Cost of diarrhoea: a household perspective from seven countries in the Global Enteric Multicentre Study (GEMS)

Abstract

Background

The burden of diarrhoeal diseases is considerable in South Asia, as well as in sub-Saharan Africa. Its economic impact is more profound in resource-limited settings like low-income and middle-income countries (LMICs). In this study, we aimed to estimate the direct and indirect costs reported by the caregivers of participants from seven LMICs.

Methods

The current study is a secondary analysis using data from the multicentre, Global Enteric Multicentre Study, which enrolled under-5 children. This prospective case–control study was conducted in seven sites, all of which were in LMICs. After adjustment for inflation, cost data were collected from the caregivers and converted into international dollar (I$). Quantile regression models were developed after adjusting for age, sex and country.

Results

This study analysed data from 4592 participants. The median (IQR) total direct cost (TDC) and total indirect cost (TIC) were I$8.4 (I$11.0) and I$10.2 (I$14.3), respectively. Statistically significant differences were found across continents for multiple variables. The highest median TDC and TIC were in Bangladesh (I$13.6 and I$23.2, respectively), while mozambique reported the lowest (I$0.4 and I$4.9, respectively), with medication accounting for 60.9% of TDC. Quantile regression analysis showed TDC was positively associated with factors like family size, urban residence, moderate-to-severe disease, caregiver education and use of rehydration methods, while treated drinking water and overweight status were negatively associated. TIC was significantly associated with seeking prior care.

Conclusions

The indirect cost of diarrhoea was higher, which indicates the impact of lost productivity due to the disease. Bolstering the healthcare financing systems, ensuring affordable medication using pricing regulation, subsidising treatment packages, promoting the water, sanitation and hygiene (WASH) initiative, promoting and practising standard case management, and timely healthcare-seeking can reduce the economic burden.

WHAT IS ALREADY KNOWN ON THIS TOPIC

• Diarrhoeal diseases place a heavy burden on both South Asia and sub-Saharan Africa.

• The economic effects are severe in low-income and middle-income countries, where resources are limited.

WHAT THIS STUDY ADDS

• The median direct cost was I$8.4 and the indirect cost was I$10.2 among the seven participating countries.

• The indirect cost was higher than the direct cost.

• Medication expenditure constituted the largest share of total direct cost, accounting for 60.9%.

HOW THIS STUDY MIGHT AFFECT RESEARCH, PRACTICE OR POLICY

• Expand oral rehydration salts use and WASH programmes is needed to prevent diarrhoeal diseases.

• Promoting timely care-seeking through public awareness is crucial.

• Strengthening healthcare financing can reduce economic burden.

Introduction

Among children under the age of 5 years, diarrhoea is still one of the leading causes of mortality and morbidity.¹ Due to being preventable and treatable with measures like oral rehydration salts (ORS), the overall mortality due to diarrhoea has seen a significant decline (approximately 55%) since the 2000s.² The burden of diarrhoeal diseases is considerable in South Asia and sub-Saharan Africa (SSA).³ The countries from these two geographical locations are responsible for 90% of the global diarrhoea deaths.⁴

Apart from mortality, the economic impact of the disease on the households and the health system is also crucial. In resource-limited settings like low-income and middle-income countries (LMICs), the effect is more profound.² Studies have proven that this economic burden even negatively impacts the households access to diarrhoeal treatment.^{2 5} Moreover, diarrhoea leads to malnutrition, which may result in death.² The disease also causes financial strains on the families as a loss of productivity is expected for the caregiver of the sick child.⁶

Several studies have aimed to estimate the household cost and the economic implications of diarrhoea among under-5 children (U5C) in various countries.⁶ A study conducted in Bangladesh reported that on average, households spend I$26.2 (a currency used for economic comparisons and are adjusted using purchasing power parity (PPP) values) in direct costs for treating diarrhoea before reporting to a diarrhoeal diseases hospital.⁶ As reported by Das et al, the total cost of invasive enteritis among U5C in Bangladesh was almost similar based on pathogens, namely Shigella (US$4.17) and Campylobacter (US$3.49).⁷ Another report based on the data from the Global Enteric Multicentre Study (GEMS) also estimated the household costs of diarrhoea by aetiology in seven countries and revealed that the household out-of-pocket (OOP) costs were higher in Mali, whereas differences due to aetiology within countries were not statistically significant.⁸ However, comprehensive data on direct and indirect costs due to diarrhoea in South Asia and SSA are scarce.

Cost-of-illness can be a useful guideline for policy-makers in prioritising, selecting and scaling up interventions.² The GEMS study was conducted on seven sites, including three countries from South Asia and four from SSA.⁹ Between 2007 and 2011, U5C with diarrhoeal disease were enrolled in the study. During enrolment, information was collected from the parents or caregivers.⁸ In this study, we aimed to estimate the direct and indirect costs of diarrhoeal illness among U5C in seven LMICs using data from the GEMS study. We also aimed to explore the factors that are associated with variation in these costs. Due to data limitations, we did not examine the likelihood of incurring any indirect costs. While the primary analysis was at the individual level, we also presented continent-wise differences descriptively to provide contextual understanding.

Methods

Data overview

We used data from the GEMS. The details of the GEMS are published elsewhere.9,11 After registration and permission, we accessed the data through ClinEpiDB, an open-access online platform for clinical and epidemiological studies.¹² GEMS was a 36-month, multicountry, age-stratified, matched case–control study that investigated diarrhoeal disease among children aged 0–59 months across seven sites in LMICs. These included three countries in South Asia (Bangladesh, India and Pakistan) and four in SSA (Kenya, Mali, Mozambique and The Gambia) (figure 1).

Figure 1. Study sites for the GEMS. (A) Three study sites in South Asia; (B) four study sites in sub-Saharan Africa. GEMS, Global Enteric Multicentre Study.

GEMS enrolled children with either moderate-to-severe diarrhoea (MSD) or less-severe diarrhoea (LSD), along with age-matched and community-matched controls, all residing in demographic surveillance system areas. Each site had designated sentinel health centres for enrolment based on standardised inclusion criteria. Clinical, epidemiological and microbiological data were collected at enrolment, and a follow-up home visit was conducted about 60 days later to assess outcomes and collect additional information. For this analysis, we used data from the GEMS-1A Case-Control study, a 12-month follow-up conducted between 2011 and 2012. Caregivers were interviewed to obtain information on household costs, healthcare-seeking behaviour and productivity loss during the follow-up visit. We included participants with non-zero direct and/or indirect costs. Detailed clinical and economic data collected during the enrolment were used for this secondary analysis.

The GEMS enrolled 14 242 participants in total, with 5844 of them being cases and 8398 being matched controls. The ‘cases’ were children aged <60 months who reported to the sentinel health centre with acute MSD disease, and controls were children without MSD. We considered 4592 children aged <60 months for our analysis. Only cases with non-zero total costs were considered following the literature review.⁷ The procedure used to select children for this analysis is described in detail in figure 2.

Figure 2. Selection of children <60 months for analysis.

Operational definitions and measurements

Outcome variables

We used the healthcare consumer’s perspective while calculating costs. In the original GEMS study, data related to the costs were directly taken from caregivers through structured interviews. This method allowed for capturing the caregivers’ OOP expenses and income loss. So, although we performed a secondary data analysis, the cost data were originally gathered from the perspective of healthcare consumers. The outcome variables of our study were total direct cost (TDC) and total indirect cost (TIC). All costs were calculated in the participant’s local currency, then adjusted for inflation and converted into international dollars (I$) for better comparison. In our analysis, we treated these variables as continuous. For the quantile regression, we focused on the median to examine the association between the outcome and the independent variables, as the data were not normally distributed.

Total direct cost

The direct cost is defined as the amount of money spent for medical management of the disease, and it includes the cost of drugs, consultation, diagnostic, patient transportation, hospital cost, food cost, etc.¹³ Studies have suggested that more than 30 cost centres can be considered for calculating direct cost. However, some cost centres may not be appropriate for every patient.^{6 13} We inquired about the expenses incurred for prior healthcare seeking and the current healthcare seeking in local currency. The expenses for current care seeking were asked after discharge from the hospital, which the caregivers self-reported. In our study, TDC was calculated by summing the cost for direct medical cost for previously sought care for diarrhoea (pharmacy, consultation irrespective of type of provider, medication cost, hospital cost, other medical expenses), direct non-medical cost for previously sought care for diarrhoea (transport cost for pharmacy, consultation irrespective of type of provider, hospital visit, buying drugs, other transportation expenses), direct medical cost for current care (consultation cost, hospitalisation cost, medication cost, diagnostic cost, other medical cost) and direct non-medical cost for current care (food expenses, hospital expenses for card, soap etc, transportation cost for the patient and other household members).

Total indirect cost

Indirect costs are expenses that arise from the consequences of an illness, rather than from direct medical treatment. These can include lost income, decreased productivity and additional expenses like home care or childcare.¹³ We calculated indirect costs using days of absenteeism from work for the caregivers. Income loss due to caregiving illness was calculated using the average daily income of the caregiver in local currency.^{7 14 15} For the loss of half a morning or afternoon, 0.25 days were considered. Similarly, for a morning or afternoon, it was 0.50 days, for a morning and afternoon, it was 1.00 days, and for less than half a morning or afternoon, it was 0.00 days. The loss of time was also calculated in the same fashion as if any other caregiver had taken care of the child. However, if a caregiver reported no income-generating activity (eg, housewives), the indirect cost was recorded as zero. For this analysis, we only considered participants reporting non-zero costs. Additionally, as we could not differentiate between true zero-indirect cost cases and pseudo-zero-indirect cost cases (caregivers not engaged in paid work), we could not develop models to identify predictors of indirect cost occurrence. Information on employment benefits, such as paid leave, was not collected. Therefore, time loss was assumed to result in income loss for all caregivers engaged in paid work.

Independent variables

The independent variables we considered for this study include age (in months), age group, number of family members, number of U5C, sex (male/female), residence (urban/rural), died (yes/no), MSD (yes/no), previously sought care (yes/no), care by licensed practitioner (yes/no), primary caregiver (mother/others), level of education of primary caregiver (informal or up to primary/ primary completed/secondary completed/postsecondary), rehydration method used (none/oral/intravenous), wealth index (poorest/poor/middle/rich/richest), treated drinking water (yes/no),¹⁶ stunting (yes/no), wasting (yes/no), underweight (yes/no), overweight (yes/no) and components of TDC (pharmacy cost/consultation cost/medication cost/healthcare centre cost/diagnostic cost/other cost). The components of TDC were only subcomponents of TDC, calculated individually and summed for total cost estimation.

Residence

We considered the location of the study site to determine whether it is urban or rural. As reported by Kotloff et al, Mali and India were considered urban settings, Pakistan was a periurban setting, and the four other countries were rural settings.⁹ For the purpose of the current analysis, we considered urban and periurban as ‘urban’ as Pakistan is the only periurban setting in the study.

MSD

We defined MSD as diarrhoea with the presence of at least one of these symptoms: sunken eyes, loss of skin turgor, intravenous hydration administered or prescribed, dysentery, admission or advised admission to hospital.⁹

Primary caregiver

In our dataset, mothers were reported as the primary caregiver in 95.9% of cases. All other caregiver types (including father, grandmother, aunt, etc) together accounted for only 4.1% of cases. Due to this distribution and for analytical clarity, we grouped the less frequent caregiver types under a single ‘other’ category.

Wealth index

It is a method that is used by various survey studies like Multiple Indicator Cluster Survey or Demographic and Health Surveys. Using this approach, the participants can be classified into five quantiles on the premise that the possession of household assets and amenities indicates the households’ relative economic position.^{17 18}

Stunting

U5C with height/length-for-age z score of <−2 SD were considered as stunted.^{19 20}

Wasting

U5C with weight-for-height z score of <−2 SD were considered as wasted.^{19 20}

Underweight

U5C with weight-for-age z score of <−2 SD were considered as underweight.^{19 20}

Overweight

U5C with body mass index-for-age z score of >1 SD were considered as overweight.^{19 20}

Adjusting for inflation

Inflation is a phenomenon where the same amount of money loses its purchasing ability over the years, mostly due to an increase in price. The Consumer Price Index (CPI) is used to track the impact of inflation.²¹ The following formula is used to adjust expenses or costs for inflation²²:

$${\displaystyle \mathrm{C}\mathrm{u}\mathrm{r}\mathrm{r}\mathrm{e}\mathrm{n}\mathrm{t}\mathrm{a}\mathrm{m}\mathrm{o}\mathrm{u}\mathrm{n}\mathrm{t}=\frac{\mathrm{C}\mathrm{P}\mathrm{I}\mathrm{f}\mathrm{r}\mathrm{o}\mathrm{m}\mathrm{r}\mathrm{e}\mathrm{c}\mathrm{e}\mathrm{n}\mathrm{t}\mathrm{p}\mathrm{e}\mathrm{r}\mathrm{i}\mathrm{o}\mathrm{d}}{\mathrm{C}\mathrm{P}\mathrm{I}\mathrm{f}\mathrm{r}\mathrm{o}\mathrm{m}\mathrm{e}\mathrm{a}\mathrm{r}\mathrm{l}\mathrm{i}\mathrm{e}\mathrm{r}\mathrm{p}\mathrm{e}\mathrm{r}\mathrm{i}\mathrm{o}\mathrm{d}}\times \mathrm{A}\mathrm{m}\mathrm{o}\mathrm{u}\mathrm{n}\mathrm{t}\mathrm{f}\mathrm{o}\mathrm{r}\mathrm{e}\mathrm{a}\mathrm{r}\mathrm{l}\mathrm{i}\mathrm{e}\mathrm{r}\mathrm{p}\mathrm{e}\mathrm{r}\mathrm{i}\mathrm{o}\mathrm{d}}$$

We adjusted all cost values to the year 2022 using country-specific CPI values. The ‘general’ CPI for each country was used to adjust for inflation. The CPIs of each centre are shown in the online supplemental table S1 and S2 .²³

Conversion of currency

For economic studies, costs are adjusted for inflation and reported in US dollars or international dollars (I$).⁶ The I$ is a hypothetical unit of currency designed to account for differences in relative prices across various contexts. For example, I$1 would purchase a comparable amount of goods and services in the country of interest as US$1 would in the USA.⁶ The current study used the following formula for conversion of the local currency, I$²¹:

$${\displaystyle \mathrm{I}\mathrm{\$}=\frac{\mathrm{A}\mathrm{m}\mathrm{o}\mathrm{u}\mathrm{n}\mathrm{t}\mathrm{i}\mathrm{n}\mathrm{l}\mathrm{o}\mathrm{c}\mathrm{a}\mathrm{l}\mathrm{c}\mathrm{u}\mathrm{r}\mathrm{r}\mathrm{e}\mathrm{n}\mathrm{c}\mathrm{y}}{\mathrm{P}\mathrm{u}\mathrm{r}\mathrm{c}\mathrm{h}\mathrm{a}\mathrm{s}\mathrm{i}\mathrm{n}\mathrm{g}\mathrm{P}\mathrm{o}\mathrm{w}\mathrm{e}\mathrm{r}\mathrm{P}\mathrm{a}\mathrm{r}\mathrm{i}\mathrm{t}\mathrm{y}(\mathrm{P}\mathrm{P}\mathrm{P})\mathrm{o}\mathrm{f}\mathrm{t}\mathrm{h}\mathrm{e}\mathrm{c}\mathrm{o}\mathrm{u}\mathrm{n}\mathrm{t}\mathrm{r}\mathrm{y}}}$$

We reported costs in international dollars (I$) to allow better comparison across countries with different price levels. This method follows economic evaluation guidelines, which recommend using I$ or 2015 USD as standard reporting currencies in multi-country studies.²² We chose I$ because it reflects local purchasing power and provides more stable estimates over time than market-based exchange rates. This approach is especially helpful when comparing costs across LMICs as the purchasing power parity (PPP) is more stable than the market-based USD.²² The PPP in 2022 for the study sites is presented in online supplemental table S1 and S2.²⁴

Statistical analysis

Data analysis was done based on the type of variable. Normality of continuous data was checked using the Shapiro-Francia test and was presented using median (IQR). The categorical data were presented using frequency and percentage. We used the Mann-Whitney U test, the Kruskal-Wallis test (Dunn test as post hoc estimation) and the χ² test to investigate the association between variables where applicable.

Quantile regression analysis was conducted at the 50th percentile (median) to identify predictor variables for TDC and TIC after adjusting for age, sex and country. Predictors were entered into the quantile regression model according to their original type. No categorisation of continuous predictors was done unless already grouped (eg, age group or wealth index). All independent variables were first tested individually to examine their direct association with cost outcomes using unadjusted quantile regression. For the adjusted models, we included age, sex and country as confounders, as both are known to influence disease severity and cost-related outcomes to control for basic demographic differences. We did not use backwards selection, as our goal was to examine the adjusted associations for the theoretically relevant predictors. All unadjusted and adjusted coefficients with their corresponding 95% CI were reported. Due to limitations in sample size and model complexity, we did not conduct stratified regression analyses for South Asia and SSA. Instead, we presented region-wise descriptive comparisons to provide contextual understanding. All statistical tests were two-sided, with a significance level of α=0.05 and were conducted using STATA V.17 (StataCorp).

Patient and public involvement

No patient or public was involved in any stage of the study.

Results

The following study analysed data from 4592 participants. The process of selecting the participants is presented in figure 2. The TDC was reported by 4592 participants, and 430 participants reported TIC. The median (IQR) TDC and TIC were I$8.4 (I$11.0) and I$10.2 (I$14.3) respectively. There was a significant (p=0.015) positive correlation (Spearman’s r=0.012), which indicates that higher TDC often coincided with higher TIC costs (figure 3).

Figure 3. Median total direct cost and total indirect cost in I$. Spearman’s rho = 0.012; p-value = 0.015; I$, International dollars.

The distribution of independent variables by continent is presented in table 1. The median (IQR) age of the participants was 14 (16) months. Most of them were male (54.6%) and resided in urban areas (59.6%). Notable differences were observed between South Asia and SSA in variables such as family size, number of children under 5, residence and caregiver education. For example, urban residence was more common in South Asia, and caregivers in SSA had lower levels of formal education. Total median (IQR) direct cost was higher in South Asia (I$10.3, I$7.6), compared with SSA (I$5.1, I$12.9) (table 1).

Table 1. Distribution of independent variables by continent (n=4592).

Variables	Totaln (%)	Continent
		South Asian (=2119) (%= 46.2)	Sub-Saharan African (=2473) (%=53.8)
Age (in months)*	14 (16)	14 (16)	14 (17)
Age group (in months)
0–11	1822 (39.7)	689 (41.0)	953 (38.5)
12–23	1524 (33.2)	684 (32.3)	840 (34.0)
24–59	1246 (27.1)	566 (26.7)	(27.5)
No. of family members*	7 (7)	6 (4)	10 (16)
No. U5C*	2 (2)	1 (1)	2 (3)
Sex (male), n (%)	2507 (54.6)	1179 (55.6)	1328 (53.7)
Residence (urban), n (%)	2736 (59.6)	1385 (65.4)	1351 (54.6)
Died (yes), n (=4351) (%)	31 (0.7)	7 (0.4)	24 (1.0)
MSD (yes), n (%)	2214 (48.2)	996 (47.0)	1217 (49.2)
Previously sought care (yes), n (%)	1355 (29.5)	908 (42.9)	407 (18.1)
Care by licensed practitioner (yes), n (=1355) (%)	175 (12.9)	168 (18.5)	7 (1.6)
Primary caregiver (mother), n (%)	4402 (95.9)	2087 (98.5)	2315 (93.6)
Level of education of primary caregiver (n=4588)
Informal or up to primary	2572 (56.1)	720 (34.0)	1852 (75.0)
Primary completed	1562 (34.1)	1088 (51.3)	474 (19.2)
Secondary completed	343 (7.5)	224 (10.6)	119 (4.8)
Postsecondary	111 (2.4)	87 (4.1)	24 (1.0)
Rehydration method used
None	753 (16.4)	91 (4.3)	662 (26.8)
Oral	3625 (78.9)	1953 (92.2)	1672 (67.6)
IV	214 (4.7)	75 (3.5)	139 (5.6)
Wealth index (n=4591)
Poorest	981 (21.4)	451 (21.3)	530 (21.4)
Poor	961 (20.9)	490 (23.1)	471 (19.1)
Middle	856 (18.6)	349 (16.5)	507 (20.5)
Rich	925 (20.2)	444 (20.9)	481 (19.5)
Richest	868 (18.9)	385 (18.2)	483 (19.5)
Treated drinking water (yes), n (%)	929 (20.2)	602 (28.4)	327 (13.2)
Stunting (yes), n (%)	1068 (23.3)	572 (27.0)	496 (20.1)
Wasting (yes), n (%)	672 (14.6)	303 (14.3)	369 (14.9)
Underweight (yes), n (%)	119 (24.4)	608 (28.7)	511 (20.7)
Overweight (yes), n (%)	389 (8.5)	145 (6.8)	244 (9.9)
Total direct cost (in I$)*	8.4 (11.0)	10.3 (7.6)	5.1 (12.9)
Total indirect cost (in I$) (n=430)*	10.2 (14.3)	11.3 (13.4)	10.2 (15.3)

^*Median (IQR) reported.

I$, international dollars; IV, intravenous; MSD, moderate-to-severe disease; U5C, under-5 children.

The distribution of the TDC in I$ is presented in figure 4. The highest median (IQR) TDC was found in Bangladesh (I$13.6, I$9.8) and the lowest was in Mozambique (I$0.4, I$1.2). Significant association was found between country and the TDC (p<0.001). A significant difference was also found between the TDC reported by the countries.

Figure 4. Total direct cost (in I$) across countries.

The pattern of total direct medical cost was different in each country (online supplemental figure 1). In most countries, the biggest part of the cost came from medication. This was especially high in India (74.6%), Mali (86.0%) and Mozambique (78.9%). But in Pakistan, consultation (41.0%) and other costs (36.8%) were the main part of the expenses. In The Gambia, almost all of the cost (87.5%) was from other items. In Bangladesh, the cost was more evenly distributed: medication (47.8%), other (22.8%) and consultation (15.3%). When we combined all countries, medication (60.9%) and other costs (24.0%) were the largest parts of the TDC (online supplemental figure 1).

The maximum median (IQR) TIC was reported by Bangladesh (I$23.2, I$14.5). The Gambia also reported a high median (IQR) TIC of I$19.7 (I$22.9). However, the minimum was reported by Mozambique (I$7.0, I$4.9). The TIC was significantly associated with the country (p<0.001) (figure 5).

Figure 5. Total indirect cost (in I$) across countries (n=430).

Univariate quantile regression models were developed to identify independent variables associated with TDC. Then the models were adjusted for age in months, sex and country. Each additional family member (p<0.001) and each additional under-5 child in the household (p<0.001) increased the median TDC by I$0.19 and I$0.78, respectively. Urban households reported I$1.56 higher median TDC than rural households. Children with MSD had I$3.65 higher median TDC compared with those with LSD. Prior healthcare-seeking was also associated with increased cost, adding I$7.18 to the median TDC. Caregivers with primary education had I$1.12 lower median TDC than those with no or informal education, while post-secondary education was associated with slightly higher costs (I$1.47). Rehydration method also influenced the cost, where oral and intravenous methods increased the median TDC by I$6.41 and I$10.33, respectively. Children with wasting and underweight had significantly higher costs, while overweight status and use of treated drinking water were associated with reduced TDC (p<0.001 for both) (table 2).

Table 2. Quantile regression analysis for association between total direct cost and predictor variables.

Variable	Unadjusted coef.		Adjusted coef.
	Coef. (95% CI: lower; upper)	P value	Coef. (95% CI: lower; upper)	P value
Age (in months)	−0.01 (−0.03; 0.02)	0.506	−0.01 (−0.03; 0.01)	0.299
Age group (in months)		Ref: 0–11
12–23	0.50 (−0.24; 1.24)	0.181	0.42 (−0.23; 1.07)	0.200
24–59	−0.32 (−1.10; 0.46)	0.419	−0.27 (−1.53; 1.00)	0.682
No. of family members	0.00 (−0.02; 0.03)	0.735	0.19 (0.17; 0.21)	<0.001
No. of U5C	−0.13 (−0.29; 0.02)	0.091	0.78 (0.67; 0.90)	<0.001
Sex		Ref: Female
Male	0.41 (−0.23;1.05)	0.212	0.24 (−0.20; −0.68)	0.290
Residence		Ref: Rural
Urban	3.45 (2.84; 4.07)	<0.001	1.56 (1.07; 2.04)	<0.001
Died (n=4351)		Ref: No
Yes	−3.22 (−6.88; 0.44)	0.085	−1.27 (−4.01; 1.46)	0.361
MSD		Ref: No
Yes	4.57 (4.13; 4.99)	<0.001	3.65 (3.21; 4.09)	<0.001
Previously sought care		Ref: No
Yes	8.21 (7.68; 8.73)	<0.001	7.18 (6.63; 7.73)	<0.001
Care by licensed practitioner (n=1355)		Ref: No
Yes	−2.59 (−4.72; −0.45)	0.018	−0.71 (−3.24; 1.83)	0.584
Primary caregiver		Ref: No
Mother	0.46 (−1.09; 2.01)	0.561	−0.12 (−1.24; 1.00)	0.832
Level of education of primary caregiver (n=4588)		Ref: Informal or up to primary
Primary completed	2.87 (2.25; 3.49)	<0.001	−1.12 (−1.66; −0.58)	<0.001
Secondary completed	2.87 (1.76; 3.98	<0.001	−0.57 (−1.46; 0.31)	0.202
Postsecondary	5.08 (3.21; 6.96)	<0.001	1.47 (−0.00; 2.94)	0.050
Rehydration method used		Ref: None
Oral	8.97 (8.25; 9.68)	<0.001	6.41 (5.53; 7.29)	<0.001
IV	10.81 (9.44; 12.19	<0.001	10.33 (8.92; 11.74)	<0.001
Wealth index (n=4591)		Ref: Poorest
Poor	0.00 (1.102; 1.02)	>0.999	−0.10 (−0.81; 0.61)	0.788
Middle	0.08 (−0.97; 1.12)	0.884	0.27 (−0.47; 1.00)	0.474
Rich	0.16 (−0.86; 1.19)	0.754	0.10 (−0.62; 0.82)	0.787
Richest	0.98 (−0.06; 2.03)	0.064	0.31 (−0.34; 1.04)	0.413
Treated drinking water		Ref: No
Yes	−1.81 (−2.67; −0.94)	<0.001	−1.46 (−2.00; −0.92)	<0.001
Stunting		Ref: No
Yes	−0.25 (−1.03; 0.51)	0.524	−0.45 (−0.96; 0.06)	0.087
Wasting		Ref: No
Yes	2.36 (1.48; 3.25)	<0.001	2.57 (1.93; 3.20)	<0.001
Underweight		Ref: No
Yes	1.23 (0.49; 1.97)	0.001	1.31 (0.79; 1.83)	<0.001
Overweight		Ref: No
Yes	−3.29 (−4.42; −2.16)	<0.001	−2.26 (−3.07; −1.45)	<0.001

Variables are adjusted for age, sex and country.

IV, intravenous; MSD, moderate-to-severe disease; U5C, under-5 children.

Adjusted quantile regression model (adjusted for age in months, sex and country) was developed to examine the association between independent variables and TIC. After adjusting, participants who had sought care before the main facility visit reported a I$3.65 increase in median TIC compared with those who did not. No significant associations were found for age, sex, residence, diarrhoea severity, death of the child and caregiver’s post-secondary education or nutritional status (table 3).

Table 3. Quantile regression analysis for association between total indirect cost and predictor variables (n=430).

Variable	Unadjusted coef.		Adjusted coef.*
	Coef. (95% CI: lower; upper)	P value	Coef. (95% CI: lower; upper)	P value
Age (in months)	−0.05 (−0.14; 0.03)	0.223	−0.05 (−0.13; 0.04)	0.301
Age group (in months)		Ref: 0–11
12–23	−0.49 (−2.82; 1.85)	0.682	1.29 (−1.61; 4.19)	0.384
24–59	−1.23 (−4.04; 1.50)	0.367	2.71 (−3.41; 8.84)	0.384
No. of family members	0.00 (−0.10; 0.10)	>0.999	0.03 (−0.07; 0.14)	0.548
No. U5C	0.05 (−0.43; 0.52)	0.848	0.12 (−0.38; 0.63)	0.629
Sex		Ref: Female
Male	−1.56 (−3.59; 0.47)	0.133	−1.48 (−3.52; 0.56)	0.154
Residence		Ref: Rural
Urban	−0.56 (−2.93; 1.82)	0.645	−1.64 (−3.97; 0.70)	0.169
Died (n=395)		Ref: No
Yes	5.08 (−1.02; 11.19)	0.103	5.61 (−1.05; 12.26)	0.099
MSD		Ref: No
Yes	0.00 (−2.13; 2.14)	>0.999	−0.67 (−3.04; 1.70)	0.580
Previously sought care		Ref: No
Yes	4.26 (2.56; 5.96)	<0.001	3.65 (1.57; 5.72)	<0.001
Care by licensed practitioner		Ref: No
Yes	2.30 (−3.12; 7.73)	0.403	4.07 (−1.23; 9.36)	0.131
Primary caregiver		Ref: No
Mother	0.00 (−4.45; 4.45)	>0.999	0.24 (−4.54; 5.02)	0.921
Level of education of primary caregiver		Ref: Informal or up to primary
Primary completed	0.00 (−2.05; 2.05)	>0.999	1.37 (−1.04; 3.79)	0.265
Secondary completed	1.09 (−3.03; 5.21)	0.603	1.65 (−2.81; 6.10)	0.468
Post-secondary	4.84 (−0.96; 10.65)	0.102	5.87 (−0.43; 12.17)	0.068
Rehydration method used		Ref: None
Oral	0.28 (−3.20; 3.76)	0.872	1.88 (−2.00; 5.76)	0.341
IV	0.77 (−3.56; 5.11)	0.727	2.78 (−1.79; 7.36)	0.232
Wealth index		Ref: Poorest
Poor	1.44 (−1.89; 4.76)	0.397	0.18 (−3.50; 3.87)	0.922
Middle	1.44 (−1.39; 4.26)	0.318	1.11 (−2.02; 4.24)	0.486
Rich	2.53 (−0.40; 5.45)	0.090	2.86 (−0.38; 6.10)	0.083
Richest	1.44 (−1.65; 4.52)	0.361	1.96 (−1.46; 5.39)	0.260
Treated drinking water		Ref: No
Yes	0.00 (−2.16; 2.16)	>0.999	−0.38 (−2.47; 1.72)	0.725
Stunting		Ref: No
Yes	0.00 (−2.42; 2.42)	>0.999	−0.42 (−2.75; 1.90)	0.720
Wasting		Ref: No
Yes	1.12 (−1.60; 3.85)	0.419	0.96 (−1.79; 3.70)	0.494
Underweight		Ref: No
Yes	1.09 (−1.11; 3.29)	0.330	0.91 (−1.24; 3.05)	0.407
Overweight		Ref: No
Yes	2.54 (−1.00; 6.09)	0.160	1.69 (−1.97; 5.35)	0.365

^*Adjusted for age, sex and country.

IV, intravenous; MSD, moderate-to-severe disease; U5C, under-5 children.

Discussion

In the past few decades, the world has seen a significant decline in the prevalence of paediatric diarrhoeal diseases. However, it remains one of the significant public health concerns, especially in LMICs.²⁵ It has been reported that children are more susceptible to the disease, with an average incidence of 3.2–12 episodes per year per child, although most of these episodes are self-limiting.²⁵ Apart from the obvious physical toll it exerts on the child, the economic strain of diarrhoeal disease on the household is by no means negligible.² In this study, we analysed the data from the GEMS study. It was a multicentre study that considered U5C as the participants. The study collected household cost-related data from the caregivers. We found that the median TIC was higher than the TDC (I$10.2 vs I$8.4). This finding signifies the often-neglected effect of diarrhoeal diseases, that is, the economic strain it puts on the family due to loss of productivity. Koopmanschap and Rutten have indicated that if any healthcare programme produces health benefits quickly, then the impact of that particular programme is more impactful.²⁶ Unfortunately, programmes that aim to reduce diarrhoeal burden often fail to consider this. While indirect cost was higher than direct cost in our analysis, this may still underestimate the true productivity loss, as time spent by unpaid caregivers was not monetised.

Our study found some significant differences regarding the independent variables between the two continents (South Asia and SSA). This indicates that although the economic situation between the countries of these two continents is often similar, there are marked differences in the culture and social dynamics.²⁷ The TDC was significantly higher in SSA (I$12.9) than South Asia (I$7.6). TIC was also higher in SSA, although the difference was insignificant. In their study, Asante et al present various findings that can account for this fact.²⁸ Although the region bears a disproportionate share of the global disease burden, the allocation of resources to healthcare is quite low. Additionally, in 2016, Africa’s per capita health spending was more than 50 times lower than the Organisation for Economic Co-operation and Development countries.²⁸ Furthermore, the financing systems for healthcare in SSA are largely dependent on high OOP payments and very low government spending.²⁸

However, when compared between countries, Bangladesh reported the highest TDC and TIC. When we consider the per capita total expenditure on health in I$, this finding becomes understandable. During 2011–2012, that is, the period of the GEMS study, Bangladesh spent I$66.8 and I$67.8, respectively, which was the second lowest among all seven participating countries.²⁹ It has been reported that OOP payments by households are responsible for nearly 67% of the total health expenditure.³⁰ Several reasons, like a poorly designed health financing system, lack of proper health insurance measures, hidden charges, understaffing, etc, are often cited as the root cause of this high OOP rate.³⁰

Cross-country differences in the composition of direct medical costs were noticeable in our study. For example, medication cost was the highest in India, Mali and Mozambique, while consultation and other costs were dominant in Pakistan and The Gambia. These differences may reflect local variations in care-seeking practices, health facility charges, drug pricing and treatment protocols. Prior studies have shown that OOP spending patterns often depend on the availability of subsidised medicines, informal healthcare providers and the structure of the national health system.^{28 30} Our study found that the overall highest proportion of TDC was due to medication, while diagnostic cost was the minimum. In developed countries, the government usually provides subsidies for pharmaceuticals. As the cost of medicines is subsidised, the direct cost is not much impacted. On the other hand, in LMICs like Pakistan, the prices of drugs are usually controlled at the level of retail pharmacy.³¹ As such, the higher proportion of medication is understandable. Steps like providing subsidies for essential medications, regulating drug prices and also enforcing laws prohibiting the sales of antibiotics without a physician’s prescription can help in reducing the direct and indirect costs.

However, we could not disaggregate the cost due to medication into types of drugs, such as antibiotics or antidiarrhoeals, because the dataset did not include this information. In future, such details would help to understand how well standard treatment guidelines are followed and how to reduce unnecessary spending.

We found that the number of family members and the number of U5C increased the median TDC. The effect of number of family members and other U5C is most likely to be indirect. Several studies have reported family size to be a significant influencer of diarrhoea among U5C.^{32 33} One possible reason is the low level of care by the parents and increased risk of transmission.³²

Residence was an important predictor of TDC, where urban households are more likely to have higher costs. The living cost in urban areas is significantly higher than in rural areas. Additionally, the cost of care at tertiary level hospitals, which are more prevalent in the urban areas, is significantly higher.³⁴

Diarrhoeal diseases can be categorised as LSD and MSD based on the level of dehydration, presence of blood in stool and hospital admission.¹¹ U5C with MSD is usually provided in-patient care. Thobari et al conducted a health facilities and community survey to estimate the economic burden of U5C in Indonesia, where they found that DTC for inpatient care was significantly higher than outpatient care. Our study reports similar findings. It has been reported that patients with inpatient care require higher professional fees, diagnostic and medication costs. Additionally, the cost becomes higher due to the presence of several specialty and subspecialty medical professionals.³⁴ Another study in Bangladesh reported similar findings, where MSD was associated with direct medical and non-medical costs.⁶ As ORS or intravenous is used as the rehydration method for MSD, the positive association between rehydration method and TDC found in our study is also understandable.

We found that children with mothers who completed primary education reported lower costs. However, when the maternal educational level is post-secondary, the cost increases. In LMICs, mothers are usually considered as the primary caregivers. Education is known to have a significant impact on the maternal care-seeking behaviour for their child’s illness.⁶ Reports suggest that when the mother is highly educated, their tendency to more frequent care-seeking also increases, leading to higher expenditure.⁶

Diarrhoea is a waterborne disease that is reported to be linked to the quality of drinking water.^{16 25 35} In a cluster randomised controlled trial, Solomon et al reported that treating drinking water considerably decreases the diarrhoeal episodes.³⁶ According to the WHO, 94.0% of diarrhoeal incidence can be prevented simply by modifying the environment, improving hygiene and sanitation, and making clean water more available.³⁶ We also found that consuming treated drinking water negatively impacted TDC.

When the child’s condition does not improve, mothers often seek additional consultations. In LMICs, the population often seeks care from informal caregivers rather than professional doctors.³⁷ However, this causes multiple negative impact on patient. These informal caregivers don’t have formal care or education which exacerbates the disease, leading to increased direct and indirect costs. Moreover, seeking care from multiple sources puts additional financial strain. Our findings also indicate that seeking previous care has caused increased TDC and TIC.

It has been suggested by multiple studies that malnutrition can increase the frequency and duration of diarrhoeal illness.^{38 39} Specially wasting is reported to be a strong predictor for prolonged illness.³⁹ Several mechanisms are suggested for explaining this phenomenon, like poor appetite, malabsorption of nutrients, hastening of intestinal transit time, etc. Our study also reported a positive association between wasting, underweight and TDC.

In this study, we used data from a large, multicentre study conducted in multiple LMICs. As such, the results of the study can be considered as generalisable. We also considered non-zero costs for both direct and indirect costs to gain valuable insights, considering the overall picture. The costs were adjusted for inflation and converted into I$ for better comparability. However, there were several limitations also. As the GEMS data were collected in 2011–2012, it may fail to capture the current trends. Additionally, cost data were reported by the caregivers, which may have caused recall bias. Differences in healthcare systems, currencies and purchasing power across regions may have complicated the direct comparisons of costs. The dataset also lacked information needed to disaggregate the costs in more detail (ie, type of drugs, diagnostic tests, etc). Another key limitation of our indirect cost estimation is that caregivers not engaged in income-generating activities, such as housewives, were assigned zero cost. This may underestimate the full productivity loss at household level, especially in settings where unpaid caregiving is common. While information on employment type or access to paid leave was not collected. As a result, time lost from caregiving was valued equally for all caregivers, even though salaried workers with leave benefits may not have experienced actual income loss.

Conclusions

Despite the global decline in paediatric diarrhoeal diseases, they persist as a critical public health issue in LMICs, imposing severe economic strains on households. This study reveals a significant disparity, with indirect costs often surpassing direct costs, underscoring the overlooked financial toll of lost productivity. Regional cost variations, driven by household size, maternal education and access to safe drinking water, highlight the necessity for tailored strategies. It is important to follow standard treatment guidelines in both health facilities and at home, because this can help reduce both direct and indirect costs of diarrhoea. Policy-makers should take steps to reduce the cost of diarrhoea for families. This includes improving health financing to lower OOP payments by implementing health insurance and by subsidising treatment packages, making medicines more affordable by regulating the price and also taking steps to stop the sales of antibiotics without proper prescription, and investing in clean water and sanitation. Early care-seeking through community awareness and promoting the use of ORS and Zinc should also be encouraged.

Data availability statement

Data are available in a public, open access repository.