Global disability-adjusted life-year estimates of long-term health burden and undernutrition attributable to diarrhoeal diseases in children younger than 5 years

1Measurement of prevalence, incidence, and mortality associated with diarrhoeal diseases is important for policy decisions, resource allocation, and targeted interventions, but it might not provide a complete understanding of the true disease burden. Evidence is increasing that childhood diarrhoea, especially in the first 2 years of life, can negatively affect nutrient absorption, leading to poor physical growth., Additional measures of long-term health loss due to diarrhoea, other than incidence and mortality, are needed. Disability-adjusted life-years (DALYs) are designed to quantify both acute and long-term health burden and are the sum of years of life lost due to premature mortality and years lived with disability. Since their introduction more than 20 years ago, DALYs have become a key metric for assessing and monitoring population health and setting priorities within a nation's health sector.

2Evidence before this studyWe searched PubMed for articles without language or publication date restrictions for (“stunting” OR “wasting” OR “growth” OR “malnutrition”) AND (“diarrhea”) AND (“prospective” OR “case control” OR “trial” OR “cohort”) originally on May 2, 2017, with an update on Jan 10, 2018. We identified 198 publications, including 17 that provided information about the association between diarrhoea and childhood growth. Many of these studies found an association between diarrhOea, measured as prevalence or cumulative exposure, and childhood growth indicators but varied in exposure and outcome definitions. We also identified several publications that reviewed this association, but none attempted to systematically quantify the burden of diarrhoea on childhood growth.Added value of this studyOur study provides a systematic, global set of estimates of the effect of diarrhoea on childhood growth impairment and the burden of impaired growth, measured in disability-adjusted life-years. We assess how diarrhoea affects the risk of subsequent infectious diseases and quantify these outcomes for each country, both sexes, and every 5 years since 1990 for children younger than 5 years. These estimates suggest that the burden of diarrhoea among children under 5 years old is multifaceted, causing increases in acute and long-term disability-adjusted life-years, and was previously under-recognised. After accounting for diarrhoea and a risk factor for childhood growth faltering, the number of DALYs due to diarrhoea among children under-5 increased by about 40%.Implications of all the available evidenceInterventions and policy to prevent diarrhoea incidence could improve child growth and infectious disease burden beyond diarrhoea exclusively. Childhood growth and infectious diseases have a complicated, circular relationship. Reducing the risk of diarrhoea might have broad effects on childhood growth and ability to thrive. Our study findings will guide public health policy for prioritising resource allocation for alleviating the burden of diarrhoeal diseases.

3Evidence before this study

4We searched PubMed for articles without language or publication date restrictions for (“stunting” OR “wasting” OR “growth” OR “malnutrition”) AND (“diarrhea”) AND (“prospective” OR “case control” OR “trial” OR “cohort”) originally on May 2, 2017, with an update on Jan 10, 2018. We identified 198 publications, including 17 that provided information about the association between diarrhoea and childhood growth. Many of these studies found an association between diarrhOea, measured as prevalence or cumulative exposure, and childhood growth indicators but varied in exposure and outcome definitions. We also identified several publications that reviewed this association, but none attempted to systematically quantify the burden of diarrhoea on childhood growth.

5Added value of this study

6Our study provides a systematic, global set of estimates of the effect of diarrhoea on childhood growth impairment and the burden of impaired growth, measured in disability-adjusted life-years. We assess how diarrhoea affects the risk of subsequent infectious diseases and quantify these outcomes for each country, both sexes, and every 5 years since 1990 for children younger than 5 years. These estimates suggest that the burden of diarrhoea among children under 5 years old is multifaceted, causing increases in acute and long-term disability-adjusted life-years, and was previously under-recognised. After accounting for diarrhoea and a risk factor for childhood growth faltering, the number of DALYs due to diarrhoea among children under-5 increased by about 40%.

7Implications of all the available evidence

8Interventions and policy to prevent diarrhoea incidence could improve child growth and infectious disease burden beyond diarrhoea exclusively. Childhood growth and infectious diseases have a complicated, circular relationship. Reducing the risk of diarrhoea might have broad effects on childhood growth and ability to thrive. Our study findings will guide public health policy for prioritising resource allocation for alleviating the burden of diarrhoeal diseases.

9The Global Burden of Diseases, Injuries, and Risk Factors Study (GBD) is a comprehensive and systematic effort to quantify deaths and incidence, prevalence, and DALYs for 264 diseases, including diarrhoea. Despite a substantial reduction of 65% in diarrhoea mortality among children younger than 5 years since 1990, the decline of diarrhoea incidence among this age group has been less pronounced than that of mortality, decreasing by 24% in the same time period and remaining the fourth-leading cause of under-5 DALYs in 2015.,,

10In this study, we leverage the Global Burden of Diseases, Injuries, and Risk Factors Study 2016 (GBD 2016) to update the burden of childhood diarrhoea, accounting for its long-term health outcomes due to undernutrition. We conceptualise this pathway by first estimating the effect per day of diarrhoea on three common measures of child growth. Undernutrition is well established as an important risk factor for infectious disease morbidity and mortality and is necessary for protein energy malnutrition (PEM). By evaluating the change in undernutrition, we attribute a fraction of all DALYs associated with undernutrition and PEM from GBD 2016 to diarrhoea. It is important to measure chronic health consequences to assess true health burden, which will in turn inform policy and decision making regarding the reduction of diarrhoeal burden through primary or secondary prevention strategies.

11This study has five components. The DALYs due to acute diarrhoea, childhood undernutrition, and PEM were previously estimated as part of GBD 2016. This analysis consists of a systematic literature review and meta-analysis, quantification of DALYs attributable to diarrhoea via growth impairment, and quantification of DALYs attributable to diarrhoea due to PEM. The final step in the estimation of the total burden of diarrhoea, accounting for long-term outcomes associated with growth impairment, is linking these results to the larger GBD framework.

12This study leveraged previously published estimates from GBD 2016. GBD provides comprehensive and internally consistent epidemiological estimates for more than 300 causes of death and disability and 90 risk factors for 195 locations, by year, sex, and four age groups younger than 5 years. Detailed descriptions of all GBD methods have been previously published.,,,, This study used two main inputs from GBD 2016. The first was the incidence of diarrhoea, estimated with use of DisMod-MR 2.1 and data from representative population surveys, health-care utilisation records, and the scientific literature. The second was the number of DALYs attributable to three childhood undernutrition indicators: low height-for-age, low weight-for-age, and low weight-for-height. Childhood undernutrition is a risk factor for three outcomes in GBD: diarrhoea, lower respiratory infections, and measles. The DALYs attributable to undernutrition were based on the prevalence of the undernutrition indicators and the association between those indicators and the outcomes. Additionally, PEM is entirely attributable to low weight among children younger than 5 years in GBD 2016. More detail is given in the appendix and the Global Burden of Diseases, Injuries, and Risk Factors Study 2015 diarrhoea, GBD 2016 causes of death, GBD 2016 non-fatal, and GBD 2016 risk factors publications.

13This extended analysis began with the hypothesis that diarrhoea is causally associated with childhood growth failure on the basis of biological plausibility and evidence from the scientific literature., We did a series of systematic reviews to establish the change in Z-scores of height-for-age (HAZ), weight-for-age (WAZ), and weight-for-height (WHZ) associated with diarrhoea among children younger than 5 years. On Aug 8, 2016, we performed detailed literature searches using the EBSCO, Embase, and PubMed databases with no restrictions on publication date or language (appendix). We updated our search on Aug 23, 2017. We also examined the reference list of a 2015 meta-analysis for additional publications and did analyses of microdata (individual-level data) that we obtained for this study. We excluded exposures that occurred in children older than 5 years and outcomes assessed at older than 12 years (appendix). We limited our analysis to studies that could assess nutritional status in terms of continuous HAZ, WAZ, and WHZ. We excluded studies that limited these assessments to dichotomised variables, such as stunting, underweight, and wasting.

14We selected these growth indicators because they are health related, are quantifiable, have sufficient literature or available microdata for meta-analyses, and are quantified in GBD 2016., We extracted both unadjusted and adjusted effect sizes from the scientific literature and did analyses of the adjusted effect sizes. We standardised the effect sizes to represent the change per day of diarrhoea. We assumed a linear relationship between diarrhoea and growth indicators. For the microdata, we did a standard panel regression accounting for child sex, age in months, and duration between measurements. We assumed that the relationship between diarrhoea and growth did not vary by age as we had insufficient data to evaluate this. We performed random-effects meta-analyses to obtain effect sizes for the change in HAZ, WAZ, and WHZ per day of diarrhoea.

15To calculate the DALYs associated with childhood undernutrition attributable to diarrhoeal diseases, we estimated the log-linear relative risk (RR) of disease and mortality for diarrhoea, lower respiratory infections, and measles per unit change in HAZ, WAZ, and WHZ on the basis of hazard ratios reported in a 2013 pooled analysis (appendix). Although hazard ratios might not be equivalent to risk ratios in many circumstances, we assume that hazard ratios can be treated similarly to risk ratios for assessment of the association between a risk and an outcome in estimation of an attributable burden. In this report, hazard ratios (HR) will be used as proxies for risk ratios and referred to as HRs. These HRs capture the relationships between undernutrition indicators and infectious disease risk. We applied the HRs to disease morbidity and mortality because of an absence of evidence evaluating disease morbidity exclusively. The HRs for stunting and wasting are adjusted to account for their covariance with each other and with underweight as described previously. We then used the effect sizes from our meta-analyses and modelled diarrhoea incidence (from GBD 2016) to calculate an outcome-specific population-attributable fraction (PAF) for the three undernutrition indicators:

16where diarrhoea incidence is the modelled number of episodes per child-year in GBD 2016 and varied by year, age, location, and sex, episode duration is the mean duration in days per diarrhoea episode duration in days per diarrhoea episode (4·3 days, 95% CI 4·2–4·4), Δzscore is the unit change per day of diarrhoea, and HR is the relative risk of each outcome per unit change in Z-score. The overall undernutrition PAF for each outcome was calculated as:

17To estimate the number of childhood undernutrition DALYs attributable to diarrhoea, we multiplied the summary PAF by the GBD 2016 DALYs for each outcome attributable to undernutrition (diarrhoea, lower respiratory infections, and measles).

18PEM in GBD 2016 is attributed entirely to WHZ and WAZ, but not HAZ. We used the prevalence of underweight and wasting from GBD 2016, defined as less than 2 SDs below the global mean weight-for-age and weight-for-height values. We used the diarrhoeal disease incidence and effect size from the meta-analysis to estimate the counterfactual prevalence of underweight and wasting in the absence of diarrhoea, defined as:

19where prevalence is the prevalence of wasting or underweight, represented as a Z-score, Δzscore is the change in WAZ or WHZ from our meta-analysis, diarrhoea incidence is the incidence per person-year, and episode duration is the mean duration of each diarrhoeal episode (4·3 days [95% CI 4·2–4·4]). We estimated the percentage change in the cumulative density from a normal distribution:

20The percentage difference between the observed and counterfactual prevalence is the PAF (appendix). We calculated PAFs independently for mild, moderate, and severe undernutrition, age group, geography, sex, and year. PEM-associated DALYs attributable to diarrhoeal diseases are the PAFs multiplied by the number of DALYs due to PEM.

21We estimated long-term DALYs due to diarrhoea for both sexes; every GBD geography; the years 1990, 1995, 2000, 2005, 2010, and 2016; and under-5 age groups. The GBD study estimates every outcome, cause, and risk at a cascade of geographies. This approach means estimation at the global, super-regional, regional, national, and sometimes subnational level. In GBD, four under-5 age groups exist. We did not estimate long-term DALYs due to diarrhoea in the neonatal age groups (0–28 days) because undernutrition in this age being due to causes other than preterm birth or other birth complications is implausible and because of the short time window for this age group. However, we consider the acute diarrhoea DALYs from each age group among children younger than 5 years, including neonates, in our analysis. We discuss percentage increase after accounting for long-term sequelae as the percentage difference between the total (acute and long-term) and acute DALYs. We assessed temporal trends by finding the relative percentage difference in DALYs per 1000 between the years 1990 and 2016. We present results for each country and for super-regions in the GBD study where the values for the super-regions are the population-weighted aggregates of the country estimates in that super-region.

22We calculated uncertainty in the final DALY estimates using 1000 draws of each input parameter, including diarrhoea incidence, the effect size from the meta-analyses, the RR of each outcome, and the number of DALYs for each outcome (PEM, lower respiratory infections, diarrhoea, and measles). Our final DALY point estimate is the mean DALY estimate, with the lower uncertainty interval defined by the 2·5 percentile of the estimate and the upper uncertainty intervals defined by the 97·5 percentile of the estimate.

23The funder of the study had no role in study design, data collection, data analysis, or writing of the report. The corresponding author had full access to all the data in the study and had final responsibility for the decision to submit for publication.

24We included 22 sources after reviewing reported values in the literature (17 sources) and incorporating individual-level data (five sources; appendix). From a series of random-effects meta-analyses, we found that per day of diarrhoea, height-for-age decreased by 0·0033 SDs (95% CI 0·0024–0·0041; p=4·43 × 10−14), weight-for-age decreased by 0·0077 SDs (0·0058–0·0097; p=3·19 × 10−15), and weight-for-height decreased by 0·0096 SDs (0·0067–0·0125; p=7·78 × 10−11; appendix).

25Diarrhoeal diseases were responsible for about 40 125 700 DALYs from diarrhoea incidence and deaths among children younger than 5 years in 2016 (table), including about 446 000 deaths (390 900–504 600) and 1105 million episodes (962 million to 1275 million).,, After inclusion of the long-term sequelae associated with undernutrition, diarrhoea was responsible for 55 778 000 DALYs, an increase of nearly 40% (39·0% [33·0–46·6]; table).

26The total burden of diarrhoea, represented as DALYs per 1000 people, is shown in figure 1. The highest rates of diarrhoea DALYs were in sub-Saharan Africa, south Asia, and southeast Asia. After accounting for long-term sequelae associated with diarrhoea, the number of diarrhoea DALYs more than doubled in some countries (table, figure 1B, appendix). Various countries in central Asia, such as Mongolia and Uzbekistan, with low acute diarrhoea burden, increased by 300% after inclusion of long-term sequelae DALYs (table). DALYs associated with diarrhoea increased by more than two-thirds in Latin America and the Caribbean, with larger increases in Andean Latin America and lower increases in Central Latin America. The southeast Asia region increased by more than half overall, with the largest increase in Cambodia and the lowest in the Maldives.

27The total diarrhoea DALYs per 1000 children younger than 5 years occurred in western (320·5 per 1000, 95% UI 258·1–389·7) and central (209·0 per 1000, 145·6–293·8 per 1000) sub-Saharan Africa, and the lowest rate occurred in Australasia (0·9 per 1000, 0·7–1·1; figure 2). The highest rate of diarrhoeal DALYs occurred in the central Sahel region of sub-Saharan Africa, including Chad (689·1 per 1000 [95% UI 513·0–890·9]), the Central African Republic (566·7 per 1000 [382·0–828·2]), and Niger (442·1 per 1000 [293·0–639·1]; appendix; table). Acute DALYs were correlated with long-term DALYs in most locations. The total rate of diarrhoea DALYs in south Asia was 77·7 per 1000 (67·1–90·2), including 69·1 per 1000 (59·7–80·0) in India and 155·2 per 1000 (112·4–208·7) in Pakistan where the burden was predominantly due to the acute DALYs. The rate of diarrhoea DALYs varied substantially among countries in north Africa and the Middle East, ranging from 2·8 per 1000 (2·0–3·8) in Kuwait to 173·4 per 1000 (107·7–259·2) in Afghanistan. In central Europe, eastern Europe, and central Asia, rates ranged from 4·2 per 1000 (2·9–5·7) in Montenegro to 149·0 per 1000 (86·9–237·7) in Tajikistan.

28About 84·7% of long-term DALYs are due to infectious diseases as a consequence of undernutrition— 87·4% in children aged 28–364 days and 79·7% in children aged 1–4 years—and PEM due to underweight and wasting accounts for the remaining 15·3% of long-term DALYs (figure 3). Most of the acute and long-term sequelae DALYs occurred in the 28–364 days age group (figure 3). The percentage increase in this age group, after accounting for the long-term DALYs compared with acute DALYs, was 50·5% (95% UI 39·8–62·0), whereas the percentage increase in total diarrhoea DALYs in the 1–4 years age group was 32·7% (26·1–40·7). The rate of total diarrhoea DALYs per 1000 was nearly six times higher in the under-1 age group (255·2 per 1000 [224·4–290·3]) than in the 1–4 years age group (42·8 per 1000 [36·6–49·8]).

29DALYs associated with diarrhoea varied over time (figure 4). At the global level, DALYs attributable to diarrhoea long-term sequelae decreased by 78·7% from 1990 to 2016 (95% UI 69·3–85·5), which was faster than the rate of decline of acute diarrhoea DALYs (70·4% [61·7–76·5]). Total diarrhoea DALYs decreased by more than 90% between 1990 and 2016 in east Asia, Tropical Latin America, and Central Latin America. The greatest reduction in this time period occurred in east Asia (96·5% [95·1–97·5]), Tropical Latin America (90·5% [88·5–93·5]), and Central Latin America (90·2% [88·1–91·9]). The rate of decline was much lower in sub-Saharan Africa (43·8% [26·9–57·2]), with the lowest overall reduction in western sub-Saharan Africa (27·8% [–2·0 to 49·6]), which was not significant. Diarrhoea DALYs increased in high-income North America (23·9% increase [–10·7 to 71·2]) and Australasia (36·6% increase [–10·3 to 105·3]), but these increases were not significant.

30By quantifying the long-term sequelae due to growth faltering, our study shows that the global burden of diarrhoea is substantially underestimated when only incidence and mortality are considered. We found that accounting for long-term sequelae increased the number of diarrhoea DALYs among children younger than 5 years by about 40%. After inclusion of these long-term sequelae, diarrhoea moves from the fifth-leading to the third-leading cause of DALYs among children younger than 5 years, surpassing malaria and neonatal encephalopathy in the number of DALYs in this age group. Most of this long-term burden of diarrhoea is associated with an increased risk of subsequent infectious disease mortality.

31We found that wasting, defined as low weight-for-height, is responsible for 40% of the undernutrition-associated diarrhoea DALYs. Among the undernutrition indicators estimated in GBD 2016, the effect of diarrhoea is larger on weight indicators, which could be reflective of short-term deficit, compared with height, which could be a lifelong deficit. This finding is intuitive as diarrhoea is associated with dehydration and immediate weight faltering. The risk of infectious disease mortality due to wasting (low weight-for-height) is much greater than that due to stunting and underweight. Children that are undernourished or unwell or have been sick recently are at a higher risk of subsequent infection and mortality than are those who are healthy., This fact presents a potentially circular, positive feedback loop in which childhood undernutrition and diarrhoea each increase the risk of the other. When possible, our analysis limited the effect of diarrhoea such that any change in childhood growth indicators occurred after an episode of diarrhoea.

32Childhood undernutrition is strongly associated with not only infectious disease mortality but also all-cause child mortality, following a dose-response relationship where the risk of mortality increases as the anthropometric measurements move further from the global mean. Undernutrition might specifically impair the innate and cell-mediated immune response, cause thymic atrophy, and upset various homoeostatic systems, such as electrolyte balance, glycaemia, and thermoregulation.,, Lower respiratory infections might also be associated with decreased physical growth, although the mechanism for growth impairment due to respiratory infections might differ from the mechanism for impairment due to diarrhoea, as enteric infections might impede nutrient absorption through inflammation or flattening of intestinal microvilli,, pathways that would not be biologically shared in respiratory infections.

33We have limited our analysis to the effects of diarrhoea on childhood undernutrition and subsequent morbidity and mortality due to infectious diseases and PEM and have assumed a causal relationship between diarrhoea and childhood growth failure. We chose the outcomes of childhood undernutrition for this study because they are quantifiable and part of GBD 2016. The ability to leverage existing estimates from GBD 2016 is a powerful and unique addition to this analysis. However, evidence exists that undernutrition in childhood might have additional long-term sequelae, such as cognitive impairment, decreased physical fitness in later life, vaccine hyporesponsiveness, Guillain-Barré syndrome, and even chronic diseases in adulthood, such as cardiovascular disease and type 2 diabetes.,,,,, Stunting, low height-for-age, might be a better measure of chronic undernutrition than weight in early childhood, particularly during the first 2 years. Although stunting might be reversible, it often leads to reduced adult stature unless compensatory growth occurs.,

34Of particular interest is the effect of physical growth stunting on cognitive development. Both cognitive development and physical growth depend on a diverse set of micronutrients and appropriate caloric intake. Some evidence also exists that enteric pathology, with or without diarrhoea, might influence cognitive development outside of the proposed causal pathway between undernutrition and cognitive development, perhaps through chronic immune upregulation and activity or inflammation, where childhood stunting is a comorbidity., Depressed cognitive potential in children might have substantial net effects on economic and social productivity.

35We defined diarrhoea as the exposure for impaired physical growth. However, evidence exists that enteric infections leading into environmental enteric dysfunction (EED), even in the absence of diarrhoea, might also be associated with growth faltering. Enteric infections could more comprehensively define physical growth faltering risk among children younger than 5 years than diarrhoea alone, but quantification of this risk and exposure is difficult. Given the limitations of available data, we believe that diarrhoea episodes provide the best proxy for total enteric infections and accurately reflect the global distribution of the associated burden. Nevertheless, additional research is needed to assess the effect of non-diarrhoea enteric infections, including EED, on child development. Additionally, enteric microbes or families of microbes, such as viruses, bacteria, protozoa, and parasites, with or without diarrhoea, might differentially mediate the relationship between enteric infection and growth impairment. These considerations illustrate that simply accounting for diarrhoea incidence, although the most comprehensive measure available, might be an imprecise metric for the risk of undernutrition due to enteric infections.

36Next steps for this work will include quantification of the effect of diarrhoeal diseases and enteric pathogen infections on cognitive development and differentiation of the effect of enteric pathogens on physical growth. We have assumed in this analysis that no difference in the effect of diarrhoea on growth exists between pathogens; however, evidence exists that some pathogens, like Cryptosporidium, might have larger effects on growth than might other pathogens., Future iterations will be strengthened with more data than currently available for diarrhoea, diarrhoeal causes, EED, childhood growth, and cognitive development.

37The burden of diarrhoea has declined rapidly in many locations in the last 25 years and might meet global targets in under-5 mortality., Treatment, such as oral rehydration solution, has probably played a major role in this progress, but our results suggest that emphasis on diarrhoea and enteric infection prevention, such as provision of safe and reliable water and sanitation and effective vaccines, could have benefits on chronic disease and could lower risk of subsequent infectious disease.,,

38Our analysis has various potential limitations. The research hypothesis in this analysis is that diarrhoea is causally associated with childhood growth failure and that growth failure is causally associated with infectious disease risk. Assessment of causality in any analysis is difficult and prone to confounding. We did a random-effects meta-analysis, which allows for heterogeneity in the observed effect size that could be due to variation in study populations, exposure, or covariates that were considered in the original analyses. Still, the results from our analysis show a significant, biologically plausible, temporally structured, and consistent relationship between diarrhoea and childhood growth. Future studies are needed to replicate and strengthen, or refute, this association.

39A core component of GBD is to produce comparable estimates of disease burden for every time period, geography, sex, and age group, even in the absence of data to inform these estimates. Our statistical methods are built on predictive validity such that covariates and Bayesian priors help to provide estimates of diarrhoea and undernutrition burden, even in locations without data. All of our modelled estimates include 95% UIs, which reflect our ability to accurately describe disease burden, even in locations without data.

40We excluded studies that reported the effect of diarrhoea on weight and height expressed as dichotomous variables. This decision was taken to make an estimate of a continuous change in weight or height, but it reduced the available sample size as many studies reported different measures of association, such as the odds ratio of stunting (HAZ of <2 SDs below the global mean) per diarrhoea episode. Furthermore, we assumed that the distribution of height and weight was normally distributed in every population. This assumption could bias our estimates of the effect of diarrhoea on PEM, depending on whether the distribution of weight, which is left-tailed on the global level, was in fact skewed or not normal in a given population.

41Our analysis revealed that, despite decades of progress, the burden of diarrhoea among children younger than 5 years remains sizeable and has been systematically underestimated. Accounting for long-term sequelae of diarrhoea, namely susceptibility to subsequent infectious disease and growth deficiencies, increased estimated diarrhoea DALYs by nearly 40%. These results call for a redoubling of efforts by the international community to prevent diarrhoea and long-term sequelae associated with childhood growth.