Effectiveness of nutrition training of health workers toward improving caregivers’ feeding practices for children aged six months to two years: a systematic review

Background Nutrition training of health workers can help to reduce child undernutrition. Specifically, trained health workers might contribute to this end through frequent nutrition counseling of caregivers. This may improve child-feeding practices and thus reduce the risk of undernutrition among children of counseled caregivers. Although studies have shown varied impacts of health workers’ nutrition training on child feeding practices, no systematic review of the effectiveness of such intervention has yet been reported. Therefore, we conducted this study to examine the effectiveness of nutrition training for health workers on child feeding practices including feeding frequency, energy intake, and dietary diversity among children aged six months to two years. Methods We searched the literature for published randomized controlled trials (RCTs) and cluster RCTs using medical databases including PubMed/MEDLINE, CINAHL, EMBASE, and ISI Web of Knowledge, and through WHO regional databases. Our intervention of interest was nutrition training of health workers. We pooled the results of the selected trials, evaluated them using the Grades of Recommendation, Assessment, Development, and Evaluation (GRADE) criteria, and calculated the overall effect size of the intervention in meta-analyses. Results Ten RCTs and cluster RCTs out of 4757 retrieved articles were eligible for final analyses. Overall, health workers’ nutrition training improved daily energy intake of children between six months and two years of age. The pooled evidence from the three studies reporting mean energy intake per day revealed a standardized mean difference (SMD) of 0.76, 95% CI (0.63-0.88). For the two studies with median energy intake SMD was 1.06 (95% CI 0.87-1.24). Health workers’ nutrition training also improved feeding frequency among children aged six months to two years. The pooled evidence from the three studies reporting mean feeding frequency showed an SMD of 0.48 (95% CI 0.38-0.58). Regarding dietary diversity, children in intervention groups were more likely to consume more diverse diets compared to their counterparts. Conclusion and recommendations Nutrition training for health workers can improve feeding frequency, energy intake, and dietary diversity of children aged six months to two years. Scaling up of nutrition training for health workers presents a potential entry point to improve nutrition status among children.


Background
Undernutrition is responsible for more than one-third of child deaths globally [1], and it is more prevalent in lowand lower-middle-income countries [2]. Poverty has remained an important underlying cause of poor nutrition status among children in these regions [3,4]. It has been cited as a factor behind food insecurity [5,6], low maternal education, poor access to healthcare [7], and burden of diseases [8], each of which mediates poor child nutrition status. A disadvantaged socioeconomic position may also feed into caregivers' poor feeding practices with regard to their children [9,10]. Poor feeding practices include low dietary diversity, feeding frequency, and energy intake [11][12][13].
Poor child feeding practices are caused by a myriad of factors. They are associated with cultural factors that may create local tendencies toward selection of lowquality complementary foods [14,15]; taboos [15]; and restrictive traditional beliefs [16]. Social factors including caregivers' poor knowledge on nutrition and lack of knowledge on food diversity in their environment may correlate with poor feeding practices [17]. Such factors may result in low dietary diversity, low feeding frequency, and low food and energy intake for children.
Caregivers' nutrition education can help to clear cultural and tradition-based misconceptions and improve their general nutrition knowledge [18]. Feeding practices can thus be improved if knowledgeable health workers treat and counsel them on proper feeding practices and monitor their progress closely [19]. Nutritionists and dieticians can, of course, perform such counseling when they are available [20][21][22]. However, health workers equipped with such specialized skills may not be in sufficient supply for routine care in many developing countries [23,24], leaving health workers who have only general nutrition knowledge to provide such care. Medical doctors, nurses, midwives, and midlevel providers are not always trained to perform such tasks and may not have adequate or practical knowledge to counsel and treat undernutrition [25]. Nutrition training for these cadres can help to bridge such knowledge gaps.
Available evidence supports that nutrition training of health workers can improve feeding practices and thus child undernutrition. Previous randomized controlled trials (RCTs) have found higher levels of nutrition knowledge and counseling behavior among health workers who received nutrition training [19,26,27]. Nutrition knowledge among caregivers improved in turn when they were frequently counseled by health workers who received nutrition training [19]. Nutrition counseling also improved caregivers' knowledge in food preparation [28,29] and healthy feeding behaviors [30]. As a result, caregivers were more likely to improve their children's feeding frequency [31], dietary diversity [32], protein, and energy intake [28,33]. Such elements of feeding practice are essential to improving children's nutrition status [34,35].
Opportunities for routine nutrition counseling can be identified in existing health care frameworks. Globally, about 88% of pregnant mothers had at least one antenatal visit while 50% had a postnatal visit to a health care facility in 2012 [36]. These represent potential opportunities for nutrition counseling by trained health workers. Only 57% of mothers had a skilled attendant at their children's birth [36]. However, for such mothers, there is also room for home-based nutrition counseling using community health workers or peers with relevant skills and proper nutrition training [27].
Nutrition training and counseling has also been successfully applied using specially selected and trained peer groups working hand-in-hand with trained health workers. For example, in a study conducted in Senegal, trained community health workers trained grandmothers, who in turn provided maternal and child health education on topics including child feeding to mothers and pregnant women [37]. Community and home-based interventions conducted by trained peers in collaboration with trained health workers have had success in areas with limited resources [27,38,39]. Pictures, posters, school rallies, street-side plays, and nutrition fairs have been commonly used to provide such forms of training or counseling. These illustrations can be ever more effective when tailored to the regional context [20].
Nutrition training for health workers has yielded varied impacts on child health, particularly regarding nutrition status. Previous systematic reviews have evaluated the impacts of nutrition counseling, maternal nutrition education [34], and complementary feeding on children's nutrition status [18,40]. However, no systematic review has been conducted to evaluate the effectiveness of nutrition training for health workers on child feeding practices, an important step to improving their nutrition status. Therefore, we conducted this systematic review to examine the effect of nutrition training for health workers on feeding frequency, energy intake, and dietary diversity of children between six months and two years of age. Our PICO (Population, Intervention, Comparator, Outcome) question was as follows: "What is the effect of nutrition training for health workers to improve caregivers' feeding practices including energy intake, feeding frequency, and dietary diversity of children under two years of age as compared to those who did not receive such training?"

Methods
In this systematic review, we included RCTs and cluster RCTs that incorporated nutrition training for health workers as an intervention of interest. We defined nutrition training for health workers as any formal nutrition course provided to health workers in the form of inservice training, continuing professional education, short courses, or seminars, aimed, for practical or research purposes, at improving the nutrition knowledge or practices of health workers. We defined our population as health workers including doctors; nurses and nurse midwives; midlevel providers including assistant medical officers, clinical officers, assistant nurses, or assistant physicians; community health workers including village health workers; and nutritionists or dietitians working in the areas where the studies were conducted. We focused our attention on three outcome variables: feeding frequency measured in the number of times the child was fed in the previous 24 hours; energy intake in kilojoules (kJ) per day; and dietary diversity, defined as the variety of food items that was fed to a child.
We developed a review protocol based on the aforementioned guidelines that was shared among the researchers before conducting the literature search. This protocol provided guidance regarding the literature search, exclusion and inclusion criteria, methods of analysis, and grading of evidence. Based on this framework, three researchers independently conducted literature searches, selection of studies, and data extraction.

Data sources for existing review
Three researchers independently searched the medical databases based on the review protocol. First, we searched for any existing review or submitted protocol on our topic listed in the Cochrane Library or the Cochrane Database of Systematic Reviews (CDSR). We also searched for similar review articles in the Abstracts of Reviews of Effects (DARE), National Institute for Health and Clinical Excellence (NICE), Educational Resources and Information Center (ERIC), and Campbell library of systematic reviews databases.

Inclusion and exclusion criteria
We included RCTs and cluster RCTs that included an explanation of nutrition training provided to health workers and child feeding practices such as dietary diversity, feeding frequency, and energy intake. We also included studies that provided information and counseling on feeding practices to caregivers of children aged six months to two years.
We excluded studies that included children aged below six months and above two years. The age range of 6 months to 2 years was chosen as the focus for this study because of its importance as the transitional period from exclusive breastfeeding to complementary feeding and family foods. Studies have shown a rapid increase in rates of undernutrition during this stage [41].
Such increases in developing countries may largely be associated with inappropriate feeding practices. Also, due to wide variations in the daily required nutritional and energy intake with age, it is difficult to generalize feeding practices for children below and above two years of age. We also excluded results covering feeding practices among children on exclusive breast-feeding regimens; on special feeding interventions including parenteral feeding and tube feeding; or subject to special feeding requirements or diagnosed with chronic conditions.

Search strategy
We conducted our literature search using medical databases including PubMed/MEDLINE, CINAHL, EMBASE, and ISI Web of Knowledge, as well as within WHO regional databases. We limited the search to a 15-year publication window (November 1st 1997 to October 30th 2012). For the PubMed databases we used the search details shown in Additional file 1. Using the Boolean combination shown in this file, we retrieved a total of 2074 articles. We used similar text words for other databases according to the database's set-up. We also conducted hand searching using the reference lists of identified articles of interest. Figure 1 shows the flow chart of the studies thus identified and the screening process for inclusion. This figure is a modified version of the PRISMA checklist [42].
We retrieved a total of 4757 articles. Of these, 2074 articles were from PubMed and 2683 from other medical databases or retrieved through hand searches of the reference lists of initially selected articles. We compared the search results of the three independent researchers and reached agreement on the final figures. On initial screening, a total of 2051 duplicate articles were excluded because they appeared in both PubMed and other databases. Furthermore, we excluded a total of 2620 articles on initial screening which did not fit the focus of our study. We then conducted a full screening on the remaining 76 articles, out of which 66 articles were further excluded for various reasons: 17 lacked information on nutrition training of health workers, 28 had no clearly defined outcome variable matching our criteria, 9 had health personnel that did not fit our criteria of health personnel, and 12 had a population with a chronic condition or special needs ( Figure 1).

Grading of evidence and risk of bias
We evaluated the quality of the selected studies using the Grades of Recommendation, Assessment, Development, and Evaluation (GRADE) technique [43,44]. GRADE quality assessment is used to evaluate studies individually, and provides an overall assessment of the pooled evidence. The quality of evidence can be categorized as "high", "moderate", "low", or "very low" based on the study design, strengths and limitations, population size, and effect size of the pooled results. Using this method, we also examined the risk of bias inherent in the individual studies included. This was also done at the outcome level.
In this study, we evaluated the quality of the evidence on energy intake and feeding frequency, arriving ultimately at a "high" quality grade (Table 1). We also conducted a risk of bias (RoB) analysis. Based on the Cochrane Handbook for Systematic Reviews of Interventions [45], five types of bias were evaluated at the study level. These were selection bias, performance bias, detection bias, attrition bias, and reporting bias. There was a serious risk of bias in two of the ten selected studies [46,47]. In Bhandari et al.'s 2001 study [46], no information on allocation concealment was available. Also, authors did not indicate whether the control group received any intervention or care, which might have introduced a performance bias. In a 2005 study by Kilaru et al. [47], meanwhile, we could not find information on allocation concealment, performance bias, or attrition bias. However, this study indicated a low risk of reporting and detection biases (Additional file 2).
Despite the indicated RoB, results of these studies are not different from the rest of included studies. We therefore determined that the said risks of bias are unlikely to have markedly influenced the overall outcome of our review.

Data synthesis and meta-analysis
Outcome variables of interest in this study were energy intake, feeding frequency, and dietary diversity, all measured for the previous 24 hours. For energy intake, three studies presented the mean values and standard deviations [48][49][50], while two other studies presented their data in the form of median values and interquartile ranges [46,51]. Because of such diversity in the reported results, we conducted a separate meta-analysis of these studies. For consistency, we converted energy intakes in kilocalories (kcal) to kJ per day in three studies [32,50,51]. For feeding frequency, three studies presented their data as the mean feeding frequency per day [48,50,52], while two studies presented their data as percentage of participants with daily feeding frequencies greater than three [31,47]. We also conducted a separate meta-analysis due to such diversity in the reported results.  Several studies were excluded from meta-analyses due to missing or incompatible data. We did not include a study by Penny et al. [32] in meta-analyses because it lacked data on standard deviations for mean energy intake. Also, this study presented results of energy intake from animal sources only; other studies provided data on total energy intake from all complimentary foods consumed in a day. We could not conduct metaanalyses for the seven studies reporting results for the dietary diversity outcome [29,32,[47][48][49]51,52] due to differences in the categorization used in each study and the diversity of foods available in the respective study areas.
We used the metan command for STATA version 12 (StataCorp, College Station, Texas, USA) to calculate the effect size in the form of standardized mean differences (SMDs) and risk ratios; to calculate their 95% CIs; to assess the statistical heterogeneity among trials; and to construct the forest plots. For the dietary diversity outcome, we compared the frequencies and percentages of foods consumed between the intervention and control group in each RCT. Though the types of foods varied across studies, they were comparable within each trial. Chi-square tests were used to compare food type consumed in intervention and control groups in the three studies.
For each study selected for analysis, we chose the comparable results reported within similar age groups for intervention and control groups to reduce the effect of heterogeneity. Some of the studies that fit our inclusion criteria also presented results of age groups beyond the set criteria. Also, some studies presented results within our age groups of interest, but followed them up for the duration of the intervention and hence presented multiple results for the same subjects. However, we selected comparable results only for children aged between 6 and 18 months. We calculated SMDs for control and intervention group within similar age groups and pooled them to calculate effect sizes for energy intake and feeding frequency. For the data presented on follow-up results, we took the results at the end of the intervention.

Energy intake per day for children aged six months to two years
The pooled evidence from the three trials reporting on mean energy intake (Table 3) showed that, nutrition training of health workers improved daily energy intake of children aged six months to two years. The pooled SMD between the intervention and control groups was 0.76 with a 95%CI of 0.63 to 0.88 in a random model. The test for overall effect gave z = 12.17, P < 0.001. For the two studies with median energy intake presented with interquartile ranges (Table 3), the pooled SMD was 1.06, with a 95% CI of 0.87 to 1.24. The test for overall effect gave z = 11.22, P < 0.001.

Feeding frequency for children aged six months to two years
The pooled evidence from the three trials reporting on mean feeding frequency (Table 4) showed that nutrition training of health workers improved feeding frequency of  Health and nutrition workers in 4 communities without nutrition training. They cared for 473 child-mother pairs. At 9 months: At 9 months: Compared to Control, Intervention group had high proportion of dietary diversity (see Table 5)   children under two years of age. The SMD between the intervention and control groups was 0.48, with a 95% CI of 0.38 to 0.58 in a random model. The test for overall effect gave z = 9.17, P < 0.001. For the two studies with feeding frequency as a percentage (Table 4) showed pooled risk ratio (RR) of 0.99, 95% CI (0.87-1.13). The test for overall effect, z = 0.09, P = 0.926.

Dietary diversity for children aged six months to two years
In the RCT conducted in Brazil [49], health workers who received nutrition training provided counseling to caregivers. Children whose caregivers were counseled by trained health workers had a higher dietary diversity compared to their counterparts (P < 0.001). Similar interventions were conducted in India [47,48,51]. In these trials, too, children in intervention groups had significantly higher dietary diversities compared to their counterparts in control groups.
Other RCTs conducted in Pakistan [29], Peru [32], and China [52] showed similar results: high percentages of children in intervention group consumed each type of food items compared to their counterparts in the control group (Table 5).

Discussion
This is the first systematic review to evaluate the effectiveness of nutrition training of health workers on child feeding practices. Previous reviews showed the effectiveness of maternal nutrition education and complimentary feeding interventions to improve child feeding practices [18] and nutrition status [34,40]. Our study helps to show a possible pathway to improve child nutrition status by starting with health worker training. We found that training of health workers can help to improve feeding practices of children between six months and two years of age. The children whose caregivers were counseled by the trained health workers had a higher mean feeding frequency, energy intake, and dietary diversity compared to their counterparts. Strong evidence thus suggests that nutrition training of health workers improves energy intake, feeding frequency, and dietary diversity of children between six months and two years of age. Such a significant outcome may be conceived of through the following pathway: First, nutrition training can increase or refresh health workers' nutrition and food sciences-related knowledge. Indeed, two RCTs conducted in Brazil [19] and India [46] found that nutrition training of health workers improved their knowledge in nutrition. Nutrition training can be used to update health workers' nutrition knowledge and to alert them to new findings pertinent to their environments [16,28,29,53]. This will enable them to address determinants of undernutrition specific to their areas, and to improve their communication, counseling, and undernutrition management skills [19,29,32,48,49]. Updated management skills including tailored counseling may also be important for the effective transfer of knowledge to the end usersin this case, the caregivers.
Second, nutrition knowledge transfer by skilled and trained health workers may be achieved when they counsel caregivers who visit health facilities [20]. Similarly, trained health workers may also access caregivers through outreach and home visits even in rural areas, and may achieve a similar outcome through such routes [47,54]. Previous RCTs showed improved nutrition knowledge and knowledge retention among caregivers counseled by health workers who received nutrition training [16,29,31,32,49,54].
Third, the counseled caregivers can serve as agents of change. Caregivers endowed with updated nutrition knowledge through frequent counseling can improve their own child feeding behaviors [16,20,28,29,31,32,[47][48][49]52,55]. Such behaviors may include food preparation hygiene, feeding frequency, proper mixing of quality foods, increased energy intake, and dietary diversity. Thus, children's growth can improve and their risk of undernutrition can be minimized [34]. Secondarily, other determinants of undernutrition such as food-borne infections can be reduced [32,56,57] and food preparation hygiene improved [52].
Nutrition counseling from trained health workers has been proven effective even in areas of limited food availability [31,34]. In such circumstances, caregivers were able to choose the right mix of foods under availability constraints. For example, in the RCT conducted in Bangladesh, about a third of families were poor and lived in food-insecure households. Despite such hardship, nutrition knowledge gained from trained health workers motivated and changed their feeding behavior. Thus, they could provide the required balance of foods to their children [31].
The findings of this review should be interpreted in light of several limitations. First, the selected studies came from different regions and there is a risk of regional variations. Such regional variations can cause differences in characteristics of participants as shown in Table 2. Also, the selected studies were conducted in the context of different health systems. In this case, the nutrition training was conducted to the health workers of different carders. For example, in Bangladeshi and India studies, training was conducted among nutritionists and other health carders including medical officers. In other settings, training was conducted among health carders available in such settings, including doctors, primary health care providers, auxiliary nurses, midwives, health assistants, and community health workers as shown in    Table 2. To minimize this limitation, we selected RCTs and cluster RCTs as these studies can minimize the effect that could have been caused by differences in intervention and control groups. Meta-analysis pools the SMDs of each study into a single effective size. This can help to reduce any discrepancies arising from variations across studies. Second, we could not conduct a meta-analysis for the dietary diversity outcome. This was due to the differences in types of foods reported in the trials included in this study. Such differences were also due to regional variations in the typical diet. Also, in all the selected studies, dietary diversity was not a primary outcome. Lack of a standard method for data collection on dietary diversity might also be a reason for such differences. To minimize the effect of variations in food type, regional, and methodological aspects, we compared the results of diets consumed within the trials. All the seven trials showed better dietary diversity for the intervention compared to the control groups. Therefore, despite the regional and methodological differences in reporting dietary diversity, all studies showed the effectiveness of the intervention on dietary diversity among children under two years of age.
Third, our results showed a significant heterogeneity among the selected studies. This might manifest in differences in training duration and qualifications of health workers, in targeted age groups, in follow-up procedures, and in regional context. We could not retrieve the training duration for all of the selected trials. However, some of the selected trials used a standardized Integrated Management of Childhood Illness (IMCI) training manual developed by WHO, while others used results of formative research conducted prior to the trial. Moreover, results from all the selected studies were consistent. Although these studies were conducted in different regional contexts, they all showed a significant improvement in feeding practices when health workers received nutrition training.
Fourth, due to time limitations, we did not register a study protocol prior to the review process. To minimize such limitations, we developed the in-house review protocol based on the pre-set guidelines before starting the evidence search. The protocol was shared among the research team and the three independent researchers who conducted the evidence search. We evaluated each step of data collection as a team to verify the scrupulous use of the protocol. To this end, we were satisfied that the original protocol was adhered to.
Fifth, our results may also not be generalizable beyond the low-and middle-income countries where the selected studies were conducted. However, based on the global nutrition situation, these are the areas with the highest burden of child undernutrition. These results may thus be especially useful to scale up the nutrition training of health workers toward improving the current child undernutrition situation.
Despite its limitations, our study also has notable strengths. This is the first systematic review to examine the effectiveness of nutrition training of health workers on child feeding practices. Second, we used the GRADE method to critically assess the quality and strength of the evidence presented. Overall, the evidence of intervention effectiveness on feeding frequency and energy intake was of high quality. Thus, the results of this systematic review may help to design policies to improve feeding practices of children through training of available health workforce cadres.
In conclusion, nutrition training for health workers can improve feeding practices for children under two years of age. Such practices include feeding frequency, energy intake, and dietary diversity. Training materials should be prepared based on the local context and should include information on how to identify foods that are available, affordable and acceptable, which is particularly important in areas of limited food availability. Moreover, trained health workers offer the prospect of an accessible and reliable information resource for local families. In this way, nutrition training for health workers can serve as an important entry point for a sustainable strategy toward improving the nutrition status of young children.

Additional files
Additional file 1: Search strategy: PubMed.
Additional file 2: Risk of bias assessment.