Global and regional trends in under-five diarrheal disease burden: impact of human development index and geographic disparities, 1990–2021

Abstract

Aim:

In this study, global and regional trends between 1990 and 2021 were examined by sex, level of development, and geography, and projected to 2025.

Background:

Diarrheal disease continues to be a leading cause of morbidity and mortality in children under the age of five (U5).

Methods:

We assessed GBD 2021 data for U5 children in 204 countries, 21 regions, and 7 super-regions with joinpoint regression to analyze time trends, a hybrid ARIMA-ETS-ANN model to project prevalence and mortality until 2025, and longitudinal multilevel modeling to determine the effect of development level. Spatial clustering in 1990 and 2021 was assessed with Local Moran's I.

Results:

In 2021, the global prevalence and mortality rates of diarrheal disease among children under five were 885.07 and 51.72 per 100,000 population, respectively. Between 1990 and 2021, global U5 mortality decreased by 80.4%, and prevalence by 71.8%. The heaviest burden remained in Sub-Saharan Africa (SSA) and South Asia (SA), though all super-regions experienced statistically significant decreases. Joinpoint regression across the entire interval (1990–2021) revealed significant overall reductions in mortality (AAPC −5.15; 95% CI: −5.19, −5.10) and prevalence (AAPC −3.98; 95% CI: −4.03, −3.94). Hybrid forecasts predict ongoing decreases through 2025, with prevalence reaching 631.3 and mortality 36.6 per 100,000 population worldwide. Multilevel models revealed steeper annual reductions in low- and medium-Human Development Index (HDI) nations, corroborated by significant time–HDI interaction terms (β=79.76 for prevalence; β=7.50 for mortality; both p<0.001), although such countries had a persisting higher absolute burden. Spatial analysis revealed enduring hotspots in SSA and SA in both 1990 and 2021, and the formation of new clusters in several high-income countries. Coldspots occurred primarily in high-income and island nations.

Conclusion:

Significant global advancements have lessened the burden of diarrheal disease in U5 children; yet, ongoing disparities attributed to unsafe water, poor sanitation and hygiene, undernutrition, and low maternal education underscore the necessity for combined water, sanitation, and hygiene interventions, rotavirus immunization, and community-based health education in high-risk areas.

Introduction

Childhood diarrheal disease is a major global public health burden, with an estimated 1.7 billion cases and 525,000 deaths each year in children under the age of five (U5) (1, 2). Though preventable and treatable, diarrheal disease continues to be the second largest cause of death of this age group after acute respiratory infections (1-3). The burden is particularly elevated in low- and lower-middle-income countries (LMICs), where scarce resources and underdeveloped health infrastructure result in over 90% of child fatalities due to U5 diarrheal disease. Most glaringly, South Asia (SA) and sub-Saharan Africa (SSA) account for a staggering 88% of these deaths (4). Childhood mortality rates due to diarrheal disease in LMICs are approximately ten-fold higher than in developed nations (5, 6).

Encouragingly, global trends in U5 diarrheal disease mortality rates declined by 69.6% from 1990 to 2017 (7). This is likely due to several factors, including improved healthcare and nutritional access, enhanced case management, increased sanitation coverage, improved prenatal and infant healthcare, and higher education levels. Socioeconomic, environmental, and behavioral factors also significantly influence diarrheal disease incidence (8). The Human Development Index (HDI), which is a composite of education, income, and life expectancy, and all are strongly associated with childhood morbidity and mortality, is another important factor (9, 10).

Knowledge of the regional differences and trends in diarrheal disease mortality and incidence is essential for policymakers and public health practitioners. It guides evidence-informed decision-making on interventions, specifically their efficacy, accessibility, affordability, and scalability at the global level. Surprisingly, current data indicate divergent regional trends. Whereas high-income North America saw an alarming rise in diarrheal disease mortality rates (125.3%) from 1990 to 2017, East Asia had a substantial reduction of 95.1% over the same interval (7).

Although many studies have examined the epidemiology of diarrheal disease in U5 children, they have largely targeted individual nations (11-14). Additionally, an in-depth understanding of temporal trends in diarrheal disease prevalence and mortality, especially at the super-regional level, is incomplete. A recent analysis based on Global Burden of Disease (GBD) 2021 data presented important descriptions of global and regional trends, key risk factors, and long-term forecasts (15). However, it did not stratify the results by development level or evaluate localized spatial trends. In contrast, the current study provides additional analytical depth by analyzing trends in diarrheal disease prevalence and mortality between 1990 and 2021 for seven GBD super-regions, stratified by sex and HDI levels. Furthermore, we detect and compare geographical hotspots and coldspots of diarrheal disease burden and forecast trends up to 2025. These analyses are intended to enhance understanding of regional differences and temporal changes in diarrheal disease burden, with a view to informing more accurate and equitable public health initiatives for disease prevention and control.

Methods

Study design and setting

A secondary data analysis was performed. Data from the latest iteration (2021) of the GBD study were utilized (16, 17). The GBD study, a large multi-institutional endeavor, produces estimates for the burden of a wide range of diseases and injuries in over 200 countries and territories. The project provides data on an enormous number of different health metrics, such as incidence (occurrence of new cases), prevalence (number of people with a condition at a given point in time), mortality (deaths due to a condition), Years of Life Lost (YLLs) to premature death, Years Lived with Disability (YLDs), and a combined measure termed DALYs (Disability-Adjusted Life Years). This information is also stratified by sex and age group, covering the years 1980 to 2021. Importantly, the present study targeted U5 children both globally and in seven super-regions: Central and Eastern Europe and Central Asia (CEEECA; 29 countries), High-income countries (HI; 35 countries), Latin America and the Caribbean (LAC; 31 countries), North Africa and the Middle East (NAME; 21 countries), SA (5 countries), Southeast Asia, East Asia, and Oceania (SAEAO; 28 countries), and SSA (46 countries). These seven super-regions follow the GBD classification system, which groups countries with similar geographic, socioeconomic, and epidemiological characteristics to ensure consistent and comparable regional analyses (18).

Participants

The GBD analysis uses a population-based strategy, examining data from various sources including national health surveys, vital registration systems, and disease surveillance reports, as opposed to enrolling individual participants directly. An advanced statistical model is used to create a representative sample that is reflective of the whole population of a given country or region. This study design ascertains complete data by avoiding participant selection bias on the basis of particular characteristics. The analysis also includes adjustments to reflect possible confounding variables that could impact health outcomes (18).

Variable

There were two main outcome variables of interest in this study: mortality rates and prevalence rates of diarrheal disease among U5 children. These were also stratified by sex and super-regions to give an overall picture of the disease burden.

Data sources

Data were retrieved from the GBD 2021 through the Global Health Data Exchange (GHDx) GBD Results Tool (http://ghdx.healthdata.org/gbd-results-tool).

Statistical analysis

Descriptive statistics

Median (interquartile range [IQR]) was utilized to report the mortality and prevalence rate by region, sex and time period (1990-1999, 2000-2009, and 2010-2021). Moreover, percent change in rates between the first and last periods was presented.

Joinpoint regression

Joinpoint regression analysis was used to detect and assess time trends in mortality and prevalence (per 100,000 population). The joinpoint model for a sequence of observations (t₁, y₁), (t₂, y₂), ..., (tₙ, yₙ) is as follows:

yi = β₀ + β₁ti + γ₁(ti − τ₁) + ... + γₖ(ti − τₖ) + εᵢ, i = 1, ..., n where ti denotes time points (years), yi represents the burden indices, τₖ (k = 1, 2, ..., K) represents the location of potential change points (K being the number of change points), β₀, β₁, γ₁, ..., γₖ are regression coefficients, and εᵢ represents the model error term. The notation (ti − τₖ)+ signifies ti − τₖ if positive and 0 otherwise. This approach facilitates the calculation of Annual Percent Change (APC) in rates between estimated change points using a log-transformed model:

APC = 100 × (exp(β₁ + γ₁ + γ₂ + ... + γⱼ) − 1)

The Average Annual Percent Change (AAPC) of each model fit is estimated as a weighted average of APCs, where segment lengths are the weights (19). The number and location of time segments were determined by the Joinpoint software using permutation tests, allowing for a statistically robust, data-driven identification of trend changes over time. Lastly, 95% confidence intervals (CIs) for APCs and AAPCs were calculated to determine statistical significance. Joinpoint software version 5.2.0 was used for this analysis (20).

Hybrid time series projections

Projections of the future burden of U5 diarrheal disease were made using a hybrid time series forecasting model combining three complementary methods: autoregressive integrated moving average (ARIMA), exponential smoothing state space models (ETS), and artificial neural networks (ANNs). The ensemble approach was taken to maximize predictive accuracy by leveraging each method's individual strengths. Model choice was informed by minimizing the Akaike Information Criterion (AIC) and Bayesian Information Criterion (BIC) (21). The hybrid model was fitted using the forecastHybrid package in R (version 4.2.1). Annual projections of incidence, mortality, and DALY rates were generated for the years 2022 to 2025, with 95% confidence intervals (22). All predicted values were then screened to guarantee epidemiological plausibility. For example, some unrealistic outputs, such as values tending to zero, were noted and interpreted cautiously.

Longitudinal multilevel analysis

To assess the contribution of development status to long-term trends in the burden of U5 diarrheal disease, a multilevel mixed-effects regression model was used. This method accounts for the hierarchical nature of the data by including random intercepts at the super-region, region, and country levels. The age-standardized incidence rate (ASIR), DALY rate (ASDR), and mortality rate (ASMR) were each modeled as outcome measures, with fixed effects for time, development status (HDI ≥ 0.79 vs. < 0.79), and their interaction (23-25). The interaction term was included to assess whether temporal trends vary by development category. The model was as follows:

Yᵢⱼₖₜ = β₀ + β₁·Timeₜ + β₂·Developmentⱼ + β₃·(Timeₜ × Developmentⱼ) + uₖ + vⱼ + wᵢ + ξᵢⱼₖₜ

where Yijkt represents the outcome for super-region i, region j, country k, and year t; and uk, vj, and wi represent random effects at the country, region, and super-region levels, respectively.

Model fitting was performed in R (version 4.1.2) using lme4 and nlme packages, with variance components estimated by restricted maximum likelihood (REML) (26). Intraclass correlation coefficients (ICCs) were estimated to measure the proportion of total variance explained by each hierarchical level, providing information on the spatial clustering of U5 diarrheal burden in global settings (27).

Spatial analysis

Spatial patterns of the distribution of diarrhea burden in 1990 and 2021 were analyzed using the Local Moran's I statistic to identify geographic clusters of similar levels of burden. Local Moran’s I was selected for its robustness in detecting localized spatial clustering and outliers. A first-order Queen contiguity matrix was used to define neighbor relationships, with countries sharing a border or a vertex considered neighbors. Spatial analyses were performed using sf, spdep, and tmap packages in R (version 4.2.1). Statistical significance was assessed by 999 Monte Carlo permutations, with p-values < 0.05 considered to represent significant spatial autocorrelation. Countries with high diarrheal disease burden and high-burden neighboring countries were defined as hotspots, while low-burden countries with low-burden neighbors were coldspots (28, 29).

Results

Epidemiological trends of diarrheal disease prevalence rates (per 100,000) among children under five

Worldwide, the prevalence of diarrheal disease in children U5 years fell significantly from 3138.81 per 100,000 in 1990 to 885.07 in 2021 (Table 1). The median rate of prevalence decreased from 2825.35 in 1990–1999 to 1385.33 in 2010–2021, representing a percent change of –65.78%. Trends in region-specific rates indicated uniform decreases: in CEEECA, prevalence reduced by –50.97% (from 1434.97 to 491.11); in HI, the reduction was modest at –4.04% (from 966.80 to 927.74); LAC recorded the biggest fall at –73.70% (from 2323.04 to 610.97); NAME indicated a –54.75% decrease (from 3086.16 to 1396.56); SA reported a –49.42% decrease (from 3142.74 to 1589.75); SAEAO indicated a –56.54% decline (from 2324.48 to 1010.21); and SSA recorded a –57.43% decrease (from 4830.13 to 2056.11).

Table 1.

Trends in diarrhea mortality and prevalence rates (per 100,000 population) among children under five years by region, sex, and time period (1990-2021)

Measure	Region	Sex	Period			Percent change
			1990-1999	2000-2009	2010-2021
Prevalence	Global	Both	2825.35 (2722.94, 2989.32)	2386.79 (2219.46, 2536.62)	1385.33 (997.87, 1789.49)	-65.78
		Female	2861.69 (2758.81, 3021.74)	2414.02 (2246.78, 2567.73)	1443.37 (1047.90, 1834.57)	-66.22
		Male	2791.36 (2689.45, 2958.93)	2361.37 (2193.94, 2507.59)	1331.00 (951.08, 1747.28)	-65.36
	CEEECA	Both	1434.97 (1354.83, 1553.98)	1024.39 (874.68, 1197.44)	491.11 (347.20, 634.58)	-50.97
		Female	1432.23 (1358.53, 1546.73)	1033.16 (878.56, 1210.64)	483.78 (344.40, 629.35)	-49.56
		Male	1437.60 (1351.29, 1560.95)	1016.08 (871.01, 1184.90)	498.02 (349.84, 639.50)	-52.32
	HI	Both	966.80 (944.08, 997.21)	1124.73 (1072.73, 1133.77)	927.74 (683.64, 1076.88)	-4.04
		Female	979.56 (960.22, 1004.79)	1113.81 (1069.84, 1121.31)	899.99 (665.42, 1059.15)	-8.12
		Male	955.26 (928.76, 990.01)	1135.14 (1075.47, 1145.62)	954.17 (700.99, 1093.78)	-0.11
	LAC	Both	2323.04 (2083.86, 2562.74)	1417.07 (1153.26, 1678.23)	610.97 (519.33, 762.78)	-73.70
		Female	2339.11 (2077.45, 2589.32)	1397.23 (1128.33, 1652.06)	591.48 (504.48, 734.95)	-74.71
		Male	2307.49 (2090.05, 2537.00)	1436.13 (1177.16, 1703.41)	629.66 (533.58, 789.45)	-72.71
	NAME	Both	3086.16 (2855.57, 3369.00)	2334.02 (2284.86, 2531.22)	1396.56 (1014.21, 1930.54)	-54.75
		Female	3024.78 (2801.31, 3293.17)	2270.06 (2230.55, 2473.34)	1431.88 (1059.14, 1917.44)	-52.66
		Male	3144.46 (2907.03, 3441.18)	2394.49 (2343.47, 2586.04)	1363.22 (971.81, 1942.92)	-56.65
	SA	Both	3142.74 (2950.42, 3496.87)	2615.15 (2430.51, 2719.90)	1589.75 (1114.01, 1955.29)	-49.42
		Female	3317.39 (3096.54, 3689.56)	2700.35 (2500.12, 2827.52)	1712.73 (1259.13, 2044.24)	-48.37
		Male	2980.93 (2815.48, 3317.77)	2536.73 (2366.45, 2620.79)	1476.62 (999.79, 1873.43)	-50.46
	SAEAO	Both	2324.48 (2212.56, 2463.49)	1837.63 (1637.17, 2002.80)	1010.21 (853.21, 1280.09)	-56.54
		Female	2279.06 (2185.32, 2407.39)	1819.06 (1622.39, 1989.29)	1019.97 (860.69, 1282.36)	-55.25
		Male	2365.22 (2236.88, 2514.16)	1854.18 (1650.36, 2014.84)	1001.45 (846.26, 1278.05)	-57.66
	SSA	Both	4830.13 (4639.68, 4993.80)	3868.10 (3577.42, 4229.57)	2056.11 (1367.60, 2809.98)	-57.43
		Female	4819.89 (4642.31, 4968.38)	3917.03 (3639.99, 4257.57)	2150.80 (1431.80, 2902.51)	-55.38
		Male	4840.11 (4637.13, 5018.64)	3820.65 (3516.63, 4202.44)	1964.17 (1305.34, 2720.03)	-59.42
Deaths	Global	Both	230.08 (208.88, 250.64)	155.16 (138.30, 174.44)	78.98 (61.59, 100.44)	-79.99
		Female	224.18 (204.62, 243.62)	153.91 (138.22, 171.74)	75.35 (58.62, 98.68)	-80.65
		Male	235.60 (212.86, 257.22)	156.32 (138.37, 176.96)	82.39 (64.38, 102.09)	-79.49
	CEEECA	Both	44.42 (40.82, 48.00)	19.11 (14.77, 25.81)	8.89 (8.12, 10.68)	-65.67
		Female	42.15 (38.86, 45.29)	17.73 (13.70, 24.39)	8.16 (7.51, 9.86)	-66.39
		Male	46.72 (42.70, 50.60)	20.41 (15.77, 27.15)	9.58 (8.69, 11.45)	-65.03
	HI	Both	1.74 (1.69, 1.99)	1.48 (1.42, 1.53)	1.02 (0.87, 1.11)	-41.29
		Female	1.52 (1.50, 1.82)	1.31 (1.28, 1.35)	0.94 (0.81, 1.02)	-38.21
		Male	1.94 (1.87, 2.15)	1.63 (1.56, 1.72)	1.10 (0.92, 1.20)	-43.28
	LAC	Both	122.72 (90.71, 160.89)	50.32 (40.13, 60.02)	23.34 (19.66, 27.37)	-80.98
		Female	112.05 (82.90, 147.71)	45.49 (36.33, 54.57)	20.59 (17.44, 24.40)	-81.63
		Male	133.05 (98.26, 173.65)	54.96 (43.77, 65.27)	25.98 (21.80, 30.21)	-80.47
	NAME	Both	139.50 (121.37, 159.89)	77.35 (60.81, 91.43)	29.23 (23.58, 36.03)	-79.05
		Female	128.98 (109.90, 151.32)	67.54 (52.54, 80.36)	25.80 (21.01, 30.70)	-80.00
		Male	149.49 (132.26, 168.04)	86.63 (68.62, 101.91)	32.46 (26.00, 41.06)	-78.29
	SA	Both	325.71 (278.45, 370.83)	165.42 (140.49, 198.21)	70.82 (48.85, 96.44)	-78.26
		Female	359.14 (309.27, 406.05)	186.75 (160.03, 222.01)	75.09 (48.99, 107.23)	-79.09
		Male	294.73 (250.00, 338.10)	145.78 (122.50, 176.30)	66.89 (48.72, 86.50)	-77.31
	SAEAO	Both	96.46 (77.75, 120.14)	43.90 (34.33, 55.73)	16.54 (12.73, 21.77)	-82.85
		Female	90.40 (72.73, 113.60)	41.68 (32.55, 52.41)	15.13 (11.77, 20.04)	-83.26
		Male	101.90 (82.24, 126.05)	45.87 (35.92, 58.70)	17.81 (13.58, 23.32)	-82.52
	SSA	Both	589.55 (541.78, 625.97)	425.01 (383.99, 474.89)	219.95 (172.80, 278.97)	-62.69
		Female	511.73 (480.40, 537.34)	393.92 (360.69, 433.25)	201.03 (160.77, 260.35)	-60.72
		Male	664.48 (601.42, 712.54)	455.15 (406.63, 515.23)	238.33 (184.46, 297.07)	-64.13

Note: Median (interquartile range [IQR]) was used to present the mortality and prevalence rate by region, sex and time period. Table also includes the percent change in rates between two periods. Percent change was calculated using the formula ((Rate in 2010-2021 – Rate in 1990-1999) / Rate in 1990-1999) × 100 and represents the relative change between the two periods. Abbreviation: Central and Eastern Europe and Central Asia (CEEECA), High-income countries (HI), Latin America and the Caribbean (LAC), North Africa and the Middle East (NAME), South Asia (SA), Southeast Asia, East Asia, and Oceania (SAEAO), Sub-Saharan Africa (SSA).

Epidemiological trends of diarrheal disease mortality rates (per 100,000) among children under five

The mortality rate due to diarrheal disease in children U5 years fell dramatically worldwide from 263.95 per 100,000 in 1990 to 51.72 in 2021, according to Table 1. The median global mortality rate fell from 230.08 in 1990–1999 to 78.98 in 2010–2021, a percent change of –79.99%. All regions recorded significant reductions: in CEEECA, –65.67% (44.42 to 8.89); HI, rates were already low and fell further by –41.29% (1.74 to 1.02); LAC experienced an –80.98% reduction (122.72 to 23.34); NAME had a –79.05% decline (139.50 to 29.23); SA registered a –78.26% reduction (325.71 to 70.82); SAEAO, the most dramatic decline at –82.85% (96.46 to 16.54); and SSA, while having the highest starting figures, managed a –62.69% reduction (589.55 to 219.95).

APC in prevalence rate of diarrheal disease in children under five years by region and sex (1990-2021)

Worldwide, the prevalence of diarrhea among U5 children exhibited a prominent decrease throughout the study period for both sexes (Figure 1, Table S1). The most significant drops were recorded in 2011–2015 (APC –7.07) and 2015–2019 (APC –9.60). Males and females both presented similar trends, with the most abrupt decline in 2015–2019 (APCs –9.87 and –9.34, respectively). Region-wise, CEEECA initially experienced an increase (1990–1993), followed by significant downturns, particularly in 2010–2015 (APC –6.98) and 2015–2018 (APC –10.24). In HI, the trend was fluctuating, with increases in 1994–2005 and the most prominent fall in 2015–2019 (APC –8.99). LAC presented steady decreases, with a peak in 2006–2009 (APC –8.88). NAME experienced even decreases, most evident in 2011–2018 (APC –9.33) and 2018–2021 (APC –10.65). SA exhibited ongoing decline, with the highest reduction in 2015–2018 (APC –11.21). SAEAO also declined steadily, barring a slight increase in 2018–2021 (APC 0.38), with the highest drop in 2012–2018 (APC –7.05). SSA presented a definite decreasing trend, particularly in 2015–2019 (APC –11.77).

Figure 1.

Joinpoint regression models analyzing trends in diarrheal disease prevalence rates across global and super regions from 1990 to 2021

APC in mortality rate of diarrheal disease in children under five years by region and sex (1990-2021)

Joinpoint regression analysis of 1990–2021 demonstrated a significant overall decrease in diarrhea-associated mortality in U5 children, with four principal declining segments: 1990–1997 (APC –3.18), 1997–2007 (–4.01), 2007–2011 (–5.72), and 2011–2021 (–7.39). These trends were similar in both sexes (Figure 2, Table S1). At the regional level, CEEECA presented with an initial increase (1990–1994; APC 3.38), followed by steep reductions, the most pronounced being during 1997–2006 (APC –10.08). HI regions witnessed fluctuating but overall declining tendencies, with steep reductions in 1990–1993 (APC –9.14) and 2018–2021 (APC –7.99). LAC demonstrated consistent declines in all periods, with the most pronounced in 1990–2001 (APC –9.74). NAME exhibited a steady downward trend, with appreciable reductions in 2005–2014 (APC –9.39) and 2017–2021 (APC –9.73), despite a temporary slowing in 2014–2017. SA demonstrated a continuous decline, with a peak in 2018–2021 (APC –13.31), while SAEAO also sustained a steady decline throughout. In SSA, the mortality decreased across four segments, with the most pronounced decline in 2017–2021 (APC –8.09)

Figure 2.

Joinpoint regression models analyzing trends in diarrheal disease mortality rates across global and super regions from 1990 to 2021

Comparing AAPC in mortality and prevalence rates of diarrheal disease in children under five years by region and sex (1990-2021)

From 1990 to 2021, significant declines (p < 0.001) in diarrhea-related mortality and prevalence rates among children under five were observed across all regions (CEEECA, HI, LAC, NAME, SA, SAEAO, SSA) and for both sexes. The global AAPC was −5.15% (95% CI: −5.19, −5.10) for mortality and −3.98% (95% CI: −4.03, −3.94) for prevalence. Declines were steeper in high-burden regions such as LAC, SA, SSA, and SAEAO compared to HI regions. In most regions, mortality declined faster than prevalence. Although sex differences were statistically significant, they were minimal. The SAEAO region showed the highest AAPC for mortality reduction, while CEEECA had the highest for prevalence (Table 2 and Table S2 in the Supplementary File).

Table 2.

AAPC in mortality rate and prevalence rate (per 100,000 population) of diarrhea disease in children under five years by region and sex (1990-2021)

Region	Sex	Prevalence rate		Death rate
		APCC (95% CI)	p-value	AAPC (95% CI)	p-value
Global	Both	-3.98* (-4.03, -3.94)	< 0.001	-5.15* (-5.19, -5.10)	<0.001
	Female	-3.83* (-3.88, -3.79)	< 0.001	-5.26* (-5.32, -5.19)	<0.001
	Male	-4.13* (-4.18, -4.09)	< 0.001	-5.08* (-5.11, -5.05)	<0.001
CEEECA	Both	-5.51* (-5.56, -5.47)	< 0.001	-5.43* (-5.50, -5.36)	<0.001
	Female	-5.49* (-5.54, -5.43)	< 0.001	-5.45* (-5.57, -5.32)	<0.001
	Male	-5.53* (-5.57, -5.49)	< 0.001	-5.41* (-5.49, -5.33)	<0.001
HI	Both	-1.40* (-1.45, -1.35)	< 0.001	-3.79* (-4.04, -3.54)	< 0.001
	Female	-1.50* (-1.55, -1.45)	< 0.001	-3.46* (-3.69, -3.18)	< 0.001
	Male	-1.31* (-1.36, -1.26)	< 0.001	-3.95* (-4.22, -3.71)	< 0.001
LAC	Both	-5.44* (-5.46, -5.43)	< 0.001	-7.61* (-7.83, -7.39)	< 0.001
	Female	-5.58* (-5.60, -5.56)	< 0.001	-7.72* (-7.97, -7.49)	< 0.001
	Male	-5.32* (-5.34, -5.30)	< 0.001	-7.52* (-7.74, -7.30)	< 0.001
NAME	Both	-4.76* (-4.83, -4.70)	< 0.001	-6.99* (-7.09, -6.89)	< 0.001
	Female	-4.63* (-4.69, -4.58)	< 0.001	-7.19* (-7.36, -7.08)	< 0.001
	Male	-4.94* (-5.00, -4.88)	< 0.001	-6.87* (-6.97, -6.78)	< 0.001
SA	Both	-4.22* (-4.30, -4.14)	<0.001	-7.54* (-7.62, -7.47)	<0.001
	Female	-3.93* (-4.03, -3.83)	<0.001	-8.00* (-8.14, -7.88)	<0.001
	Male	-4.52* (-4.59, -4.46)	<0.001	-7.05* (-7.12, -7.00)	<0.001
SAEAO	Both	-3.52* (-3.56, -3.48)	< 0.001	-7.71* (-7.77, -7.66)	< 0.001
	Female	-3.42* (-3.46, -3.37)	< 0.001	-7.76* (-7.81, -7.71)	< 0.001
	Male	-3.59* (-3.62, -3.55)	< 0.001	-7.66* (-7.72, -7.62)	< 0.001
SSA	Both	-4.76* (-4.81, -4.73)	< 0.001	-4.82* (-4.89, -4.76)	< 0.001
	Female	-4.60* (-4.64, -4.56)	< 0.001	-4.54* (-4.60, -4.48)	< 0.001
	Male	-4.92* (-4.97, -4.89)	< 0.001	-5.00* (-5.05, -4.95)	< 0.001

Abbreviation: Central and Eastern Europe and Central Asia (CEEECA), High-income countries (HI), Latin America and the Caribbean (LAC), North Africa and the Middle East (NAME), South Asia (SA), Southeast Asia, East Asia, and Oceania (SAEAO), Sub-Saharan Africa (SSA), Average Annual Percent Change (AAPC), Confidence Interval (CI).

Projected trends in diarrheal disease burden among children under five by sex and region through 2025

As shown in Table 3, Figure 3, and Figure 4, the global burden of diarrheal disease among U5 children is projected to decline between 2022 and 2025. Prevalence is expected to drop from 821.9 (95% CI: 762.0–869.3) to 631.3 (243.5–844.5), and mortality from 47.5 (44.1–51.2) to 36.6 (24.9–46.5) per 100,000 population. Regional disparities persist, with SSA bearing the highest burden by 2025, while HI regions remain lowest. Sex-based differences continue, with females generally showing higher prevalence but lower mortality. Full annual projections are available in Table S3 (Supplementary File).

Table 3.

Projected prevalence and mortality of diarrheal disease among children under 5 by sex and super-region (2022–2025)

Measure	Location	Year	Female (95% CI)	Male (95% CI)	Both (95% CI)
Prevalence	Global	2022	881.50 (836.25, 921.57)	788.03 (752.82, 822.37)	821.93 (762.03, 869.29)
		2025	731.98 (431.12, 886.98)	670.81 (432.88, 808.94)	631.26 (243.49, 844.51)
	SAEAO	2022	845.81 (810.02, 890.09)	829.35 (797.33, 872.56)	835.45 (798.35, 880.75)
		2025	755.64 (456.15, 1014.79)	743.47 (448.50, 989.14)	745.33 (435.42, 1001.43)
	CEEECA	2022	243.23 (207.30, 284.88)	247.29 (220.20, 280.57)	244.80 (213.99, 282.56)
		2025	136.94 (-49.39, 264.79)	141.81 (-8.02, 248.82)	136.82 (-27.02, 254.89)
	HI	2022	608.68 (575.26, 670.18)	637.23 (607.87, 676.00)	623.10 (583.63, 690.75)
		2025	555.96 (336.82, 698.75)	573.36 (340.85, 787.25)	568.58 (339.69, 737.90)
	LAC	2022	462.04 (438.49, 483.05)	488.15 (469.00, 502.22)	475.08 (454.49, 492.66)
		2025	448.46 (325.35, 590.71)	466.78 (371.20, 568.04)	456.32 (347.69, 577.54)
	NAME	2022	709.71 (631.95, 783.02)	731.73 (638.43, 796.89)	712.87 (638.60, 789.05)
		2025	492.29 (185.10, 743.82)	633.52 (266.16, 988.63)	550.57 (249.02, 799.74)
	SA	2022	1067.94 (962.71, 1182.92)	776.45 (702.72, 866.99)	890.41 (796.96, 978.21)
		2025	890.54 (478.02, 1320.08)	570.27 (133.35, 938.61)	643.84 (231.22, 960.55)
	SSA	2022	1081.28 (998.65, 1167.20)	987.61 (938.97, 1043.76)	1033.53 (984.90, 1092.58)
		2025	831.15 (323.66, 1246.76)	759.12 (412.50, 998.86)	794.05 (436.54, 1030.04)
Mortality	Global	2022	44.41 (40.83, 48.49)	50.06 (46.82, 54.10)	47.49 (44.07, 51.19)
		2025	32.03 (17.03, 43.53)	39.41 (29.97, 49.53)	36.59 (24.95, 46.51)
	SAEAO	2022	10.90 (9.50, 12.32)	12.31 (11.01, 13.84)	11.66 (10.32, 13.13)
		2025	10.73 (2.53, 21.02)	12.08 (3.78, 22.59)	11.48 (3.42, 22.10)
	CEEECA	2022	7.51 (5.44, 9.71)	7.94 (5.37, 11.24)	7.80 (5.58, 10.63)
		2025	7.73 (-0.13, 15.73)	7.61 (-2.92, 16.17)	7.89 (-0.82, 16.46)
	HI	2022	0.63 (0.52, 0.75)	0.71 (0.57, 0.85)	0.67 (0.54, 0.79)
		2025	0.49 (0.17, 0.81)	0.54 (0.18, 0.93)	0.50 (0.22, 0.86)
	LAC	2022	15.08 (8.40, 21.38)	18.56 (10.95, 25.71)	16.88 (9.73, 23.69)
		2025	13.99 (-5.99, 29.96)	17.08 (-6.24, 34.91)	15.65 (-6.08, 32.43)
	NAME	2022	15.82 (12.39, 19.14)	18.43 (14.17, 22.28)	17.32 (13.58, 20.65)
		2025	12.28 (-0.30, 21.81)	13.78 (-2.86, 24.34)	13.32 (-0.29, 22.09)
	SA	2022	30.47 (24.22, 37.86)	34.12 (28.48, 42.46)	32.26 (26.83, 39.97)
		2025	21.86 (-6.87, 47.09)	25.39 (-2.05, 44.19)	23.27 (-3.75, 44.20)
	SSA	2022	121.72 (106.01, 137.51)	135.95 (118.65, 156.27)	129.22 (115.06, 146.84)
		2025	93.36 (49.07, 131.08)	95.52 (53.91, 146.23)	94.69 (58.04, 136.23)

Note: Projected diarrhea prevalence and mortality rates (per 100,000) among children under 5 were estimated by sex and GBD super-region using a hybrid time series forecasting model that combined ARIMA (Autoregressive Integrated Moving Average), ETS (Exponential Smoothing State Space), and ANN (Artificial Neural Network) methods. Forecasts were generated for the years 2022 through 2025, and each estimate was accompanied by a 95% confidence interval (CI). Super-region abbreviations used in the table include: SAEA-O (Southeast Asia, East Asia, and Oceania), CEEECA (Central Europe, Eastern Europe, and Central Asia), HI (High-Income), LAC (Latin America and Caribbean), NAME (North Africa and Middle East), SA (South Asia), and SSA (Sub-Saharan Africa).

Figure 3.

Trends and forecasts of diarrhea prevalence among children under five years by sex across global burden of disease super-regions (1990–2025)

Figure 4.

Trends and forecasts of diarrhea mortality among children under five years by sex across global burden of disease super-regions (1990–2025)

Long-term impact of human development index on diarrhea burden trend

Table 4 indicates that while both prevalence and mortality decreased worldwide, the slope of decrease was much steeper in less developed nations. More developed nations had lower baseline prevalence (β = −754.66; 95% CI: −850.39 to −658.92; p < 0.001) but slower yearly decreases (interaction β = 79.76; 95% CI: 76.03 to 83.48; p < 0.001). The same pattern held for mortality (interaction β = 7.50; 95% CI: 7.05 to 7.94; p < 0.001). Figure 5 depicts these disparities, with steeper drops but higher overall rates in less developed areas. Substantial proportions of variance were accounted for at the country (43% for prevalence; 51% for mortality) and super-region levels (28% and 36%, respectively), indicating geographic disparities.

Table 4.

Multilevel longitudinal analysis of diarrhea burden (prevalence and mortality) in children under five by development status (1990–2021)

Metric	Predictors	Parameters	Estimates (95% CI)	P-value
Prevalence	Intercept	------	3028.23 (2285.30, 3771.16)	<0.001
	Time	------	-91.46 (-93.30, -89.63)	<0.001
	Development status	More developed vs. less developed	-754.66 (-850.39, -658.92)	<0.001
	Time * Development status	Time * (More developed vs. less developed)	79.76 (76.03, 83.48)	<0.001
	Random effects
		$\mathrm{Residual\; Variance}{(\mathrm{\sigma }}^2)$	247455.56 (238217.16, 256890.57)
		Country level (${\tau }_{00\mathit{Country}:\mathit{Region}:\mathit{Super\; Region}})$	540200.47 (438259.90, 675563.47)	<0.001
		Region level (${\tau }_{00\mathit{Region}:\mathit{Super\; Region}}$)	310273.93 (115504.74, 823886.43)	<0.001
		Super Region level (${\tau }_{00\mathit{Super\; Region}})$	829060.07 (209480.54, 2796451.04)	<0.001
	ICC (%)
	Country level	43.02
	Region level	16.10
	Super Region level	28.03
Mortality	Intercept	------	201.49 (111.55, 291.43)	0.004
	Time	------	-7.88 (-8.10, -7.66)	<0.001
	Development status	More developed vs. less developed	-62.30 (-73.69, -50.91)	<0.001
	Time * Development status	Time * (More developed vs. less developed)	7.50 (7.05, 7.94)	<0.001
	Random effects
		$\mathrm{Residual\; Variance}{(\mathrm{\sigma }}^2)$	3500.92 (3370.22, 3634.41)
		Country level (${\tau }_{00\mathit{Country}:\mathit{Region}:\mathit{Super\; Region}})$	9903.77 (8068.52, 12312.77)	<0.001
		Region level (${\tau }_{00\mathit{Region}:\mathit{Super\; Region}}$)	130.57 (0.00, 1782.40)	<0.001
		Super Region level (${\tau }_{00\mathit{Super\; Region}})$	14088.88 (4659.44, 43859.66)	<0.001
	ICC (%)
	Country level	51.00
	Region level	0.47
	Super Region level	35.85

Note: A multilevel longitudinal analysis was conducted to evaluate the effects of time, development status (more developed vs. less developed countries), and their interaction on diarrhea prevalence and mortality among children under five (U5) from 1990 to 2021. A linear mixed-effects model was applied, with random intercepts specified at the country, region, and super-region levels to account for the hierarchical data structure. Fixed-effect estimates were reported along with 95% confidence intervals (CIs) and p-values. Variance components were estimated for each level, and the proportion of total variance explained at each level was expressed using Intraclass Correlation Coefficients (ICCs).

Figure 5.

Temporal trends in predicted diarrhea prevalence (a) and mortality (b) among children under five by human development index (HDI) classification, 1990–2021

Shifting spatial patterns of diarrhea prevalence and mortality among children under five (1990–2021)

Figure 6 illustrates the spatial distribution of diarrhea prevalence. In 1990, significant hotspots were concentrated in SSA (e.g., Nigeria, Chad, Ethiopia), SA (e.g., Bangladesh), and parts of the Caribbean, with extremely high rates in Angola (6,030) and Madagascar (6,402 per 100,000). Coldspots appeared in countries like Austria, Bhutan, and Sri Lanka. By 2021, while hotspots persisted in areas such as South Sudan and the Philippines, new clusters emerged in some developed countries (e.g., Japan, Belgium). Coldspots in 2021 remained primarily in HI and island nations, including Australia, Brunei, and Singapore.

Figure 6.

Global distribution and regional disparities in the burden of diarrheal disease, with emphasis on high-risk zones in sub-Saharan Africa

In 1990, hotspots of diarrhea-related mortality among U5 children were concentrated in SSA, notably in Nigeria (1,024), Chad (1,212), and Niger (1,364 per 100,000), with significant spatial clustering (p < 0.05). Other high-burden areas included Bangladesh and Ethiopia. Coldspots, by contrast, were found in countries like Mauritius, Zimbabwe, and Libya. By 2021, although mortality declined globally, hotspots persisted in SSA, particularly in Chad (560), South Sudan (449), and Lesotho (299), reflecting ongoing disparities. Meanwhile, new coldspots emerged in countries like Equatorial Guinea, Seychelles, and Sudan, indicating progress in reducing child diarrhea mortality. See Tables 4–7 in the Supplementary File for details.

Discussion

This study assessed global and regional trends in U5 diarrheal disease burden from 1990 to 2021. Overall, the burden declined substantially, with global prevalence decreasing by ~60% and mortality by ~80%. By 2010 to 2021, prevalence fell below 1,400 cases and mortality below 100 deaths per 100,000 U5 children. Girls generally had higher prevalence but lower mortality than boys. Notably, annual mortality reductions outpaced declines in prevalence. At baseline (1990 to 1999), SSA, SA, and NAME bore the highest burden (prevalence >3,000 and mortality >100 per 100,000), yet also showed the greatest improvements in the last decade, achieving over 50% reduction in prevalence and over 60% in mortality. However, these regions still lagged behind HI and CEEECA, which maintained the lowest rates throughout the study. Encouragingly, all seven super-regions exhibited declining trends. The fastest annual mortality reductions (AAPC > –7.5%) occurred in SAEAO, LAC, and SA, while CEEECA and LAC had the steepest prevalence declines (AAPC > –5%). Although sex disparities were modest, they remained statistically significant. Most regions experienced their greatest declines between 2015 and 2021. These findings were corroborated by hybrid model projections, which indicated continued reductions through 2025, particularly in SSA and SA. Spatial analyses revealed persistent hotspots in SSA and emerging clusters in some HI countries, possibly due to improved detection or internal disparities. Lastly, multilevel models confirmed that while developed countries had lower initial burden, less developed countries experienced steeper declines, underscoring both progress and enduring global inequalities.

Our findings indicated that girls had greater prevalence, but boys experienced greater mortality. Although some studies corroborate the lack of a substantial sex difference (30), others indicate that such differences may be due to differential care-seeking or sociocultural treatment prioritization patterns (31). On the other hand, the global reduction is consistent with previous research attributing it to better water, sanitation, and hygiene (WASH) services, nutrition, and rotavirus vaccination, with WASH alone responsible for as much as 88% of the reduction in mortality among U5 children (32-38). Likewise, research has consistently underscored the protective effects of exclusive breastfeeding, maternal hygiene, safe water sources, and improved sanitation infrastructure in reducing the risk for diarrheal disease (39, 40).

Moreover, the reduction in mortality from U5 diarrheal disease dropped faster than the decrease in prevalence, citing ongoing difficulties in breaking the transmission cycle of diarrheal pathogens. This discrepancy is likely due to improved clinical care and greater access to oral rehydration therapy (ORT) and zinc supplementation, which reduce deaths but do not prevent new infections. (41). High prevalence remains strongly linked to poverty, poor sanitation, and low maternal education. Even where infrastructure has improved, behavioral and environmental deficits continue to sustain transmission (31, 42-44). Additional biological and maternal risk factors, such as child undernutrition, lack of exclusive breastfeeding, and maternal history of diarrhea, further elevate disease risk (45-49). Pathogen-specific studies have identified enterotoxigenic and enteropathogenic E. coli as major contributors, often with high antibiotic resistance, while rotavirus remains the most common global cause. However, shifting genotype patterns following vaccine rollout underline the need for ongoing surveillance and vaccine adaptation (37, 50). Persistently high burden in certain regions highlights the urgent need for sustained, context-sensitive investments in infrastructure, behavior change, and policy interventions, particularly in vulnerable areas like flood-prone communities where fecal contamination exacerbates disease spread. (43). Taken together, these findings emphasize that while clinical interventions have been effective in reducing mortality, long-term control of diarrheal disease requires a multifaceted strategy that combines targeted public health policies, improved infrastructure, and sustained behavioral and environmental change, especially in high-risk settings.

Despite global progress in reducing the burden of diarrheal disease, SSA continues to experience persistently high prevalence and mortality among U5 children. Numerous studies highlight the complex interplay of socioeconomic, environmental, and behavioral factors that sustain this burden. He et al. demonstrated a strong inverse association between wealth quintile and diarrhea prevalence, emphasizing that children from lower-income households are significantly more vulnerable. Their analysis across 36 countries revealed that lack of access to improved water (RR = 1.05) and sanitation (RR = 1.04) remain key drivers of infection (42). Environmental factors such as exposure to contaminated water, poor waste management, and animal access to water sources exacerbate this vulnerability, especially in flood-prone regions (43). In Ghana, Bandoh et al. found that despite reported improvements in WASH infrastructure, diarrhea prevalence remained high, suggesting that the quality and sustainability of these interventions may be inadequate (44). The role of maternal education and hygiene practices is also critical; Solomon et al. and Edward et al. reported that poor maternal handwashing practices and lack of treated drinking water significantly increased diarrhea risk (51, 52). Moreover, Ahmed et al. identified four modifiable risk factors including unclean cooking fuel, delayed breastfeeding, household poverty, and low maternal education which collectively accounted for 34% of diarrhea cases in SSA (31). These findings highlight the complex and entrenched nature of the problem, emphasizing the need for integrated, context-specific interventions that combine infrastructure development with education and behavior change to achieve lasting reductions in prevalence and mortality in SSA.

HI and CEECA regions consistently report the lowest diarrheal disease burden among U5 children, largely attributable to strong health systems, advanced infrastructure, and high standards of living. HI countries reported only 490 diarrheal deaths among U5 children in 2021, according to Black et al. (53), in stark contrast with the hundreds of thousands from low-income regions. In these environments, the key pathogens responsible for diarrhea mortality, including rotavirus and norovirus, are largely contained through routine vaccination and expedient clinical management. Whereas, for example, rotavirus is still a prevalent pathogen worldwide, its burden has dropped significantly in HI countries since the introduction of vaccines into national immunization schedules (37). Even norovirus contributes more significantly in these parts of the world compared to LMICs, where bacterial pathogens such as Shigella and E. coli play a greater role. Improved housing conditions, stringent food safety regulations, access to treated drinking water, and universal health coverage also contribute to the lower burden in HI and CEECA by collectively reducing incidence and case fatality. Despite disparities in access that persist worldwide, the experience of these regions highlights the value of combining vaccination, hygiene, and public health infrastructure for sustained control of diarrheal illness. Ongoing surveillance remains critical to track emerging genotypes and assure high vaccine coverage, particularly among disadvantaged populations within otherwise low-burden settings.

To perform our analysis, we used a combined forecasting model that included ARIMA, ETS, and artificial neural networks to predict diarrhea cases and deaths in children under five between 2022 and 2025. This approach allowed us to use the strengths of each method, resulting in reliable predictions with low uncertainty. Hybrid models like ARIMA–ETS–ANN are preferred over single models for their lower forecast uncertainty and combined strengths, with many studies showing improved accuracy in public health forecasting (54-56). We also examined differences between boys and girls and included all seven global regions from the GBD study, providing detailed insights for regional health planning. In comparison, Chu et al. (2024) used a different method called the Bayesian age-period-cohort model, which was good for long-term trends but only gave global-level estimates and showed very wide uncertainty ranges. For example, their predicted death rate in 2035 ranged from 0 to over 118 per 100,000, making it less practical for local decision-making (40). Our projections showed a continued global decline in both cases and deaths through 2025, with the highest burden remaining in SA and SSA, and the lowest in HI regions.

One of the key innovations of our analysis is the application of longitudinal multilevel modeling, which, in contrast to standard regression techniques, explicitly models the hierarchical nature of global health data. This approach permitted us to estimate diarrheal disease prevalence and child mortality while accounting for nested variation at the country, region, and super-region levels. To our knowledge, this represents one of the first such applications to GBD-based data on diarrheal disease, allowing for a more detailed understanding of both temporal trends and geographic disparities. The analysis showed that although the global burden decreased overall, less-developed countries saw steeper annual declines yet still maintained higher absolute rates. Interestingly, more than 40 percent of the variability in prevalence and more than 50 percent in mortality was due to differences at the country level, underscoring residual geographic heterogeneity. These results illustrate the added value of multilevel modeling in global health research and underscore the necessity of geographically targeted interventions to hasten progress in low-development environments.

In spatial epidemiology, hotspots are locations in which high values cluster spatially more intensively than would be expected by chance, and coldspots are locations with notably low values that cluster spatially. Using Local Moran's I and Z-scores, our spatial analysis identified distinct patterns for the prevalence of and mortality due to diarrhea among children below the age of five in 1990 and 2021. In the case of mortality, we observed hotspot consistency across most SSA countries, such as Chad and South Sudan, signifying locations of persistently high risk. Coldspots for mortality were observed in countries such as Seychelles, Mauritius, and Libya, representing more favorable conditions. For prevalence, hotspots were initially concentrated in SSA and parts of SA, while in 2021, new clusters were observed in more developed regions, potentially reflecting heterogeneity within-country or differences in reporting. The shift suggests new spatial trends in disease reporting or health access. Overall, our study is the first to assess spatial clustering of diarrhea burden over time using GBD data, offering critical information for geographically targeted interventions in both high-burden and emerging-risk areas.

Limitation

While this study offers a comprehensive assessment of global and regional trends in the burden of diarrheal disease among U5 children, it is not without limitations. First, our reliance on secondary data from the GBD 2021 study means that all findings are based on modeled estimates, which, while robust, are subject to certain assumptions and methodological constraints. The accuracy and completeness of underlying data sources can vary substantially by country, especially in low-income settings where vital registration and disease surveillance systems may be limited. This introduces potential biases and should be considered when interpreting trends. To address this, the GBD employs rigorous statistical modeling techniques and extensive data triangulation to harmonize estimates across heterogeneous sources. Additionally, our study mitigated these constraints by applying robust analytic techniques, including joinpoint regression to capture temporal shifts, hybrid forecasting to enhance predictive validity, and multilevel modeling to account for nested data structures and contextual variation. It is also important to consider the uncertainty intervals presented with each estimate, as they reflect the inherent variability and potential imprecision in the data. While subnational heterogeneity and individual-level determinants could not be captured, we used sex- and region-stratified analyses and spatial autocorrelation methods to reveal important disparities. By integrating these methodological safeguards, the study aimed to ensure credible, policy-relevant insights despite the inherent constraints of secondary data analysis.

Conclusion

This research presents a comprehensive and methodologically robust analysis of under-five diarrheal disease burden at global, regional, and national levels from 1990 to 2021. Using joinpoint regression, hybrid forecasting, spatial analysis, and longitudinal multilevel modeling, we offer a high-resolution image of trends, inequities, and projections. While declines in global prevalence and mortality are encouraging, persistent geographic and socioeconomic inequities, particularly in SSA and SA, indicate that important challenges remain. In the future, the global response must shift from reactive treatment to proactive prevention through sustained investment in WASH infrastructure, universal vaccination against rotavirus, and locally adapted, community-based health education. Strengthening data systems, equity-promoting policy, and long-term financing are critical to bridging the remaining gaps. Preventable U5 diarrhea deaths can be averted, but only through collective, inclusive, and context-specific action.

Conflict of interests

The authors declare that they have no competing interests.