALRI: acute lower respiratory infection.
Facility‐based studies
One facility‐based study, Schroeder 2015, reported insufficient information to estimate an intervention effect and therefore was not included in Analysis 1.1 above. Schroeder 2015 narratively reported that mothers in the intervention arm delayed the introduction of complementary foods compared with mothers in the control arm (study author‐reported P < 0.051). This study used intervention delivery strategies that included printed materials (educational brochures and reminder postcards containing intervention messages) and telephone calls.
1b. Duration of exclusive breastfeeding
Four studies measured the effect of educational intervention on the duration of exclusive breastfeeding (de Oliveira 2012; Penny 2005; Vitolo 2005; Wen 2011), of which three reported sufficient data for inclusion in a meta‐analysis (Penny 2005; Vitolo 2005; Wen 2011). We conducted the analysis for duration of exclusive breastfeeding (≥ four months of age) using the generic inverse variance approach in RevMan 5 (Review Manager 2014), to allow for inflating the standard error of Penny 2005 (see below). The average RR, pooled across both community‐ and facility‐based studies was RR 1.58 (95% CI 0.77 to 3.22; 3 studies, 1544 children; Tau2 = 0.30, I2 = 80%; very low‐quality evidence; Analysis 1.2; summary of findings Table for the main comparison). We further investigated the impact of the ICC value on the pooled intervention effect in a sensitivity analysis (See Sensitivity analysis and Figure 5).
Sensitivity analysis 3. Comparison of different ICC (primary outcomes), outcome: 3.1 duration of exclusive breastfeeding (≥ 4 months of age)
The intervention delivery strategy in Wen 2011 was counselling and social support, Vitolo 2005 included dietary counselling, printed materials (brochures with key messages; simple, coloured leaflet with food pictures depicting a healthful meal), while that of de Oliveira 2012 included counselling sessions and promotional materials like booklets and flip charts. In Penny 2005 the intervention involved group sessions for caregivers of children of similar ages, demonstrations of the preparation of complementary foods, the use of flip charts and single‐page recipe fliers.
Community‐based studies
Pooled results
Three studies examined community‐based educational intervention (de Oliveira 2012; Vitolo 2005; Wen 2011). Only two studies reported data that could be combined in a meta‐analysis (Vitolo 2005; Wen 2011). The pooled estimate of effect suggests that educational intervention increased the duration of exclusive breastfeeding by 132% (average RR 2.32, 95% CI 1.45 to 3.73; 2 studies, 1167 children; Tau2 = 0.00, I2 = 0%; low‐quality evidence; Analysis 1.2; summary of findings Table for the main comparison).
Individual study results
de Oliveira 2012 reported insufficient information to be included in the meta‐analysis. The authors reported the median duration of exclusive breastfeeding: 2.9 months (interquartile range 1.0 to 4.7) in the intervention arm and 1.3 (interquartile range 0.6 to 3.0) in the control arm, (study author‐reported P = 0.001, no further detail available).
Facility‐based studies
Only one facility‐based study reported the effect of educational intervention on the duration of exclusive breastfeeding (Penny 2005). After we retrospectively accounted for clustering (using the approximate approach outlined above in the Unit of analysis issues section), and assuming an ICC of 0.02, the estimate of intervention effect was compatible with both a decrease and an increase in the duration of exclusive breastfeeding (RR 0.95, 95% CI 0.70 to 1.29; 1 study, 377 children; low‐quality evidence; Analysis 1.2; summary of findings Table for the main comparison).
1c. Adequacy of complementary foods
Eighteen studies reported the outcome of adequacy of complementary foods. However, the types of foods, measures and methods of assessment reported were too diverse to be combined in a meta‐analysis. Several studies reported a dietary diversity score or infant/child feeding index, or both, but it was not sufficiently clear from the reports whether they were based on comparable criteria or food groups and so we considered it inappropriate to combine them in a meta‐analysis. We provide a narrative summary of the individual study findings below.
Community‐based studies
Thirteen community‐based studies reported findings for the outcome of adequacy of complementary foods fed to children (Aboud 2008; Aboud 2009; Aboud 2011; Bhandari 2001; Bhandari 2004; Campbell 2013; Daniels 2012; Negash 2014; Reinbott 2016; Shi 2010; Tariku 2015; Vazir 2013; Vitolo 2005). We categorised outcomes into those that focused on the adequacy of nutrient intake/diversity of complementary food (i.e. quality), and the volume and frequency of adequate complementary food (i.e. quantity).
Adequacy of nutrient intake/diversity of complementary food
Eleven community‐based studies reported an outcome related to the adequacy of nutrient intake/diversity of complementary food (Aboud 2008; Aboud 2009; Aboud 2011; Bhandari 2001; Bhandari 2004; Campbell 2013; Negash 2014; Reinbott 2016; Shi 2010; Vazir 2013; Vitolo 2005). One study reported energy intake only (Bhandari 2001), and one study reported details for responsive feeding only (Daniels 2012).
Although we were unable to combine the studies in a meta‐analysis, due to the manner in which the results were reported, 10 of the 11 study authors reported intervention effect estimates or sufficient details of at least one relevant outcome. One study reported insufficient detail (Bhandari 2001).
Aboud 2008 reported the mean number of times specific foods were eaten (in 24 hours) for separate foods and asserts, "eggs, fruit, vegetables and carbohydrates were more often reportedly given to the children of caregivers in the complementary feeding intervention, and biscuits/sugar more often given to controls" (p 282). Intervention effect estimates at follow‐up could be calculated from Table 3 of their report for consumption of: rice (MD 0.07, 95% CI −0.12 to 0.26); dal (MD −0.12, 95% CI −0.31 to 0.07); fish (MD −0.15, 95% CI −0.42 to 0.12); egg (MD 0.19, 95% CI 0.07 to 0.31); fruit (MD 0.28, 95% CI 0.06 to 0.50); vegetables (MD 0.57, 95% CI 0.26 to 0.88), cows' milk (MD 0.12, 95% CI −0.19 to 0.43); carbohydrate (MD 0.32, 95% CI 0.08 to 0.56); and biscuits (MD −0.30, 95% CI −0.60 to 0.00). All food types are study author‐reported.
Aboud 2009 also reported the mean number of times specific foods were eaten (in 24 hours) for separate foods; rice (MD −0.11, 95% CI −0.32 to 0.10); dal (MD 0.09, 95% CI −0.05 to 0.23); fish (MD 0.07, 95% CI −0.27 to 0.41); egg (MD 0.06, 95% CI −0.07 to 0.19); fruit (MD 0.22, 95% CI 0.10 to 0.34); vegetables (MD −0.42, 95% CI −0.86 to 0.02); cows' milk (MD 0.09, 95% CI −0.14 to 0.32); carbohydrate (MD −0.21, 95% CI −0.44 to 0.02); and biscuits (MD 0.10, 95% CI −0.24 to 0.44). All food types are study author‐reported. Aboud 2009 also reported a mean dietary diversity score for each group, which can be used to calculate an unadjusted difference in means; MD 0.32 (95% CI 0.05 to 0.59) in favour of the complementary‐food intervention group.
Aboud 2011 did not provide sufficient information to estimate an intervention effect for the adequacy of nutrient intake/diversity of complementary food. Study author‐reported findings stated that, "of the 7 critical food categories, 20 control children ate a mean of 2.96 foods and the children in the intervention group ate 3.07 foods" (p e1195). In addition, they stated that group differences were non significant at postintervention and follow‐up. The study authors also reported that dietary diversity scores increased for all groups from pre‐testing (mean = 2.61) to follow‐up (mean = 3.03) (study author‐reported P < 0.001). No further information was reported to allow estimation of relative effect.
Bhandari 2004 reported energy intake (Kj/24 hours) from all foods at nine months of age (MD 531.00 Kj/24 hours, 95% CI 398.24 to 663.76) and at 18 months (MD 1230.00 Kj/24 hours, 95% CI 1052.50 to 1407.50).
The types of food consumed (24‐hour recall) were also reported at nine months of age and 18 months of age. At nine months of age foods consumed were: commercially‐available bread (RR 6.77, 95% CI 3.11 to 14.71); home‐made bread (RR 1.03, 95% CI 0.93 to 1.14); rice (RR 3.08, 95% CI 1.48 to 6.39); potatoes (RR 2.40, 95% CI 1.44 to 3.99); legumes (RR 2.68, 95% CI 1.77 to 4.06); any milk (i.e. breastmilk or non‐breastmilk) (RR 1.11, 95% CI 1.05 to 1.17); meat or egg (RR 4.47, 95% CI 0.22 to 92.81); vegetables (RR 3.35, 95% CI 1.55 to 7.22); fruits (RR 1.36, 95% CI 0.97 to 1.91). At 18 months of age foods consumed were: commercially‐available bread (RR 2.16, 95% CI 1.54 to 3.01); home‐made bread (RR 0.95, 95% CI 0.90 to 1.01); rice (RR 1.09, 95% CI 0.68 to 1.73); potatoes (RR 1.31, 95% CI 1.04 to 1.66); legumes (RR 1.24, 95% CI 0.99 to 1.56); any milk (i.e. breastmilk or non‐breastmilk) (RR 1.03, 95% CI 1.00 to 1.05); meat or egg (RR 4.53, 95% CI 0.22 to 94.07); vegetables (RR 1.08, 95% CI 0.85 to1.36); and fruits (RR 1.11, 95% CI 0.95 to 1.30). All food types are study author‐reported.
Intakes of cereal legume gruels or mixes (RR 3.52, 95% CI 2.44 to 5.06), milk cereal gruels or milk cereal mixes (RR 3.20, 95% CI 2.36 to 4.32), undiluted milk (RR 3.02, 95% CI 2.42 to 3.78), addition of butter/oil (RR 17.42, 95% CI 4.23, 71.70), and recommended snacks (RR 1.31, 95% CI 1.15 to 1.49) were also reported by study authors to be higher in nine‐month‐old children in the educational intervention communities. Similar patterns were seen at 18 months of age, but the study authors reported that differences between the two groups were less pronounced for cereal legume gruels than those at nine months of age, possibly because these foods are commonly given at this age. Estimates are based on raw means, SDs and percentages, as reported in the original paper.
Campbell 2013 reported a 24‐hour dietary recall outcome at postintervention for average daily consumption of: fruits (MD 10.99, 95% CI −6.09 to 28.06); vegetables (MD 4.53, 95% CI −4.38 to 13.43); non‐core drinks (MD −2.21, 95% CI −13.71 to 9.30); non‐core sweet foods such as chocolate, candy and cakes (MD −3.69, 95% CI −6.41 to 20.96); non‐core savoury foods such as crisps and savoury biscuits (MD −1.01, 95% CI −2.82 to 0.80); and water consumption (MD 24.17, 95% CI −9.85 to 58.20). All food types, effect estimates and 95% CIs are study author‐reported.
Negash 2014 reported the raw mean dietary energy intake (kcal) at postintervention, from which we calculated the MD with 95% CIs (MD 160.00, 95% CI −24.31 to 344.31). They also reported mean protein intake (g) for each intervention group (MD 7.10 g, 95% CI 1.56 to 12.64); mean fat intake (g) (MD −0.60 g, 95% CI −10.35 to 9.15); carbohydrate intake (g) (MD 32.00 g, 95% CI 3.18 to 60.82); and iron intake (mg) (MD 9.70 mg, 95% CI 4.19 to 15.21). It is not stated whether nutrient intakes are based on a 24‐hour recall.
Reinbott 2016 assessed dietary diversity (RR 1.16, 95% CI 1.04 to 1.30), and minimum acceptable diet (RR 1.26, 95% CI 1.07 to 1.48). They also reported that a 24‐hour dietary diversity score was calculated using a seven‐food‐group score, the child dietary diversity score (RR 0.20, 95% CI 0.00 to 0.40), and reported individually for: grains (RR 1.01, 95% CI 0.99 to 1.03); roots and white tubers (RR 1.87, 95% CI 1.50 to 2.34); legumes, nuts and seeds (RR 1.03, 95% CI 0.86 to 1.24); dairy products (RR 0.75, 95% CI 0.57 to 0.98); flesh foods namely meat, poultry, fish and offal (RR 1.02, 95% CI 0.95 to 1.09); eggs (RR 1.28, 95% CI 1.09 to 1.50); pro‐vitamin‐A‐rich foods such as yellow‐ and orange‐fleshed roots and tubers, orange‐fleshed fruits, and dark green leafy vegetables (RR 1.17, 95% CI 1.03 to 1.33); other fruits and vegetables (RR 1.12, 95% CI 1.01 to 1.25); fats and oils (RR 1.02, 95% CI 0.92 to 1.14); and sugary foods and crisps (RR 0.93, 95% CI 0.86 to 1.00). All food types are study author‐reported.
Shi 2010 reported findings for diversity of complementary foods at three time points: when child was six months, nine months and 12 months of age. RR greater than 1 suggested the educational intervention increased the consumption of the food. They reported whether the child had ever been fed at six months: bread, rice or noodles (RR 1.37, 95% CI 1.22 to 1.54); roots or tubers (RR 1.06, 95% CI 0.78 to 1.43); yellow or orange foods (RR 1.25, 95% CI 1.02 to 1.53); green leafy vegetables (RR 1.78, 95% CI 1.33 to 2.38); beans, peas or lentils (RR 2.22, 95% CI 1.61 to 3.04); fruits (RR 1.16, 95% CI 1.05 to 1.28); eggs (RR 1.27, 95% CI 1.14 to 1.41); meats (RR 2.84, 95% CI 1.91, 4.21); and cooking oils/fats (RR 1.92, 95% CI 1.40 to 2.63).
At nine months the findings were: bread, rice or noodles (RR 1.03, 95% CI 1.00 to 1.06); roots or tubers (RR 1.10, 95% CI 1.00 to 1.21); yellow or orange foods (RR 1.15, 95% CI 1.07 to 1.24); green leafy vegetables (RR 1.20, 95% CI 1.12 to 1.30); beans, peas or lentils (RR 1.43, 95% CI 1.28 to 1.59); fruit (RR 1.04, 95% CI 1.01 to 1.07); eggs (RR 1.04, 95% CI 1.00 to 1.07); meats (RR 1.61, 95% CI 1.44 to 1.81); and cooking oils/fats (RR 1.19, 95% CI 1.09 to 1.29).
At 12 months the findings were: bread, rice or noodles (RR 1.01, 95% CI 0.99 to 1.04); roots or tubers (RR 1.23, 95% CI 1.13 to 1.34); yellow or orange foods (RR 1.27, 95% CI 1.18 to 1.37); green leafy vegetables (RR 1.11, 95% CI 1.05 to 1.17); beans, peas or lentils (RR 1.37, 95% CI 1.24 to 1.51); fruits (RR 1.03, 95% CI 1.00 to 1.06); eggs (RR 1.08, 95% CI 1.03 to 1.12); meats (RR 1.67, 95% CI 1.49 to 1.86); and cooking oils/fats (RR 1.21, 95% CI 1.13 to1.30).
Vazir 2013 reported the percentage of each group who consumed the following foods, at nine and 15 months: rice (9 months: RR 1.17, 95% CI 1.09 to 1.27; 15 months: RR 2.92, 95% CI 1.89 to 4.51); goat's liver (9 months: RR 13.42, 95% CI 4.97 to 36.27; 15 months: RR 2.92, 95% CI 1.89 to 4.51); goat's meat (9 months: RR 4.85, 95% CI 2.33 to 10.07; 15 months: RR 1.33, 95% CI 1.01 to 1.75); poultry (9 months: RR 2.65, 95% CI 0.72 to 9.83; 15 months: RR 1.98, 95% CI 1.37 to 2.85); banana (9 months: RR 1.58, 95% CI 1.27 to 1.97; 15 months: RR 1.28, 95% CI 1.11 to 1.48); buffalo milk (9 months: RR 0.99, 95% CI 0.97 to 1.01; 15 months: RR 1.13, 95% CI 1.00 to 1.27); egg (9 months: RR 3.14, 95% CI 2.22 to 4.44; 15 months: RR 1.37, 95% CI 1.16 to 1.61); spinach (9 months: RR 17.90, 95% CI 4.38 to 73.20; 15 months: RR 1.42, 95% CI 1.06 to 1.90); pulses (9 months: RR 1.01, 95% CI 0.98 to 1.03; 15 months: RR 1.25, 95% CI 1.12 to 1.40); and added fat (9 months: RR 2.10, 95% CI 1.56 to 2.83; 15 months: RR 1.42, 95% CI 1.06 to 1.90). Median nutrient and energy intakes were also reported.
Vitolo 2005 reported the relative effect of caregiver educational intervention on the consumption of energy‐dense food at 12 to 16 months of age. RR less than 1 suggested the educational intervention reduced the consumption of energy‐dense food; candies (RR 0.85, 95% CI 0.74 to 0.98); soft drinks (RR 0.88, 95% CI 0.79 to 0.99); table sugar (RR 0.98, 95% CI 0.93 to 1.03); honey (RR 0.65 95% CI 0.50 to 0.84); cookies (RR 0.79, 95% CI 0.71 to 0.89); chocolate (RR 0.72, 95% CI 0.60 to 0.86); salty snacks (RR 0.86, 95% CI 0.76 to 0.97); lipid‐dense foods group (RR 0.62, 95% CI 0.49 to 0.80); and sugar‐dense foods group (RR 0.46, 95% CI 0.31 to 0.68). The effect estimates are study author‐reported (Vitolo 2012 in Vitolo 2005). At two to three years' follow‐up (when children were aged three to four years old), the study authors also reported a Health Eating Index score (MD 3.52, 95% CI 1.18 to 5.88) (Vitolo 2010 in Vitolo 2005). For the outcome of 'good diet' (Healthy Eating Index score > 80), the study‐author‐reported RR was 2.12 (95% CI 1.09 to 4.12).
With regards to consumption of specific foods and nutrients at the two‐to‐three‐year follow‐up time point, the study authors reported the following MDs for the following food types: grains (MD ‐0.11, 95% CI 0.60 to 0.38); meats (MD 0.10, 95% CI −0.56 to 0.75); vegetables (MD 0.53, 95% CI 0.10 to 0.95); fruits (MD 0.87, 95% CI 0.15 to 1.59); milk (MD 0.34, 95% CI −0.20 to 0.88); total fat* (MD 0.07, 95% CI −0.32 to 0.46); sodium* (MD 0.91, 95% CI 0.15 to 1.66); cholesterol* (MD −0.31, 95% CI −0.69 to 0.07); saturated fat* (MD 0.33, 95% CI −0.43 to 1.09). (*Lower scores indicate a greater intake.)
Volume and frequency of adequate complementary food
Seven community‐based studies reported outcomes related to the volume and frequency of adequate complementary food (quantity). Intervention effect estimates were either reported by study authors or could be estimated by the review authors in all of these studies.
For the outcome 'total mouthfuls' for Aboud 2008, we calculated an unadjusted MD of 1.45 (95% CI −0.74 to 3.64). For the outcome percentage child self‐fed mouthfuls, we calculated a follow‐up MD of 16.42 (95% CI 3.32 to 29.52).
Aboud 2009 also reported that the mean number of mouthfuls per meal consumed by children at follow‐up did not differ, with an overall MD of −0.39 (95% CI −4.62 to 3.84). The mean number of self‐fed mouthfuls as a percentage of total mouthfuls was 47.8 (± 42.4) in the intervention group compared with 32.2 (± 41.0) in the control group (study author‐reported); MD 15.60 (95% CI 3.83 to 27.37). The results of the ANCOVA, as reported by study authors, was d = 0.37 P = 0.01.
Aboud 2011 reported mean number of mouthfuls per meal for control and two active intervention groups. Here, we combined the two active arms of the intervention (it was a three‐armed study) to allow this comparison to be made; MD 5.76, 95% CI 2.10 to 9.42. Also reported was the mean number of self‐fed mouthfuls which, as a percentage of the total for each group, favoured the intervention: MD 10.19 (95% CI −0.20 to 20.58).
Bhandari 2004 reported mean meal frequency within a 24‐hour period at nine months of age (MD 0.50, 95% CI 0.31 to 0.69) and 18 months of age (MD 0.50, 95% CI 0.33 to 0.67) in favour of the intervention. Study author‐reported P values for the comparisons were < 0.01.
Campbell 2013 also reported the effect of the educational intervention on prevalence of any (versus none) non‐core food and drink consumption at postintervention (mean child age = 18 months). For non‐core drink intake the study authors reported an odds ratio (OR) of 0.81 (95% CI 0.51 to 1.30), for sweet snack intake an OR of 0.69 (95% CI 0.43 to 1.10), and for savoury snack intake an OR of 1.25 (95% CI 0.87 to 1.81). These effect estimates are not adjusted for covariates.
Reinbott 2016 reported the minimum meal frequency (as defined by WHO) as a RR of 1.04 (95% CI 0.98 to 1.10). The study authors also reported the following results from a linear regression of seven‐day food frequency, adjusted for age of child, wealth and maternal education: fish (B (beta) = 0.73, SE (standard error)(B) = 0.36, 95% CI 0.02 to 1.44, P = 0.05), pro‐vitamin‐A‐rich roots and tubers (B = 1.11, SE(B) = 0.25, 95% CI 0.62 to 1.60, P < 0.001), and dark green leafy vegetables (B = 1.15, SE(B) = 0.33, 95% CI 0.51 to 1.80, P = 0.001). Other categories of food frequencies were not reported.
Shi 2010 reported meal frequency (semi‐solid or solid foods) at three time points: six, nine and 12 months of age. At six months of age the MD was 0.57 (95% CI 0.34 to 0.80) and at nine months of age the MD was 2.72 (95% CI 2.35 to 3.09). Incomplete data were reported for the 12‐month outcome and we were unable to calculate an effect estimate for this time point.
Facility‐based studies
Amongst the facility‐based studies, Koehler 2007 reported compliance with food‐based recommendations and standardised daily nutrition scores. It was not possible to estimate an intervention effect from the published paper.
Olaya 2013 assessed the frequency and number of portions of each food consumed. Study author‐reported findings for the mean number of portions (per week) of each food consumed were reported in box and whisker plots for meat, red meat, vegetables, fruit, follow‐on formula milk, cows' milk, legumes, and sugar and sweetened foods (frequency). We have not extracted effect estimates from this plot. Olaya 2013 also reports the proportion of infants consuming recommended food groups, at the recommended frequency per week for the following food groups: meat (all types) (RR 1.65, 95% CI 1.10 to 2.46); red meat (RR 1.48, 95% CI 1.17 to 1.87); vegetables (RR 2.45, 95% CI 1.43 to 4.20); fruit (RR 1.59, 95% CI 1.19 to 2.12); and legumes (RR 1.44, 95% CI 0.91 to 2.26). Study authors also reported the MD for iron and zinc status between the intervention and control groups at six and 12 months of age (six months: ferritin = MD 24.69, 95% CI 221.8 to 12.4 mg/L; zinc = MD 3.65, 95% CI 28.8 to 16.0 mg/dL. 12 months: ferritin = MD 6.31, 95% CI 2.7 to 15.4 mg/dL; zinc = MD 24.23, 95% CI 217.9 to 9.4 mg/dL).
Adequacy of complementary food outcomes reported in Penny 2005 included eating nutrient‐dense, thick foods at lunch (a recommended complementary feeding practice) (six months: intervention 48 (31%) of 157 versus control 29 (20%) of 147; difference between groups 19 (11%), P = 0·03); achieving dietary requirements for energy (8 months: intervention 30 (18%) of 170 versus control 45 (27%) of 167, P = 0·04; 12 months: 64 (38%) of 168 versus 82 (49%) of 167, P = 0·043); dietary iron intake from complementary foods (8 months: intervention 155 (91%) of 170 versus control 161 (96%) of 168, P = 0.047; 9 months: 152 (93%) of 163 versus 165 (99%) of 167, P = 0.003); and dietary zinc intake from complementary foods (9 months: intervention 125 (77%) of 163 versus control 145 (87%) of 167, P = 0·012). Effect estimates and P values are as reported by Penny 2005. Unadjusted mean energy and nutrient intakes from complementary foods (24‐hour recall) were reported in a figure, but we were not able to estimate an intervention effect for the outcomes
It was not possible to estimate intervention effect estimates from Schroeder 2015. The study authors reported that the "intervention group was less likely to use infant cereal (P < 0.001) or stage 1 vegetables (P < 0.05) as the first complementary food. Also, the intervention group offered significantly less soda (P < 0.006), sweetened tea (P < 0.01), punch (P < 0.02), or cows' milk (P < 0.001) than the control group" (p 3). A comparison between six and 24 months indicated that the control group increased consumption of unsweetened drinks (P < 0.04) and of vitamin supplements (P < 0.04) relative to the intervention group, as reported by study authors. Parents in the intervention group exerted more dietary restriction on their child (P < 0.01) and were more active in monitoring child feeding (P < 0.05) than those in the control group.
1d. Hygiene practices
Six community‐based studies reported the impact of educational interventions on hygiene practices (Aboud 2009; Aboud 2011; Bhandari 2004; Negash 2014; Shi 2010; Tariku 2015), of which only one was an individually‐randomised study (Negash 2014) and five were cluster‐randomised studies.
There was considerable variation in the definition of the outcome of hygiene practices across studies; for example, washing a child's hands before feeding (Aboud 2009; Bhandari 2004; Shi 2010), washing a child's hands with soap (Aboud 2011; Tariku 2015), washing of the caregivers' hands before feeding or food preparation (Bhandari 2004; Negash 2014; Shi 2010; Tariku 2015), and handwashing after defecation (Negash 2014). Where a study reported more than one handwashing outcome, we chose the outcome relating to handwashing before feeding and prioritised caregiver handwashing for the meta‐analysis. The intervention delivery strategies included group education sessions, demonstrations/practicals of meal preparation, role play with infants, use of printed materials (posters, flip books, feeding‐recommendation cards, picture books), home visits, women's group meetings, village rallies, debates, side plays and nutrition fairs.
Community‐based studies
Pooled results
Four studies provided sufficient data for inclusion in a meta‐analysis, having retrospectively accounted for clustering (assuming an ICC of 0.02) (Aboud 2009; Aboud 2011; Bhandari 2004; Shi 2010). We conducted a random‐effects meta‐analysis using the generic inverse variance approach in RevMan 5 (Review Manager 2014), and explored the impact of the ICC in the sensitivity analyses in Figure 6. Having accounted for clustering, there was moderate‐quality evidence that educational intervention increased caregiver‐reported handwashing before feeding by an average of 38% (Analysis 1.3: average RR 1.38, 95% CI 1.23 to 1.55; 4 studies, 2029 participants; Tau2 = 0.00, I2 = 0%; summary of findings Table for the main comparison).
Sensitivity analysis 3. Comparison of different ICC (primary outcomes), outcome: 3.2 hygiene: handwashing before feeding
Single study results
Two studies were not included in the meta‐analysis (Negash 2014; Tariku 2015), as neither reported sufficient information to calculate an intervention effect estimate. Negash 2014 narratively reported that handwashing before feeding and after defecation had decreased in the intervention group but remained unchanged in the control group. We could not calculate an effect estimate for either study, due to a lack of clarity around the numbers randomised. Tariku 2015 reported, "regarding to the hand washing practice, the proportion of mothers who would wash their hands after intervention significantly increased for all Kebeles [administrative district] compared to pre‐intervention, but no significant differences were found in the proportion of hand washing practices. For the use of soap to wash their child’s hand, there were significant difference between the Traditional intervention and Control Kebeles (p = .005); and between the Health Belief Model intervention and Control Kebeles (p = .001)" (p 8). Note, the study authors reported P values only, and we were unable to estimate an intervention effect estimate due to insufficient information reported in the paper.
Facility‐based studies
None of the facility‐based studies reported the effect of educational intervention on hygiene practices.
2. Adverse events
One study investigated the compliance with, and acceptability of, the intervention (Olaya 2013). They reported a 74% compliance rate with the recommendations of the intervention. Only one out of the 38 mothers felt that the recommendations were not helpful. On the affordability of recommended complementary foods, 83.8% of the mothers could afford the recommended complementary food while six mothers found the foods too expensive. The recommended complementary food was tolerated by all infants in the study and there were no reported adverse effects.
Secondary outcomes
1. Growth
Fourteen studies reported growth outcomes (Aboud 2008; Aboud 2009; Bhandari 2001; Bhandari 2004; Campbell 2013; Daniels 2012; Negash 2014; Olaya 2013; Penny 2005; Reinbott 2016; Saleem 2014; Schroeder 2015; Shi 2010; Vazir 2013). Of these, we were able to combine eight quantitatively in at least one of the growth meta‐analyses. Four studies were not included in the meta‐analyses because they included age ranges or reported growth data at time points that were insufficiently similar to other studies (Aboud 2008; Aboud 2009; Negash 2014; Saleem 2014). They are reported below under the heading 'Individual study results'. Campbell 2013 was not included in the meta analysis because the study reported body mass index (BMI) only and we could not combine this with other measures of growth. Reinbott 2016 reported mean height‐for‐age (HAZ) and mean weight‐for‐age (WAZ) z scores, rather than stunting, wasting or underweight outcomes. The results from Campbell 2013 and Reinbott 2016 are also reported below.
The 14 studies moreover reported growth outcomes at various time points. However, we had a priori selected time points of six and 12 months of age because these mark the half and first year of an infant's life respectively. Thereafter, we chose to analyse growth parameters at six‐monthly intervals (18 and 24 months of age), since the rate of growth reduces after infancy.
Pooled analysis results
We conducted the meta‐analysis using the generic inverse variance approach, to allow for inflating the standard error of Penny 2005, Schroeder 2015 and Shi 2010. For all growth outcomes, we assumed an ICC = 0.05. Overall, the body of evidence for all growth outcomes was considered low quality. See summary of findings Table 2.
For attained weight (kg), the pooled results for the three studies that recruited women during pregnancy (Bhandari 2001; Shi 2010;Vazir 2013) are compatible with both a reduction and an increase in attained weight at six months of age, relative to control (MD 0.03 kg, 95% CI −0.10 to 0.17; 3 studies, 1221 children; Tau2 = 0.00, I2 = 0%; very low‐quality evidence). This was also observed at 12 months of age (MD 0.06 kg, 95% CI −0.04 to 0.15; 5 studies, 2464 children; Tau2 = 0.00, I2 = 0%; very low‐quality evidence), 18 months of age (MD 0.10 kg, 95% CI −0.14 to 0.35; 2 studies, 1402 children; Tau2 = 0.02, I2 = 52%; very low‐quality evidence), and at 24 months of age (MD −0.14 kg, 95% CI −0.36 to 0.08; 2 studies, 920 children; Tau2 = 0.00, I2 = 0%; low‐quality evidence). See Analysis 2.1.
For the outcome of mean height/length (cm), findings from the meta‐analysis are indicative of both a harm and a benefit of educational intervention relative to the control intervention, at all four time points assessed (see Analysis 2.2). Summary effect estimates were similar at six months of age (MD 0.16 cm, 95% CI −0.21 to 0.52; 3 studies, 1221 children; Tau2 = 0.00, I2 = 0%; very low‐quality evidence), 12 months of age (MD 0.32 cm, 95% CI 0.11 to 0.52; 5 studies, 2464 children; Tau2 = 0.00, I2 = 0%; low‐quality evidence), 18 months of age (MD 0.58 cm, 95% CI −0.22 to 1.38; 2 studies, 1402 children; Tau2 = 0.21, I2 = 61%; very low‐quality evidence), and 24 months of age (MD −0.13 cm, 95% CI −0.58 to 0.32; 2 studies, 920 children; Tau2 = 0.00, I2 = 0%; low‐quality evidence).
Individual study results
Six studies could not be included in the meta analyses (Aboud 2008; Aboud 2009; Campbell 2013; Negash 2014; Reinbott 2016; Saleem 2014).
Aboud 2008 reported mean attained weight (kg) at five months postintervention in each group (MD 0.46 kg, 95% CI 0.07 to 0.85) and weight gain (kg) (MD 0.34 kg, 95% CI 0.12 to 0.56). Aboud 2008 also reported effect sizes for weight (d = 0.28) and weight gain (d = 0.48). It was not feasible to combine this study in the meta‐analysis due to the different age groups studied (aged 12 to 24 months at baseline).
Aboud 2009 reported two growth outcomes: WAZ (MD 0.01, 95% CI −0.24 to −0.26) and child's attained weight (kg) (MD 0.01 kg, 95% CI −0.29 to 0.31). Again, it was not feasible to combine this outcome due to the different age groups studied (aged 8 to 20 months at baseline).
It was not possible to calculate intervention effect estimates for Negash 2014. The only information available was study author‐reported, "control and intervention children had similar gains in weight (˜ 0.9 kg) and height (˜ 4 cm)" (p 483).
Saleem 2014 measured the following infant growth outcomes: weight, length, mid upper‐arm circumference (MUAC), stunting, wasting, and underweight at four time points. They reported weight, length and MUAC at follow‐up in a figure, all of which favoured the intervention group (P values = 0.001, 0.002 and 0.001 respectively). We have not extracted effect estimates from this plot. They also reported the reduction of stunting and underweight as OR 8.36 (95% CI 5.6 to 12.42) and OR 0.75 (95% CI 0.4 to1.79), favouring the intervention compared to the control group (adjusted OR).
2. Incidence of malnutrition among participants
Pooled analysis results
We report the findings of the meta‐analyses for the outcome of nutritional status measures in Analysis 2.3. Five studies reported stunting, defined as HAZ ≤ −2 SD (Bhandari 2001; Bhandari 2004; Kang 2017; Olaya 2013; Penny 2005). Two studies reported usable data for wasting, defined as WHZ ≤ −2 SD (Bhandari 2001; Kang 2017). Three studies reported usable data for the outcome of underweight, defined as WAZ ≤ −2 SD (Bhandari 2004; Kang 2017; Olaya 2013). For the outcome of stunting, the 95% CIs for the effect estimate are suggestive of both a harm and a benefit of educational intervention, relative to the control intervention (average RR 0.89, 95% CI 0.74 to 1.06; 5 studies, 3487 children; Tau2 = 0.00, I2 = 0%; low‐quality evidence). For the outcome of wasting, 95% CI are again suggestive of both a benefit and harm of the complementary feeding intervention relative to control (average RR 0.79, 95% CI 0.48 to 1.30; 2 studies, 2000 children; Tau2 = 0.00, I2 = 0%; low‐quality evidence). Three studies were included in the analysis for underweight (Bhandari 2004; Kang 2017; Olaya 2013). Again, 95% CIs for the average RR were compatible with both an increase and decrease in the outcome (average RR 0.99, 95% CI 0.68 to 1.44; 3 studies, 2900 children; Tau2 = 0.00, I2 = 0%; low‐quality evidence).
Individual results for studies that could not be included in the meta‐analyses are presented below.
Individual study results
Daniels 2012 reported HAZ (MD −0.02, 95% CI −0.19 to 0.15), WAZ (MD −0.13, 95% CI 0.27 to 0.01). They also reported rapid weight gain (OR 1.5, CI 95% 1.1 to 2.1) (control put on more weight, more rapidly).
Saleem 2014 reported MUAC, stunting, wasting, and underweight at four time points. They reported weight, length and MUAC at follow‐up in a figure, all of which favoured the intervention group (P values = 0.001, 0.002 and 0.001 respectively). We have not extracted effect estimates from this plot. They also reported the reduction of stunting and underweight as OR 8.36 (95% CI 5.6 to 12.42) and OR 0.75 (95% CI 0.4 to1.79), favouring the intervention group compared to the control group (adjusted OR).
Reinbott 2016 reported unadjusted means for the following nutritional status outcomes: HAZ (MD −0.06, 95% CI −0.20 to 0.08), WHZ (MD 0.00, 95% CI −0.13 to 0.13) and WAZ (MD −0.02, 95% CI −0.15 to 0.11).
3. Morbidity
Morbidity was measured by episodes of diarrhoea. We were unable to conduct a meta‐analysis for this outcome due to differences in the ways it was measured and reported. Four studies evaluated the effect of educational intervention on diarrhoea (Bhandari 2001; Bhandari 2004; Reinbott 2016; Vitolo 2005). Vitolo 2005 reported a beneficial effect of educational intervention on the incidence of diarrhoea, with the number of events reported as 46 in the intervention arm and 98 in the control arm. Numbers were not provided.
Bhandari 2001 reported that the intervention had no effect on diarrhoea episodes and prevalence: nutritional counselling group (study author‐reported episodes per child in the intervention group = 6.9 (± 3.2), prevalence per 100 d 14.6 (± 12.0); episodes per child in the visitation/control group = 6.7 (± 3.4), prevalence per 100 d 13.2 (± 9.8)). Diarrhoea prevalence at 12 months of age as reported by Bhandari 2004 was 16.8 in the intervention arm versus 13.1% in the control arm (study author‐reported P = 0.174).
Reinbott 2016 reported a decrease in the prevalence of diarrhoea in the past two weeks in the intervention and control groups between the baseline (control 41.6%, intervention 36.9%) and impact survey (control 26.2%, intervention 27.9%).
See Table 5 for details of the effect of the intervention on diarrhoea as reported by the study authors.
4. Mortality
None of the included studies reported or evaluated the effects of educational intervention on infant/child mortality.
5. Hospitalization
Only one, community‐based study measured the effect of educational intervention on hospitalisation (Vitolo 2005). The study reported that the number of days spent hospitalised was nine days in the intervention arm and 15 days in the control group.
See Table 6 for details of the effect of the intervention on hospitalisation as reported by the study authors.
| Study | Result |
| Intervention = 9 days, control = 15 days |
6. Change in knowledge
Eight of the included studies reported positive outcomes of the intervention on the knowledge of caregivers (Aboud 2008; Aboud 2009; Aboud 2011; Negash 2014; Penny 2005; Shi 2010; Vazir 2013; Yin 2009). More intervention mothers recalled the intervention messages at follow‐up, could recall the recommended feeding practices and messages accurately, gave correct responses to questions on complementary feeding practices, and had higher knowledge scores. We were unable to combine the results in a meta‐analysis due to differences in the measures of knowledge that were used in the various studies. We present the study authors' report on the effect of the intervention on knowledge outcomes in Table 7.
| Study | Result (trial authors' judgement) |
| More intervention mothers recalled messages (5 out of 8 message categories P < 0.0001), especially hygiene (washing hands before eating), responsive feeding and talking to the child during the meal | |
| More intervention mothers recalled messages at follow‐up | |
| Mothers in the intervention group recalled more messages at follow‐up, especially pertaining to hygiene, self‐feeding, responding, stimulating, and foods to feed. Of 8 messages, control mothers recalled a mean of 0.59 (SD 1.0) and mothers in the intervention group recalled a mean of 2.37 (SD 1.5) | |
| Knowledge of complementary feeding in the intervention group rose from 5.8 (± 2.1) at baseline to 7.1 (± 1.0) at end line (P < 0.001), whereas scores for the control group stayed unchanged at 6.3 (± 1.6) at both time points | |
| Caregivers in the intervention group were more knowledgeable of key feeding practices and messages. | |
| At 6, 9, 12 and 18 months of age, after the implementation of the intervention, more caregivers in the intervention group responded correctly to the questions on feeding practices than those in the control group (statistically significant results for all questions) | |
| Educational messages to the intervention groups were significantly associated with changed maternal knowledge/beliefs about foods that are good for infants at ages 9 and 15 months. The percentage of mothers who had more knowledge regarding recommended foods from animal sources, such as egg and liver, and responded positively on selected appropriate foods to be given to infants, was higher, both at 9 and 15 months, in the intervention groups but this was not seen in the control group | |
| After being educated with feeding guideline on infants and young children, the knowledge of infants' mothers was greatly improved and KAP scores of the mothers after intervention were higher than at baseline (F = 183.556, P = 0.006); the percentage of correct answers on nutrition knowledge in the intervention groups was significantly higher than that of the control group. At six months of intervention, the KAP scores of intervention group 1 (12.0) and intervention group 2 (11.6) were higher than that of the control group (10.5) (least significant difference? (LSD) t = 5.96, P < 0.001; LSD t = 4.25, P < 0.001) |
KAP: knowledge, attitude and practice; SD: standard deviation
Sensitivity analyses
We conducted sensitivity analyses for the primary outcomes only. We re‐ran all analyses assuming a fixed‐effect model. The conclusions remained unchanged.
We investigated the impact of assuming an alternative ICC on the summary effect estimates for the following primary outcomes: duration of exclusive breastfeeding (≥ four months of age) and hygiene practices (predominantly defined as washing hands before feeding). For both outcomes we compared the impact on the pooled summary estimates using ICCs of 0.01, 0.05 and 0.10. For the outcome of duration of exclusive breastfeeding, only three studies were included in the meta‐analysis, and a single study was adjusted (Penny 2005). Increasing the ICC to 0.10 did not impact the results for this outcome (see Analysis 3.1). For the outcome of hygiene practices (handwashing before feeding), results remained in favour of educational intervention (see Analysis 3.2).
For the main analyses, we included studies according to intention‐to‐treat principles for dichotomous outcomes, and assumed that all study dropouts (regardless of allocation) had not experienced the 'event'. For complementary food introduced before four to six months, 149 participants dropped out of the intervention arms and 184 dropped out from the control arms. In the main analysis, we assumed that these participants had not introduced complementary foods. In the sensitivity analysis, therefore, we examined the impact of assuming dropouts had introduced complementary food before six months. The pooled average RR and 95% CI are very slightly attenuated towards the null (RR 0.89, 95% CI 0.81 to 0.97; Analysis 3.1), however, conclusions remained unchanged.
For duration of exclusive breastfeeding, 122 participants dropped out of the intervention arms and 160 dropped out from the control arms. In the main analysis it was assumed that these participants had not exclusively breastfed for at least four months. In the sensitivity analysis, we assumed that dropouts had been exclusively breastfed for four months or longer. The pooled average RR and 95% CI are attenuated towards the null (RR 1.00, 95% CI 0.85 to 1.18; Analysis 3.2). However, due to the extent of the uncertainty in the main analysis (RR 1.58, 95% CI 0.77 to 3.22), our conclusions for this outcome remain unchanged.
For improved hygiene practices (handwashing before feeding), 181 participants dropped out of the intervention arms and 150 dropped out from the control arms. (Note, for Shi 2010, we assumed the 110 dropouts had occurred equally between the control and intervention arms.) In the sensitivity analysis, we assumed that dropouts used appropriate hygiene practices before feeding their infant. Conclusions for this outcome also remain unchanged (RR 1.30, 95% CI 1.17 to 1.46; Analysis 3.3).
Discussion
Summary of main results
The review sought to assess the effectiveness of educational interventions for improving complementary feeding practices and other related health and growth outcomes in young children. We identified a total of 23 studies, 19 of which were community‐based studies and four were facility‐based studies. Overall, the evidence available suggests that educational interventions improve complementary feeding practices marginally; there was little evidence of an effect for growth patterns or nutritional status.
Effect of educational intervention on complementary feeding practices
There was a small positive effect of educational interventions on the time of commencement of complementary feeding by the caregivers of the children. However, the studies that were included in the meta‐analysis were all conducted in high‐income and lower‐ and upper‐ middle‐income countries (de Oliveira 2012; Edward 2013; Vitolo 2005; Wen 2011). The studies conducted in the lower‐middle‐ and low‐income countries did not report on time of commencement of complementary feeding hence there was no information to report. Edward 2013 showed greater benefit of the intervention in delaying the onset of complementary feeding than other studies. This may have been due to the mentorship model employed in the study as community doulas were used to deliver the educational intervention to adolescent mothers. These doulas had also been teenage mothers and were sufficiently familiar with the ethos and environment of the participants.
The focus of most of the included studies seemed to be on the adequacy (quality and quantity) of complementary foods fed to infants. Eighteen of the 23 included studies reported on this outcome in ways that were too varied to be combined for any form of analysis. All of the studies, however, reported improvements in the quality and quantity of complementary foods as indicated by the conclusions of the study authors. This showed that most caregivers in the intervention arms complied with the intervention messages irrespective of the fact that the studies did not provide complementary foods as part of the interventions. A possible explanation for this improvement is that most of the studies were conducted after undertaking formative research to identify gaps and resources available in these locations. This made the intervention messages culturally appropriate and enhanced the acceptability or affordability (or both) of the interventions, since most of the recommended foods were readily available in the intervention settings. This strengthens the evidence that educational interventions without the provision of foods are effective in improving complementary feeding practices. Although standard measures for accessing infant and young child feeding have been developed (e.g. the WHO minimum acceptable diet, minimum dietary diversity, minimum meal frequency), only one, recently conducted study put them to use (Reinbott 2016). This made it difficult to assess the adequacy of foods fed to infants using these indicators in a meta‐analysis and, as such, in this review we assessed adequacy of food fed to children based on results reported in the individual studies and the study authors' conclusions.
Educational interventions showed positive effect on the duration of exclusive breastfeeding for studies conducted in the community (Vitolo 2005; Wen 2011), but showed no effect on the studies conducted in health facilities (Penny 2005). The test for subgroup differences between community and facility‐based studies suggested a difference in treatment effect by setting.
Analysis of the (mostly community‐based) studies that reported hygienic practices showed that educational interventions had a weak positive effect on hygiene practices (Aboud 2009; Aboud 2011; Bhandari 2004; Shi 2010). One study conducted in sub‐Saharan Africa reported that educational intervention had a negative effect on hygiene practices (not included in meta‐analysis; Negash 2014). Although the study authors did not report on water availability in the study area, it is well established that this can threaten compliance with recommended hygiene practices. Interestingly, all of the studies that reported on hygiene practices were conducted in the community.
The effect of educational interventions in preventing diarrhoea showed mixed results. Four community‐based studies reported this outcome and only one study recorded a clearly beneficial effect of educational interventions in reducing the episodes of diarrhoea in the intervention group. The other three studies found no clear effect on the incidence and prevalence of diarrhoea.
Educational interventions were effective in reducing the days spent in the hospital in one community‐based study. Other studies did not report this outcome.
Educational interventions were also effective in improving the knowledge of caregivers in all of the included studies. Although we were unable to pool the results in a meta‐analysis, the study authors reported that caregivers in the intervention groups were able to recall the intervention messages at follow‐up, recall recommended feeding practices and messages accurately, and had higher knowledge scores.
None of the studies reported any clear adverse effects of the interventions.
Effect of educational intervention on growth
The studies included in the meta‐analyses did not show an effect of educational intervention on growth parameters. The test for differences in the weight of the children taken at baseline and at 6, 12, 18 and 24 months did not show any statistical difference. The analysis showed similar findings for height/length and for underweight, stunting and wasting.
Of the studies not included in the quantitative analysis, three showed a positive effect of educational intervention on growth parameters, while the other two did not suggest a positive effect of educational intervention.
Although the study authors measured growth parameters at various time points, we only included growth parameters at 6, 12, 18 and 24 months of age in the meta‐analysis. This is because 6 and 12 months of age mark the half and first year of an infant's life respectively, and since the rate of growth reduces after infancy, we choose a six‐monthly interval thereafter (18 and 24 months of age).
We found no studies evaluating or reporting the effects of educational interventions on mortality.
Overall completeness and applicability of evidence
Of the 23 studies included in this review, five were conducted in high‐income countries: Australia (Campbell 2013; Daniels 2012; Wen 2011), Germany (Koehler 2007) and the USA (Schroeder 2015). Six were conducted in upper‐middle‐income countries: Brazil (de Oliveira 2012; Vitolo 2005), China (Shi 2010; Yin 2009), Colombia (Olaya 2013), and Peru (Penny 2005). Eight were conducted in lower‐middle‐income countries, including Bangladesh (Aboud 2008; Aboud 2009; Aboud 2011), Cambodia (Reinbott 2016), India (Bhandari 2001; Bhandari 2004; Vazir 2013), and Pakistan (Saleem 2014). Three studies were conducted in a low‐income country: Ethiopia in tropical Africa ((Kang 2017; Negash 2014; Tariku 2015). The location of one study was not stated in the study report (Edward 2013).
Eight of the 23 studies were conducted in urban settings (Daniels 2012; de Oliveira 2012; Edward 2013; Koehler 2007; Olaya 2013; Schroeder 2015; Vitolo 2005; Wen 2011), two in peri‐urban settings (Penny 2005; Saleem 2014), one in an urban slum (Bhandari 2001), and 11 in rural settings (Aboud 2008; Aboud 2009; Aboud 2011; Bhandari 2004; Kang 2017; Negash 2014; Reinbott 2016; Shi 2010; Tariku 2015; Vazir 2013; Yin 2009). One study report stated that the study was conducted in local government areas but did not state clearly whether the setting was urban, semi‐urban or rural (Campbell 2013). Community‐based studies were well distributed among the high‐ and middle‐income countries but health facility‐based studies were conducted mainly in the high‐ and upper‐middle‐income countries.
The findings of these studies could be applied across the social groups because the studies were conducted in high‐, upper‐middle‐ and lower‐middle‐income countries. However, it is important to note that the studies from low‐income settings were all from the same country in sub‐Saharan Africa (Ethiopia), consequently while the findings of this study could be applied in the high‐, lower‐upper‐ and lower‐middle‐income countries, the same cannot be said of the low‐income countries where the three studies in this classification were conducted in the same country (Ethiopia).
The participants included in the studies, mother/caregiver‐child pairs, were also properly suitable for the review since the children included in the studies ranged from birth to 24 months of age and this age bracket includes the time frame for the onset of complementary feeding. Most of the outcomes were measured on children while mothers/caregivers received the educational intervention. The intervention delivery mechanisms and promotional activities are also assessed as applicable across settings since they generally included group sessions/meetings, demonstration and practical sessions, the use of flip charts, picture books and brochures. These strategies are easily reproducible across settings irrespective of income classification or development rating.
The intervention messages were also culturally appropriate and incorporated locally available foods in recommendations on the types of foods and food groups to be fed to children of complementary feeding age. This encouraged the mothers/caregivers to use resources locally available to them and increased the acceptability of the intervention. This was evident by the rate of compliance and, in one of the studies, the mothers contributed the cooking materials used in the nutrition sessions. The messages also included key aspects of adequate complementary feeding such as recommendations on the duration of breastfeeding, continued breastfeeding in addition to complementary foods, dietary diversity, consistency of complementary foods, hygiene and feeding based on satiety cues.
In general, the majority of the interventions were delivered to groups of women (typically the mothers) or caregivers in their own homes. Interventions used a mixture of interactive sessions, demonstrations of correct practice, imitation, role plays, group discussions, peer support, story telling, picture books and village rallies amongst others. Reporting of exact intervention content was mostly poor; for example, replication of interventions from reported detail may not be possible. In the same vein, an appraisal of the educational approaches used in the studies is most likely not feasible. Nothwithstanding that this review did not set out to evaluate the education models/approaches used in implementing the studies, participatory approaches, such as the 'Trials of Improved Practices' (TIPs) and other formative research procedures, are believed to yield higher levels of acceptability for the interventions being implemented.
The studies also measured key child‐feeding indicators and outcomes, which are generally measurable across settings and, as such, can be easily applied and replicated.
Quality of the evidence
We assessed the quality of evidence using the GRADE approach (Guyatt 2008). The evidence that educational interventions improve complementary feeding practices (time of introduction of complementary foods) is considered to be of moderate quality (summary of findings Table for the main comparison), while that of growth outcomes is considered to be of low to very low quality (summary of findings Table 2). Most of the studies were at unclear risk of selection bias due to unclear allocation concealment. In addition, some of the studies were at high or unclear risk of performance and detection bias since they did not blind or describe the blinding of participants, personnel and outcome assessors. Most of the studies favoured the intervention arms, although the results of the meta analysis showed some imprecision.
Consequently, further research is likely to have an important impact on our confidence in the estimate of effect and may change the estimate for improved complementary feeding practices. We are very uncertain about the estimates of effects for the growth outcomes, which indicate that evidence is insufficient to confirm that education is an effective intervention for improving the growth of infants, while further research is very likely to have an important impact on our confidence in the estimate of effect for nutritional status and is likely to change the estimate.
Potential biases in the review process
This review attempted to assess the effect of educational interventions on a broad spectrum of topical aspects of complementary feeding. It is the only Cochrane Review that has evaluated the effectiveness of education on four key aspects of complementary feeding across the globe. Other non‐Cochrane reviews have assessed the effectiveness of education and other complementary feeding interventions on complementary feeding and growth in low‐income countries (Imdad 2011; Lassi 2013; Shi 2011), while Dewey 2008 assessed the effectiveness of complementary feeding interventions in general in low‐income countries. Our search strategy was highly sensitive and we did not apply any language restrictions. We also included published data and contacted study authors for unpublished data.
As shown in the 'Risk of bias' assessment, one potential bias in the review process was that a number of included studies were unable to blind participants and personnel, as such we cannot rule out the possibility of detection bias and its effect on the results in the intervention groups. We were also unable to retrieve the full texts of 10 studies we believe might qualify for inclusion in this review (see Studies awaiting classification). Due to the limited number of studies we were able to include in our meta‐analyses, we did not conduct the planned sensitivity analyses to detect the effect of excluding studies with missing data, unpublished studies, and studies with high risk of bias on the overall results of the meta‐analysis.
Some studies in our analysis either did not account for the effect of clustering in their analysis, or reported raw (unadjusted) estimates. As such, we followed section 16.3.4 and 16.3.5 of the Cochrane Handbook for Systematic Reviews of Interventions for calculating the effective sample size and incorporating cluster studies in the meta‐analysis (Higgins 2011). These are approximate methods and results should be interpreted accordingly.
Agreements and disagreements with other studies or reviews
The effectiveness of educational interventions for improving complementary feeding practices in low‐income countries has been previously studied by Shi 2011. The findings of this review agree with that of Shi 2011, although it was limited to low‐income countries. On the effect of educational interventions on growth, the findings of this review are similar to those of Imdad 2011 and Lassi 2013, notwithstanding that the studies were also undertaken in low‐income countries. In general, the review by Dewey 2008 found educational interventions to be an effective strategy for promoting appropriate complementary feeding in low‐income countries.
Theoretical model: educational interventions for improving complementary feeding practices
Footnotes
aGSM: global system for mobile communication.
Risk of bias graph: review authors' judgements about each risk of bias item presented as percentages across all included studies
Risk of bias summary: review authors' judgements about each risk of bias item for each included study
Sensitivity analysis 3. Comparison of different ICC (primary outcomes), outcome: 3.1 duration of exclusive breastfeeding (≥ 4 months of age)
Sensitivity analysis 3. Comparison of different ICC (primary outcomes), outcome: 3.2 hygiene: handwashing before feeding
Comparison 1 Educational intervention versus no educational intervention for improving complementary feeding practices (ICC = 0.02), Outcome 1 Complementary food introduced at appropriate age.
Comparison 1 Educational intervention versus no educational intervention for improving complementary feeding practices (ICC = 0.02), Outcome 2 Duration of exclusive breastfeeding (≥ 4 months old).
Comparison 1 Educational intervention versus no educational intervention for improving complementary feeding practices (ICC = 0.02), Outcome 3 Hygiene practices: community‐based intervention.
Comparison 1 Educational intervention versus no educational intervention for improving complementary feeding practices (ICC = 0.02), Outcome 4 Knowledge.
Comparison 2 Educational intervention versus no educational intervention for improving complementary feeding practices: growth outcomes (ICC = 0.05), Outcome 1 Weight.
Comparison 2 Educational intervention versus no educational intervention for improving complementary feeding practices: growth outcomes (ICC = 0.05), Outcome 2 Height/length.
Comparison 2 Educational intervention versus no educational intervention for improving complementary feeding practices: growth outcomes (ICC = 0.05), Outcome 3 Nutritional status (underweight, stunting, wasting).
Comparison 3 Sensitivity analyses for dropouts (primary outcomes), Outcome 1 Sensitivity analysis: introduction of complementary food.
Comparison 3 Sensitivity analyses for dropouts (primary outcomes), Outcome 2 Sensitivity analysis: duration of exclusive breastfeeding (dropouts as responders).
Comparison 3 Sensitivity analyses for dropouts (primary outcomes), Outcome 3 Sensitivity analysis: hygiene practice (dropouts as responders).
| Educational intervention versus no educational intervention for improving complementary feeding practices | ||||||
| Patient or population: children of complementary feeding age Comparison: no educational intervention | ||||||
| Outcomes | Illustrative comparative risks* (95% CI) | Relative effect | Number of participants | Quality of the evidence | Comments | |
| Assumed risk | Corresponding risk | |||||
| No educational intervention | Educational intervention (ICC = 0.02) | |||||
| Age at introduction of complementary foods | Study population | RR 0.88 | 1738 | ⊕⊕⊕⊝ | ‐ | |
| 661 per 1000 | 581 per 1000 | |||||
| Moderate | ||||||
| 746 per 1000 | 656 per 1000 | |||||
| Duration of exclusive breastfeeding (≥ 4 months of age) | Study population | RR 1.58 | 1544 | ⊕⊝⊝⊝ | ‐ | |
| 129 per 1000 | 204 per 1000 | |||||
| Moderate | ||||||
| 0 per 1000 | 0 per 1000 | |||||
| Duration of exclusive breastfeeding (≥ 4 months of age): community‐based intervention | Study population | RR 2.32 | 1167 | ⊕⊕⊝⊝ | ‐ | |
| 40 per 1000 | 92 per 1000 | |||||
| Moderate | ||||||
| 0 per 1000 | 0 per 1000 | |||||
| Duration of exclusive breastfeeding (≥ 4 months of age): facility‐based intervention | Study population | RR 0.95 | 377 | ⊕⊕⊝⊝ | ‐ | |
| 426 per 1000 | 405 per 1000 | |||||
| Moderate | ||||||
| 0 per 1000 | 0 per 1000 | |||||
| Hygiene practices: community‐based intervention | Study population | RR 1.38 | 2029 | ⊕⊕⊕⊝ | ‐ | |
| 546 per 1000 | 754 per 1000 | |||||
| Moderate | ||||||
| 0 per 1000 | 0 per 1000 | |||||
| *The basis for the assumed risk (e.g. the median control group risk across studies) is provided in footnotes. The corresponding risk (and its 95% CI) is based on the assumed risk in the comparison group and the relative effect of the intervention (and its 95% CI) | ||||||
| GRADE Working Group grades of evidence | ||||||
| aWe downgraded the quality of the evidence by one level due to serious risks of bias; the method of sequence generation, allocation concealment and blinding of outcome assessors was unclear or not undertaken in some of the studies | ||||||
| Educational intervention versus no educational intervention for improving complementary feeding practices: growth outcomes | ||||||
| Patient or population: children of complementary feeding age Comparison: no educational intervention | ||||||
| Outcomes | Illustrative comparative risks* (95% CI) | Relative effect | Number of participants | Quality of the evidence | Comments | |
| Assumed risk | Corresponding risk | |||||
| No educational intervention | Educational intervention (ICC = 0.05) | |||||
| Weight (at 6 months of age) | ‐ | The mean weight at 6 months of age in the intervention groups was 0.03 kg higher (0.10 lower to 0.17 higher) | ‐ | 1221 | ⊕⊝⊝⊝ | ‐ |
| Weight (at 12 months of age) | ‐ | The mean weight at 12 months of age in the intervention groups was 0.06 kg higher (0.04 lower to 0.15 higher) | ‐ | 2464 | ⊕⊝⊝⊝ | ‐ |
| Height/length (at 6 months of age) | ‐ | The mean height/length at 6 months of age in the intervention groups was 0.16 cm higher (0.21 lower to 0.52 higher) | ‐ | 1221 | ⊕⊝⊝⊝ | ‐ |
| Height/length (at 12 months of age) | ‐ | The mean height/length at 12 months of age in the intervention groups was 0.32 cm higher (0.11 to 0.52 higher) | ‐ | 2464 | ⊕⊕⊝⊝ | ‐ |
| Nutritional status: stunting (H/LAZ ≤ −2 SD) | 199 per 1000 | 177 per 1000 | RR 0.89 | 3487 | ⊕⊕⊝⊝ | ‐ |
| Nutritional status: wasting (WH/LZ ≤ −2 SD) | 400 per 1000 | 316 per 1000 | RR 0.79 | 2000 | ⊕⊕⊝⊝ | ‐ |
| Nutritional status: underweight (WAZ ≤ −2 SD) | 138 per 1000 | 136 per 1000 | RR 0.99 | 2900 | ⊕⊕⊝⊝ | ‐ |
| *The basis for the assumed risk (e.g. the median control group risk across studies) is provided in footnotes. The corresponding risk (and its 95% CI) is based on the assumed risk in the comparison group and the relative effect of the intervention (and its 95% CI) | ||||||
| GRADE Working Group grades of evidence | ||||||
| aWe downgraded the quality of the evidence by two levels due to very serious risks of bias; the method of sequence generation, allocation concealment and blinding of outcome assessors was unclear or not undertaken in most of the studies | ||||||
| Measures of treatment effect | Event rate outcomes In this review, it is possible that some outcomes (e.g. diarrhoea, hospitalisation, malnutrition) may have been recorded as counts where the event can occur multiple times to the same participant. Where study data allow (i.e. data are available on both events and person‐years at risk), we will calculate rate ratios for count outcomes. However, study authors can report count data in a number of ways. As such, our strategy will be to extract count data in the form as reported by the original authors. For example, if study authors have reported the outcome using a rate ratio, we will extract it as such. If study authors have reported the outcome as dichotomous, we will extract it as a dichotomous outcome, noting the potential disadvantages of doing so. Multiple outcome data It is possible that studies will summarise outcomes in several ways, for example, both as a continuous and dichotomous measure. For the primary outcomes, if person‐years at risk are available, our preference will be to analyse count data as a rate ratio. However, if sufficient information is not available, and the event is common, we will analyse count data as if it were continuous. We consider the continuous measure to be clinically reasonable and preferable to dichotomising the primary outcomes. If neither of these approaches is suitable, we will extract the data as if it were dichotomous, ensuring that we classify all participants into one of two possible groups only. |
| Unit of analysis issues | Multiple intervention groups Studies with more than two intervention arms can pose analytical problems in a meta‐analysis. For example, it is important to avoid 'double‐counting' of participants. Where studies may have two or more active arms to be compared against a control, or two control conditions versus an experimental condition, we will combine similar interventions to generate a single pair‐wise comparison for the meta‐analysis. If interventions are not similar, we will split the 'shared' comparator into two groups and include as two comparisons. |
| Dealing with missing data | If we are unable to retrieve missing dichotomous data, we will conduct an available‐case analysis. We plan to undertake a sensitivity analysis assuming that participants who withdrew from either arm after randomisation experienced a negative event. In common with many public health educational interventions, dropouts are often due to perceived difficulties with the intervention or information contradictory to existing beliefs or community norms (among other reasons). As such, it is not realistic to consider a 'best case' sensitivity analysis where all dropouts successfully adhered to the intervention, for weaning practice. We will analyse missing continuous data on a completers basis, including only those participants with a final assessment. Where we are unable to obtain the missing SDs from the study authors, we will calculate them from P values, t values, confidence intervals, or standard errors, where these have been reported. If this is not possible, and only a minority of studies are missing SDs, we will impute the SD using other studies in the meta‐analysis. We will also report the extent of the missing data, describe the attrition for each study in the 'Risk of bias' tables, and discuss the possible impact of this missing data on the results of the review. We will perform a sensitivity analysis to assess the impact of the inclusion of studies with missing data on the findings of the review (Deeks 2017, Section 9.7). |
| Assessment of reporting biases | We will try to minimise publication bias by doing a comprehensive search of multiple sources and databases, and by including studies of good methodological quality and data from unpublished and ongoing studies (Sterne 2017, Section 10.3). If we have a sufficient number of included studies (at least 10), we will use outcome data to run a funnel plot regression to investigate the possibility of publication bias (Sterne 2017, Section 10.4). Funnel plot asymmetry could be due to publication bias, poor methodological quality, true heterogeneity, or a real relationship between study size and effect size or chance. We will further investigate publication bias by comparing the data extracted from published and unpublished studies in a sensitivity analysis (Sterne 2017, Section 10.4.4) |
| Subgroup analysis and investigation of heterogeneity |
|
| Sensitivity analysis | We will conduct a sensitivity analysis in order to detect the effect of excluding studies with missing data, unpublished studies, and studies with high risk of bias (judged using Cochrane’s tool for assessing risk of bias (Higgins 2017)) on the overall results of the meta‐analysis. In this analysis, we will explore the possible effects of marked differences between included studies. We will also undertake a fixed‐effect meta‐analysis to determine the robustness of the results from the random‐effects meta‐analysis |
| SD: standard deviation. | |
| Study | Promotional activity | Message content | Ways information was collected/outcome measure assessed | Intervention providers | Delivery (e.g. mechanism, medium, intensity, fidelity) |
|
|
|
| Peer educators | During weekly group sessions | |
|
|
|
| Peer educators | Group training sessions held weekly | |
|
|
|
| Peer educators | Group training sessions held weekly | |
| Counselling sessions using a nutritional counselling guide book | Not described |
| Trained nutritionists | Monthly counselling sessions | |
|
|
|
|
| Counselling on complementary feeding conducted as follows:
| |
|
| Intervention materials incorporated 6 purpose‐designed key messages (for example, “Color Every Meal With Fruit and Veg,” “Eat Together, Play Together,” “Off and Running”) within a purpose‐designed DVD and written materials |
| Dietician | 6 x 2‐h sessions delivered quarterly at first‐time parents’ group regular meeting | |
|
| Messages in:
|
|
| Interactive group sessions at a choice of days and times, and at the same child health centres as those used for measurements | |
|
|
|
|
| The counselling sessions occurred in the maternity ward close to the time for hospital discharge and at 7, 15, 30, 60, and 120 days after the birth at the mother's home | |
|
| Doulas discouraged the introduction of solid food during the early months of life for both breast‐fed and formula‐fed infant |
| Doulas |
| |
|
| Mothers discussed messages around:
with the operators |
| Female operators | During group nutrition education sessions | |
|
|
|
|
| The counselling was carried out during education sessions in the community | |
|
|
|
| Trained community nutrition promoter (CNP) together with local NGO conducted the nutrition education sessions | The 7 nutrition education sessions were held 2–4 hours weekly or biweekly depending on the availability of the participants | |
|
|
| Unclear | 2 female research assistants (with at least 14 years of schooling) and 2 female community health workers (with at least 10 years of schooling) | Interventions were offered in participants' homes | |
|
| Not described |
| Healthcare providers in the intervention areas |
| |
|
|
| Interviews using questionnaires |
|
| |
|
|
|
| High‐school‐educated village women who were themselves mothers | Home visits | |
|
|
Intervention messages were based on the “Ten steps for healthy feeding for Brazilian children from birth to 2 years of age” |
| Trained field workers who were undergraduate students in nutritional sciences | Home visits | |
| Counselling sessions on infant feeding practices, infant nutrition and active play, family physical activity and nutrition, as well as social support |
|
| Trained research nurses | Home visits | |
|
| Mothers were educated with feeding guideline on infants and young children
| ‐ | Experts in maternal and child nutrition | ‐ | |
| NGO: non‐governmental organisation; TN: study number | |||||
| Study | Promotional activity | Message content | Ways information was collected/outcome measure assessed | Intervention providers | Delivery (e.g. mechanism, medium, intensity, fidelity) |
|
| Nutrition counselling was based on the Dietary Schedule for the First Year of Life (Dietary Schedule) recommended by the Nutrition Committee of the German Pediatric Society. Recommendations of the schedule include:
|
| Counsellors | Telephone calls and printed materials | |
|
| Guidelines focused on the following 3 main messages that were emphasised at all study visits:
Mothers were offered specific advice on the number of portions of meat that should be given; mothers were also advised to include chicken liver and heart as affordable forms of meat, and suggestions were given for the preparation of recommended foods. Mothers were also advised to give fruit and vegetables daily |
| Researchers | Clinic visits | |
|
|
|
| Health workers | Health facility | |
|
| The intervention was based on the modules of Growing Leaps and Bounds, a set of educational materials developed by a group of experts and funded by the Dannon Institute. These materials aim at:
While the brochures emphasise a few key points, they also provide detailed advice on infant feeding practices, physical activity, and developmental milestones related to eating patterns |
|
| Paediatric visits at 1, 2, 4, 6, 9, 12, 15, 18, and 24 months of age and at annual visits thereafter up to 5 years of age | |
| S/N: study number | |||||
| Study | Interventions |
| Intervention group 1 (RFS): 6 weekly sessions on responsive parenting (feeding and stimulation) in addition to the regular programme Intervention group 2 (RFS plus Sprinkles): 6 weekly sessions on responsive parenting (feeding and stimulation) in addition to the regular programme and 6 months of a food powder fortified with minerals and vitamins Control: regular programme | |
| Intervention group 1: received a milk‐based cereal and nutritional counselling Intervention group 2: monthly nutritional counselling alone Intervention group 3: visitation group (used as the control group in the study) | |
| Intervention group 1: were offered a telephone hotline 3 times per week, open for 2 hours each time Intervention group 2: received additional written information on the Dietary Schedule distributed in 3 parts, each dealing with the diet in the coming period Intervention group 3: were offered additional personal telephone counselling | |
| Intervention group 1: the complementary feeding group (CFG) received the integrated child development services plus the World Health Organization recommendations on breastfeeding and complementary foods Intervention group 2: the responsive complementary feeding and play group received the same intervention as the CFG plus skills for responsive feeding and psychosocial stimulation Control: routine Integrated Child Development Services ‐ standard of care | |
| RFS: responsive feeding and stimulation | |
| Study | Result |
| The incidence and prevalence of diarrhoea and ALRI were not significantly affected by either intervention Nutritional counselling group: episodes per child 6.9 (± 3.2), prevalence per 100: d 14.6 (± 12.0) Visitation group: episodes per child 6.7 (± 3.4), prevalence per 100: d 13.2 (± 9.8) | |
| The reported prevalences of common illnesses in the previous 7 days did not differ in the 2 groups | |
| Diarrhoeal illness in the past 2 weeks (%) Baseline: intervention = 36.9%, control = 41.6% Impact: intervention = 27.9%, control = 26.2% | |
| Number with event: intervention = 46, control = 98 | |
| ALRI: acute lower respiratory infection. | |
| Study | Result |
| Intervention = 9 days, control = 15 days |
| Study | Result (trial authors' judgement) |
| More intervention mothers recalled messages (5 out of 8 message categories P < 0.0001), especially hygiene (washing hands before eating), responsive feeding and talking to the child during the meal | |
| More intervention mothers recalled messages at follow‐up | |
| Mothers in the intervention group recalled more messages at follow‐up, especially pertaining to hygiene, self‐feeding, responding, stimulating, and foods to feed. Of 8 messages, control mothers recalled a mean of 0.59 (SD 1.0) and mothers in the intervention group recalled a mean of 2.37 (SD 1.5) | |
| Knowledge of complementary feeding in the intervention group rose from 5.8 (± 2.1) at baseline to 7.1 (± 1.0) at end line (P < 0.001), whereas scores for the control group stayed unchanged at 6.3 (± 1.6) at both time points | |
| Caregivers in the intervention group were more knowledgeable of key feeding practices and messages. | |
| At 6, 9, 12 and 18 months of age, after the implementation of the intervention, more caregivers in the intervention group responded correctly to the questions on feeding practices than those in the control group (statistically significant results for all questions) | |
| Educational messages to the intervention groups were significantly associated with changed maternal knowledge/beliefs about foods that are good for infants at ages 9 and 15 months. The percentage of mothers who had more knowledge regarding recommended foods from animal sources, such as egg and liver, and responded positively on selected appropriate foods to be given to infants, was higher, both at 9 and 15 months, in the intervention groups but this was not seen in the control group | |
| After being educated with feeding guideline on infants and young children, the knowledge of infants' mothers was greatly improved and KAP scores of the mothers after intervention were higher than at baseline (F = 183.556, P = 0.006); the percentage of correct answers on nutrition knowledge in the intervention groups was significantly higher than that of the control group. At six months of intervention, the KAP scores of intervention group 1 (12.0) and intervention group 2 (11.6) were higher than that of the control group (10.5) (least significant difference? (LSD) t = 5.96, P < 0.001; LSD t = 4.25, P < 0.001) | |
| KAP: knowledge, attitude and practice; SD: standard deviation | |
| Outcome or subgroup title | No. of studies | No. of participants | Statistical method | Effect size |
| 1 Complementary food introduced at appropriate age Show forest plot | 4 | 1738 | Risk Ratio (M‐H, Random, 95% CI) | 0.88 [0.83, 0.94] |
| 1.1 Community intervention (≥ 6 months old) | 3 | 1490 | Risk Ratio (M‐H, Random, 95% CI) | 0.86 [0.80, 0.93] |
| 1.2 Community intervention (≥ 4 months old) | 1 | 248 | Risk Ratio (M‐H, Random, 95% CI) | 0.92 [0.83, 1.02] |
| 2 Duration of exclusive breastfeeding (≥ 4 months old) Show forest plot | 3 | 1544 | Risk Ratio (Random, 95% CI) | 1.58 [0.77, 3.22] |
| 2.1 Community‐based intervention | 2 | 1167 | Risk Ratio (Random, 95% CI) | 2.32 [1.45, 3.73] |
| 2.2 Facility‐based intervention | 1 | 377 | Risk Ratio (Random, 95% CI) | 0.95 [0.70, 1.29] |
| 3 Hygiene practices: community‐based intervention Show forest plot | 4 | 2029 | Risk Ratio (Random, 95% CI) | 1.38 [1.23, 1.55] |
| 4 Knowledge Show forest plot | 2 | 399 | Mean Difference (IV, Random, 95% CI) | 1.29 [0.33, 2.25] |
| Outcome or subgroup title | No. of studies | No. of participants | Statistical method | Effect size |
| 1 Weight Show forest plot | 7 | Mean Difference (Random, 95% CI) | Subtotals only | |
| 1.1 Mean weight (kg) at 6 months old | 3 | 1221 | Mean Difference (Random, 95% CI) | 0.03 [‐0.10, 0.17] |
| 1.2 Mean weight (kg) at 12 months old | 5 | 2464 | Mean Difference (Random, 95% CI) | 0.06 [‐0.04, 0.15] |
| 1.3 Mean weight (kg) at 18 months old | 2 | 1402 | Mean Difference (Random, 95% CI) | 0.10 [‐0.14, 0.35] |
| 1.4 Mean weight (kg) at 24 months old | 2 | 920 | Mean Difference (Random, 95% CI) | ‐0.14 [‐0.36, 0.08] |
| 2 Height/length Show forest plot | 7 | Mean Difference (Random, 95% CI) | Subtotals only | |
| 2.1 Height/length (cm) at 6 months old | 3 | 1221 | Mean Difference (Random, 95% CI) | 0.16 [‐0.21, 0.52] |
| 2.2 Height/length (cm) at 12 months old | 5 | 2464 | Mean Difference (Random, 95% CI) | 0.32 [0.11, 0.52] |
| 2.3 Height/length (cm) at 18 months old | 2 | 1402 | Mean Difference (Random, 95% CI) | 0.58 [‐0.22, 1.38] |
| 2.4 Height/length (cm) at 24 months old | 2 | 920 | Mean Difference (Random, 95% CI) | ‐0.13 [‐0.58, 0.32] |
| 3 Nutritional status (underweight, stunting, wasting) Show forest plot | 5 | Risk Ratio (Random, 95% CI) | Subtotals only | |
| 3.1 Stunting (HAZ ≤ ‐2 SD) | 5 | 3487 | Risk Ratio (Random, 95% CI) | 0.89 [0.74, 1.06] |
| 3.2 Wasting (WHZ ≤ ‐2 SD) | 2 | 2000 | Risk Ratio (Random, 95% CI) | 0.79 [0.48, 1.30] |
| 3.3 Underweight (WAZ ≤ ‐2 SD) | 3 | 2900 | Risk Ratio (Random, 95% CI) | 0.99 [0.68, 1.44] |
| Outcome or subgroup title | No. of studies | No. of participants | Statistical method | Effect size |
| 1 Sensitivity analysis: introduction of complementary food Show forest plot | 4 | 1738 | Risk Ratio (M‐H, Random, 95% CI) | 0.89 [0.81, 0.97] |
| 2 Sensitivity analysis: duration of exclusive breastfeeding (dropouts as responders) Show forest plot | 3 | 1544 | Risk Ratio (Random, 95% CI) | 1.00 [0.85, 1.18] |
| 3 Sensitivity analysis: hygiene practice (dropouts as responders) Show forest plot | 4 | 2029 | Risk Ratio (Random, 95% CI) | 1.30 [1.17, 1.46] |
