Associated factors, post infection child growth, and household cost of invasive enteritis among under 5 children in Bangladesh

Both Campylobacter- and Shigella-induced invasive enteritis are common in under-5 Bangladeshi children. Our study aimed to determine the factors associated with Campylobacter and Shigella enteritis among under-5 children, the post-infection worsening growth, and the household cost of invasive enteritis. Data of children having Shigella (591/803) and Campylobacter (246/1148) isolated from the fecal specimen in Bangladesh were extracted from the Global Enteric Multicenter Study (GEMS) for the period December 2007 to March 2011. In multiple logistic regression analysis, fever was observed more frequently among shigellosis cases [adjusted OR 2.21; (95% CI 1.58, 3.09)]. Breastfeeding [aOR 0.55; (95% CI 0.37, 0.81)] was found to be protective against Shigella. The generalized estimating equations multivariable model identified a negative association between Shigella and weight-for-height z score [aOR − 0.11; (95% CI − 0.21, − 0.001)]; a positive association between symptomatic Campylobacter and weight-for-age z score [aOR 0.22; (95% CI 0.06, 0.37)] and weight-for-height z score [aOR 0.22; (95% CI 0.08, 0.37)]. Total costs incurred by households were more in shigellosis children than Campylobacter-induced enteritis ($4.27 vs. $3.49). Households with low-level maternal education tended to incur less cost in case of their shigellosis children. Our findings underscore the need for preventive strategies targeting Shigella infection, which could potentially reduce the disease burden, associated household costs, and child growth faltering.

Characteristics of Shigella-positive children having moderate-to-severe diarrhea (MSD) in Bangladesh. About 44.3% of the Shigella-positive children with MSD were aged 24-59 months and half of them were female. Giardia was found to be the most common co-pathogen (Table 1) in the Shigella positive children compared to the Shigella negative children. A significantly higher proportion of under-five children with MSD and associated Shigella infection were stunted, wasted and underweight. The duration of diarrhea before coming to the facility was less. The children had more frequent visible blood in the stool, less vomiting, and more often presented with a history of fever during admission. Regarding the inclusion criteria for MSD, the children more commonly had dysentery and required hospital admission. Caregivers of the Shigella-positive children practiced handwashing less frequently before nursing the child/preparing baby food and after cleaning the child; one-fourth of them belonged to middle and upper middle-class families. Shigella-positive children were less often breastfed. Their stool examinations reported the habitual presence of fecal red blood cell (RBC) and mucus more frequently in comparison to the stool specimens of Shigella-negative children.
Characteristics of Campylobacter-positive children having MSD. About 65% of the Campylobacterpositive children with MSD were aged 0-11 months and 42% of them were female; less often had malnutrition; mostly presented with visible blood in stool and fever during admission, less frequently presented with vomiting. Among the inclusion criteria of MSD, more children had dysentery and required less hospital admission. They had a large family size, and more under-5 children in the house; more frequently had toilet facility at the house; practiced handwashing more commonly before nursing the child/preparing baby food; less often washed hand with soap. Eighteen percent of children were from wealthy families and were more commonly breastfed. Their stool examinations reported the frequent presence of fecal RBC and mucus compared to the stool specimens of Campylobacter-negative MSD children. EAEC was found more often as co-pathogens (Table 1) in the Campylobacter-positive children.
Multiple logistic regression reveals that MSD children with Shigella were significantly associated with the presence of blood in stool [ Child growth with fecal Shigella and Campylobacter. In the unadjusted model, the mean heightfor-age z score (HAZ) was found to be elevated at the endline but the mean weight-for-age z score (WAZ) and weight-for-height z score (WHZ) were reduced in the endline (Fig. 1) among the fecal Shigella and Campylobacter-positive under-5 children.
In Table 3, the findings of generalized estimating equations (GEE) modeling have been presented. A significant association between Shigella infection and WHZ was shown in the unadjusted GEE model. After adjusting for other covariates, namely: age, gender, MSD, breastfeeding status, mother's education, number of people regularly sleeping in the house, number of under-5 children at house, handwashing material, hand washing before nursing a child and after cleaning the child, access to toilet facility, the main source of drinking water, wealth index, co-pathogens Cryptosporidium and Giardia, comorbidity (pneumonia, diarrhea, dysentery, malaria, typhoid), the multivariable model revealed a significant negative association between WHZ [Coefficient: − 0.11 (95% CI − 0. 21 (Table 3). The analysis was replicated without adjusting the significant co-pathogens (Cryptosporidium and Giardia) and no difference in the adjusted mean growth was found (Table S3) Campylobacter-mediated enteritis cases, but the total direct cost was similar in both the Shigella-mediated and Campylobacter-mediated enteritis cases (Table S4).

Determinants of costs.
The relationship between the wealth index, gender, education, age, severity and duration, household direct medical costs and overall costs were investigated (Table 4). There was no difference in total costs for both the Shigella-mediated and Campylobacter-induced enteritis. Both enteritis indicated a higher cost with higher levels of maternal education, especially for household total costs in the case of Shigella-mediated enteritis (p = 0.046). Higher household total costs with a greater duration of hospital stay were observed in both cases (p < 0.05) ( Table 4).

Discussion
In our study, children with Shigellosis and Campylobacter infections presenting with dysentery were compared to fecal Shigella and Campylobacter negative children. Our findings were comparable to other studies 9-11 , despite being unable to exclude the co-pathogens responsible for dysentery in shigellosis and Campylobacter infections. Campylobacter-positive children reported a significantly lower incidence of fever at admission in comparison to Campylobacter-negative children. However, fever was more common in the case of Shigella-positive children compared to Shigella-negative children. It was probably due to mild, often self-limiting Campylobacter infections that needed only supportive treatment 12 . This observation was similar to a study conducted in the north of Israel 12 . Henceforth, fever on admission associated with dysentery will be helpful for the clinicians to differentiate between shigellosis and Campylobacter infections among under-5 children.
In our study, Shigella-positive children less often presented with sunken eyes. Findings from a study conducted in a large urban diarrhea treatment facility in Bangladesh reported frequent presentation of shigellosis with some or severe dehydration in children [13][14][15] . In the case of Campylobacter infection, we observed no association with inclusion criteria of MSD and our findings were consistent with a study among Canadian children, where dehydration was not reported to be a common presenting feature of Campylobacter-mediated enteritis 16 and a similar finding was observed in a large waterborne outbreak of Campylobacter jejuni in Norway 17 .
In our analysis, breastfeeding was found to be a protective factor for shigellosis. Very little is known about the impact of breastfeeding on Shigella-related diarrheal diseases. Another research from Bangladesh studied the children up to the first three years of age and found that breastfed children up to 35 months of age had a higher  www.nature.com/scientificreports/ level of immunity against severe shigellosis 18 . In rural Mozambique, breastfeeding was also found protective for diarrhea caused by Shigella 19 .
We observed a significant negative association of Shigella-mediated enteritis with weight-for-height z score. In other studies, researchers have indicated a similar type of results 7,20 . Malnourished children have been found to present with longer duration of illness and deep ulcerations in the colon. Both acute and prolonged episodes of shigellosis may result in extensive loss of blood from the colonic ulcerations 21 . Thus, in addition to the other Table 2. Results of multiple logistic regression after exploring the associated factors of Shigella and Campylobacter infections in MSD children aged less than 5 years in Bangladesh. Fever: measured at least 38 °C or parental perception, RBC, red blood cell; aOR, adjusted odds ratio; CI, confidence interval.  www.nature.com/scientificreports/ effects of diarrhea, shigellosis results in loss of serum protein that, for children on a marginal diet, must be compensated by increased protein intake for optimal growth to occur 22,23 . This fecal protein loss may have been partially responsible for growth faltering in children with shigellosis. However, in our study, we found no effect of shigellosis in limiting linear growth.
In several studies, researchers observed an association between Campylobacter infection and reduced weight gain as well as reduced linear growth 20,24 . However, in our study, we did not find any association of asymptomatic Campylobacter with child anthropometry indices. There was a positive association between symptomatic Table 3. Association of fecal Shigella and Campylobacter with a child's HAZ, WAZ, and WHZ: results of generalized estimating equations modeling (dependent variable-HAZ, WAZ, and WHZ). Coef., coefficient; CI, confidence interval; HAZ, height-for-age z score; WAZ, weight-for-age z score; and WHZ, weight-forheight z score. a Adjusted for age, gender, MSD, breastfeeding status, mother's education, number of people regularly sleep in the house, number of under-5 children at house, hand washing before nursing a child and after cleaning the child, handwashing material, main source of drinking water, available toilet facility, wealth index, copathogen: Cryptosporidium and Giardia, and comorbidity (malaria, typhoid, pneumonia, diarrhea, and dysentery). www.nature.com/scientificreports/ Campylobacter infection with child growth. This may be because of the treatment of symptomatic episodes with antibiotics. Thus, antibiotic treatment may be a confounding factor in estimating the true effect of Campylobacteriosis on child growth. Similar findings were observed in a systematic review and meta-analysis of Campylobacter infection 25 . In our study, stunting, wasting, or underweight, in any form, were not associated with invasive enteritis caused by Campylobacter or Shigella. Since the risk of diarrheal disease among severely malnourished children may be higher compared to that in the well-nourished children, our population may be less than ideally suited to disentangling this impact, limiting our ability to assess whether the association between Campylobacter and growth during enrollment was mediated by the nutritional deficit. In our study population, Cryptosporidium and Giardia were prevalent and they are known to influence the growth of children [26][27][28] . However, their impact can also be eliminated in this situation, as we controlled their effect during GEE modeling for both Campylobacter and Shigella infections. The analysis was repeated to those children not infected with Cryptosporidium and Giardia. The difference in adjusted mean growth was measured in terms of the HAZ, WAZ, and WHZ score, which was observed to be almost the same. Other studies which used a single assessment of nutritional status to establish a possible link between Shigella or Campylobacter with malnutrition have not been able to distinguish between the growth effect of Shigella or Campylobacter-mediated enteritis or the increased vulnerability of malnourished children to infection.
Medical costs differed by sex, with direct costs being higher for girls suffering from Shigella enteritis, and higher for boys suffering from Campylobacter enteritis, with no difference in overall costs between Shigella and Campylobacter-mediated enteritis. Given the evidence revealed in the literature that household spending on health care, food, and education sometimes favors boys over girls, these findings warrant further exploration 29,30 . We also observed evidence of lower total costs for children with lower maternal education levels in the case of Shigella-mediated enteritis. Literate mothers incurred higher costs for the treatment of Shigella infections in comparison to illiterate mothers. This was more likely because of illiteracy to curtail or prolong care-seeking. This brings with it the danger that delayed care leading to more adverse outcomes among illiterate mothers' children. We do not have any ready explanation for this observation but further studies may address these issues.
Total medical costs were amplified by the increased duration of hospital stay in cases of both Campylobacter and Shigella-mediated enteritis. Another study from Northern Ghana also reported higher hospital costs in inpatients than those who received outpatient treatment 31 .
Unbiased sampling following a standard protocol 32 , a large sample size 33 , and high-quality laboratory performance 32 were the strengths of our analysis. In this study, we aimed to determine the factors associated with both symptomatic and asymptomatic Shigella and Campylobacter infections among under-5 children. A single home follow-up visit approximately 60 days after enrollment was a valuable addition to the results of this research, which allowed us to understand the growth outcomes of children during the vulnerable times of their lives.
Nevertheless, our study has several limitations warranting a careful interpretation of the results when explaining these findings. It includes the inability to determine the relationship between maternal age and BMI, gestational age, and birth weight data for child growth failure. Due to a limited number of samples, we could not conclude the differential effects of Campylobacter and Shigella by species. This study did not evaluate the antimicrobial susceptibility patterns. Additionally, the cost of adverse outcomes and mental effects (such as distress and tiredness) of diarrheal disease caregivers have not been clarified in the current study. Moreover, the study was conducted in a sub-district of Bangladesh, so the results may not be generalizable for the whole country.
In conclusion, the use of clinical predictors may make it possible to target appropriate empiric antimicrobial therapy for children most likely to have invasive enteritis in resource-constrained settings. Our findings underscore the need for preventive strategies targeting Shigella, which could potentially reduce the disease burden and its sequelae such as child growth faltering during the first 5 years of life. Results also indicate the economic burden of households. Appropriate coping mechanisms may be undertaken to alleviate this burden. This may have public health implications particularly in the case of households with illiterate mothers or childhood invasive enteritis, mainly in the case of girls.

Method
Study site. Related data were extracted from the Global Enteric Multicenter Study (GEMS), Bangladesh site database 34 . The location of the GEMS Bangladesh site was in a rural community, situated in the Mirzapur subdistrict of Tangail, Bangladesh. Details about the study site have been reported elsewhere 34-36 . Study design and study participants. The design and methodology of the GEMS were mentioned earlier 36 . Briefly, data were extracted from cases and controls enrolled at the GEMS Bangladesh site, a threeyear research during December 2007 and March 2011. GEMS was a prospective matched case-control study conducted for 36 months at 7 sites where demographic surveillance systems (DSS) regularly updated censused populations. The sampling frame comprised children aged < 60 months residing within each site's DSS area. Children brought to sentinel health centres serving each DSS-respondent-children were assessed to match with the inclusion criteria for MSD irrespective of their socioeconomic status. Every fortnight, 8-9 cases per age stratum (0-11, 12-23, and 24-59 months) per site were targeted for enrolment. Within 14 days of each case enrolled, they undertook to enroll 1-3 randomly selected age-and sex-matched controls from the same or nearby communities [36][37][38] . The research had a well-defined standardized protocol for recruitment 36 . The published 37 , working hypothesis 35 , epidemiology 36 , clinical 39 , laboratory 36 , and statistical methods 40  Collection of stool samples and stool microbiology. Stool specimens for the GEMS were examined for every child at the time of enrolment using the GEMS laboratory procedure protocol 42-44 . Shigella and Campylobacter spp. isolation. Tests for isolation of both Campylobacter and Shigella spp.
used in GEMS have been described elsewhere 43 . Inclusion criteria for MSD. Every child was evaluated for diarrhea and study enrollment eligibility. The episode had to be current (initiated after around 7 days without diarrhea), acute (initiated within the previous 7 days), and at least one of the following characteristics for moderate-to-severe diarrhea (MSD) had to be met: sunken eyes (confirmed by parent or caretaker as more than usual; loss of skin turgor (abdominal skin pinch with slow [≤ 2 s] or very slow [> 2 s] recoil); intravenous rehydration administered or prescribed; dysentery (visible blood in loose stools); or hospitalized with diarrhea or dysentery 50 .

Variable of interest
Breastfeeding status. Breastfed referred to both exclusive and partially breastfeed children.

Socio-demographic information.
This involved data from the participant's household (defined as a group of people sharing a cooking fire) which included mother as a primary caretaker, education of mother (illiterate or literate), and size of household (including the number of children < 5 years of age, number of people regularly sleeping in the house). The explanatory variables were known to be building materials (cement or non-cement), the practice of handwashing (before nursing or preparing baby food; after handling animals, and cleaning a child), access and the main source of drinking water (tube well and non-tube well water), water treatment (water treatment method of drinking water available or not), improved sanitation facilities (an available toilet facility for disposal of human fecal waste or not), pets on the premises (sheep, goat, rodent/fowl, cow, dog, and cat), and methods for hand washing (water with soap or without soap).
Wealth index. Based on the wealth index quintiles (poor, lower middle, middle, upper middle, and rich), households were categorized into socio-economic status (SES) to determine potential associated factors for disease as well as indicators for constructing a wealth index for each site 36,51 .
Duration of hospital stay. The outcome was described by using a total duration of hospital stay (less than 4 days and ≥ 4 days).  Child growth. Only case-control (Campylobacter and Shigella positive and negative) sets of data on both enrollment and follow-up HAZ, WAZ, and WHZ for participants enrolled in GEMS were included in our study 36 . We used weighted means of baseline and endline HAZ, WAZ, and WAZ (n = 648 vs. 640) for Shigellapositive children; and HAZ, WAZ, and WAZ (n = 673 vs. 660) for Campylobacter-positive children irrespective of MSD from enrolment to follow-up respectively from GEMS Bangladesh site.
Household cost. Direct medical costs, direct non-medical costs, indirect costs, and overall costs per study child (fecal Campylobacter and Shigella positive MSD) were analyzed for care-seeking from the medical facilities for the treatment of a given episode of MSD, all of which were converted to the current US dollar rate. Direct medical expenses were classified as both informal and formal expenses, with the former representing treatment given by a local healer or pharmacist and the latter combining all health centers, hospitals, and licensed practitioners. Direct non-medical costs were split down by transport and other costs, while indirect costs were either time costs or other costs.
Statistical methods. Considering mean and standard deviation (SD) for continuous variables and frequency as a percentage to summarize the data for categorical variables, we reported the child, maternal, and household-level characteristics. Student's t test for continuous variables was performed to compare the mean differences, and changes in proportions were compared by the Chi-squared (χ 2 ) test. Since Shigella and Campylobacter infection were binary indicators, we performed multiple logistic regression analyses to identify the significantly associated factors of Shigella and Campylobacter infections in children aged < 5 years for Shigella and Campylobacter positive children having MSD. The covariates were adjusted for multiple logistic regression models using a stepwise forward selection method if associated with p value < 0.25 in the simple model 52 , whereas other relevant variables such as age and sex were adjusted for a p value < 0.25 due to biological as well as public health importance as more traditional levels such as 0.05 can fail in identifying variables known to be of importance. All covariates were included in the subsequent models to obtain an adjusted final model. Adjusted odds ratios (aOR) with a 95% confidence interval (CI) as the strength of the associations were determined from multiple logistic regression. The relationship of explanatory variables (presence of fecal Campylobacter and Shigella) with the continuous outcome variables (HAZ, WAZ, and WHZ) were examined longitudinally using generalized estimating equations (GEE) with exchangeable correlation and identity link function 53,54 . The variance inflation factor (VIF) was calculated to assess multicollinearity and no variable with a VIF > 5 was identified. Some cases incurred no expenditures for both medical and total costs, and the remainder created a right-skewed distribution. For MSD cases, descriptive statistics (means and SD) for costs were estimated. The analysis of variance (ANOVA) was used to independently assess the results of the economic status of households, maternal education, sex of a child, age, and duration of illness. We determined the strength of the association by estimating the coefficients and their 95% CIs. A probability of less than 0.05 was assumed statistically significant. All data were analyzed using STATA version 15.0 IC (College Station, TX: Stata Corp LLC).
Ethical consideration. The ethical committees and the respective research review boards at the University of Maryland School of Medicine and the committees overseeing each site and their collaborating partners from other institutions approved the clinical protocol, consent forms, case report forms, field methods, and other supportive materials prior to the start of the study. All methods were performed in accordance with the relevant guidelines and regulations. The signed informed consent for the inclusion of children in the study was obtained from the parents/guardians of the children (both sick MSD cases and healthy controls).