Am. J. Trop. Med. Hyg., 102(5), 2020, pp.

1124–1130
doi:10.4269/ajtmh.19-0769
Copyright © 2020 by The American Society of Tropical Medicine and Hygiene

Effect of Improved Water Quality, Sanitation, Hygiene and Nutrition Interventions on Respiratory
Illness in Young Children in Rural Bangladesh: A Multi-Arm Cluster-Randomized Controlled Trial
Sania Ashraf,1,2* Mahfuza Islam,1 Leanne Unicomb,1 Mahbubur Rahman,1 Peter J. Winch,2 Benjamin F. Arnold,3
Jade Benjamin-Chung,3 Pavani K. Ram,4 John M. Colford, Jr.,3 and Stephen P. Luby1,5
1
International Centre for Diarrhoeal Disease Research, Bangladesh (icddr,b), Dhaka, Bangladesh; 2Department of International Health, Johns
Hopkins Bloomberg School of Public Health, Baltimore, Maryland; 3Division of Epidemiology and Biostatistics, School of Public Health, University
of California Berkeley, Berkeley, California; 4School of Public Health and Health Professions, University at Buffalo, Buffalo, New York; 5Division of
Infectious Diseases and Geographic Medicine, Stanford University, Stanford, California

Abstract. Acute respiratory infections cause mortality in young children. We assessed the effects of water, sanitation,
hygiene (WASH) and nutritional interventions on childhood ARI. Geographic clusters of pregnant women from rural
Bangladesh were randomly assigned to receive 1) chlorinated drinking water and safe storage (W); 2) upgraded sanitation
(S); 3) handwashing promotion (H); 4) combined water, sanitation, and handwashing (WSH); 5) nutrition intervention
including lipid-based nutrient supplements; 6) combined WSH plus nutrition (WSHN); or 7) no intervention (control).
Masking of participants was not possible. Acute respiratory illness was defined as caregiver-reported persistent cough,
panting, wheezing, or difficulty breathing in the past 7 days among index children, those born to enrolled women. We
assessed outcomes at 12 and 24 months of intervention using intention to treat. Compared with children in the control
group (ARI prevalence, P: 8.9%), caregivers of index children reported significantly lower ARI in the water (P: 6.3%,
prevalence ratio (PR): 0.71; 95% CI: 0.53, 0.96), sanitation (P: 6.4%, PR: 0.75, 95% CI: 0.58, 0.96), handwashing (P: 6.4%,
PR: 0.68, 95% CI: 0.50, 0.93), and the combined WSH+N arms (P: 5.9%, PR: 0.67, 95% CI: 0.50, 0.90). Those in the nutrition
(P: 7.4%, PR: 0.84, 95% CI: 0.63, 1.10) or the WSH arm (P: 8.9%, PR: 0.99, 95% CI: 0.76, 1.28) reported similar ARI
prevalence compared with control children. Single targeted water, sanitation, and hygiene interventions reduced reported
respiratory illness in young children. There was no apparent respiratory health benefit from combining WASH interventions.

INTRODUCTION especially in already malnourished children.8 Combined school-

based interventions that improved water quality and sanitation
Acute respiratory infections (ARIs) are a leading cause of were associated with reductions in respiratory illness.9,10 These
morbidity and mortality in young children globally.1 Acute overlapping risk factors suggest that combining interventions that
respiratory infection and pneumonia cause the majority of improve nutrition and with those that improve water quality,
hospitalizations and death among children younger than 5 sanitation, and hygiene conditions in resource-poor settings
years especially in low-income countries.2 Risk factors for could lead to larger reductions in childhood illness compared with
pneumonia include low birth weight, malnutrition, low ex- each component alone.11
clusive breastfeeding rates, poor handwashing, crowding, Although the impact of improved WASH and nutrition on
use of solid fuels, and low maternal education, all of which are childhood respiratory health has been studied, their impact has
common in poor households.2 In low-income settings, ef- not been directly compared with each other individual in-
fective interventions include immunization against re- tervention or an intervention that combined WASH and nutrition
spiratory pathogens (measles, Haemophilus influenzae type in the same study population. Because combined interventions
B, and pneumococcus) and reducing indoor air pollution.3 are often more difficult and expensive to implement, compared
However, poor environmental conditions that support with single interventions, determining the relative health effects
transmission of respiratory pathogens can worsen childhood of each can help identify cost-effective strategies. We aimed to
morbidity.4 Water, sanitation, hygiene (WASH) interventions assess whether the effect of single WASH and nutrition inter-
that improve these conditions, therefore, have the potential ventions reduced respiratory illness in young children when
to reduce respiratory illness by interrupting pathogen delivered alone or in combination in the WASH Benefits trial in
transmission. rural Bangladesh.
Handwashing with water and/or soap effectively interrupts
transmission of respiratory pathogens through droplets and
fomites.5 Older observational studies noted reductions in child METHODS
mortality from pneumonia after the introduction of improved water
quality through centralized drinking water interventions.6 Nutrition Study design. The WASH Benefits Bangladesh study was a
interventions such as promoting exclusive breastfeeding or de- community-based cluster-randomized trial conducted in rural
livering vitamin A that boost a child’s immunity can alleviate villages in Gazipur, Kishoreganj, Mymensingh and Tangail
childhood morbidities including respiratory illness.7 Water, sani- districts. The study design and rationale were published earlier
tation, hygiene and nutrition interventions that reduce diarrheal (See Consolidated Standards of Reporting Trials checklist in
disease morbidity can also reduce pneumonia by preventing supporting documents).12
compromised immune responses or micronutrient deficiencies, It included six intervention arms and a double-sized control
arm. In Bangladesh, the unit of randomization was a group of
compounds visited by a single local promoter and separated
* Address correspondence to Sania Ashraf, Department of Interna-
by at least a 1-km buffer region to minimize the risk of spillover
tional Health, Johns Hopkins Bloomberg School of Public Health, between clusters. The clusters were block randomized into
615 N Wolfe St., Baltimore, MD 21205. E-mail: saniashraf@[Link] either one of the six intervention arms or the control arm.13

1124
 IMPACT ON CHILDHOOD RESPIRATORY ILLNESS 1125

The study protocol was approved by the Research and specific days in the last week. The study team did not collect
Ethical Review Committee at the International Centre for data on clinical signs of the severity of the respiratory syn-
Diarrhoeal Disease Research, Bangladesh (PR-11063), the drome. In this study, we used combinations of these three
University of California, Berkeley (2011-09-3652), and the In- reported symptoms to assess childhood respiratory illness.
stitutional Review Board at Stanford University (25863). Our main outcome of interest was a 7-day prevalence of acute
Participants. Research assistants screened rural com- respiratory illness (ARI), defined as caregiver-reported symptoms
pounds to identify eligible pregnant women in their first or second of persistent cough or panting, wheezing, or difficulty breathing (1
trimester who did not plan to move in the next 24 months. or 2) in the 7 days before the interview.14 In secondary analyses, we
Pregnant women who lived close to each other were enrolled into explored the impact of these interventions using alternate com-
the study following written informed consent from the compound binations of the measured symptoms: 7-day prevalence of only
head, the woman, and guardians of children younger than 3 panting, wheezing, or difficulty breathing (2) and ARI plus fever ([1
years. The children born to the enrolled pregnant women were or 2] and 3). Panting, wheezing, or difficulty breathing encompass
considered “index” children. We followed the closed cohort symptoms of asthma, bronchiolitis, or, occasionally, bacterial
longitudinally and measured symptoms of illness at 12 and pneumonia. Although this might reflect chronic conditions such as
24 months after initiating the intervention. asthma, these symptoms reflect a burden and a risk factor for
Randomization and masking. Blocks of eight adjacent respiratory illness morbidity. Exploring ARI plus fever could be
clusters were randomized into 1) chlorinated drinking water and indicative of more severe respiratory infection. Respiratory illness
safe water storage, 2) sanitation, 3) handwashing, 4) combined and these definitions were not prespecified for this trial.
water + sanitation + handwashing (WSH), 5) nutrition, 6) combined Trained field surveyors who were not involved in the delivery
nutrition + WSH, or the 7) nonintervention control group. The of the interventions interviewed the mother of the index child to
control arm was double sized to improve precision of estimates collect data on respiratory symptoms. We included caregiver-
when compared with multiple arms. An offsite investigator (B. F. reported abrasion or bruising as negative control outcomes.15
A.) used a random number generator to block randomize these Outcomes were measured approximately 12 and 24 months
clusters. This trial was designed as a pair-matched, cluster- following intervention roll out. These outcome evaluations
randomized trial. This was a geographically pair-matched design were spread out over the entire year because of the long du-
meaning any comparison between two arms is pair-matched ration of enrollment. Intervention adherence was assessed by
within the randomization block. The participants were unaware of a separate team at regular intervals using structured obser-
their intervention group assignment until after the baseline survey vations and objective measures.16
and randomization. Because the intervention included distribution Statistical analyses. The sample size calculation for this
of products and related promotion by community health pro- trial was based on primary outcomes, diarrhea, and child
moters, masking of the subjects or the data collectors was not growth. It assumed a relative risk of diarrhea of 0.7 or smaller,
possible. The research team who implemented the intervention with 10% prevalence in the control group, and a difference of
was separate from the data collection team. The analysis was 0.15 length for age Z-score between the intervention and
carried out using re-randomized uninformative assignments to control groups, adjusting for repeated measures within clus-
enable masked statistical analyses from raw datasets. Results ters. Other assumptions were type I error (α) of 0.05, power
were unmasked once statistical analysis was completed. (1−β) of 0.8, and a 10% dropout after baseline. The control arm
Procedures. The interventions were described in detail was double sized to account for multiple hypothesis tests.12
previously.12,13 Interventions were delivered at the household We conducted an intention to treat analysis in which we
level or the compound level and included 1) chlorine tablets and compared each intervention arm against the control arm.
safe storage vessel; 2) upgrades to dual-pit latrines with water Because the nutrition intervention provided supplements only
seals for all households in the compound and provision of child to the index child, we restricted the analysis to index children
potties and sani-scoops to index households; 3) handwashing for all arms. We conducted two subgroup analyses 1) stratified
stations with soapy water detergent and bottles near the by child gender because male children might be more sus-
kitchen and the latrine delivered to index households; 4) age- ceptible to respiratory illness3 and 2) by survey round (Year 1
appropriate nutrition from birth to 24 months including a supply and Year 2) to examine differences in intervention effects on
of lipid-based nutrient supplements (6–24 months) in addition the prevalence of reported respiratory illness overtime. We
to exclusive breastfeeding and maternal and infant nutrition also compared the impact on outcomes between combined
recommendations to mothers and the index child; 5) combined WSH and individual arms and the nutrition plus WSH (WSHN)
WSH; and 6) combined WSH plus nutrition (WSH+N). Local arm and WSH and the nutrition-only arm.
women from the community were recruited and trained as The analysis followed the procedures used in the primary
promoters who conducted household visits and community outcome analysis (pre-registered analysis protocol https://
discussions to promote the interventions based on a behavior- [Link]/wvyn4/). The pair-matched design ensured that the
change strategy. These promotions included interactive ses- calendar time of the measurements (season) was balanced
sions to develop collaborative solutions with the participants to across treatment groups.
continue their improved practices. The promoters were paid a We used a generalized log linear regression model to estimate
monthly stipend of approximately USD 20. Control arms did not the effect of each intervention compared with the control group. To
receive any hardware/products or promoter visits. estimate adjusted prevalence ratios (PRs), we included pre-
Outcomes. In this study, we assessed the impact on re- specified covariates that were associated with the outcome based
spiratory outcomes in index children as reported by the pri- on a likelihood ratio test (P < 0.2). Potential covariates included field
mary caregiver. We asked the primary caregiver to recall if the staff who collected data, including the month of measurement,
index child had the following symptoms: 1) persistent cough, household food insecurity, child age, child gender, mother’s age,
2) panting/wheezing/difficulty breathing, or 3) fever during mother’s height, mother’s education level, number of children < 18
1126 ASHRAF AND OTHERS

years in the household, number of individuals living in the com- households reported that they did not face food insecurity.
pound, distance in minutes to the primary water source, household Symptoms for respiratory illness was assessed for 4,747 index
roof, floor, wall materials, and household assets. Analyses were children at the 12-month follow-up (mean age: 0.73 years, SD:
carried out with R, version 3.2.4 (R Foundation for Statistical 0.14) and 4,667 index children (mean age: 1.87 years, SD:
Computing, Vienna, Austria) and STATA 13.0 (Stata Corp LP, 0.17) at the 24-month follow-up.
College Station, TX). This trial achieved high adherence to all interventions.13,16
The trial is registered with [Link], NCT01590095. All measures suggested marked differences in promoted be-
Independent data safety monitoring boards in Bangladesh haviors from the control group at both Year 1 and Year 2, with
oversaw the trial. adherence over 75% in the single intervention group and the
combined intervention groups.
RESULTS Compared with the control group (8.9%), the reported
prevalence of ARI in index children was lower in the water
Field-workers recruited participants from 5,551 com- (6.3%; PR: 0.71, 95% CI: 0.55, 0.91), sanitation (6.4%; PR:
pounds to form 720 clusters of pregnant women. Between 0.72, 95% CI: 0.56, 0.92), handwashing (6.4%; PR: 0.68, 95%
May 2012 and July 2013, we randomly allocated clusters to CI: 0.52, 0.88), and the combined WSH+N arms (5.9%; PR:
one of six interventions or the double-sized control arm 0.66, 95% CI: 0.51, 0.86) at 1-year and 2-year follow-ups
(Figure 1). Loss to follow-up included no live births (n = 361), (Figure 2). Notably, the impact observed in WSH+N was
death of index child (n = 235), relocation (n = 375), withdrawal similar to that in the single WSH arms. In our study, the re-
(n = 296), and absence during assessments (n = 182) (Figure 1). ported ARI prevalence in index children from the nutrition
Treatment groups were balanced at baseline on demographic (7.4%; PR: 0.82, 95% CI: 0.64, 1.04) or the combined WSH
characteristics, household composition, facilities and prac- arm (8.9%; PR: 0.99, 95% CI: 0.79, 1.23) was not significantly
tices relating to the use of cooking fuel, drinking water, lower than those in the control group.
handwashing, and sanitation (Table 1). Specifically, the aver- Prespecified adjusted analyses resulted in similar effect
age number of household members was five. A majority used a estimates of interventions on reported ARI in index children
shallow tube well for their drinking water (74%). On average, across all measures (Supplemental Table 1). Children in single
less than a third of the households owned hygienic latrines WASH and combined WSH plus nutrition arms had lower
with functional water seals (29%). Availability of water or soap prevalence of reported ARI than those randomly assigned to
was low near the toilet or the kitchen. Overall, 69% of the combined WSH group (Supplemental Table 2).

FIGURE 1. Summary of participant enrollment, randomization, retention, and analysis populations for respiratory outcomes, that is, index
children.
 IMPACT ON CHILDHOOD RESPIRATORY ILLNESS 1127

TABLE 1
Baseline characteristics across intervention arms
Water + sanitation + Nutrition + water + sanitation +
Control Water Sanitation Handwashing handwashing Nutrition handwashing

No. of households N = 1,382 N = 698 N = 696 N = 688 N = 702 N = 699 N = 686

Maternal
Age (years) 23.6 (5.0) 23.7 (5.2) 23.7 (5.2) 23.8 (5.5) 24.3 (5.5) 23.7 (5.1) 23.8 (5.5)
Years of education 5.9 (3.4) 5.8 (3.4) 5.8 (3.5) 5.8 (3.3) 5.9 (3.3) 5.8 (3.5) 5.6 (3.5)
Paternal
Years of education 4.9 (4.0) 4.9 (4.1) 5.0 (4.2) 4.6 (4.1) 5.0 (4.2) 4.8 (4.0) 4.7 (3.9)
Works in agriculture 414 (30%) 224 (32%) 204 (29%) 249 (36%) 216 (31%) 232 (33%) 207 (30%)
Household
Number of people 4.7 (2.3) 4.6 (2.2) 4.7 (2.1) 4.7 (2.2) 4.7 (2.1) 4.7 (2.2) 4.7 (2.1)
Has electricity 784 (57%) 422 (60%) 408 (59%) 405 (59%) 426 (61%) 409 (59%) 412 (60%)
Has a cement floor 145 (10%) 82 (12%) 85 (12%) 55 (8%) 77 (11%) 67 (10%) 72 (10%)
Acres of agricultural land 0.15 (0.21) 0.14 (0.20) 0.14 (0.22) 0.14 (0.20) 0.15 (0.23) 0.16 (0.27) 0.14 (0.38)
owned
Drinking water
Tube well as primary 1,038 (75%) 500 (72%) 519 (75%) 482 (70%) 546 (78%) 519 (74%) 504 (73%)
water source
Stored water observed at 666 (48%) 353 (51%) 341 (49%) 347 (50%) 304 (43%) 301 (43%) 331 (48%)
home
Sanitation
Daily defecation in the open
Adult men 97 (7%) 39 (6%) 52 (8%) 64 (9%) 54 (8%) 59 (9%) 50 (7%)
Adult women 62 (4%) 18 (3%) 33 (5%) 31 (5%) 29 (4%) 39 (6%) 24 (4%)
Children aged 8 to < 15 53 (10%) 25 (9%) 28 (9%) 43 (15%) 30 (10%) 23 (8%) 28 (10%)
years
Children aged 3 to < 8 267 (38%) 141 (37%) 137 (38%) 137 (39%) 137 (38%) 129 (39%) 134 (37%)
years
Children aged 0 to < 3 245 (82%) 112 (85%) 117 (84%) 120 (85%) 123 (79%) 128 (85%) 123 (88%)
years
Latrine
Owned 750 (54%) 363 (52%) 374 (54%) 372 (54%) 373 (53%) 377 (54%) 367 (53%)
Concrete slab 1,251 (95%) 644 (95%) 610 (92%) 613 (93%) 620 (93%) 620 (94%) 621 (94%)
Functional water seal 358 (31%) 183 (31%) 177 (30%) 162 (28%) 152 (26%) 183 (31%) 155 (27%)
Visible stool on slab or 625 (48%) 350 (53%) 332 (52%) 335 (52%) 289 (44%) 331 (51%) 298 (46%)
floor
Owned a potty 61 (4%) 27 (4%) 28 (4%) 35 (5%) 27 (4%) 36 (5%) 30 (4%)
Human feces observed
In the house 114 (8%) 65 (9%) 56 (8%) 70 (10%) 48 (7%) 58 (8%) 49 (7%)
In the child’s play area 21 (2%) 6 (1%) 6 (1%) 8 (1%) 7 (1%) 8 (1%) 7 (1%)
Handwashing
Within six steps of latrine
Has water 178 (14%) 83 (13%) 81 (13%) 63 (10%) 67 (10%) 62 (10%) 72 (11%)
Has soap 88 (7%) 50 (8%) 48 (8%) 34 (5%) 42 (7%) 32 (5%) 36 (6%)
Within six steps of kitchen
Has water 118 (9%) 51 (8%) 51 (8%) 45 (7%) 61 (9%) 61 (9%) 60 (9%)
Has soap 33 (3%) 18 (3%) 14 (2%) 13 (2%) 15 (2%) 23 (3%) 18 (3%)
Nutrition
Household is food 932 (67%) 495 (71%) 475 (68%) 475 (69%) 482 (69%) 479 (69%) 485 (71%)
secure*
Data are expressed in n (%) or mean (SD). Percentages were estimated from slightly smaller denominators than those shown at the top of the table for the following variables because of missing
values: father works in agriculture, open defecation, latrine has a concrete slab, latrine has a functional water seal, visible stool on the latrine slab or floor, ownership of child potty, observed feces in
the house or child’s play area, and handwashing variables.
* Assessed by the Household Food Insecurity Access Scale.

In secondary analyses, we observed a similar impact using the but there was no significant difference in the PRs across
more specific outcome, where reported ARI plus fever in index gender (Table 2). We found no differences in the effect of in-
children was lower in the water treatment (3.4%, PR: 0.65, 95% terventions in Year 1 versus Year 2 except in the water and
CI: 0.46, 0.93), sanitation (2.8%, PR: 0.54, 95% CI: 0.37, 0.78), WSH+N arms, where the impact on ARI and ARI plus fever was
handwashing (3.3%, PR: 0.63, 95% CI: 0.44, 0.91), and the higher in Year 2 (Supplemental Table 3). The prevalence of ARI
combined WSH+N (2.9%, PR: 0.58, 95% CI: 0.40, 0.84) than varied over the intervention period (Supplemental Figure 1).
those in the control arm (5.7%) (Figure 2). However, we observed
a reduction in the reported prevalence of panting, wheezing, or DISCUSSION
difficulty breathing only among index children from the hand-
washing (198%, PR: 0.58, 95% CI: 0.36, 0.91) and WSH+N In this cluster-randomized trial, reported respiratory illness
arms (2.4%, 95% CI: 0.44, 1.03) compared with the control (ARI) among index children was significantly lower in house-
group (3.5%) holds that received the sanitation intervention that included
In subgroup analyses, we found that the prevalence of these regular promotion plus individual latrines, potties, and scoops
respiratory illness symptoms was lower in females than males, (28% lower); or chlorinated drinking water intervention (30%),
1128 ASHRAF AND OTHERS

FIGURE 2. Intervention effects on the 7-day prevalence of respiratory illness in index children; 1- and 2-year assessments combined in
Bangladesh. Acute respiratory illness (ARI) defined as mothers’ reports of persistent cough or panting, wheezing, or difficulty breathing in the
past 7 days among index children. Data are prevalence ratios compared with the children in the control group, with 95% CIs. C = control; H =
handwashing; S = sanitation; W = water; WSH = combined water, sanitation, and handwashing; WSHN = water, sanitation, handwashing, and
nutrition.

handwashing intervention alone (32%); or all in combination consistently reduced with simple handwashing interventions,
along with nutritional supplements (34%) than those in control as promoted within this trial. This study additionally demon-
households (prevalence: 8.8%). Children randomly assigned strates the effectiveness of the handwashing intervention,
to nutrition interventions or combined water, sanitation, and where homemade soapy water with free detergent refills was
hygiene interventions did not experience fewer respiratory promoted with free handwashing stations near the latrine
illnesses than children in the control arms. and kitchen.
For handwashing interventions, these findings reinforce Prior studies report mixed results on the impact of sanitation
well-known protective effects of handwashing on respiratory and water interventions on respiratory illness.17,18 Our study
illness by interrupting pathogen transmission through findings add to the literature that demonstrate a reduction in
hands.5 Viral infections that are predominantly spread respiratory illness in children from sanitation interventions.19
by fomite contact through hands, eyes, or noses are Given that we observed a significant reduction in diarrheal

TABLE 2
Unadjusted respiratory outcome PRs by child gender, interventions vs. control, among index children in Bangladesh: 1- and 2-year follow-up
combined
Males Females

Outcome/arm N Prev PR (95% CI) N Prev PR (95% CI) Interaction, P-value

Cough or difficulty breathing (ARI)

Control 1,131 9.81 Ref 1,157 7.78 Ref –
Water 602 7.64 0.76 (0.53, 1.06) 606 4.85 0.64 (0.42, 0.96) 0.45
Sanitation 591 6.94 0.69 (0.48, 0.98) 585 5.81 0.75 (0.51, 1.09) 0.73
Handwashing 578 6.40 0.64 (0.43, 0.96) 584 5.65 0.72 (0.48, 1.09) 0.65
WSH 618 9.22 0.94 (0.68, 1.29) 576 8.51 1.08 (0.78, 1.50) 0.54
Nutrition 593 9.11 0.90 (0.68, 1.18) 566 5.65 0.71 (0.48, 1.07) 0.23
WSH+nutrition 559 6.26 0.63 (0.41, 0.96) 638 5.64 0.72 (0.50, 1.03) 0.61
Panting, wheezing, or difficulty breathing
Control 1,131 4.07 Ref 1,157 2.94 Ref –
Water 602 4.15 0.93 (0.55, 1.56) 606 2.31 0.82 (0.43, 1.55) 0.74
Sanitation 591 3.89 0.89 (0.54, 1.50) 585 1.20 0.43 (0.19, 0.97) 0.13
Handwashing 578 2.60 0.56 (0.33, 0.30) 584 1.37 0.53 (0.25, 1.15) 0.92
WSH 618 4.21 0.99 (0.57, 1.17) 576 3.13 1.05 (0.61, 1.81) 0.89
Nutrition 593 3.88 0.86 (0.46, 1.54) 566 1.94 0.71 (0.36, 1.42) 0.63
WSH+nutrition 559 2.86 0.62 (0.32, 1.20) 638 2.04 0.73 (0.39, 1.13) 0.76
Fever and ARI
Control 1,131 5.75 Ref 1,157 4.41 Ref –
Water 602 4.32 0.75 (0.46, 1.24) 606 2.48 0.53 (0.30, 0.94) 0.33
Sanitation 591 3.05 0.52 (0.30, 0.90) 585 2.56 0.56 (0.31, 1.00) 0.88
Handwashing 578 3.46 0.60 (0.35, 1.03) 584 3.08 0.67 (0.36, 1.25) 0.77
WSH 618 4.05 0.89 (0.44, 1.08) 576 4.69 1.09 (0.33, 0.74) 0.22
Nutrition 593 5.40 0.92 (0.62, 1.34) 566 3.00 0.65 (0.36, 1.16) 0.29
WSH+nutrition 559 3.58 0.53 (0.30, 0.95) 638 2.35 0.64 (0.41, 0.98) 0.66
ARI = acute respiratory illness; PR = prevalence ratio; WSH = water, sanitation, and handwashing.
 IMPACT ON CHILDHOOD RESPIRATORY ILLNESS 1129

diseases in the sanitation arm in this trial, it is possible that conditions was masked to the intervention assignment.
the children potentially benefitted from lower respiratory ill- Therefore, respiratory illness measured through caregiver-
ness through stronger immune systems and adequate reported symptoms is subject to courtesy bias. The
micronutrient levels.8,13 We also observed lower fecal in- direction of courtesy bias in households that receive inter-
dicator bacteria in food and stored water in the improved ventions is known to inflate health impact when outcome is
water arm in this trial, suggesting reduced contamination based on the caregiver-reported prevalence of disease.27
along direct transmission pathways.20 Reduction in ARI from Respiratory illness unlike diarrheal disease is less likely to
improved water quality requires further investigation into be directly linked to our interventions by the study respon-
exposure and transmission of relevant waterborne patho- dents. Moreover, courtesy bias would not be expected to
gens. Further research into reliable objective biomarkers that affect reports in the single water treatment, sanitation, and
can be used in community-based studies could improve handwashing promotion arms, but not in the combined
pneumonia diagnoses in respondents with cough or non- arms. We also found no evidence of bias using negative
specific symptoms.21 control outcomes in this study, suggesting that differential
Malnourished children are at a higher risk of infection in- outcome reporting bias was unlikely.13
cluding respiratory illness.2 The WASH Benefits trial de- Second, in the absence of clinical assessments of symp-
livered lipid nutrient supplements (LNS) for children between toms, we defined our primary outcome (ARI) broadly as cough
6 and 24 months while promoting breastfeeding practices or panting, wheezing, or difficulty breathing. This did not allow
and providing micronutrient-rich complementary food. us to detect changes in more severe respiratory illness such as
Children in the nutrition intervention groups were taller and pneumonia or allow us to compare our estimates with studies
had higher weight-for-height Z scores than the children in that use the WHO definition of pneumonia. We detected
control households, indicating better nutritional status.13 In somewhat stronger effects in the most specific assessment of
analyses published elsewhere, compared with the control ARI plus fever, suggesting that these interventions likely impact
group, children in the nutrition arms (N and WSHN) had severe respiratory illness such as pneumonia (Figure 2). Re-
higher prevalence of meeting the minimum dietary diversity ported symptoms such as cough, panting, or wheezing or
score through complementary feeding, which was promoted shortness of breath in young children are nonspecific and can
alongside LNS.22 In a subsample at 3 months, 51–55% of indicate noninfectious causes such as asthma. However, these
women reported exclusive breastfeeding their children in the symptoms reflect a burden of illness to a child’s immune system
last 24 hours compared with 18% in the control group.23 We and may increase the risk of pneumonia. The impact on ARI is
did not observe a significant reduction in reported re- likely to reflect a genuine interruption of respiratory pathogen
spiratory illness in children from households that received transmission because the effect was consistent when assess-
nutrition supplements. When compared with children in ing more specific respiratory illness (reported fever plus ARI) in
control households, those in the single nutrition arm had an the water, improved sanitation, handwashing, and WSH+N in-
18% lower prevalence of reported respiratory illness (ARI), terventions compared with the control households.
but the difference was not significant in this trial. A non- Water, sanitation, and hygiene interventions that achieved
significant reduction in caregiver-reported respiratory illness high uptake reduced respiratory illness in young children in
morbidity in children following LNS is consistent with results rural Bangladesh. The same benefit was observed when wa-
from other studies.24,25 We, however, report a significant ter, sanitation, and hygiene interventions were successfully
reduction in reported respiratory illness when nutrient sup- integrated with nutrition interventions. We did not find any
plements were delivered in households that also received additive benefit of combining multiple components of WSH in
improved water, sanitation, and hygiene interventions. Im- this study. These findings provide further support for multiple
proving nutritional status of young children may be in- health benefits of water, sanitation, handwashing, and nutri-
sufficient to impact respiratory illness in highly contaminated tion interventions.
environments.
We have no satisfying explanation for why the combined Received October 16, 2019. Accepted for publication January 23,
WSH package did not reduce ARI similar to individual W, S, 2020.
and H components. One hypothesis could have been that Published online February 24, 2020.
implementation of several interventions together resulted in
Note: Supplemental material, tables, and figure appear at www.
lower adherence in the combined arm, but this was not cor- [Link].
roborated by measures of adherence16 or in patterns ob-
served in other infectious disease outcomes.13 Notably, we Acknowledgments: icddr,b acknowledges the time of the study par-
ticipants and the dedication of the field team to achieve the aims of this
found a significant reduction in reported ARI and fever plus ARI study.
in combined WSH+N households, suggesting implementation
Financial support: This research was funded by Global Development
of or adherence to a more complex, combined intervention as grant OPPGD759 from the Bill & Melinda Gates Foundation to the
not a limiting factor here. In any case, our results contribute to University of California, Berkeley, CA. S. P. L., S. A., M. I., B. F. A., and
findings from other studies that did not detect additive bene- J. M. C. report grants from the Bill & Melinda Gates Foundation during
fits to child health from combining WSH interventions.26 Our the conduct of the study. P. K. R. reports grants from Leland Stanford
failure to detect added benefits from combined WASH inter- University during the conduct of the study for support to the WASH
Benefits project. M. R. reports grants and nonfinancial support from
ventions over single interventions suggests that future studies the Bill & Melinda Gates Foundation (through a subcontract from UC
or programs should consider single targeted interventions to Berkeley) during the conduct of the study.
be cost effective. Disclaimer: The funder reviewed the design of the study but was
This study has several limitations. Neither the respondent not involved in data collection, analyses, or interpretation of the
nor the data collector who surveyed the household results.
1130 ASHRAF AND OTHERS

Authors’ addresses: Sania Ashraf and Peter J. Winch, Department of 12. Arnold BF et al., 2013. Cluster-randomised controlled trials of
International Health, Johns Hopkins Bloomberg School of Public individual and combined water, sanitation, hygiene and nutri-
Health, Baltimore, MD, E-mails: sashraf2@[Link] and pwinch@ tional interventions in rural Bangladesh and Kenya: the WASH
[Link]. Leanne Unicomb, Infectious Diseases Division, In- benefits study design and rationale. BMJ Open 3: e003476.
ternational Centre for Diarrhoeal Disease Research, Bangladesh 13. Luby SP et al., 2018. Effects of water quality, sanitation, hand-
(icddr,b), Dhaka, Bangladesh, E-mail: leanne@[Link]. Mahfuza washing, and nutritional interventions on diarrhoea and child
Islam and Mahbubur Rahman, Centre for Communicable Diseases growth in rural Bangladesh: a cluster randomised controlled
(CCD), International Centre for Diarrhoeal Disease Research, Ban- trial. Lancet Glob Health 6: e302–e315.
gladesh (icddr,b), Dhaka, Bangladesh, E-mails: mi_sheuli@[Link] 14. Feikin DR, Olack B, Bigogo GM, Audi A, Cosmas L, Aura B, Burke
and mahbubr@[Link]. Benjamin F. Arnold, Jade Benjamin- H, Njenga MK, Williamson J, Breiman RF, 2011. The burden of
Chung, and John M. Colford, Division of Epidemiology and Bio- common infectious disease syndromes at the clinic and
statistics, University of California Berkeley, Berkeley, CA, E-mails: household level from population-based surveillance in rural and
[Link]@[Link], jadebc@[Link], and jcolford@berkeley. Urban Kenya. PLoS One 6: e16085.
edu. Pavani K. Ram, School of Public Health and Health Professions, 15. Arnold BF, Ercumen A, Benjamin-Chung J, Colford JM, 2016.
University of Buffalo, Buffalo, NY, E-mail: pkram@[Link]. Brief report: negative controls to detect selection bias and
Stephen P. Luby, Division of Infectious Diseases and Geo- measurement bias in epidemiologic studies. Epidemiology 27:
graphic Medicine, Stanford University, Stanford, CA, E-mail: sluby@ 637–641.
[Link]. 16. Parvez SM et al., 2018. Achieving optimal technology and be-
havioral uptake of single and combined interventions of water,
This is an open-access article distributed under the terms of the
sanitation hygiene and nutrition, in an efficacy trial (WASH
Creative Commons Attribution (CC-BY) License, which permits un-
benefits) in rural Bangladesh. Trials 19: 358.
restricted use, distribution, and reproduction in any medium, provided
17. Huda TMN, Unicomb L, Johnston RB, Halder AK, Yushuf Sharker
the original author and source are credited.
MA, Luby SP, 2012. Interim evaluation of a large scale sanita-
tion, hygiene and water improvement programme on childhood
REFERENCES diarrhea and respiratory disease in rural Bangladesh. Soc Sci
Med 75: 604–611.
1. Liu L, Oza S, Hogan D, Perin J, Rudan I, Lawn JE, Cousens S, 18. Arnold B, Arana B, Mäusezahl D, Hubbard A, Colford John MJ,
Mathers C, Black RE, 2015. Global, regional, and national 2009. Evaluation of a pre-existing, 3-year household water
causes of child mortality in 2000–13, with projections to inform treatment and handwashing intervention in rural Guatemala. Int
post-2015 priorities: an updated systematic analysis. Lancet J Epidemiol 38: 1651–1661.
385: 430–440. 19. Patil SR, Arnold BF, Salvatore AL, Briceno B, Ganguly S, Colford
2. Rudan I et al., 2013. Epidemiology and etiology of childhood JM Jr., Gertler PJ, 2015. The effect of India’s total sanitation
pneumonia in 2010: estimates of incidence, severe morbidity, campaign on defecation behaviors and child health in rural
mortality, underlying risk factors and causative pathogens for Madhya Pradesh: a cluster randomized controlled trial. Plos
192 countries. J Glob Health 3: 010401. Med 11: 10–12.
3. Jackson S, Mathews KH, Pulanic D, Falconer R, Rudan I, 20. Ercumen A et al., 2018. Effects of single and combined water,
Campbell H, Nair H, 2013. Risk factors for severe acute lower sanitation and handwashing interventions on fecal contami-
respiratory infections in children: a systematic review and meta- nation in the domestic environment: a cluster-randomized
analysis. Croat Med J 54: 110–121. controlled trial in rural Bangladesh. Environ Sci Technol 52:
4. Humphrey JH, 2009. Child undernutrition, tropical enteropathy, 12078–12088.
toilets, and handwashing. Lancet 374: 1032–1035. 21. Sungurlu S, Balk RA, 2018. The role of biomarkers in the diagnosis
5. Aiello AE, Coulborn RM, Perez V, Larson EL, 2008. Effect of hand and management of pneumonia. Clin Chest Med 39: 691–701.
hygiene on infectious disease risk in the community setting: a 22. Jannat K, Luby SP, Unicomb L, Rahman M, Winch PJ, Parvez SM,
meta-analysis. Am J Public Health 98: 1372–1381. Das KK, Leontsini E, Ram PK, Stewart CP, 2019. Complementary
6. Sedgwick W, Macnutt J, 1910. On the Mills-Reincke phenomenon feeding practices among rural Bangladeshi mothers: results from
and Hazen’s theorem concerning the decrease in mortality from WASH Benefits study. Matern Child Nutr 15: e12654.
diseases other than typhoid fever following the purification of 23. Lin A et al., 2019. Effects of water, sanitation, handwashing, and
nutritional interventions on environmental enteric dysfunction
public water-supplies. J Infect Dis 7: 489–564.
in young children: a cluster-randomized controlled trial in rural
7. Bhutta ZA, Das JK, Walker N, Campbell H, Rudan I, Black RE,
Bangladesh. Clin Infect Dis ciz291.
2013. Interventions to address deaths from childhood pneu-
24. Mangani C, Ashorn P, Maleta K, Phuka J, Thakwalakwa C, Dewey
monia and diarrhoea equitably: what works and at what cost?
K, Manary M, Puumalainen T, Cheung YB, 2014. Lipid-based
Lancet 381: 1417–1429.
nutrient supplements do not affect the risk of malaria or re-
8. Ashraf S, Hamidul Huque M, Kenah E, Agboatwalla M, Luby SP,
spiratory morbidity in 6-to 18-month-old Malawian children in a
2013. Effect of recent diarrhoeal episodes on risk of pneumonia
randomized controlled trial 1–3. J Nutr 144: 1835–1842.
in children under the age of 5 years in Karachi, Pakistan. Int J 25. Bendabenda J, Alho L, Ashorn U, Cheung YB, Dewey KG, Vosti
Epidemiol 42: 194–200. SA, Phuka J, Maleta K, Ashorn P, 2016. The effect of provid-
9. Trinies V, Garn JV, Chang HH, Freeman MC, 2016. The impact of a ing lipid-based nutrient supplements on morbidity in rural
school-based water, sanitation, and hygiene program on ab- Malawian infants and young children: a randomized controlled
senteeism, diarrhea, and respiratory infection: a matched- trial. Public Health Nutr 19: 1893–1903.
control trial in Mali. Am J Trop Med Hyg 94: 1418–1425. 26. Fewtrell L, Kaufmann RB, Kay D, Enanoria W, Haller L, Colford JM,
10. Patel MK et al., 2012. Impact of a hygiene curriculum and the 2005. Water, sanitation, and hygiene interventions to reduce
installation of simple handwashing and drinking water stations diarrhoea in less developed countries: a systematic review and
in rural Kenyan primary schools on student health and hygiene meta-analysis. Lancet Infect Dis 5: 42–52.
practices. Am J Trop Med Hyg 87: 594–601. 27. Wood L, Egger M, Gluud LL, Schulz KF, Jüni P, Altman DG, Gluud
11. Schlaudecker EP, Steinhoff MC, Moore SR, 2011. Interactions of C, Martin RM, Wood AJ, Sterne JA, 2008. Empirical evidence of
diarrhea, pneumonia, and malnutrition in childhood: recent bias in treatment effect estimates in controlled trials with dif-
evidence from developing countries. Curr Opin Infect Dis 24: ferent interventions and outcomes: meta-epidemiological
496–502. study. BMJ 336: 601–605.