Risk factors for malnutrition among school-aged children: a cross-sectional study in rural Madagascar

Nutrition continues to be a key global development agenda under the Sustainable Development Goals (SDGs) after the completion of the Millennium Development Goals (MDGs) era. Despite a critical need for addressing malnutrition among all populations, the monitoring indicators currently in SDG2 and previously in MDG1 are consistently focused on children under 5 years of age [1, 2]. While admitting children under 5 years of age serve not only as the highest priority sub-population but also as a leading proxy for entire populations in monitoring nations’ nutritional status, it is necessary to capture broader age groups. Scaling Up Nutrition (SUN), a global multisectoral movement on nutrition, addresses malnutrition in a more comprehensive manner, by expanding its target monitoring groups beyond children under 5 years of age (children 13–15 years of age, women 15–49 years of age, and adults ≥18 years of age) [3]. Yet, schoolchildren aged 5–14 years remain excluded from the monitoring frameworks of these major global development agendas as if they were a neglected sub-population. As a result, school-aged children are rarely targeted in nutrition surveys, despite significant impacts of nutritional status on their health, cognition, educational achievements and future economic productivity [4, 5].

Malnutrition among children under 5 years of age has been regularly monitored in all countries since the WHO’s launch of the standard growth chart in 1978 [6]. Yet, it was just 2007 when WHO launched the standard growth chart for children and adolescents aged 5–19 years, by applying prevalence of stunting, underweight, and not wasting but thinness as the malnutrition indicators [7]. The long-term absence of the technical tools for monitoring the three malnutrition dimensions until 2007 might have discouraged researchers from conducting studies on malnutrition among school-aged children. Recently, World Bank started highlighting a need for addressing malnutrition among school-aged children as a foundational investment a country should make, by launching the Human Capital Measurement Project [5]. Thus, a global momentum is being created toward more studies on malnutrition among school-aged children.

For over 20 years, Madagascar has been challenged by consistently higher prevalence of stunting among children under 5 years of age (around 50%) for decades. The latest data (49.2% as of 2016) is the fifth highest in the world [8]. To address the country’s long-lasting high malnutrition prevalence, the Government of Madagascar increasingly focuses on multi-sectoral coordination for greater coverage and impacts of the national nutrition program, by launching the third National Nutrition Action Plan (PNAN III). PNAN III also calls for multisectoral interventions for school-aged children, e.g. community-based deworming, school feeding, household food security, and crop diversification [9]. As a subcomponent policy of PNAN III, National School Feeding and Nutrition Strategy (SNANS) specifies the school-based nutrition interventions. Nevertheless, even basic malnutrition indicators among school-aged children (i.e. prevalence of stunting, underweight, thinness, overweight and obesity) have been neither analyzed nor reported. It is generally difficult for stunted children to catch up during 24–59 months of age [10]. Therefore, the malnutrition status among children under 5 years of age in Madagascar implies the equally high malnutrition prevalence among those ≥5 years of age, due to chronic food insecurity and inadequate feeding practices in the country [11].

In view of unknown but presumably higher malnutrition among school-aged children, the Malagasy Ministry of National Education (MoNE), in collaboration with Japan International Cooperation Agency (JICA), has been committed to strengthening the community-based school feeding program, by developing and piloting its effective and sustainable model. Basic framework of the program defines community-based school feeding as regular provision of meals on school days using locally either produced, procured or donated food items through a school management committee. This study provides the program with the baseline data which not only inform its detailed design but also will be compared with end-line data in the future.

This study is aimed at estimating malnutrition prevalence among children enrolled in rural public primary schools in Madagascar, where school-feeding is implemented, as the pre-intervention baseline data. It further attempts to identify the possible risk factors for malnutrition among them. This study is expected to contribute to informing the design of the community-based school feeding program, by making evidence-based recommendations on it.

A cross-sectional household survey was conducted, to estimate malnutrition prevalence (i.e. stunting, underweight, thinness, overweight and obesity) among school-aged children in Antananarivo-Avaradrano district, Analamanga region, Madagascar. Note that stunting and underweight among children 5–19 years of age are defined in the same manner as those among children under 5 years of age (i.e. by employing z-scores for height-for-age and weight-for-age). Yet, thinness among children 5–19 years of age is differently defined from wasting among children under 5 years of age, though both are aimed at measuring acute malnutrition in common. Thinness is defined as those having a z-score < − 2 for BMI-for-age, while wasting is defined as those having a z-score < − 2 for weight-for-height [7]. Moreover, overweight and obesity among children 5–19 years of age are defined as those having z-scores > 1 and > 2 respectively for BMI-for-age, while definitions of overweight and obesity among children under 5 years of age are based on z-scores for weight-for-height.

Of 404 children sampled, five were dropped due to parental refusal or absence at their households despite successful contacts with children at schools. It was found that six of 399 children were under 5 years of age. Thus, those six children were excluded from the analysis. As a result, data collected from 393 children and their household members were analyzed. WHO growth reference does not cover children older than 120.8 months of age in z-score calculation for weight-for-age. Therefore, 11 children categorized into the age group were excluded from estimation of prevalence of underweight [16].

The overall prevalence rates of stunting, underweight and thinness were 34.9% [95% CI: 30.4–39.7%], 36.9% [95% CI: 32.2–41.9%] and 11.2% [95% CI: 8.5–14.7%], respectively (Table 1 [7, 8, 18‐20] and Fig. 1). Of 393 children assessed in this study, 123 (31.2%) suffered from multiple forms of malnutrition (Fig. 2). Note that 19 children (4.8%) suffered from all three forms of undernutrition. Prevalence rates of overweight and obesity were 1.0% [95% CI: 0.4–2.6%] and 0% [95% CI: 0.0–1.0%], respectively. Thus, overnutrition among school-aged children in Antananarivo-Avaradrano district need not be addressed as a major public health issue. For this reason, further analyses were conducted only for stunting, underweight and thinness in search for their possible risk factors.

The background variables were compared between stunted and not stunted children, between underweight and not underweight children, and between thin and not thin children. Of 51 background variables, five produced significant associations (P <  0.05) with whether being stunted in bivariate analyses (Tables 2 and 3). Similarly, seven and three background variables produced significant associations (P < 0.05) with whether being underweight and with whether being thin, respectively. Moreover, multicollinearity was not confirmed, as only VIF < 10 was detected among them. Therefore, all these background variables with significant associations were employed as the independent variables for logistic regressions.

The results of logistic regressions are shown in Table 4. In the logistic regression on being stunted, four of five independent variables produced significant adjusted odds ratios (ORs) (P < 0.05), i.e. ‘Seller, trader or commercial business’, ‘Sweets’, ‘Age’, and ‘Total number of household members’. In the logistic regression on being underweight, four of seven independent variables produced significant adjusted ORs (P < 0.05), i.e. ‘Seller, trader or commercial business’, ‘Age’, ‘Total number of household members’, and ‘Household dietary diversity score ’. In the logistic regression on being thin, two of three independent variables produced significant ORs (P < 0.05), i.e. ‘Age’ and ‘Total number of household members’.

Prevalence of stunting among children aged 5–14 years (34.9%) was lower than national prevalence of stunting among children under 5years of age (49.2%) [8]. Its data as of 2008–2016, during which the children assessed in this study used to be under 5 years of age, were consistently around 50%, too [18]. Yet, the data specific to Antananarivo-Avaradrano district as of 2008–2016 is not available. We should assume that prevalence of stunting during their under-five childhood was at the same level as the current figure, as catch-up in height-for-age during 24–59 months of age is generally unexpectable [10, 14].

Prevalence of underweight among children aged 5–14 years (36.9%) was at the same level as national prevalence of underweight among children under 5 years of age (36.0%). Prevalence of thinness among children (11.2%) was lower than those of stunting and underweight. Absence of national and provincial prevalence of thinness makes it difficult to compare this result with others. This is because prevalence of thinness has never been reported from Madagascar. Prevalence of thinness is a malnutrition indicator applicable exclusively for 5–19 years of age. Therefore, the indicator is rarely assessed. Note that a majority of nutrition surveys target not school-aged children and older age groups but exclusively children under 5 years of age and their mothers. It is not surprising to learn that the relatively new indicator is generally not attractive enough for health ministries to monitor its data, as it has not been employed as an indicator for global agenda such as SDGs [1].

‘Age’ produced significant ORs > 1 (P < 0.05) in all three logistic regression models for stunting, underweight and thinness (Table 4). This is consistent with the increase in prevalence of stunting, underweight and thinness according to age (Fig. 3). Earlier studies reported that z-score of height-for-age either remains low or even further drops during the period from two to 14 years of age in low- and middle-income countries. It is only adolescence (i.e. 14–20 years of age) when catch-up in height-for-age could occur after the initial 2 years of life [10, 14]. Since there remain 4–5 years for the current first and second graders to enter that age group, impact of strengthening school feeding on height-for-age might not be observable during their primary education enrollments. Nevertheless, adequate food availability and dietary diversity over a sufficient period time including primary education period should be ensured [10, 14]. Thus, strengthening of school feeding will be an essential investment and its evaluation should be conducted not prematurely in a few years but at an appropriate time, i.e. when and/or after first and second graders complete primary education. To increase the likelihood of post-under-two catch-up in both height-for-age and weight-for-age, childhood gastrointestinal infections [14] (schistosomiasis in particular) [27‐29] should be adequately prevented and controlled. Madagascar has the world’s fifth highest prevalence of schistosomiasis (52.1% as of 2010) [30]. WHO recommends that all school-aged children be treated by mass praziquantel administration once a year in schistosomiasis high endemic countries with prevalence > 50% [31] such as Madagascar. Though 152 of 399 children (38.1%) assessed in this study underwent school-based deworming, their prevalence of all three forms of undernutrition was not significantly different from those not having undergone it (Table 2). Irregular and sporadic implementations of school-based deworming program might have made deworming less effective. Of 10 primary schools visited, eight implemented deworming interventions before. Yet, at three of eight primary schools, more than 3 years have passed since the last deworming implementation.

Also, ‘Total number of household members’ produced significant ORs > 1 (P < 0.05) in logistic regression models for stunting, underweight, and thinness (Table 4). This indicates that, the greater number of members a household has, the higher likelihood of being stunted, underweight and thin its schoolchild has. There are two possible reasons for this. First, per-capita food availability at larger households should become smaller. Members of larger households naturally can access and consume a less amount of foods than those of small households, when comparing households at the same income level. In addition, the average number of household members in the lowest quintile income group is the greatest of all quintile income groups, in Madagascar [32]. This synchronously makes it difficult for larger households to ensure per-capita food availability. Second, Malagasy households generally tends to allocate a less amount of foods inadequately meeting age-specific energy requirement to young and older children [33]. Thus, nutritional deficiencies among schoolchildren in this study might be more severe than among their parents. Having assumed this, school-feeding should play a key role as the critical nutrient source supplementary to children’s inadequate diet at household level derived from inappropriate intrahousehold food distribution.

‘Household dietary diversity score (HDDS)’ produced significant OR < 1 (P < 0.05) in the logistic regression model for underweight (Table 4). This indicates that children consuming a fewer number of food groups are likely to be underweight. Inadequate intake of animal-based protein ingredients (‘Meats’, ‘Milk and milk products’) were significantly associated with being stunted and underweight (P < 0.05) in bivariate analyses (Table 2), though these variables did not produce significant OR (P < 0.05) (Table 4). This may imply that there might be a need for improvement of dietary diversity through using particularly animal-based protein ingredients for school meals. Those having ‘had lunch through school feeding program’ on the previous day of anthropometric measurements had significantly greater HDDS than those not having (P < 0.05: Mann-Whitney U test). Yet, its mean difference between the two groups was only 1.12 (=7.17–6.05). In fact, ‘had lunch through school feeding program’ was not associated with being stunted, underweight and thin. These support a need for improvement in school feeding program by increasing the number of food groups in school meals.

This study has limitations in accuracy in schoolchildren’s individual dietary data, as parental 24-h diet recalls and school feeding program records were sources of the data. Yet, monotonous dietary pattern in rural Malagasy households is generally likely to ensure a certain generalizability of 24-h dietary not only over a period of time but also to all household members. Note that parents of 384 children in this study (97.7%) stated that dietary pattern of last 4 days remained the same. Moreover, earlier studies found that parents are reliable reporters at least of their children’s in-home dietary intake [34].

Prevalence rates of stunting and underweight among 5–14 years of age were as high as those among children under 5 years of age. Difference in the definitions between thinness among children 5–14 years of age and wasting among children under 5 years of age makes it difficult to compare their prevalence rates.

‘Age’ and ‘Total number of household members’ were identified as the common possible risk factors for stunting, underweight and thinness among schoolchildren aged 5–14 years. Adequate food availability and dietary diversity over a sufficient period time including primary education period should be necessary for increasing likelihood of catch-up expectable during adolescence. Malagasy households have strong preference for rice and cannot afford to buy meats except on the days when they receive income [35]. School feeding program, therefore, should be able to more feasibly procure animal-protein ingredients than households, by leveraging economies of scale in food procurement.

Nevertheless, household feeding practices should improve in parallel, as school feeding can cover only 172 days (Malagasy standard number of school days) out of 365 days (47%) at a maximum in a year. Efforts on feeding practices need to be made both at school and household levels up to 14 years of age and beyond. Note that this study is expected to contribute to evidence-based menu planning for the school feeding program, by serving as the first study that reports risk factors and profiles for malnutrition of school-aged children in Madagascar.