How do breastfeeding workplace interventions work?: a realist review

Background Women are representing an increasing share of the labor force, thus, raising the need to accommodate breastfeeding working mothers at the workplace. While there is an emerging body of evidence supporting the positive influence of workplace lactation programs on breastfeeding outcomes, there is a lack of literature on the mechanisms underlying those interventions. Aims of this realist review were three-fold: to uncover underlying mechanisms, determine who benefits the most from such interventions and important contextual factors influencing uptake. Methods Purposive bibliographic searches on Medline, Web of Science Core Collection, CINAHL, Global Health, LILACS, Global Index Medicus, Business Source Complete, Proquest Dissertations and Theses and Open Access Theses and Dissertations were conducted to identify relevant publications. Included publications (qualitative and quantitative) described interventions aiming to improve the breastfeeding behavior of working mothers, that were initiated by the employer, reported on breastfeeding outcomes and had a clearly defined workplace. Publications only focusing on maternity leave or that were not published in English, Spanish, Portuguese or German were excluded. A realist approach was followed to identify how workplace interventions work, who benefits the most and the important contextual factors. Results The bibliographic search yielded a total of 4985 possible publications of which 37 publications were included in the realist analysis. Effective workplace breastfeeding interventions activate three mechanisms: 1) awareness of the intervention, 2) changes in workplace culture, manager/supervisor support, co-worker support and physical environments, and 3) provision of time. Contextual factors such as the distance between the workplace and the infant and the type of workplace may influence the degree of activation of the underlying mechanisms for programs to positively impact breastfeeding outcomes. Conclusions In order to be effective, workplace breastfeeding interventions need to: raise awareness of the intervention(s) available among working mothers as well as their work environment, change the workplace culture, foster manager/supervisor support and co-workers support, provide enough time and adequate space and facilities for women to breastfeed or express breastmilk during the workday. Supplementary Information The online version contains supplementary material available at 10.1186/s12939-021-01490-7.


Background
In the last decades, the number of women entering the labor force has steadily increased and women are representing a larger share of the labor market than ever before [1]. Globally, women's labor force participation rate is 48.5%, which demonstrates their growing and significant contribution to their various national economies [2]. Because of women's growing presence in the labor market, there is an increasing number of employers seeking to accommodate the needs of working women who choose to have children and want to breastfeed. Indeed, employers already recognizing the importance of breastfeeding, have offered various levels of lactation support especially for women working in the formal but not the informal sector [3,4].
It is well established that breastfeeding is associated with numerous short-and long-term health benefits for the breastfeeding mother and the breastfed child. As such, breastfed children have lower risk for morbidity and mortality from infectious diseases, increased intelligence scores, and a reduction in risk for overweight and perhaps diabetes in later life [5][6][7]. For mothers, breastfeeding is associated with lower risk for breast cancer, ovarian cancer and type 2 diabetes [8].
Despite the various benefits of breastfeeding and the WHO/UNICEF recommendation for exclusive breastfeeding for 6 months, globally only about 44% of infants < 6 months of age were exclusively breastfed in 2019 [9]. This is still below the goal of at least 50% by 2025 defined in the Global Nutrition Targets 2025 by UNICEF and WHO [10]. Maternal employment is often cited as a major barrier to breastfeeding [11] and returning to work is associated with early cessation of breastfeeding [12,13].
Previous research has shown a positive association between workplace lactation support and interventions with higher breastfeeding rates and duration of breastfeeding [14]. Relatively low-cost interventions such as lactation rooms and nursing breaks may reduce absenteeism and improve workplace performance, commitment and retention, while also improving breastfeeding outcomes [11,15]. Also, mothers receiving other types of workplace support such as provision of electric breast pumps, access to lactation professionals, and refrigerators for storing their breastmilk in the workplace were more likely to initiate and continue breastfeeding after returning to work [14,16]. Availability of employer-sponsored childcare and flexible schedules can also increase an employee's likelihood of success with breastfeeding [17]. Therefore, more and more countries and organizations are introducing measures to support working mothers in reaching their breastfeeding goals [18], thus, enabling mothers to better combine their work requirements and their infant feeding goals.
While there is an emerging body of evidence supporting the positive influence of workplace lactation programs on breastfeeding outcomes [14][15][16]19], there is a lack of literature on the mechanisms underlying those interventions. An understanding of these mechanisms is crucial in learning how to operationalize, implement and disseminate robust and effective lactation programs. Systematic reviews focus on the outcome and the level of outcome, thus on statistical inferences, but do not investigate the underlying mechanisms or context of the intervention leading to those outcomes. However, in order to find how interventions work in different contexts, statistical inferences are insufficient. Disentangling the underlying mechanisms linking intervention, context and outcome are needed to support the policymaking and implementation processes under 'real world' conditions. Thus, in addition to probability considerations, plausibility and adequacy considerations [20] need to be taken into account in order to understand how workplacebased breastfeeding interventions actually work.
Given the need for an approach that accounts for context, the objective of this study was to follow a realist approach to better understand how worksite related breastfeeding programs work across different contexts. We followed the realist review approach as it does not focus on making statements about the strength of quantitative vs. qualitative study designs but rather it integrates and values the different perspectives offered by them and enable the researcher to unravel and understand underlying pathways. Furthermore, workplace breastfeeding interventions meet the seven criteria of complex service interventions that can best be examined through the lens of a realist review [21] (Table 1).
In order to inform policy makers and employers about workplace breastfeeding interventions, the specific questions we aimed to answer though this realist review were: 1) How do breastfeeding interventions at the workplace work?, 2) Who benefits the most from such interventions across different contexts?, and 3) What are important contextual factors which determine whether different mechanisms produce their intended breastfeeding outcomes? databases without subject indexing, we achieved high specificity by searching for the concepts of breastfeeding and workplace in titles and author-provided keywords. The Medline search is provided in Table 3; the remaining searches are provided in Additional file 1.
The results from the search of all bibliographic databases were deduplicated in EndNote X9 and imported to Covidence systematic review software (Veritas Health Innovation, Melbourne, Australia. Available at www. covidence.org). Both software packages were used in the versions licensed at Yale University.
In order to be included, the publications must have described interventions that aimed to or could be expected to improve the breastfeeding behavior of working mothers and that were initiated by the employer, its representative, an employer-like persona or the work supervisor of the parent(s). Furthermore, the study must have reported on breastfeeding outcomes (quantitative or qualitative) and the workplace must have been clearly defined (either real or virtual space). Any data that described breastfeeding behavior were considered as breastfeeding outcomes. Therefore, studies reporting on Table 1 The 7 Criteria of Complex Service Interventions and Their Evaluation Through Realist Reviews (after [21]) Criteria of complex service intervention and their evaluation through realist reviews Application to workplace breastfeeding programs 1) Public health interventions hypothesize that after implementing an intervention, the condition will be improved. The review needs to closely follow those underlying theories.
Workplace breastfeeding programs are expected to improve breastfeeding rates among participating working mothers.
2) Public health interventions only accomplish their goals by active input from individuals. It is therefore essential to not only rely on controlled study designs but to also review the actions of involved stakeholders to be able to explain the success or failure of the intervention.
In order to have a successful workplace breastfeeding intervention, active input from several individuals is needed e.g. communicate the availability of a breastfeeding/lactation room, active support from co-worker to take over responsibility while the breastfeeding women is on her pumping break, etc.
3) Public health interventions involve a long and complex pathway which includes the operational changes needed to implement the intervention and the uptake and adherence to the program by the participants. Thus, the review needs to consider the entire implementation chain, determine the needed intermediate outputs for a successful final outcome and define processes and blockage points by including a variety of publication types.
In order to understand workplace breastfeeding interventions, the review needs to analyze who initiated the program, how it was developed, how it was implemented and who actively participated in the implementation and how as well as who adheres to the offered breastfeeding intervention. 4) Public health interventions are non-linear. The review therefore needs to determine and account for the different influences of the different parties by not only including controlled trials.
Within the implementation chain of a workplace lactation program, different stakeholders influence each other. Depending on their power, the workplace breastfeeding program differs e.g. workplace lactation programs might be more extensive if employee have strong representatives compared to situations in which the employers strictly follow governmental regulations. 5) Public health interventions are embedded in multiple social systems. Such differences are best uncovered by considering finding from qualitative study designs.
The wish for privacy in a lactation room in a woman dominated environment may be different then in a male dominated work environment. 6) Public health interventions differ depending on the context and the understanding of the implementing stakeholder. The review needs to help to understand what similar terms mean to different stakeholders by disentangling underlying implementation mechanisms and contexts behind interventions labelled the same way but in reality being different from each other.
A workplace lactation room can range from simply referring to an empty room vs. a fully equipped lactation room with an accompanying policy regulating breaktime and flexible working hours.

7) Public health interventions are open systems with feedback loops.
Hence changes in overall environment as a result of interventions need to be considered across different contexts.
By successfully implementing a workplace lactation policy, the overall work-based environment may change and as a consequence, the policy itself strengthens. quantitative breastfeeding outcomes (e.g. exclusive breastfeeding rate, exclusive breastfeeding duration, breastfeeding cessation, duration of any breastfeeding) as well as on qualitative breastfeeding outcomes were included. No date limit for publications nor limitation on types of data collected in the study (qualitative vs. quantitative data) were applied as inclusion or exclusion criteria.
Publications that only focused on maternity leave and that were not published in English, Spanish, Portuguese or German were excluded. Publications only focusing on maternity leave interventions were excluded because the review's focus is workplace breastfeeding interventions that can be fully influenced by the employer. Maternity leave interventions are often regulated, at least to some extend by governments, and are thus, not under the sole control of the employer. Publications that were published in languages other than English, Spanish, Portuguese or German are listed in a separate supplementary table (Additional file 2). Literature reviews were excluded from the analysis.
The inclusion criteria as well as the maternity leave exclusion criteria were applied to the title-abstract screening stage. Articles initially selected for full-text screening were screened in detail for all inclusion and exclusion criteria. Exclusions at the full-text screening stage were grouped into following reasons: 1) intervention was not initiated by the employer, 2) breastfeeding outcomes not reported, 3) article focused on maternity leave, 4) article not published in English, Spanish, Portuguese or German, and 5) other reasons. All screening rounds were conducted by two reviewers following a consensus approach (KL and VT). Discrepancies were resolved through discussion and consultation with a third reviewer (RPE).
Non-peer-reviewed documents (e.g., conference papers, press releases, etc.) were not included. When such documents were retrieved in the database searches, the document's authors were contacted for more information when deemed necessary. Dissertations were handled as peer-reviewed articles and authors were not contacted.
Goal of the data extraction was to find contextmechanism-outcome (CMO) patterns that can potentially explain the relationship between the intervention and the outcome by understanding the underlying mechanism in the context in which the intervention takes place. Focusing the data extraction on contexts, mechanisms and outcomes is important to identify reasons behind program successes or failures instead of just identifying successful and unsuccessful interventions [21]. Data extraction was conducted by the two reviewers. The data points extracted from peer-reviewed articles were: year of publication, type of study/publication, country, implementation (type, components/activities and approach), context (full-/part-time employment, type of work position, length of maternity leave, etc.), sample size, implementation period, intervention population and implementation outcomes (acceptability, adaptation, appropriateness, cost, feasibility, fidelity, penetration and sustainability). Additionally, breastfeeding outcomes were extracted to better understand the influence of the intervention on breastfeeding outcomes.
For the development of a CMO framework, we defined the intervention types, intervention category, context categories as well as the breastfeeding outcome for each included study. The CMO framework and findings were summarized and discussed in a narrative synthesis.

Description of articles found through bibliographic search
The bibliographic database search yielded a total of 4879 possible documents. After removing a total of 1522 duplicates, 3357 citations were screened at the titleabstract phase in Covidence. From these 3357 citations, 156 were eligible for full-text screening. Separately, the search on Open Access Theses and Dissertations (OATD) resulted in a total of 106 possible documents. Because of technical issues with the exportation of search results, the 106 documents from OATD were handled outside of Covidence during the title-abstract screening, thus, only one reviewer screened the OATD search results. A total of 10 OATD documents were found to be eligible for the full text screening. After removing 6 documents that had already been screened in Covidence as a result of the other searches, 4 OATD documents were imported into Covidence for the fulltext screening. This led to a total of 160 articles that were eligible for the full-text screening which was conducted by two reviewers. At the full-text screening phase a total of 123 articles were excluded, leaving 37 articles for analysis. Articles were excluded for the following reasons: intervention was not initiated by the employer (32 articles), no report of breastfeeding outcomes (49 articles), not published in English, Spanish, Portuguese or German (4 articles) or for other reasons (38 articles). Subcategories under "other reasons" included articles which could not be delivered via interlibrary loan (10 articles), literature reviews (7 articles) and duplications discovered during full text screening (2 articles). Only one reason for exclusion could be recorded in Covidence. An overview of the screening process is depicted in Fig. 1. The 37 articles that were included in the realist analysis came from 11 countries. The majority of publications were from studies conducted in the United States of America (19 articles [22][23][24][25][26][27][28][29][30][31][32][33][34][35][36][37][38][39][40];). The other publications came from studies in Taiwan (5 articles [41][42][43][44][45];), Brazil (4 articles [46][47][48][49];), Thailand (2 articles [50,51];), Australia [52], China [53], Ethiopia [54], Indonesia [55], Mexico [56], Spain [57] and Turkey [58] (each 1 article). The 37 articles resulted from 33 single studies published over a period of 35 years; from 1985 to 2020. An overview of the included studies is given in Tables 4, 5, 6, 7, 8 and 9.
The intervention sample in all but one study, was conformed by employed women. Sahip and Turan [58] described the effects of a workplace health program for expectant fathers in Turkey. The intervention was delivered by specially trained workplace physicians and consisted of six 3-4 h sections: 1) health during pregnancy, 2a) pregnancy nutrition, 2b) birth, 3) communication techniques, 4) infant health care and feeding, 5) fatherhood, and 6) family health after birth. Children of fathers

Realist analysis
To better understand how workplace breastfeeding interventions work and how their influence on breastfeeding outcomes differs in different contexts, we categorized the interventions described in the analyzed articles into 4 types of intervention and 15 intervention categories (Table 10). To develop a CMO framework (Fig. 2), we analyzed the outcomes per intervention category. We concentrated our analysis on the outcomes on changed breastfeeding behavior as well as changes in perceived workplace breastfeeding culture (e.g., how common breastfeeding at the workplace is), manager/ supervisor support, co-worker support and in the physical environment. It is to mention, that workplace breastfeeding outcomes can also lead to additional outcomes that were out of scope of this review such as job satisfaction [22,31]. Our analysis revealed three potential mechanisms or factors that might explain how workplace interventions influence breastfeeding habits at the workplace: 1) having more time to breastfeed during work, 2) awareness of the intervention among mothers, supervisors and coworkers and 3) change in culture, management/supervisor support, co-worker support and/or physical environment. Having more time to breastfeed can be directly achieved for example by providing lactation breaks or flexible working hours or indirectly for example by changing the physical environment (e.g., providing lactation rooms [43,54]) or co-worker support (e.g., by raising the understanding of co-worker through a communication strategy such that the co-worker support flexible work schedule). The mechanism of awareness about the intervention acts directly on the mother as well as indirectly through raising awareness of others; e.g. supervisors and co-workers who can then direct the mother to the lactation program. As long as the mother is not aware of the intervention, she cannot use it, and thus, her breastfeeding habit at the workplace will not change due to the intervention [25,43,44]. Supervisor awareness was also important in order to be able to advise their pregnant employees of corporate lactation programs. Therefore, changes in culture, management/supervisor support, coworker support and/or physical environment are mechanisms and associated with changes in breastfeeding habits of employed mothers [23,31,33,35,36,38,41,42].
The analysis of outcomes based on their intervention category also led to the identification of important contextual factors. The context in which the intervention is implemented determines whether it is sufficient to affect breastfeeding outcomes. We identified the following contexts: breastfeeding−/family-friendly business designations [32,41], distance between workplace and infant [50,53,54], flexibility of work schedule/workload [38], marital status [24], maternal education level [24,31,57], maternal health conditions [38], number of offered lactation services [26,28,34], parity [57], race/ethnicity [24,26], shift work [29,40,[42][43][44], time of registration to program [26,34], type of employment (full-time vs. parttime) [26], type of salary (fixed vs. hourly) [24,39], type of workplace [24,[42][43][44]. As an example of how different work-related contexts require different implementation mechanisms, or, lead to different outcomes for similar interventions, we will concentrate on the context of distance between workplace and infant as well as the contexts of flexibility of work schedule/workload, shift work and type of workplace.
The sole implementation of a workplace breastfeeding intervention was not always sufficient to cause a change in breastfeeding habits at the workplace. We found that a large distance between the workplace and the location of the infant as well as commute times are a major barrier for continued breastfeeding after returning to work. Mothers whose children are living far apart from them described the large distance between them and their infants as a reason to stop breastfeeding after return to work; despite the presence of lactation break policies [50,53], lactation facilities [50] or even the offer of a drop-off service that brings expressed breastmilk to the local bus and van station from where the breastmilk is transported to rural areas, where the infant is living [50]. Underlying reasons seem to be different: Mothers in Thailand report that they fear that their breastmilk will be spoiled during the long transportation and that grandmothers, who are caring for the infant, lack the knowledge how to handle frozen/expressed breastmilk [50], in China the legal provision of one-hour feeding break for working mothers of infants aged younger than 1 year is insufficient for mothers whose children are not near the workplace since the provision of lactation facilities is not legally required, thus, requiring mothers to travel to their infant to feed them [53]. While the example in Thailand shows that mechanisms outside the work environment are influential, the example in China      Outcomes Differences in use of breast-pumping breaks according to age (p = 0.0459), maternal education (p < 0.0001), husband's education (p = 0.0002), worksite (p < 0.0001; clean room uses less often, office uses more often), shift work (p < 0.0001; shift workers use less often), NS difference according to work hours per day (p = 0.5164) Positive association between the use of lactation breaks after returning to work and awareness about lactation rooms (p = 0.0173), breast-pumping breaks policy (p < 0.0001), support from supervisor (p < 0.0001) and coworkers (p < 0.0001), encouragement to use lactation room from environmental health nurses (p < 0.0001) and provision of lactation consultant by employer (p = 0.0074) Negative association between the use of lactation breaks after returning to work and feeling of embarrassment (p = 0.0046), perception of inefficiency (p < 0.0001) and believing that lactation breaks would affect supervisor's assessment of performance (p = 0.0079) OR for intention to use breast-pumping break after returning to work:      • 58% of mothers who pumped in a bathroom did not BF for as long as intended • 29% of mothers who did not pump in a bathroom BF for as long as intended • p < 0.001 Significant association between BF intention and perceptions of workplace lactation support with BF duration, p < 0.001 Significant association between perception of workplace lactation support BF duration for ≥6 months, p < 0.001 Significant association between utilization of workplace lactation support and BF duration for ≥6 months, p < 0.001   shows that the sole provision of time can be insufficient if the physical environment does not support breastfeeding.
Other contexts in which the sole implementation of breastfeeding interventions at the workplace are insufficient, are workplaces with a busy and/or inflexible work schedule, workplaces other than an office, and type of work such as shift work. Professions such as nurses or physicians have a work schedule that is often influenced by external factors which contributes to inconsistent scheduling. In such a context, the sole provisions of lactation breaks and lactation facilities are inefficient if mothers perceive their co-workers and supervisors as not supportive of them taking breastfeeding breaks [38].
Inflexible environments (e.g., fabric/production workplaces) and workplaces without a specific office as well as shift work are also contexts in which the sole provision of lactation breaks and lactation facilities are insufficient. Mothers working in a clean room (a room that is maintained free of contaminants) in two Taiwanese manufacturing companies had lower breastfeeding rates than their female colleagues having an office space in the same company [42][43][44]. Tsai and colleagues found that office workers and non-shift workers used the lactation breaks and facilities available to them more often than workers of the clean room and shift workers. To be able to use the offered lactation breaks, clean room workers needed to fully change from their clean room suits into normal clothes and then dress back into their clean room suits. This may have left less time to pump, thus, making the available lactation break times too short and less effective for breastfeeding. Chen et al. found that among women who were aware of lactation breaks and facilities, office workers had showed a higher likelihood for continued breastfeeding after returning to work than clean room workers indicating that in this context, the awareness mechanism was not mediating the breastfeeding outcomes. Given the rigid work schedule of clean room and shift workers, it is more likely that among them the intervention of breastfeeding breaks and facilities may not increase the time available for breastmilk expression.
Performance bonuses for manufacturing workers may inhibit the activation of support from co-workers, and thus, the activation of the time-releasing mechanism needed for breastfeeding and expressing breastmilk. Performance bonuses for most Taiwanese manufacturer workers are based on group productivity and individual performance [44], thus, breastfeeding mothers working in the clean room dependent on their co-workers to take over their duties while they are breastfeeding in order to keep their own performance bonuses and the ones of their co-workers, which may be unrealistic. Therefore, in order to increase breastfeeding rates among manufacturing and shift workers, workplace lactation interventions need to involve the network of co-workers to indirectly enable breastfeeding workers to use lactation breaks and facilities.

Discussion
To our knowledge, this is the first realist review identifying potential mechanisms underlying the impact of workplace breastfeeding interventions as well as the effect of different contexts on the influence on breastfeeding outcomes of such interventions. Using a realist review approach, we were able to integrate findings from a plethora of study designs including qualitative studies.
An innovative aspect from our review is that we were able to develop a pragmatic context-mechanismoutcome (CMO) framework. This framework shows that contextual factors such as long distances between the workplace and the infant will hinder the influence of the intervention at improving breastfeeding outcomes among working mothers. Specifically, our CMO framework identified three mechanisms at work that need to be activated for an intervention to be effective: 1) the awareness of workers, supervisors and co-workers about the availability of entitlement to a given intervention, 2) changes in: perceived breastfeeding culture at the workplace, including manager/supervisor and co-workers support and adequate physical environments, and 3) having time to breastfeed or express breastmilk during work time.
Positive associations between workplace lactation support and interventions have been shown previously [11,[14][15][16][17]19]. However, systematically understanding how such interventions work has been a major gap. Furthermore, examining how contexts mediate or moderate the influence of workplace breastfeeding interventions across different workforce groups had not been previously done, as far as we know. Our review is impactful because it provides this information at a time when an increasing number of women participate in the labor force [1,2]. Our review indeed identified mechanisms that if properly taken into account when designing interventions may empower mothers to not have to decide between choosing the best nutrition for their infant or working, thus, helping close a major inequity gap affecting working women with infants globally.
While it is important to lessen breastfeeding inequalities between working and non-working mothers, it is as important to lessen these inequalities among working women. In the course of our analysis, we found that the studies represented a wide spectrum of maternal demographics including age, education level, race and ethnicity, income level, and marital status. Overall, higher breastfeeding rates and longer breastfeeding duration in the workplace were associated with higher maternal education [24, 25, 28, 31, 42-44, 54, 55], higher income levels [25,28], being White [24][25][26] and being married or living with a partner [24,28]. The included studies did not allow us to examine the underlying pathways that may have explained differences in implementation approaches as a result of differences in socioeconomic and demographic contexts because awareness and/or uptake of the interventions were seldomly examined as a function of the afore mention characteristics. Therefore, future research is needed to elucidate how best to tailor work-based breastfeeding interventions to different contexts.
The available data did also not allow us to determine underlying pathways which may explain differences in breastfeeding outcomes of work-based intervention as a function of type of employment (full-time vs. part-time employment). This is because the study authors did not examine awareness and/or uptake of the intervention as a function of the employment status which would be needed to be able to determine differences in awareness and/or uptake of the intervention among women with different employment status as pathway for the differences seen in breastfeeding outcomes among women with different employment status. One study showed a significant association between part-time employment and higher breastfeeding rates at 6 months while two other studies did not find significant associations [34,39]. Possible explanations of these inconsistent findings are the heterogeneous use of the term "part-time" and • Maternal comfort to ask for lactation accommodation, p = 0.00 • Maternal perception of supportive co-worker, p = 0.00 • Maternal comfort to take lactation breaks, p = 0.00 • Maternal comfort to adjust break schedule to meet pumping needs, p = 0.00 Negative association for BF duration: • Maternal perception that number of hours worked made it difficult to combing BF and working, p = 0.00 • Maternal perception of insecure job, p = 0.04 NS association for BF duration: Maternal perception of supportive manager, p = 0.75 BF breastfeeding, BM breastmilk, EBF exclusive breastfeeding, N/A not applicable, NS not significant, RTW return to work the heterogenous group of part-time working mothers. For example, it is clear that the needs for breastfeeding support of a mother working 8 h a week are likely to be different from the needs of a mother working 40 h a week. However, it is unclear how the needs of a mother working 40 h a week would compare with the needs of a mother working 34 h a week, either as a four-day working week or distributed across all weekdays. Moving forward, instead of using unclear terms like "part-time" employment, future research should consider evaluating workplace breastfeeding interventions based on actual hours worked per day and days worked per week.
While this realist review integrated evidence from a plethora of study designs allowing for the examination of how workplace breastfeeding interventions work across various contexts and uncovering potential pathways for impact, it is not without limitations. Firstly, the review did not include studies that were solely focusing on maternity leave benefits. Rather, it focused on lactation interventions for mothers returning to the workplace. We made this decision because previous work has documented the positive impact of extended duration of maternity leave on breastfeeding outcomes [59,60] and maternity leave policies are beyond the sole domain of the employers. Secondly, in order to bring focus to the review, we limited the search to studies reporting on breastfeeding outcomes. This limitation omitted publications reporting on other outcomes of workplace breastfeeding interventions such as job satisfaction or health costs for employers, employees and/or society. And lastly, as recommended for realist reviews, we used a purposive rather than a comprehensive screening strategy. Thus, it is possible that a relevant paper could have been available in the databases we searched, but was not retrieved by our queries, as we screened only those papers where the workplace context was explicit in either the title and author keywords or subject headings. To mitigate this risk, we searched in multiple subjectindexed databases, on the reasoning that a paper which was poorly indexed in Medline may have been better indexed in CINAHL or Global Health. Because of time limitations, we did not conduct grey literature searches as well as citation chaining. It is possible that these omissions led to the introduction of biases. Nevertheless, we think that this is unlikely because the vast majority of grey literature about workplace breastfeeding interventions are technical guidelines on how to implement specific interventions such as lactation rooms [61], and do not report on breastfeeding outcomes, thus would not have passed the eligibility criteria for inclusion of the present review. Since we searched a plethora of databases, we are confident that our search picked up the vast majority of eligible publication, thus, we estimate the risk of bias introduction due to the missing citation chaining as minimal.
Our review strongly calls for more mixed methods work-based breastfeeding intervention research in low-and middle-income countries, that also includes the very large number of women working in the informal economy. Of the 37 included studies, only one study was conducted in a low-income country [54] as defined by the World Bank [62]. This is unfortunate, as out of the approximately 7.7 billion people in the world in 2019, 6.5 billion people lived in low-and middle-income countries, and 670 millions lived in low-income countries [63]. None of the included studies of this review focused solely on informal employment and the majority included only formally employed women. While formally employed mothers can be protected by laws and regulations, such as mandated maternity leave, informally employed women may need to depend on other mechanisms that support their informed decisions about infant feeding [53]. Therefore, there is a profound inequity in the selection of settings where the work-based breastfeeding research has been conducted, as well as the type of employment included in those studies (formal vs. informal economy). This is unacceptable given the very high proportion of women employed in the informal sector in low-and middle-income countries [64]. How the policy design and program implementation mechanisms need to differ for delivering effective work-related breastfeeding interventions targeting women employed in the informal vs. the formal economy, still need to be elucidated using qualitative and quantitative implementation research approaches.

Conclusion
Workplace breastfeeding interventions work through raising awareness among employees, supervisors and coworkers, changes in workplace breastfeeding culture, including knowledge, attitudes, and support from managers/supervisors and co-workers, and improvements in the physical environment, alongside with the time release needed by working mothers while at work for breastfeeding or extracting breastmilk. In order to better address breastfeeding inequities affecting working mothers, workplace breastfeeding interventions need to be tailored according to several contextual factors including socioeconomic and demographic characteristics of the mothers or end users. The evidence of this review clearly shows that workplace breastfeeding interventions cannot follow a one-size-fits-all approach, but rather should be tailored for the contextual factors underlying the different working conditions for mothers globally.
Additional file 1. Search strategies of all searched databases.
Additional file 2. Publications not published in English, Spanish, Portuguese, or German.