Chronic malnutrition is highly prevalent among children under age 5 globally, with the greatest burden affecting children in low- and middle-income countries (LMICs) in Africa and Asia [1]. Stunting, defined as height- or length-for-age (HAZ/LAZ) less than 2 standard deviations below the population standard mean [2], is an indicator of chronic malnutrition [3]. Fifteen percent of all deaths and 21% of disability-adjusted-life-years in children under 5 years have been attributed to stunting [4]. Stunting also has long-term consequences, including impaired cognitive development, increased risk of non-communicable disease in adulthood, and decreased economic productivity [5].

Although the etiology of chronic malnutrition is multi-faceted, an estimated 13.5% of global stunting prevalence is attributable to diarrheal disease [6]. A meta-analysis of longitudinal studies in 5 LMICs reported a child’s odds of stunting at 24 months of age increased by 16% with every 5% increase in incidence of diarrhea (odds ratio 1.16 [95% confidence interval (95% CI) 1.07, 1.25]) [7]. In addition, children in seven LMICs across Africa and Asia who experienced moderate-to-severe diarrhea (MSD) lost significantly more height/length for age z-score (HAZ/LAZ) in the 2–3 months following the episode than age- and village-matched controls [8].

Addressing linear growth faltering in children with MSD may be an important step towards reducing stunting and its long-term consequences. This may be particularly true for those under 24 months of age, as this is the critical time period in which most growth faltering occurs [9] and during which interventions are likely to be effective. However, it is unclear which groups of children are at highest risk. In addition, few interventions have been successful at mitigating the nutritional consequences of diarrhea [10]. Identifying risk factors for post-MSD linear growth faltering can inform which groups of children should be prioritized for inclusion in trials of potential interventions, and, once an effective intervention has been identified, to optimize the effectiveness of intervention delivery within programs by targeting children at high risk of growth faltering.

Using data from children under 24 months old with MSD enrolled in a previous large diarrhea etiology study (the Global Enteric Multicenter Study, or GEMS), we sought to identify determinants of linear growth faltering in the 60–90 days following presentation with MSD. We evaluated the frequency and severity of linear growth faltering in this population and identified the clinical, host, and socioeconomic factors associated with faltering in linear growth during the short-term follow-up period. We also developed and validated a predictive model and risk scoring tool for estimating an individual child’s risk of short-term growth faltering following MSD.

Among the 9439 children with MSD who were enrolled in the GEMS study, 2205 children aged ≥ 24 months and 1031 children with a missing or implausible LAZ value were excluded. This resulted in 6203 surviving children under 24 months of age included in the analysis (Fig. 1). Median age of included children was 11 months (interquartile range 7–16) and 43.2% were female (Table 1). Distribution across the 7 sites was similar to that in the parent study: 705 (11.4%) in The Gambia, 1172 (18.9%) in Mali, 410 (6.6%) in Mozambique, 961 (15.5%) in Kenya, 1195 (19.3%) in India, 993 (16.0%) in Bangladesh, and 767 (12.4%) in Pakistan. Approximately 22% (n = 1352) of children presented with dysentery, 94.4% of whom were given or prescribed an antibiotic in the health facility (whereas 75.4% of children without dysentery were given an antibiotic). Thirty percent presented with fever and 19.8% were hospitalized at presentation. Approximately, one in four children presenting with MSD were stunted at presentation and one in five were wasted. Approximately 43.2% (n = 2681) of these children under 24 months of age experienced a subsequent diarrhea episode during the follow-up period, per caregiver report at the follow-up visit. Fig. 1

Flowchart of included subjects

In this post hoc analysis of children with MSD enrolled in the GEMS study, we found that over one-fifth of children under 24 months had linear growth faltering at ~ 60-days following the MSD episode. We identified several risk factors for linear growth faltering, including age, fever, general IMCI danger sign, and nutritional status. We found that some of these factors yielded reasonable predictive value to identify children likely to experience severe linear growth faltering following MSD. We found that most children who experienced linear growth faltering were not stunted at MSD presentation. Stunting status at diarrhea presentation may not identify all children who are at risk for linear growth declines following an episode of MSD. Using these other clinical factors to predict linear growth faltering may result in earlier and more complete identification of children who are on a trajectory of linear growth declines, comparing to using only stunting status at diarrhea presentation to predict post-diarrhea growth declines and thus may be useful for targeting interventions to prevent stunting.

Patterns of linear growth faltering followed patterns determined by age and baseline LAZ. The older the child, or the lower the LAZ value at enrollment, the lower the probability that the child with lose LAZ over the ensuing 2–3 months. Growth in early life is rapid and decreases as the child ages [20, 21]. Correspondingly, risks of linear growth faltering decrease as children age, with the highest risk occurring before 12 months. Our findings are consistent with previous work noting the substantial losses of LAZ in early life [9] and suggest that interventions may confer the most benefit within this critical period. This growth pattern also underscores methodological considerations for analyses of linear growth faltering. Children in the youngest age groups have the highest growth velocity and therefore have the greatest opportunity to lose or gain LAZ. Growth faltering in older children may have different underlying etiologies and health implications than that occurring in infancy, and so we have restricted the analysis to children under 24 months of age and controlled for age and baseline LAZ in the analysis. Challenges in ascertaining and interpreting losses in linear growth by age group highlight the need for research to assess the clinical relevance of different magnitudes of loss in LAZ by different age groups.

Similar to the patterns of LAZ loss by age, children with LAZ values that are already low (below 0) were less likely to lose more LAZ. We have described how losses in LAZ increase consistently with higher LAZ, consistent with patterns observed elsewhere [22]. While our results suggest stunting status or low LAZ values may not identify children who are at risk for further linear growth deterioration, children who are already stunted are at high risk of the health and cognitive detriments associated with chronic malnutrition [5]. Whether health consequences of loss in LAZ are differential by stunting status remain unclear. A modest loss in LAZ may prove to have more health consequences in already stunted children than a loss of greater magnitude in non-stunted children at diarrhea presentation.

We identified host, clinical, and environmental characteristics that were significantly associated with short-term linear growth faltering. Acute malnutrition (measured either by MUAC or WLZ) was significantly associated with subsequent growth faltering. This could also reflect dehydration status, and this finding may reflect that the more dehydrated children had more severe intestinal injury and absorptive capacity, and thus more likely to experience linear growth faltering. Ponderal growth has previously been found to be associated with linear growth. A longitudinal analysis of birth cohorts from the USA, Ghana, and Honduras reported that WLZ was positively correlated with length gain [23], as did a cohort study of Jamaican 9–24-month-old stunted children [24]. Additionally, a study in the West Indies reported that severely malnourished children needed to attain ≥ 85% WLZ before they began to gain LAZ [25]. These studies suggest ponderal growth may precede linear growth, as weight loss reflects a lack of available nutrients needed to sustain linear growth. It is also possible the higher risks of severe linear growth faltering we observed in acutely malnourished children may be due to higher rates of subsequent diarrhea episodes during the follow-up period. Previous research has reported higher incidence of diarrhea in acutely malnourished children [26, 27], though we did not have data on diarrhea beyond 14 days of follow-up to examine this hypothesis. Acutely malnourished children presenting with MSD may thus be an easily identifiable population who may benefit from nutritional interventions that protect against linear growth faltering.

Presentation with fever was associated with linear growth faltering as has been shown previously [28]. Fever may be a sign of more severe intestinal inflammation and injury, as is often seen in bacterial diarrhea, which may be associated with linear growth faltering. This is supported by the finding that children with MSD who were hospitalized at presentation were at higher risk of linear growth faltering than those who were not. Finally, the presence of any IMCI danger sign at MSD presentation was also associated with a loss of more LAZ. Studies have demonstrated the potential of IMCI programs for improving quality of care and child survival [29–31]. However, a Cochrane review of the effectiveness of IMCI programs reported little to no benefit on stunting or wasting [30] which could reflect the lack of effective interventions for improving nutritional status upon identification of high-risk children.

In our analysis, children presenting with dysentery had lower risks of linear growth faltering than those with non-dysenteric MSD. This finding was unexpected and differs from that of other studies that found dysentery, or specific pathogens known to cause dysentery, to be associated with risk for linear growth faltering [32–34]. Our detection of a reduced risk associated with dysentery may be related to clinical management. WHO guidelines recommend antibiotics for dysentery [35], and in our data, children presenting with dysentery were more likely to receive an antibiotic than those without. It is unclear whether antibiotic management of MSD alters growth [36]; some research has reported growth-promoting effects of antibiotic treatment on length and weight in children in LMICs [37, 38]. Clinical trial data will be needed for evaluating the effectiveness of antibiotic management of MSD for protecting against subsequent linear growth faltering.

We found that children in lower wealth quintiles had the highest rates of linear growth faltering. Poverty is a well-established underlying cause of childhood stunting. There are large disparities in stunting rates by wealth quintile within LMICs, with child stunting rates in lowest wealth quintiles as much as 13 times higher than in the highest [39]. Socioeconomic factors are the most consistently identified correlates of stunting [40], and it has been estimated that every 10% increase in national gross domestic production per person would result in a 6% decrease in stunting prevalence [41]. Economic development may be influential in protecting children with MSD against linear growth faltering [42]. We found that children in households without access to improved defecation facilities experienced greater losses in LAZ, though results were not consistent for both outcomes and for our analyses of water source. Greater exposure to environmental pathogens may place children at higher risk of linear growth faltering, as pathogen-specific diarrhea [33, 34] and asymptomatic pathogen carriage [43–45] have been found to be associated with linear growth faltering. Unimproved WASH may also contribute to environmental enteric dysfunction (EED), which is strongly associated with linear growth faltering and thought to play a central role in stunting [46, 47]. However, WASH interventions have not yielded consistent benefits. While a review of stunting in 137 LMICs using Global Burden of Disease data reported unimproved sanitation to be a leading cause of stunting [6], a Cochrane review reported only modest benefits of WASH on child length but limited availability and quality of evidence [48]. Large clinical trials of WASH interventions did not detect a benefit on child growth [49–51].

When considering which risk factors best predicted likelihood of severe linear growth faltering, age, stunting, wasting, fever, and presence of any IMCI danger sign emerged as the most important. The predictive ability of the model was fair to moderate and could benefit from further research to improve the predictive ability of the model, perhaps by including data not available in GEMS, such as birth weight or HIV status information. External validation would further improve the model. The risk score model performed better than any individual predictive factor, suggesting that the combination of these factors is more useful for identifying children at risk of severe linear growth faltering than any of these variables individually. We identified the risk score cut-point that maximizes sensitivity and specificity, but the cut-point used in practice should be weighed against the costs or negative consequences of potential interventions. This predictive model uses only easily collected clinical data routinely documented at diarrhea presentation, and such a risk score could be useful for identifying children at highest risk for inclusion in trials of interventions to reduce linear growth faltering and ultimately may prove useful in determining how to best apply successful interventions once benefit is demonstrated, by identifying high-risk children who stand to benefit from such an intervention or be monitored more closely following MSD.

There have been few studies to our knowledge that identify risk factors of linear growth faltering in children following an episode of MSD. Our study contributes data on this important topic, using a large, multi-country cohort with a rigorous study design and data collection practices. There are several limitations to our analysis as well. Data on birth size, HIV status, and previous and subsequent diarrhea episodes were not available in the parent study, which may be relevant to this secondary analysis. Our analysis assessed short-term effects (2–3 months) only. It has been reported that catch-up growth is possible following a diarrhea episode if no subsequent diarrhea episodes are experienced [52], and it is possible that some of the growth deficits we observed were transient. The risk and predictive factors we have identified for short-term losses in LAZ may or may not be the same factors associated with longer term growth declines. However, we found that a substantial proportion of these children presenting with MSD experienced a repeated diarrhea episode in the subsequent 50–90 days, and this additional growth insult may have precluded catch-up growth for this subset, who may have continued on a linear growth decline. This short-term period also represents a very vulnerable period, as children’s risk of death was more than 8-fold in this period following a MSD episode compared to healthy controls [8]. Longer follow-up studies will be important for assessing sustained linear growth deficits associated with diarrhea, as well as other health outcomes associated with linear growth faltering at different ages. The cut-off of 0.5 LAZ for our definition of severe linear growth faltering is arbitrary, and the clinical implications of this magnitude of loss are unclear. Additionally, all definitions used implicitly assume the impact of LAZ loss is the same, irrespective of age or enrollment LAZ. We adjusted for age and LAZ at baseline in our analysis, but difficulties remain with interpreting the health detriments of these outcomes.

Children presenting with MSD that are acutely malnourished (or more dehydrated), under 12 months of age, presenting with more severe disease (as indicated by hospitalization, presence of fever, or IMCI danger signs), and those living with limited access to improved sanitation may be at higher risk of linear growth faltering following MSD. To identify children for inclusion in further trials and to guide clinical decision-making for close monitoring of high-risk children or targeting an intervention once an effective intervention has been identified, age, nutritional status, and signs of disease severity may be useful to identify children at highest risk.