Introduction
Although great progress has been made in tackling undernutrition at a global level, recent estimates indicate that among children under the age of 5 years, 22% are stunted and 7% wasted worldwide, and this now increasingly co-exists with rising levels of overnutrition [
1]. Peru, like many other countries in Latin America, has historically been marked by persistent high rates of stunting (low height-for-age) but over the last decade it has succeeded in reducing the prevalence more than half (from 31% in 2000 to 13% in 2016 in children under 5 years) [
2,
3]. However, the country has some of the world’s largest rural-urban disparities in nutritional outcomes and recent findings suggest that rural-urban and socio-economic inequalities in stunting have increased [
3‐
6]. The prevalence of overweight in young Peruvian children has remained relatively stable since the turn of the century (at 10–12% in children under 5 years); although this varies greatly by region, and there has been a secular increase in overweight levels in adults [
3,
7].
Early undernutrition may manifest itself in deficits in weight and/or linear growth. Stunting is the result of long term or chronic undernutrition (in utero or after birth) and/or infection hindering a child’s height potential, while wasting reflects a more immediate result of acute undernutrition or infection [
8,
9]. Research is nonetheless increasingly suggesting that these two forms of undernutrition should not be viewed independently as they share many common risk factors, can co-occur, and repeated/chronic wasting could lead to stunting [
10,
11]. Underweight is influenced by both weight and height and can reflect being wasted and/or stunted [
8]. Inadequate nutrition during early periods of growth has been associated with changes in physiological and metabolic functioning and lower fat-free and fat mass and potentially fat mass distribution. This may in turn have consequences on physical health (e.g increased metabolism-related disease), mental health, cognition and social wellbeing over the lifecourse [
8,
12,
13].
In contrast, overweight has been associated with diets and other behavioural and environmental factors that favour fat deposition, but it may also be associated with micronutrient deficiencies and hunger in certain contexts [
9,
14]. Overweight in childhood and adolescence has been associated with increased risk of diabetes, cardiovascular disease and cancer [
15], especially among those undernourished in earlier life [
16]. Overweight may also co-exist with stunting at the individual and household levels [
16]; both may also track into adulthood and affect offspring growth and wellbeing [
8,
17]. In Peru however, the mother and child double burden of malnutrition, mostly characterised by overweight mothers with stunted children within the same households, has dropped from 19 to 12% between 1992 and 2012. With reported increases in maternal overweight, this decline is mostly due to changes in stunting during that period [
18].
Different forms of malnutrition emerge within a broader sociocultural, environmental, economic, and political context which impacts food security and environment, eating practices and lifestyles, and access to health, sanitation, and hygiene services. Peru’s recent history of conflict in the 1980s and 1990s, which led to a great increase in rural-urban migration and urban informal settlements and exacerbated pre-existing high levels of poverty and food insecurity, shifted the context towards more malnutrition. Peru is also regularly affected by the El Niño–Southern Oscillation which severely disrupts regional weather patterns, causing natural disasters such as floods, food insecurity and disease, which may in turn contribute to deficits in child growth [
19]. Future changes in weather due to global warming are predicted to increase the intensity and frequency of such events and population displacement and migration [
20].
Studies from India, Mexico and Tanzania which explored the impact of migration on child growth in low- or middle-income countries, found that children born to internal (within-country) migrant mothers may have different nutritional outcomes than their non-migrant counterparts [
21‐
26]. Notably, children of rural-urban migrant women tended to have a lower risk of undernutrition and a higher risk of overweight compared to the children of rural non-migrant women but the opposite when compared to children of urban non-migrant women [
21‐
26]. Research into the impact of internal migration on children found that children who migrated with their parents (as opposed to children born after maternal migration) had lower survival [
27]; however, it is unclear if this pattern may also be the case for other health outcomes. There have also been conflicting reports on whether children of more recent migrants are more vulnerable to different forms of malnutrition [
22,
25].
Studies on adults in Peru show inequalities in anthropometric outcomes by internal migration status and suggest that moving to urban areas is associated with an increased risk of becoming overweight or obese [
3,
28]. Findings of the Young Lives cohort of Peruvian children born in the early 2000s indicate that children of migrant women had lower levels of stunting and underweight than those of non-migrant mothers and that rural-urban migration might be associated with better child nutrition [
29]. It is nonetheless uncertain how different pathways of internal migration are associated with growth and child nutritional health in this region and how this may vary by child migration status. Previous work from Peru suggests that geographic, socio-economic and prenatal factors are important determinants of child nutrition and may help explain how child nutrition varies by migration pathway [
4‐
6,
30].
This study uses nationally representative Demographic and Health Surveys conducted in Peru between 1991 and 2017 and aims to quantify and assess:
-
Aim 1: secular trends in child growth and nutritional status in Peru over time by maternal adult internal migration status (1991–2017).
-
Aim 2: associations of maternal adult internal migration and child growth and nutritional status in the 2017 survey, considering in turn time since migration, child timing of birth and type of urban area of current residence.
-
Aim 3: associations of maternal adult internal migration and the mother and child double burden of malnutrition (within children and between mothers and children living in the same households) in the 2017 survey.
The conceptual model hypothesizes that migration to urban areas will raise nutritional measures (reducing undernutrition but increasing overweight) while migration to a rural area will have the opposite effect; but these effects may be reduced after controlling for common pre-existing factors associated with both maternal adult migration and offspring growth and nutritional health and they may also differ by migration characteristics.
Discussion
This paper sought to explore how maternal adult internal migration was associated with child growth and nutritional status in Peru. Addressing our first aim, we found secular changes in both child growth and nutritional status over time in Peru, however these changes for the most part occurred regardless of maternal adult internal migration status. Broadly across time, children of migrant women in urban areas had poorer linear growth and greater levels of stunting but lower levels of overweight compared to urban non-migrants, with the opposite pattern in rural areas. Addressing our second aim, more in-depth analyses of recent data from 2017 showed that these differences were partly explained by confounding factors associated with maternal migration and child growth and nutritional status. Differences remained, however, suggesting that women’s internal migration may influence offspring anthropometry. Addressing our third aim, we found no differences in the risk of mother and child double burden of malnutrition between migrant and non-migrant women neither within rural nor urban areas. We discuss these findings in greater detail below and then consider both future implications and what mechanisms may underlie the patterns found.
Changes over time
Between 1991 and 2017 there was a steady increase in HAZ, a smaller increase in WAZ (mostly in the 1990s) and a smaller decrease in WHZ in Peruvian children under 5 years. This was accompanied by a large decline in stunting across time and a small decline in underweight (in the 1990s), with little change in the prevalence of overweight. These secular trends reflect the findings shown in previous DHS analyses [
3,
6]. Reductions in stunting have been attributed to multi-sectorial poverty-reduction efforts implemented during that period [
2,
6].
WHZ was overall higher than reference data for all children across the DHS and, although WHZ and overweight did not increase over time in DHS children under 5 years, we have seen increasing levels of overweight and obesity at older ages in Peru and across the Latin America region [
28,
37,
38]. Furthermore, there appeared to be a mean shift towards positive WAZ for urban children over time. The positive WHZ trend may reflect historically low WHZ which could have slowed linear growth, subsequently allowing for recovery of WHZ using energy saved by slowing linear growth [
10]. However, as Peru continues through its nutrition transition and food environments and behaviours change in the region [
6,
37,
38], the prevalence of overweight will likely increase as the children increase in age. Indeed, previous findings in the Young Lives Peru cohort study of children indicate that this will likely happen both as the individual children grow older and across birth cohorts over time, particularly for those in living in urban areas [
37].
While there were improvements overall, differences persisted between maternal adult internal migration groups over time for linear and weight growth and stunting, whereby urban areas showed better growth and lower prevalence of undernutrition, but higher prevalence of overweight compared to rural areas with migrants falling between urban and rural non-migrant groups. Children of urban-urban migrant mothers were the most like urban non-migrants while children of rural-rural migrants were the most like rural non-migrants. Children of rural-urban and urban-rural migrant women had outcomes between these two groups (urban-urban and rural-rural) but nonetheless displayed differences; the rural-urban group being more similar to the rural-rural group and the urban-rural group to the urban-urban group.
For WHZ and overweight, there was some evidence of a widening of differences by maternal migration groups over time driven by increases in urban non-migrants and decreases or no change in all other groups. This contrasts with findings in Peruvian adults which showed that obesity had increased across most migration groups and particularly in rural areas [
6,
28]. It is therefore possible patterns would have been different in older children.
Associations of maternal internal migration and child anthropometry in 2017
Further analyses carried out on the DHS for 2017 (addressing aim 2) showed that there were significant differences between mothers in different adult internal migration groups in terms of markers of early maternal socio-economic and nutritional environment (reflected in maternal schooling, age, ethnicity, height and child birth order). When these were considered, the predicted differences in child growth and nutritional health by maternal migration status were reduced but patterns were maintained for the most part. This supports findings that maternal phenotype and capital are important determinants of offspring growth and health [
39‐
41] and justified our more detailed analyses that adjusted for such factors where data were available.
Overall, after adjusting for confounders, children in rural areas remained more likely to be worse off than children in urban areas in terms of linear growth and undernutrition (mostly stunting) but better off in terms of lower overweight prevalence. Among rural-dwelling children, having a migrant mother, particularly an urban-rural one, was associated with greater linear growth and lower likelihood of undernutrition (mostly stunting) with no differences for overweight or by timing of birth (reflecting rural or urban birth).
In contrast, in urban areas, having a migrant mother, particularly a rural-urban one, was associated with lower weight growth and overweight prevalence compared to non-migrants but there were no differences for linear growth and stunting. This was particularly the case if a child was born rurally and migrated subsequently to an urban setting; whereas children of urban-urban mothers and children of rural-urban mothers who were born after migration showed similar mean WHZ and overweight prevalence.
There was little evidence of any difference between children of recent and of long-term migrant mothers after taking confounders into account. Previous research in India had found that greater duration since maternal rural-urban migration was associated with higher levels of child undernutrition [
22]; others have however argued that it very much depends on the context and that longer urban exposure is likely to increase obesogenic exposures [
25]. Indeed, in adults in Peru, a longer time in urban residence was associated with a higher probability of being obese [
28]. The small sample sizes in our analyses and young ages of the children may explain why we found no associations.
Longitudinal research is needed to explore whether as children of internal migrant women age, they become more like their urban non-migrant peers or instead retain anthropometric differences associated with maternal migration from rural areas. The data on mothers suggests that an earlier rural exposure is not necessarily protective against becoming overweight later, indeed migrant mothers in 2017 had higher levels of overweight than non-migrants within both urban and rural areas, with the highest across all groups found in rural-urban migrant mothers. However, the obesity prevalence was higher in the urban non-migrants (at 32%, compared to 26 and 22% in urban-urban and rural-urban migrant mothers respectively). Research on adults in the Peru Migrant Study (up to 2012–13) shows there may be variations by type of obesity with rural-urban migrants having the highest risk of central obesity [
42].
For those living in urban areas, differences between migrants and non-migrants may differ by the type of area of residence; for example, children of migrants in capital/large cities and towns had better linear growth than non-migrants but not in small cities where patterns were in the opposite direction but not significant. The lack of significance here and for other outcomes may be due to the small sample sizes. It is therefore important to consider level of urbanization in child health and nutrition research and policy, not only because of variations in outcomes but also because the needs of migrant and non-migrant may vary.
The mother & child double burden of malnutrition
The double burden of malnutrition analyses (aim 3) showed that mothers and children in rural areas had the highest risk of the double burden of malnutrition, mostly characterised as an overweight or obese mother with a stunted child, though there were no differences by maternal migration status after adjusting for confounders. Another study conducted in the Brazil DHS from 2006 found no differences in the mother and child double burden of malnutrition between rural and urban areas [
43]; it is nonetheless possible that differences have emerged since.
Indeed, research by Popkin et al. shows that, with changes in global food systems and health behaviours, the double burden of malnutrition in LMICs is increasingly being concentrated in rural areas [
16]. This highlights the need for social programmes in rural areas to consider both under- and over-nutrition at the household level, as it has previously been suggested that such programmes in Latin America remain primarily focussed on tackling undernutrition in children [
36]. A study exploring the mother and child double burden of malnutrition in the DHS for Peru over time showed that while there had been an overall decline in prevalence, there had been an increase in undernourished children with overweight or obese mothers [
44].
Potential mechanisms
Together, the results of this paper suggest that maternal migration may impact offspring growth and nutritional health in different ways; this may be due to both exposure to rural or urban environments and the process of migrating itself. While migrating to cities may lead to new or better opportunities for employment and education and higher standards of living, it may also come with a loss of social support, barriers to accessing services, and stressful and hazardous work and living conditions [
45‐
50]. Differences in these factors could explain variations in child health and survival by migration status [
26,
51,
52].
It has recently been suggested that pollution may play a role in stunting [
53] and previous research in Peru found that migrants in Greater Lima had a 10-fold increase in exposure to larger particulate matter (PM2.5) compared to non-migrants [
54]. We did not find higher levels of stunting in children of migrants compared to non-migrants in larger cities but further investigation should compare child growth in areas with different levels of pollution within cities and at different ages. Changes in altitude with migration could potentially also play an important role in Peruvian child growth [
55] and should also be further explored. Social networks during migration are likely important as well, for example in Indonesia it was found that rural-urban migration only negatively impacted adult mental health if it occurred alone rather than with family [
47]. We did not however have information on these factors in the DHS.
Research suggests that stunting is mainly established by 24 months and therefore that interventions should ideally intervene with this critical window [
39]. However more recently it was suggested that there may be opportunities beyond that with some interventions in LMICs showing improvements in height at later ages [
39]. As they become older, it is possible that migrant and non-migrant children catch up with each other in stature, although differences will likely remain [
56]. Research comparing highland and lowland (migrant) Peruvian children found differences in body lengths that were maintained throughout the ages studied up to 14 years [
55]. Furthermore, early stunting, catch-up growth and rapid weight gain (which can follow early undernutrition) could predispose to greater fat mass later on [
57,
58]. Low stature in children may reflect a low homeostatic metabolic capacity which could predispose children to ill health in the future, particularly when exposed to urban obesogenic environments (high metabolic load) [
59]. Based on higher levels of overweight at older ages in Peru and research on women in the DHS showing increasing level of obesity with time since migration, the burden of obesity in Peruvian children will likely increase as they age and, in the case of migrants, acculturate to their new environment [
28,
42,
60,
61].
While moving from rural to urban areas may improve access to food, it may also lead to food insecurity for some, as well as overeating, poorer quality food, changing food habits and/or sedentary behaviours, which may lead to obesity and related disease [
26,
60,
62‐
67]. In Peruvian adults it was found that rural-urban migration was associated with an increased risk of becoming obese, developing inflammation, and metabolic disease, but not hypertension, type 2 diabetes and higher HbA1c or mental health disorders [
49,
60,
68‐
70], perhaps reflecting benefits of an early rural exposure, as we discuss further on. The risk was increased with greater time in urban place of destination and possibly greater age at migration, although results were conflicting for the latter [
60,
68,
69].
The negative impacts of urban migration are likely to be greater in slums or informal settlements where barriers to healthier foods may be higher [
45,
71‐
73]. Findings from the Young Lives cohort of Peruvian children has shown that greater stunting is associated with poor access to sanitary facilities and greater food insecurity [
61]. Qualitative research carried out in Peru’s ‘young towns’, peri-urban towns created informally by migrants, indicated that lack of green spaces and community resources, safety and the unhealthy food environment created barriers to good nutritional health [
74].
An early exposure to particular gut microbes (‘Old Friends’) and various organism in the rural environment may influence immune system regulation and inflammatory processes and lead to rural and urban differences in child health [
75]. Research has suggested that gut health may be important for linear growth in early life [
39]. Rural to urban migration may possibly influence nutritional health through these pathways as well. Rook et al. found that an early rural (low-income country) exposure confers some benefits to immune regulation and various aspects of physiological health, thus reducing the risk of developing chronic inflammatory and psychiatric conditions after migration to a high-income urban environment [
76,
77]. However, in contrast to the findings on metabolic health in other research, later migration was associated with greater benefits [
76]. These effects of an early rural exposure may possibly explain why children of rural-urban migrants appeared to have lower levels of overweight. Internal rural to urban migration may impact other aspects of child health such as body composition, epigenetic ageing, inflammation, and hormonal profiles which are not available in existing data and would warrant further investigation.
Strength and limitations
The DHS framework provides data that is collected using piloted and well-designed surveys following best practice for study design and data collection. As the surveys are carried out consistently through DHS/ENDES, this offers a large and nationally representative dataset that is comparable over time. The surveys also offer a wealth of information on both children and their mothers. This allowed us to explore migration and child nutritional outcomes as well as several maternal, socio-economic and environmental covariates. Three child nutritional outcomes are available, and the z scores are calculated by DHS using WHO 2006 reference data for all DHS surveys allowing for comparison across years.
The anthropometric data for these outcomes are collected by trained professionals using widely used and appropriate equipment. Interviewers verified some of the current information such as current residence making the data unlikely to be biased. Response rates were high for all surveys and the level of missing data was low for the measures used in this research.
The use of self-report for past area of residence and time in current area of residence could nonetheless have introduced some bias into the results (such as recall bias, particularly with increased time since migration). Different perceptions of what should be classified as a town or city when recalling previous area of residence could also have introduced bias. However, the relationship between maternal adult internal migration and child nutritional outcomes was relatively similar over time in different groups of women. Analyses were repeated using a measure of lifetime internal migration, and the patterns of association were like those using adult internal migration. Interviewers verified current residence making these data unlikely to be wrong. As sensitivity analyses, the statistical analyses carried out in DHS 2017 were repeated for the previous DHS survey and the results were similar.
The categories of migration used in the paper bring together a number of possible pathways of migration (when considering capital, city, town and countryside and previous and current areas of residence) which could not be explored in the regression models due to lack of power and which were therefore simplified into categories based on urban and rural residence instead, similarly to other research. However further exploration of the data (not shown in the paper) indicated that there were no differences by previous urban area of residence, only by current urban area of residence (which we therefore considered in our stratified analyses).
It is also likely that the association of maternal adult internal migration and child growth and nutritional health would have differed by previous geographic region, which was not available in the DHS. For example, children of rural-urban mothers who had migrated from coastal regions may have differed from those who had migrated from highland regions. The inclusion of ethnicity and maternal height as confounders should have allowed to adjust for this to some extent, as we found that current geographic region was associated with both, but there may have remained uncaptured differences. Additional analyses were carried out exploring associations of maternal adult internal migration and child outcomes by current region and patterns remained similar after adjusting for confounders (not shown in the paper).
The cross-sectional design of the DHS and the reliance of self-reported data means that the relationship between maternal migration and child nutritional outcomes cannot be demonstrated to be causal. Longitudinal data such as the Young Lives study would have allowed to explore causality and patterns as children age; however, the small sample and limited data did not allow us to address our research aims. Unmeasured confounding in the regression analyses may also have led to false conclusions both regarding the association between migration and child nutritional outcomes.
There may additionally be some ‘selection’ or ‘healthy migrant’ effects, whereby we see better-off or healthier individuals in the baseline population migrating. This may explain why we see better growth and nutritional health in migrants in some of our results. There is however limited evidence of these effects with internal migration; with most research to date having focussed on international migration. While mothers who have migrated from rural to urban areas do not appear taller than their non-migrant counterparts, which would have suggested better early socio-economic circumstances and/or health, they are more educated and from different ethnic backgrounds which could suggest greater wealth and/or greater potential for adaptation to the new setting. If there are indeed unobserved differences such as these which could not be accounted for then our findings may have overestimated the true effect of migration on health. This may nonetheless depend on the context of migration; it is possible that for mothers who migrate in situations of emergency (e.g conflict, climate, natural disaster) we would see a larger impact on child health compared to those who migrate for marital or other socio-economic reasons. It is likely that a lot of the women in our sample migrated for reasons related to marriage and this may have led to better socio-economic conditions and a change in environment which we were not able to capture with DHS data.
Conclusion
While Peru has made great progress in improving child growth and reducing undernutrition overall, this varies by maternal and child internal migration history. Similarly to studies in other settings, we show that migrating may create unique nutritional profiles combining the risks and benefits of both rural and urban life and that both migration and rural/urban exposures are important in shaping child physical development and health in Peru. Being the child of a migrant in an urban area, regardless of whether urban or rural in origin, appears to be associated with lower weight growth and risk of overweight and even more so if the child was born before or during migration. In rural areas however, being the child of a migrant, particularly an urban-rural migrant was associated with better linear growth and reduced undernutrition. Nonetheless, even considering migration pathways, we still see large differences in growth and nutritional health between rural and urban children.
Although the prevalence of overweight, and inequalities therein, are relatively low in Peruvian children under 5 years, other research at older ages indicates these outcomes will likely increase, as will associated disease. It is unclear whether rural-urban migration may impact children’s physical development and physiology in other ways, and further investigations should explore other differences, for example in body composition and hormonal profiles. Further research into which factors mediate the association between maternal migration and child nutritional health would also help identify areas amenable to policies and interventions to improve child health in settings with high internal migration like Peru.
Interventions to improve child nutritional health could take into consideration maternal and child internal migration history, as well as different aspects of malnutrition which could exist within the same groups of children or the same households. As Peru continues to go through its nutritional and economic transition, we may see further declines in stunting, but a greater burden of obesity and metabolic disease emerge. Nonetheless, with recent developments related to the SARS-Cov-2 pandemic and as the climate crisis evolves, there could also be a widening of socio-economic inequalities and reversal of any progress made so far [
19,
20,
78].
Publisher’s Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.