A scoping review of social‐behaviour change techniques applied in complementary feeding interventions

Abstract Education and other strategies to promote optimal complementary feeding can significantly improve practices, but little is known about the specific techniques successful interventions use to achieve behaviour change. We reviewed the literature for complementary feeding interventions in low‐/middle‐income countries (LMIC) published since 2000. We systematically applied a validated taxonomy mapping process to code specific behaviour change techniques (BCTs) used in each intervention; effectiveness ratios for each BCT were estimated. Sixty‐four interventions met inclusion criteria, were abstracted, BCTs identified, and coded. Dietary diversity was the most commonly assessed component of complementary feeding, and interpersonal communication, either individually or in groups, was the most commonly used delivery platform. Of the 93 BCTs available for mapping, the 64 interventions included in this review applied a total of 28 BCTs. Interventions used a median of six techniques (max = 13; min = 2). All interventions used “instruction on how to perform the behaviour.” Other commonly applied BCTs included “use of a credible source” (n = 46), “demonstration of the behaviour” (n = 35), and “providing information about health consequences” (n = 30). Forty‐three interventions reported strategies to shift the physical or social environment. Among BCTs used in >20 interventions, five had effectiveness ratios >0.8: “provision of/enabling social support”; “providing information about health consequences”; “demonstration of the behaviour”; and “adding objects to the environment” namely, food, supplements, or agricultural inputs. The limited reporting of theory‐based BCTs in complementary feeding interventions may impede efforts to improve and scale effective programs and reduce the global burden of malnutrition.


| INTRODUCTION
Breastmilk is recognized as the optimal food for infants, but beginning around 6 months of age, breastfed infants cannot meet their nutrient needs from human milk alone. Therefore, appropriate complementary feeding practices are essential to optimize child growth and development especially from 6 to 24 months of age when children are at high risk of undernutrition (Dewey, Lutter, Martines, & Daelmans, 2001).
Appropriate complementary feeding encompasses the provision of (a) age-appropriate amounts of hygienically prepared food each day in terms of meals per day and amount of food served per meal; (b) foods of adequate thickness to ensure energy density; and (c) a sufficiently diverse diet, consisting of four or more food groups daily, to ensure micronutrient requirements are met (Dewey, 2001;WHO, 2008a;WHO, 2008b). Despite considerable efforts, progress to improve complementary feeding practices in low-/middle-income countries (LMIC) has been slow. For too many infants, complementary foods are of inadequate nutritional quality, introduced too early or too late, or are provided in insufficient quantity or frequency. From 1990 to 2010, less than one third of children aged 6-24 months received adequate dietary diversity and only about half received a sufficient number of meals each day (Lutter et al., 2011); 20% were fed a diet that met minimum meal adequacy requirements. These numbers remained relatively unchanged from 2010 to 2016 (J. M. White, Begin, Kumapley, Murray, & Krasevec, 2017).
Initiation and maintenance of sustained behaviour change is foundational to improving complementary feeding practices and child nutrition. Recent systematic reviews of behaviour change programs designed to improve complementary feeding practices found that although practices can be changed for the better, the degree of improvement and impacts on child nutrition are often small (Dewey & Adu-Afarwuah, 2008;Graziose, Downs, O'Brien, & Fanzo, 2018;Heidkamp, 2017;Lamstein et al., 2014;Lassi, Das, Zahid, Imdad, & Bhutta, 2013). Those achieving significant changes used formative research to identify critical barriers and facilitators of optimal practice of behaviour and applied these findings to shape processes and strategies to address behavioural change at the individual or societal level.
Effective interventions also clearly delineated the impact pathway of the intervention, specifying the steps from behaviour change activities to nutritional impacts. Of concern, however, is the relatively limited use or explicit specification of individual behaviour change theory in the design and implementation of complementary feeding interventions (Briscoe & Aboud, 2012;Pelto, Martin, van Liere, & Fabrizio, 2016). Use of theory facilitates rigorous intervention design, allowing programmers to explicate the proposed mechanisms through which an intervention is hypothesized to change behaviour. Use of theory also enables comparison of interventions and aids intervention replication (Breuer, Lee, De Silva, & Lund, 2016;Mayne & Johnson, 2015).
Beyond these requirements is the need to replicate interventions to both verify impacts and enable scale up. However, replication proves challenging for complementary feeding interventions. Intervention design frameworks for behaviour change and implementation details that allow bridging of activities to specific behaviour change are rarely reported in manuscripts and reports. Recently, the Behaviour Change Wheel framework was developed to better support intervention designers, implementers, and evaluators (Michie, Atkins, & West, 2014;. The Behaviour Change Wheel draws on the Capabilities, Opportunities, and Motivations framework for behaviour change (COM-B). The COM-B model specifies three behavioural constructs-capability, opportunity, and motivation (COM)-that interact for initiation and maintenance of behaviour change (-B). Each component of the model (C, O, and M) can be further subdivided into 15 theoretical domains using the Theoretical Domains Framework (TDF; Cane, O'Connor, & Michie, 2012;French et al., 2012). Nine intervention functions, identified by literature review and expert consensus, encompass the vast majority of possible intervention categories that map to the COM-B and subsequent TDF to promote behaviour change. From these broad categories, framework developers applied and iterative DELPHI approach to derive and validate a taxonomy of 93 hierarchically organized behaviour change techniques (BCTs; Abraham & Michie, 2008;Michie et al., 2013;Michie et al., 2015). BCTs, often referred to as the "active ingredients" of a behaviour change intervention, are specific, irreducible actions that can be observed and replicated. In the Behaviour comparison across behaviour change interventions. Specific BCTs that are part of an intervention can be identified through coding components according to a taxonomy and mapped for each intervention.
Mapping then allows researchers to identify, categorize, and synthesize BCTs across interventions.
Because behaviours and the interventions designed to change them are often complex, identifying BCTs not only allows researchers to identify the component of an intervention that has an effect on behaviour but also supplies a standard for comparing causal mechanisms across interventions.
Identification and mapping of BCTs has been applied to numerous health care and population level behaviour change interventions (Michie et al., 2014), including diet-related behaviour change interventions for obesity reduction in North America and Europe (Martin, Chater, & Lorencatto, 2013;Tate et al., 2016). However, this approach has had limited application to child nutrition in low-/middle-income (LMIC) settings (Aboud & Singla, 2012). Such taxonomy mapping would not only support the identification of effective techniques for adoption and scale up and but also highlight less utilized techniques that may hold promise but require further research. For purposes of this scoping review, we applied the BCT taxonomy-v1 (BCTTv1; Michie & Abraham, 2008;Michie et al., 2013) to identify and map BCTs reported in complementary feeding interventions in LMIC.

| Search strategy
To identify commonly reported BCTs for complementary feeding interventions, we systematically reviewed Pubmed, PsychInfo, and Web of Science databases and hand searched relevant previous reviews. Systematic reviews of complementary feeding interventions have been previously conducted to evaluate intervention effectiveness and were used as launching off points for snowball identification of potentially relevant publications. Two reviews in particular (Graziose et al., 2018;Lamstein et al., 2014)  between January 2016 and December 2017. Searches were conducted in PubMed, Embase, and PsychInfo and included the following search terms: "complementary feeding" OR "child feeding" OR "infant feeding" AND "behavior change" or "behavior change communication" OR "nutrition education" OR "social marketing" OR "social and behavior change" AND "behaviours" or "practices" OR "stunting" OR "growth" OR "underweight" OR "anemia" OR "wasting." A total of 396 titles were returned from this search. The first author reviewed titles for relevance and then further screened abstracts.

| Inclusion/exclusion criteria
Studies and programs were eligible for inclusion if they aimed to shift complementary feeding behaviours (diet diversity, meal frequency, food thickness, consumption of animal source foods, meal volume, consumption of targeted food groups, and responsive feeding) among children 6-24 months using social and behaviour change (SBC) strategies, including direct nutrition education. Nutrition-sensitive strategies (agriculture and women's empowerment) and strategies that provided supplements or special foods/crops were also eligible if they included an explicit nutrition education or SBC component and aimed to improve complementary feeding practices in the target population.
To be eligible, exposure to nutrition education/SBC activities needed to occur at least monthly and for a duration of at least 3 months. Only studies/programs in LMIC published in English between January 2000 and December 2017 were included. Because the review focused on the type of intervention techniques used, we did not exclude studies based on evaluation design, though evaluation design was taken into account when examining BCTS in relation to program effectiveness.
Due to the detailed nature of the information needed about the intervention we excluded conference abstracts for which no full-text article could be located.

| Data abstraction, analysis, and summary measures
Data from all full-text studies and reports that met inclusion/exclusion criteria were abstracted into a standardized form. We abstracted key variables with regard to the study identifiers and context, study design and limitations, intervention details, evaluation strategy, and outcomes evaluated. Additionally, information on the theory of change or program impact pathway, intervention design approach, behavioural theory, and whether the intervention was informed through formative research was abstracted if provided. Program activities were detailed by targeted behaviours, key messages, activity format (Lamstein et al., 2014), activity location, duration, and intensity. Outcomes were categorized as follows: knowledge, attitudes, and beliefs; antecedent outcomes (self-efficacy, motivation, intentions); behavioural/practice outcomes; and health outcomes. Data were abstracted at the programmatic level; if multiple articles described the same program, their information was compiled and reported as a single program.
Multicountry programs were abstracted by country of implementation when sufficient country-specific information was provided (e.g., Alive & Thrive programs in Ethiopia, Bangladesh, and Vietnam were abstracted as separate programs).

| BCT coding and taxonomy mapping
The primary objective of this review was to understand how complementary feeding interventions aim to change behaviour, irrespective of whether they were successful. This approach differs from previous reviews focusing on whether complementary feeding interventions changed behaviour. To achieve this, we applied the taxonomy of BCTs Disagreements in coding were discussed and resolved with the first author. Of note, in coding environmental opportunity-related BCTs (i.e., 12.1 restructuring the physical environment and 12.5 adding objects to the environment), we coded food, supplements, rations, and other items that could be immediately consumed by the child with minimal preparation as 12.5. Additionally, we coded agriculture and kitchen inputs used to improving the home food environment or the ability of the household to prepare optimal foods as both 12.1 and 12.5 because they were objects added to the environment that served to restructure the environment.

| Assessing BCTs in relation to effectiveness
To assess BCTs in relation to effectiveness, we further examined interventions that evaluated outcomes related to complementary feeding practices. We limited this assessment to quasi-experimental or randomized controlled studies and that included at least baseline and endline data. We then divided trials into effective versus noneffective for four WHO-recommended complementary feeding outcomes-diet diversity, minimum acceptable diet, feeding frequency, and thick porridges-and macronutrient and/or micronutrient intake data. We did not include timely introduction of complementary foods at 6 months in estimation of effectiveness ratios because this practice more closely links with exclusive breastfeeding. The environmental, social, and behavioural determinants of delaying the introduction of complementary foods to 6 months maintaining optimal exclusive breastfeeding (EBF) likely differ from those of providing optimal complementary feeding; thus, BCTs for EBF and optimal complementary feeding likely differ. The criteria for assessing effectiveness in improving practices was defined as a significant difference (P < .05) in these behaviours at follow-up between groups. We developed a percentage effectiveness "ratio," which represented the ratio of the number of times each BCT was a component of an intervention in an effective trial divided by the number of times they were a component of all trials, including noneffective trials (Martin et al., 2013).
We acknowledge that a primary limitation of the BCT taxonomy approach to coding and estimation of effectiveness ratios relies on the level of detail authors provide in manuscripts and publically available documents. We used identified manuscripts and reports as resources for additional snowball searching to locate additional details on interventions and programs; however, it was beyond the team's capacity to contact implementers for curricula, implementation details, and intervention activities not made publically available. Given this limitation, it is possible that BCTs are underreported potentially biasing effectiveness ratio estimates. Vietnam (Bruyeron et al., 2010;Dickey et al., 2002;Nguyen et al., 2014); two each in Cambodia (Olney, Talukder, Iannotti, Ruel, & Quinn, 2009;Reinbott et al., 2016) and Nepal (Cunningham et al., 2017;Osei et al., 2017); and one each in Indonesia (S. White et al., 2016), Pakistan (Zaman, Ashraf, & Martines, 2008), and Egypt (Brasington et al., 2016). Target reach ranged from less than 1,000 for smaller scale research and pilot programs to over one million targeted for large-scale, government-implemented programs (i.e.,

| Description of included interventions
Madagascar and India). Feeding frequency was the next most common outcome (n = 34).
Of the five possible platforms for SBC delivery, all but one (Sarrassat et al., 2015) used interpersonal counselling (IPC) strategies whether individual IPC only (n = 16), group IPC without individualized IPC (n = 18), or a combination of both (N = 27). Eighteen reported using media such as radio spots, television, publicly displayed print materials, and radio/video listening groups, all of which were delivered in conjunction with either individual or group IPC or both with the exception of one (Sarrassat et al., 2015) that used media only. Only five reported using policy advocacy/enforcement.

| Commonly reported BCTs
Of 96 (Table 1)   community workshops/skits about benefits of behaviour/ consequences of not practicing recommended behaviours 6.1 Demonstrating the behaviour (n = 35). Often in the form of cooking demonstrations, especially of specific foods for diet diversity, enriched, or thick porridge. Typically conducted as part of mother groups or community events. Sometimes used for demonstration of responsive feeding practices 9.1 Credible source (n = 46). Use of a respected community member to provide information or encourage behaviour change. Most often, this was a trained community health frontline worker; some interventions used physicians, elected community leaders, or religious leaders 12.2 Restructuring the social environment (n = 21). Engagement of influencers (i.e., community leaders, fathers, mother-in-laws/ elder women) to shift gender, social or cultural norms that hinder optimal IYCF; includes women's education/literacy and empowerment activities 12.5 Adding objects to the environment (n = 29). Food/ micronutrient supplements, food rations, or special foods (CSB, food products) or agricultural inputs Note. BCTs included in this table were identified in >20 interventions. Abbreviations: BCT, behaviour change technique; LMIC, low-/middle-income countries.

| Estimating effectiveness ratios
We next sought to assess whether patterns emerged between the type and/or quantity of BCT used and intervention effectiveness. Of the 64 unique interventions identified, six (Alive & Thrive India, n.d.; Fernald et al., 2016;Gelli, Becquey, et al., 2017;Gelli, Margolies, et al., 2017;Kadiyala et al., 2016;Kulwa et al., 2014) were in progress and had not reported evaluation data as of August 2018. Forty-two were randomized controlled or quasi-experimental studies, and one was a propensity score matched impact evaluation; the remaining interventions applied evaluation strategies that consisted of pre/post designs with no control group (n = 15), cross-sectional surveys (n = 1), or were missing baseline data (n = 1) and were excluded from these analyses. Of the 42 with adequate evaluation strategies, 31 had data on at least one of five WHO-defined indicators for optimal complementary feeding indicators (diet diversity, meal frequency, and thick porridge) or macronutrient/micronutrient intakes. Those not reporting on these indicators were excluded from further analysis; they reported on anthropometry (n = 5), responsive feeding (n = 1), offering or consumption of specific foods (n = 3), used summative indices for complementary feeding (n = 2), or did not specify the complementary feeding indicator with sufficient detail to classify (n = 1).
Tables 2 and 3 present the results from these analyses and depict those BCTs applied across interventions by complementary feeding outcome indicator. In general, there was insufficient data to qualitatively surmise associations with specific outcomes as many BCTs had fewer than five applications per outcome. When we collated data across outcomes and examined ratios for the seven BCTs with more than 20 instances of use (Table 1)

| DISCUSSION
Several scoping exercises have explored whether complementary feeding education and promotion can achieve behaviour change (Graziose et al., 2018;Lamstein et al., 2014 When mapping the most commonly used BCTs to intervention functions specified in the Behaviour Change Wheel framework (Michie et al., 2014), complementary feeding interventions in LMIC depend predominantly on two of nine intervention functions, namely, education to increase knowledge/understanding (BCTs 4.1, 5.1, 6.1, and 9.1). Although many interventions specified the use of demonstrations (BCT 5.1), it is likely that these techniques were more educational in nature than skill-building as they did not report providing opportunities for participants to practice learned skills or receive feedback on their progress with learning the skill. Enablement strategies including enhanced social support (BCT 3.1), adding items to the environment (BCT 12.5), or restructuring the social or physical environment (BCT 12.1) were reported; however, they were less frequently reported than education focused approaches. In their mapping of the BCTs used in effective paediatric obesity prevention and treatment, Martin et al. (2013)    Although we recognize the need for context-specific approaches, it is likely there are valuable lessons to be learned from sharing of effective behaviour change strategies between those working to prevent young child overweight and obesity in high-income countries and those working to improve infant feeding in LMIC. Platforms that promote the valuing and sharing of experiences and lessons learned between these two, often siloed groups, are needed to optimize child diets and growth in all contexts.
We should note that in this systematic review, interventions typically compared the impacts of a multicomponent package against receipt of against an existing standard of care, for example, infant and young child feeding counselling provided by health care workers as part of routine clinic visits or nothing at all. Given that many of the interventions had multiple components and used a mix of delivery platforms, it is unclear which intervention components contributed to behaviour change. Lack of clarity regarding the relative effectiveness of specific intervention components precludes identifying the most optimal package in a given context-one that is efficient, costeffective, and scalable. Additional research is needed to test intervention components against one another to identify combinations of components in a given context that produce the most optimized package. One promising approach in the field of behavioural intervention research that enables such testing is the Multiphase Optimization Strategy (Collins, Murphy, Nair, & Strecher, 2005). This approach has been applied, for example, in the fields of HIV (Collins, Kugler, & Gwadz, 2016;Gwadz et al., 2017), smoking cessation Collins et al., 2011), and to develop optimized behavioural interventions. Note. Data are presented as ratio of effective studies using BCT/all studies using BCT (Martin et al., 2013). BCTS with ratios >0.80 are highlighted.
Abbreviation: BCT, behaviour change technique; DDS, diet diversity score; MDD, minimum diet diversity; MAD, minimum adequate diet. A second limitation to our work was the stringent coding strategy we applied, per the protocol outlined by the BCT taxonomy training.

| Limitations
This protocol entailed mapping only those BCTs with sufficient detail to adequately and correctly code. In many cases, intervention details were insufficiently described in peer-reviewed manuscripts and publically available reports. Despite additional snowball searching, we were unable to locate details for approximately half of the interventions included in this review. As such, we likely underreport BCTs, having coded as absent those techniques that intervention designers would argue were present but were not reported. For example, 45 interventions used group education. One could assume that group-based education could be designed and implemented to enable social support for a given set of practices; however, many of these interventions lacked sufficient detail to all.
This lack of reporting detail has emerged as problematic for other health behaviours, including exclusive breastfeeding (Gosselin, 2016). reporting on group behaviour change interventions. However, greater enforcement of these guidelines is needed by both donors/funders of interventions and publishers. As well, given the small number of studies for many of the BCTs, we argue there is a need for rigorously designed and documented behaviour change research in the field of complementary feeding to identify those techniques that contribute to the initiation and maintenance of behaviour change not only over time but also from one child to the next.

| Conclusions
In their review of behaviour change interventions for infant feeding in LMIC, Briscoe and Aboud noted "behaviours may be difficult to change because they are habitual, normative and preventive. Habitual behaviours are difficult to change because they are performed automatically without much thought; normative behaviours bear the weight of tradition and approval; and preventive behaviours often lack a salient immediate outcome" (Briscoe & Aboud, 2012, p. 590). Global reviews of progress on infant feeding practices reflect these challenges (Lutter et al., 2011;J. M. White et al., 2017). As Atkins and Michie note, diet change programs have typically taken an "ISLAGIATT-It seemed like a good idea at the time-approach" to intervention design (Atkins & Michie, 2015). Others have similarly reflected on the lack of theoretical foundations, formative research, theory of change, or systematic intervention design frameworks used in the development of SBC strategies for infant feeding (Graziose et al., 2018;Iannotti et al., 2017;Lutter et al., 2011;Pelto et al., 2016). Given the slow progress to date on improving infant feeding practices, coupled with current global goals to reduce both stunting and child overweight, the lack of reporting for interventions and the use of unsystematic, atheoretical, or nonevidence-based approaches to design complementary feeding behaviour change interventions could be viewed as inefficient at best and unethical at worst. Our findings speak to the urgent need for greater public sharing of intervention details among the global IYCF community-inclusive of both high and LMIC-especially related to design, underlying theory of change, implementation, and lessons learned. Such public sharing is critical for the identification, verification, replication, and scale up of effective change strategies.

This study was funded by Bill & Melinda Gates Foundation and the
Tata-Cornell TARINA program.

CONFLICTS OF INTEREST
The authors declare they have no conflicts of interest.

AWG and
LG conceptualized the research question and analytical approach. EW and SS abstracted all data and mapped behaviour change techniques. AWG wrote the first and subsequent drafts of the article. All authors contributed to critically revising the article and gave final approval of the version to be published.