WPS6702 Policy Research Working Paper 6702 A Randomized, Controlled Study of a Rural Sanitation Behavior Change Program in Madhya Pradesh, India Sumeet R. Patil Benjamin F. Arnold Alicia Salvatore Bertha Briceno John M. Colford, Jr. Paul J. Gertler The World Bank Sustainable Development Network Water and Sanitation Program November 2013 Policy Research Working Paper 6702 Abstract Poor sanitation and open defecation are thought to be a Nevertheless, the Total Sanitation Campaign successfully major cause of diarrhea and intestinal parasite infections increased toilet coverage by 19 percent in intervention among young children. In 1999, India launched the villages compared with control villages (41 percent v. Total Sanitation Campaign with the goal of achieving 22 percent), while reported open defecation decreased universal toilet coverage in rural India by 2012. This by 10 percent among adults (74 percent v. 84 percent). paper reports on a cluster-randomized, controlled trial The intervention also led to some improvements in that was conducted in 80 rural villages in Madhya water quality and protozoan infection, but consistent Pradesh to measure the effect of the program on toilet improvements were not observed across multiple child access, sanitation behavior, and child health outcomes. health outcomes (diarrhea, helminth infections, child The study analyzed a random sample of 3,039 households growth). However, the exposure period was likely to and 5,206 children under five years of age. Field staff have been too short to result in any benefit of the collected baseline measures of sanitation conditions, sanitation interventions on child health. Given the behavior, and child health, and re-visited households 21 large improvements in toilet construction documented, months later. The analysis finds that implementation an additional follow-up survey with a longer period of of the program activities was slower than the original exposure would yield valuable information on the effects timeline (only 35 percent of villages were triggered of improved sanitation conditions on health outcomes. more than six months before the follow-up survey). This paper is a product of the Water and Sanitation Program (WSP), Sustainable Development Network. WSP is a multi- donor partnership created in 1978 and administered by the World Bank to support poor people in obtaining affordable, safe, and sustainable access to water and sanitation services. The paper is part of a larger effort by the World Bank to provide open access to its research and make a contribution to development policy discussions around the world. Policy Research Working Papers are also posted on the Web at http://econ.worldbank.org. The corresponding author may be contacted at srpatil@neerman.org. The Policy Research Working Paper Series disseminates the findings of work in progress to encourage the exchange of ideas about development issues. An objective of the series is to get the findings out quickly, even if the presentations are less than fully polished. The papers carry the names of the authors and should be cited accordingly. The findings, interpretations, and conclusions expressed in this paper are entirely those of the authors. They do not necessarily represent the views of the International Bank for Reconstruction and Development/World Bank and its affiliated organizations, or those of the Executive Directors of the World Bank or the governments they represent. Produced by the Research Support Team A Randomized, Controlled Study of a Rural Sanitation Behavior Change Program in Madhya Pradesh, India Sumeet R. Patil 1 Benjamin F. Arnold 2 Alicia Salvatore 3 Bertha Briceno 4 John M. Colford, Jr. 5 Paul J. Gertler 6 JEL codes: I15, I18 Key words: India, Madhya Pradesh, Total Sanitation Campaign, Randomized Controlled Trial, Impact Evaluation, Toilets, Open Defecation, Diarrhea, Parasite Infection, Weight, Height, Growth, Anemia Sector Board: WAT 1 University of California, Berkeley and Network for Engineering and Economics Research and Management (NEERMAN, India), corresponding author. 2 Division of Epidemiology, School of Public Health, University of California, Berkeley. 3 Community Partnership Program, Stanford Cancer Institute, Stanford University. 4 World Bank, Water and Sanitation Program. 5 Division of Epidemiology, School of Public Health, University of California, Berkeley. 6 Haas School of Business and School of Public Health, University of California, Berkeley. Acknowledgements We would like to thank Dr. Sandipan Ganguly (National Institute of Cholera and Enteric Diseases) and Dr. Nilesh Amritkar (Envirocare Pvt Ltd) for support with sample collection and analysis. Drs. Lia Fernald and Patricia Kariger provided support developing the anthropometry and anemia measurement protocols and training the field teams. We would further like to thank members of WSP, Eduardo Perez, Ajith Kumar, Upneet Singh, and K. Arokiam for information about the program and support throughout the study. Members of GfK Mode, Dr Urmil Dosaj, Dr Prem Talwar, Mr Subhabrata Guha, and Mr Ashutosh Rana, helped coordinate the field activities. Finally, we are indebted to the many field staff that administered the surveys and collected measurements for this study. This paper is a product of the Water and Sanitation Program (WSP), Sustainable Development Network. WSP is a multi-donor partnership created in 1978 and administered by the World Bank to support poor people in obtaining affordable, safe, and sustainable access to water and sanitation services. The paper is part of a larger effort by the World Bank to provide open access to its research and make a contribution to development policy discussions around the world. Policy Research Working Papers are also posted on the Web at http://econ.worldbank.org. For more information, please email the corresponding author, Sumeet Patil, at srpatil@neerman.org. 2 / 41 1. Introduction Open defecation is thought to be a major cause of the persistent worldwide burden of diarrhea and intestinal parasite infection among children < 5 years old [1]. Reducing open defecation requires access and use of improved sanitation facilities, which are defined as facilities that prevent human feces from re-entering the environment; in 2010, an estimated 47% of the world population did not have access to improved sanitation facilities [2]. India alone accounts for a quarter of the world’s deaths from diarrheal diseases (387 million) [3], a third of those without improved sanitation (814 million) and nearly 60% of those who continue to practice open defecation (626 million)—the lowest rung on the sanitation ladder [2]. Observational studies of interventions that contain human feces and prevent them from entering the environment have been shown to reduce diarrheal disease [4,5] and intestinal parasite infections [6–8]. However, much of the research has focused on the provision of sewerage systems in urban centers, and to date there have been no published, randomized trials to measure the effect of sanitation provision on diarrheal disease or intestinal parasite infections in young children. Furthermore, in rural areas of low-income countries, the provision and maintenance of networked sewerage is usually prohibitively expensive in the short term. As a result, most government and donor financing in the rural sanitation sector has focused on the provision of non-networked toilets or latrines. One approach to trigger large-scale toilet construction is Community-Led Total Sanitation (CLTS) [9]. Under CLTS, an external facilitator leads community exercises with the aim of changing community norms around open defecation. The activities highlight the magnitude of open defecation, they incorporate elements to elicit shame and disgust, and they attempt to mobilize community action for building toilets and leading the community to be “Open Defecation Free” (ODF). Facilitators encourage communities to come up with their own latrine designs that incorporate low cost, locally available materials. Individuals typically contribute their own time and materials to build toilets; for this reason, CLTS programs use relatively little public funds and rely heavily on private investment. Since the early inception of CLTS in Bangladesh in the late 1990s, the practice has rapidly spread throughout the rest of Asia and Sub-Saharan Africa [9]. The Indian government launched the Total Sanitation Campaign (TSC) in 1999 with a goal of reaching universal rural sanitation coverage by 2012. India’s TSC used principles of CLTS to motivate private toilet construction by attempting to change community norms around open defecation; TSC also provided financial incentives for local governments to achieve high levels of coverage, and subsidies for households to offset the capital costs of toilets. 3 / 41 In 2006, the World Bank’s Water and Sanitation Program (WSP) launched the Global Scaling Up Rural Sanitation program 7 to support local and national governments in India, Indonesia, and Tanzania to provide sanitation at scale. The program focused on learning how to combine the approaches of CLTS, behavior change communications, and social marketing of sanitation to generate sanitation demand and strengthen the supply of sanitation products and services at scale. In India, WSP supported efforts in rural Madhya Pradesh (MP) and Himachal Pradesh (HP) to strengthen TSC implementation. WSP incorporated a cluster-randomized impact evaluation in Madhya Pradesh to gather rigorous evidence of the effectiveness of the TSC. 8 The evaluation was designed to fill three research gaps in the rural sanitation sector that, if filled, would help guide future implementation practice: (1) lack of randomized evaluations; (2) lack of objective outcome measures (including health outcomes); and (3) lack of evaluations of large-scale programs. The objectives of the evaluation were to measure the effect of the TSC on toilet coverage, open defecation, water quality, and child health (diarrheal disease, intestinal parasite infections, anemia, and growth). We hypothesized that the program would increase toilet coverage and reduce open defecation. We further hypothesized that the program would improve water quality by reducing the quantity of feces in the environment that could contaminate shallow groundwater aquifers or water distribution networks. The reduced load of enteric pathogens in the environment may also reduce transmission through flies, which are well-established vectors for enteric pathogens [10–12]. Conditional on improvements in these intermediate outcomes, we hypothesized that children < 24 months at enrollment in intervention villages would have a lower prevalence of diarrhea, intestinal parasite infections, and anemia when measured after the intervention. Finally, we hypothesized that the program would improve average weight-for-age and height-for-age in these young children as a result of fewer symptomatic and asymptomatic enteric infections [13– 17]. Our research contributes to a small body of literature on the effectiveness of community demand driven, large-scale sanitation programs. Despite widespread delivery of large-scale sanitation programs that incorporate CLTS-like approaches, we are aware of few published, independent evaluations of them. A randomized controlled trial of TSC in Orissa, India found that the program – a combination of CLTS tools for behavior change and hardware subsidies to poorer households – led to a 19 percentage point increase in toilet ownership (32% v. 13%) in 2006 after 5-7 months of implementation [18]. In Zimbabwe, an observational study of the Participatory Hygiene and Sanitation Transformation (PHAST) campaign that combined elements of CLTS with subsidized hardware increased toilet ownership from 2% to 43% [19]. In 2004, a CLTS campaign in Ethiopia that constructed over 89,000 latrines found that 87% of 160 7 For more information on Global Scaling Up Rural Sanitation, see www.wsp.org/scalingupsanitation. 8 The study was initially designed for two independent sites in Madya Pradesh (MP) and in Himachal Pradesh (HP). Due to delays on baseline survey preparation and the impossibility to hold implementation in the control areas in the state of HP, it was decided that the final impact evaluation only included the state of MP. 4 / 41 randomly selected households had completed their latrines and that 90% (144/160) of the latrines were in use (the study provided no information about toilet coverage before and after the campaign) [20]. Taken together, these studies suggest that sanitation behavior change interventions result in toilet construction, but fall short of achieving “total” sanitation (or, 100% open defecation free communities). Two recent working papers that study India’s TSC have attributed large improvements in child height-for-age to the TSC. However, one study used administrative monitoring data collected by the project implementation teams which has been discredited due to vast over reporting compared to data from the recent census [43]. The second study from Maharashstra is a cluster-randomized trial [49]. While the study design is robust and the findings are important, the effects on height-for-age seem too large given the study included many children beyond the age of growth faltering (24 months) and only 8 percentage points in toilet construction [see 50]. Furthermore, there are no results published in peer reviewed journals on the health effects of TSC-style programs even though improved health has been a central benefit used to motivate the elimination of open defecation [22]. 2. Background and Description of the Program The government of India launched the Total Sanitation Campaign (TSC) in 1999, and at the time of this evaluation it was one of the largest rural development programs in India. The TSC included subsidies for and promotion of individual household latrines, school sanitation and hygiene education, Anganwadi 9 toilets, and community sanitation complexes. The TSC also supported rural sanitary marts and production centers. The TSC included several unique features that distinguished it from earlier approaches to rural sanitation in India, including: • A campaign-mode implementation approach • A focus on demand generation through behavior change and communication campaigns • A shift from high subsidy to low subsidy regimes • A flexible menu of technology options • An award, called the Nirmal Gram Puraskar (NGP), which is awarded to communities who are Open Defecation Free (ODF), defined as 100% of the population with access to a private- or shared toilet, and that meet all of the other total sanitation requirements. The award includes money and formal recognition to help incentivize entire communities. TSC funding and program implementation was managed at the district level 10 and program activities were implemented at the community (Gram Panchayat) level. Within each district, 9 Anganwadi – courtyard shelter in Hindi – is a child-care and mother-care center that is sponsored by the Indian government under integrated child development scheme (ICDS) program. These centers focus on children aged 0-6 years. 10 Indian administrative set up consists of central government, then state government, then districts, then blocks, then gram panchayats (GPs). Each GP consists of one or more villages. Within each village there can be one or more habitations or neighborhoods. 5 / 41 TSC implementation was coordinated within Blocks, which are one administrative level above individual communities. Block TSC offices coordinated latrine mason training, CLTS and social mobilization activities, and community outreach efforts. Block TSC offices also maintained official records of funds released to communities by the district, expenditures under the TSC, and toilet construction progress. WSP’s Scaling Up Rural Sanitation program in India supported the TSC program in two states: Himachal Pradesh and Madhya Pradesh. WSP supported local authorities in the creation of an enabling environment to facilitate TSC activities, the development of local capacities at the district level and the use of monitoring information to assess progress towards TSC goals. In Madhya Pradesh, the department of rural development was responsible for implementing the TSC. A concurrent program in Madhya Pradesh named Nirmal Vatika (“Clean House” in Hindi) provided additional support in the form of household subsidies and toilet construction materials during the evaluation period. TSC and Nirmal Vatika together provided at least Rs 4200 (US $80) to Below Poverty Line (BPL) households and at least Rs 2000 (US $40) to non- BPL households to support toilet construction. These costs were determined by the government to be adequate to construct a single toilet with proper sub-structure including a slab and water seal, and a superstructure for the toilet. BPL households were expected to complete the superstructure or toilet room (4 walls, a roof and a door) using locally available materials. In addition, community level ODF awards (named Nirmal Gram Puruskar) are provided for GPs ranging from Rs 50,000 (US $1,000) for GPs with populations ≤1000 to Rs 500,000 (US $10,000) for GPs with populations of 10,000 or more. Awards for achieving ODF status were also given at the Block level (maximum award of Rs 2,000,000 or US $40,000) and at the District level (maximum award of Rs 5,000,000 or US $100,000). In addition to usual government efforts, WSP supported training for triggering activities in the 20 intervention GPs in each district to help build the capacity of local government staff and program volunteers. 3. Evaluation Design and Methodology The evaluation used a cluster-randomized, controlled trial to evaluate the impacts of TSC in the Dhar and Khargone districts of MP. The study randomized communities within each district into two groups: an intervention group who received the TSC program and a control group who did not receive TSC activities until after the evaluation. The evaluation conducted household surveys to measure household and child characteristics both before and after TSC implementation. 3.1 Sample Size The study was originally designed to have 80% power to detect a reduction in diarrhea prevalence among children < 5 years old of 4.5 percentage points assuming 15% prevalence in the control group (or a 30% relative reduction) with a two-sided alpha of 5%. This led to a 6 / 41 design with 40 clusters per arm and 25 households with children < 5 per years cluster. After the baseline survey, we decided to additionally power the study to detect differences between groups in height-for-age Z-scores based on a hypothesis published on the possible effects of improved hygiene and sanitation on child growth [15]. We reviewed measures of variability and within- cluster correlation of height-for-age Z-scores (SD=2.09, intra-class correlation=0.17), and chose to increase the within-cluster sample sizes from 25 to 38 households to ensure the study had 80% power to detect differences of +0.2 Z in height-for-age. 3.2 Sample Selection and Randomization Village selection: Due to the large-scale implementation of TSC in Madhya Pradesh at the beginning of the study in 2009, we did not include a random sample of villages and instead enrolled villages in administrative areas that were willing to participate in the study. The evaluation selected Gram Panchayats (hereafter referred to as “villages”) 11 in three stages. First, the state government and WSP selected two districts out of the 50 in Madhya Pradesh – Dhar and Khargone – who were willing to work with WSP to implement the TSC. Second, the evaluation included 11 of 13 blocks from Dhar district and 8 of 9 blocks from Khargone; the remaining blocks were excluded from the evaluation because block officials were unwilling to maintain control villages for the length of the study. Third, in each block the government identified villages where they were indifferent about implementing the TSC program. We selected 40 villages in each district with between 2 and 8 villages from each block (larger blocks contributed more villages). Household selection: In each selected village the field team listed and mapped 200 households and used systematic random sampling to select 25 households with at least 1 child < 24 months of age in the baseline. The field team focused survey efforts in the most populous parts of each village. We visited the same households again at follow-up, and expanded the study to include 38 households per cluster due to sample size considerations (described above). Because we conducted the follow-up survey 21 months after the baseline, the eligibility criteria for newly enrolled households was that they had at least one child between the ages of (0 + 21 =) 21 months and (24 + 21 =) 45 months and were living in the village at the time of the baseline survey to be commensurate with the eligibility criteria for the original cohort. Randomization: The 80 study villages (Gram Panchayats) were the independent units and were randomly allocated in equal proportion to intervention or control. The randomization was stratified at the administrative block level. We stratified the randomization because TSC implementation was coordinated at the block level and we wanted to ensure that the treatment arms were evenly allocated between districts and geographically stratified within districts. The randomization took place in a public lottery in each of the 19 blocks. In each lottery, eligible 11 A Gram Panchayat could include one large village or a group of villages; on average, a Gram Panchayat included 2 to 3 villages in the study region. 7 / 41 village names for a block were written in pieces of paper and the block TSC coordinator blindly selected half of the papers as intervention villages. Overall, we assigned a total of 20 villages in each district to the TSC intervention and 20 to control (40 villages per arm). 3.3 Outcome Measurement The evaluation measured outcomes using a combination of survey questionnaires, field staff observations, environmental samples, child measurements, and specimen collection. Questionnaires used in the follow-up survey were the same as those used in the baseline survey with some additional questions to measure program exposure and program outcomes. The study collected the following sources of information. Household Questionnaire: The household questionnaire collected information about household membership and demographics, income, assets, dwelling characteristics, TSC program exposure, water and sanitation infrastructure, sanitation- and hygiene-related behaviors, mortality, and exposure to health interventions. Interviewers conducted standardized observations of dwelling hygienic conditions and sanitation and handwashing facilities when visiting households. Caregiver reported diarrhea: For children < 5 years old, field staff administered a health calendar to each child’s primary caregiver, which was modeled on a previously published design [23]. The calendar measured daily symptoms back from the morning of the interview for 14 days (baseline survey) and 7 days (follow-up survey); the recall period was shortened for the second survey after we determined that recall errors were unacceptably large beyond 7 days [24]. We defined diarrhea as ≥3 loose or water stools in 24 hours, or a single stool with blood/mucus [25]. We additionally measured respiratory symptoms (cough, congestion, difficulty breathing) and child health symptoms not plausibly influenced by the TSC program (bruising/abrasions, itchy skin/scalp) to serve as negative control outcomes [26] because we were concerned that caregiver- reported symptoms could be subject to differential reporting bias in this unblinded trial [27,28]. Anthropometry: We measured eligible children for height, weight, and mid-upper arm circumference (MUAC) using a standardized anthropometry protocols [29,30]. Pairs of anthropometrists measured child length/height to the nearest 0.1 cm using a portable stadiometer (Seca); children < 24 months were measured in the recumbent (lying) position and older children were measured standing. Weight was measured to the nearest 0.1 kg using an electronic scale (Tanita); children unable to stand were weighed in their caregiver’s arms and the caregiver’s weight measured separately. MUAC was measured to the nearest 0.1 cm using a pediatric measuring tape. All measurements were collected in duplicate and we used the average of the two measurements in the analysis. We excluded observations if the two measurements differed by > 10% (N=21 for height, N=85 for weight, N=23 for MUAC). We translated the anthropometric measurements into Z-scores using the WHO 2006 growth standards and a publicly available Stata algorithm [31]. 8 / 41 Anemia testing: If the caregiver provided informed consent, trained field staff cleaned a fingertip with disinfection liquid, and then collected a finger prick blood sample using a spring loaded, disposable lancet from children in study households between the ages of 6 and 60 months. A drop of blood was placed in a micro-cuvette, and the hemoglobin concentration was measured (HemoCue, HemoCue Ltd, Angelholm, Sweden) and reported to the caregiver. We classified children as severely anemic if their hemoglobin concentration was < 7.0 grams per deciliter (g/dl), moderately anemic if their hemoglobin concentration is 7.0 - 9.9 g/dl, and mildly anemic if their hemoglobin concentration is 10.0 - 11.9 g/dl [32]. Parents of children who were severely anemic were advised to visit the nearest health facility for medical attention. Water Sample Collection and Analysis: Field staff collected 100 ml water samples from a random sample of approximately 400 households in each group to collect water samples from stored drinking water and from the household’s designated water source. At the time of sample collection, field staff recorded information about the water source, storage and treatment (if any). Water samples were stored and transported in ice boxes for analysis within 24 hours to a field lab and tested for E. coli using membrane filtration within 36 hours of collection. The lower and upper limits of detection were 1 and 200 colony forming units (CFU) per 100 ml. Samples below the lower limited of detection were imputed at 0.5 (1/2 the limit of detection [33]), and samples beyond the upper limit of detection were imputed at the limit of detection (200). Child Stool Collection and Analysis: At follow-up field staff collected stool from a random sample of 1,150 households (from an eligible list with replacement sampling) and collected stool from the youngest child over 21 months of age (<24 months at baseline). The field team left a user-friendly stool collection kit (sterilized spatula, paper, container) with the child’s mother with verbal and written instructions on how to collect the samples. They advised mothers to note the consistency, color, and presence of blood in stool. The field staff then returned the following day to collect the stool samples. At the time of collection, field staff observed the sample for consistency and presence of blood, and added 10% formalin preservative. The preserved samples were stored in a shaded, cool and dry place and transported to the National Institute for Cholera and Enteric Diseases (NICED, Kolkata). NICED tested the samples within 30 days of collection for soil transmitted helminths (Ascaris lumbricoides, Trichuris trichuria, Ancylostoma duodenale and Necator americanus) and tapeworm helminths (Hymenolepis nana, Taenia sp., Diphyllobothrium latum) using the Kato-Katz technique [34]. A separate aliquot was analyzed to determine the presence of protozoans (Giardia lamblia, Cryptosporidium sp., Entamoeba histolytica) using a commercially available Enzyme-Linked Immunosorbent Assay (ELISA) kit (TechLab) [35,36]. If a child tested positive for one of the protozoan infections using ELISA, the result was confirmed using isolated DNA from the ELISA positive samples and PCR-restriction fragment length polymorphism (RFLP) methods for genotyping local isolates of Giardia (β- giardin) [37], Cryptosporidium (18s rRNA) [38], and E. histolytica (SSU rRNA) [39]. Community Questionnaire: The community questionnaire was administered at the village- level. We conducted surveys of the Gram Panchayat secretary (main administrative official), 9 / 41 elected head of the village, schools, Anganwadis, masons, and doctors (if available). Since our study sample was restricted to households with children < 24 months at enrollment, in the follow-up survey we added a rapid assessment from a random sample of households in the villages to measure sanitation coverage, quality of toilet construction, availability of hand washing place, type of drinking water source, and waste management practices. Block Questionnaire: In each administrative block we conducted a survey to assess the inputs in TSC program in terms of finance, material, manpower and their associated cost. The purpose of this questionnaire was to conduct a detailed cost benefit analysis at a later stage (not included in this report) and to obtain official records of program implementation and status. 3.4 Field Implementation Protocols and instruments used for data collection were designed by the WSP global impact evaluation team. These were revised, adapted, translated, and piloted by the India principal investigators. All data collection activities were conducted in Hindi. Study protocols and instruments are available from the corresponding author upon request. GfK Mode, Ltd. was contracted to conduct the fieldwork. The baseline survey was conducted between May and July, 2009, and the follow-up survey was conducted between February and April, 2011. The follow-up survey was conducted in a slightly different season due to concerns that the TSC program would soon expand into control villages. The India principal investigators, global team experts and GfK Mode researchers trained the field teams. Six field teams, each with four interviewers, 2 anthropometrists, 1 microbiology sampler, 2 supervisors, and one editor conducted the fieldwork at follow-up. Two field executives and one field coordinator handled oversight of the work in Madhya Pradesh. 3.5 Statistical Analysis We compared the baseline characteristics of the randomized treatment groups to determine if any observed characteristics were imbalanced by chance despite randomization. Due to highly comparable groups at baseline and the large increase in our within-cluster sample between baseline and follow-up, our analysis focused on group comparisons post-intervention (using follow-up measures only). To evaluate the potential for attrition (loss to follow-up) between baseline and follow-up to bias our estimates, we compared baseline characteristics of those present at follow-up with those lost to follow-up; we also compared the balance of baseline characteristics across treatment groups for individuals who were present at both baseline and follow-up to determine whether attrition was differential by treatment group. Our parameter of interest for all outcomes was the mean difference between randomized groups and we conducted the analysis using individuals as they were randomized (Intention to Treat; ITT). We estimated differences between groups using the following linear regression model: 10 / 41 Yijk = α + βTj + δXij +bk + εijk (1) Where, Yijk is the outcome for individual i in village j and block k, Tj is the intervention indicator (1 for intervention, 0 for control); Xij are individual and cluster level characteristics used in adjusted analyses; bk are indicator variables for block since randomization was stratified at the block level; and εijk is the error term. The parameter β estimates the ITT difference between randomized treatment groups. In adjusted analyses, we included the following covariates to improve precision: whether the household head had attended school, whether the household belonged to a Scheduled Caste or Tribe (SCST), child age, child sex. Additionally, the adjusted models included any baseline characteristics found to be imbalanced between randomized groups. To further assess differential impacts of the program by important population subgroups, we re-estimated the effect of the intervention for households with and without a toilet at baseline, and households above and below the official poverty line. Since we would expect behavior and child health outcomes to be correlated within villages, all estimates used Huber-White robust standard errors clustered at the village level [40]. We did not adjust P-values for multiple comparisons. 4. Results 4.1 Enrollment, Baseline Balance, and Attrition Table 1 lists the number of households in each district and block at the baseline and follow- up in intervention and control groups as well as the percentage attrition in the sample. At baseline the evaluation enrolled 1,954 households and 3,390 children < 5 y in 80 villages with complete interviews. Of the 1,954 households enrolled, 1,655 were located at the 21-month follow-up measurement (15% attrition). There was no significant difference in attrition rates between the intervention (16%) and control (15%) groups. At follow-up, the study enrolled an additional 1,384 randomly sampled households that met the original enrollment criteria. Thus, the follow-up sample included 3,039 households and 5,209 children < 5 y in the final analysis. Table 2 summarizes baseline household and child characteristics by randomized group. With few exceptions, randomization balanced the groups well with respect to measurable characteristics. However, there were a few important imbalances that are worth noting. First, 89% of the households in the intervention group had access to JMP improved water sources compared to 80% of households in the control group. In contrast, a larger fraction of control households had soap and water at handwashing locations used after defecation than in intervention households (44% v. 54% for intervention and control). On average, more children were found to be anemic in control group than the treatment (88% v. 93% for intervention and control, but this 11 / 41 difference is unreliable because of measurement errors 12) and the Z-score for height-for-age was also imbalanced (-1.38 v. -1.81 for intervention and control). Given these baseline differences between groups in access to JMP-defined improved water sources, handwashing conditions, and Z-score for height-for-age , we included the baseline village-level averages of these variables as covariates in adjusted models. The internal validity of a longitudinal or panel study relies on maintaining low attrition during the follow-up period. In our study 15% (299 of 1,954) of the baseline households were not located in the follow-up for reasons of permanent migration, seasonal migration, and inability to find a household after repeated visits. This 15% attrition rate was not large enough to affect the power of our estimates. Of 299 households, 154 were in the intervention group and 145 were in the control group. Temporary or seasonal migration accounted for 81 percent of the lost households, which indicates that lost households may be economically different than those that remained in the study. Table 3 compares characteristics between intervention and control groups for the household samples that is lost at follow-up and that remains intact during follow- up survey. The attrition was not differential by randomized treatment group based on observable characteristics (Table 3). Characteristics remained balanced between intervention and control groups in intact households similar to those noted in Table 2. 4.2 Implementation and Exposure to TSC Activities Implementation of TSC is carried out by local governments. Assessing and documenting TSC activities is challenging in the absence of a robust monitoring system for TSC. A comparison of official TSC data with recent figures from the Census suggests that official information consistently overestimates toilet construction coverage 13. In India, toilet construction triggers the release of TSC funds for the subsidies that the government provides to below poverty line (BPL) households. The existing system tracks the funds released and utilized at different levels and toilet construction coverage as reported by the GP secretary. The monitoring system does not track project inputs in terms of behavior change or social mobilization efforts which are key components of TSC. Even the reported toilet construction levels are not objective and validated measurements. We explored a variety of sources to assess the extent to which activities have been conducted and beneficiaries report being exposed to the TSC. First, we conducted a survey of TSC coordinators at the block level to assess TSC activities. The TSC coordinator is the official 12 In baseline anemia was tested with children under 2 years but ovr 6 month of age. Due to high heat during summer months the Hemocue machines malfunctioned. We were able to collect data from only 27 out of 40 GPs and 329 children from the control group, and from 24 GPs out of 40 and 293 children from the intervention group. This high level of data loss and unreliability of hemocue measurements in high temperature reduces our confidence on the results presented, especially when the statistical significance level is a = 0.1. 13 In MP, 86.9% of rural households still lack s toilet; a negligible improvement from 91.1% households without toilet as per 2001 census. As per TSC M&E records; however, almost 82% of the target households who lacked toilet in 2003 baseline are now provided an individual household latrine (IHL). 12 / 41 responsible to maintain records of TSC related activities. Second, to assess recollection of TSC activities and exposure to tools and information related to TSC, we surveyed Gram Panchayat (village) GP secretaries, school teachers, and Anganwadi (health centres) workers in the village, as well as a random sample of households in each village. This village survey selected randomly 40 to 80 households from each village (depending on its size) for a rapid questionnaire to assess their sanitation conditions. 14 The information provided by TSC Block coordinator and GP secretary indicate that implementation of TSC activities was far slower than the original timeline. The program originally envisioned that implementation would be completed 18 months after our baseline survey. However, as depicted in Figure 2, TSC coordinators reported that social mobilization and behavior change activities were being carried out even close to the follow-up survey at 21 months. Figure 2 summarizes the timeline of the latest CLTS triggering relative the follow-up survey based on information for 39 of the 40 intervention villages from the block coordinators (we were unable to complete a coordinator interview in 1 village). We found that 35% (14/40) of villages were triggered >6 months before the follow-up survey, but that 50% (20/40) were triggered 1 month before the follow-up survey or later. It is possible that the villages were triggered multiple times in the past, but the survey could not capture retrospectively a complete record of TSC activities in the village. Block coordinators were also asked to give an estimate of when they believe the intervention village would become ODF, which is summarized in Figure 3. For 63% (25/40) of intervention villages, the block coordinators estimated the villages would become ODF 6-months or longer after the follow-up survey. Table 4 provides additional results from the block coordinator interviews. We found that the most recent TSC-related activities were carried out after, during, or 1-3 months prior to follow-up survey in February 2011. Post-triggering follow-up visits to the village were reported in 11 of the 40 intervention villages at the time of our follow-up survey. While block coordinator interviews also identified that the behavior change aspects of TSC were being implemented, their official records indicated that the villages were close to having 100% toilet construction coverage as reported in Table 4. Official records of toilet construction would be indicative of the pace at which TSC funds were available and released to the local government authorities. On the other hand, the information from additional sources (GP secretaries, school teachers, Anganwadi workers in the village, and the rapid assessment from random sample of households) confirmed that TSC activities translated in a higher recollection and knowledge of TSC in the intervention areas compared to the control. Results from these additional sources are summarized in Table 5. 14 Note that unlike the main household survey, this sample was not restricted to households with children < 24 months at enrollment; it is thus more representative of the village conditions as a whole. 13 / 41 According to village secretary interviews, villages in the Dhar district had higher levels of exposure to CLTS activities and study villages reported higher levels of implementation using funds from TSC or Nirmal Vatika. According to information collected from school teachers, the schools in the treatment areas were exposed to more CLTS tools and there was a substantial increase in school toilet coverage in comparison to control villages (67% v. 47%). Anganwadi workers in both districts are almost similarly exposed to CLTS tools with statistically significant higher exposure in intervention villages. Across a broad set of measures we found that households in intervention villages were more aware of CLTS triggering, had higher knowledge of TSC, and experienced more personal visits to convince them to build and use toilets (Table 7, Panel 1). The percentage of households reporting that they changed their sanitation behavior as a result of personal visits was higher in the intervention group (15%) than in the control group (6%). The rapid household survey of the entire village indicated that while awareness of TSC was similar in intervention and control villages, 36% of households in intervention villages versus 20% of household in control villages had a usable toilet. Furthermore, an additional 19% of households in intervention villages had toilet under construction versus just 7% in control villages. Our estimates of household toilet coverage are much lower than the official TSC statistics of 47% to 97% toilet coverage in intervention villages as maintained by the block coordinator. In the main household survey (households with children < 24 months at enrollment), we found that the households in intervention villages in both districts had better exposure in terms of personal visits to promote sanitation, self reported change in sanitation related behavior following personal visits, and exposure to CLTS triggering tools, but the toilet construction levels were dramatically lower than official records (Table 5). Although the 21-month follow-up survey closely followed implementation activities in some intervention villages, the evaluation faced the risk of contamination of control villages. Prior to planning the follow-up survey, the discussions with the project implementation staff confirmed TSC related activities had started in 10 of the 40 control villages. 15 In addition, the government announced the roll out of activities in the remaining TSC control villages within few months of the planned follow-up survey. The objective of the evaluation was re-aligned to focus on immediate impacts of TSC as opposed to longer term impacts that could be expected after a longer program exposure (e.g., child development outcomes). Taken together, the information from block coordinator surveys and exposure to TSC related activities measured at the village shows clear evidence of more social mobilization and behavior change activities as well as toilet construction in intervention villages compared to control 15 Activities were starting in 10 control villages: Kharadi, Ladvi, Aasngaon, Bagfal, Umari (Khargone District) and Mourad, Bhil Barkheda, Kuradiya, Marol, Bajattakhurd (Dhar district). 14 / 41 villages, although some social mobilization and behavior changes were ongoing more recently than anticipated. In what follows we present the effects of the different exposure to TSC between treatment and control areas on outputs and intermediate outcomes along the value chain. We present some analysis by subgroups of interest and we finalize with a discussion of main findings. 4.3 Toilet Construction, Sanitation Behavior, and Health Outcomes Secular Changes from Baseline to Follow-up (Table 6) Table 6 includes means of key variables between follow-up and baseline in the intervention and control groups. Between baseline and follow-up surveys, the household demographic and economic variables have remained stable. Household access to JMP improved water sources was nearly universal at follow-up in intervention (97%) and control (95%) villages, and water quality, measured by E. coli indicator bacteria in stored drinking water, also improved over time in both groups. Diarrhea prevalence was lower at follow-up in both groups in part because of the aging cohort and possibly because the follow-up survey took place in a different season than baseline. Toilet Coverage and Sanitation-related Behavior (Table 7, Panel 2) The TSC program increased JMP improved toilet coverage by 19 percentage points in intervention villages compared to control villages (41% v. 22%). At follow-up, JMP improved toilet coverage in the intervention villages ranged from 5% to 79% (median 42%, IQR 27%, 57%), and in control villages it ranged from 0% to 74% (median 26%, IQR 5%, 36%). Although fewer households in intervention villages were likely to report daily open defecation compared to control villages for adult men (73% v. 83%), adult women (74% v. 84%), and children < 5 y (84% v. 89%), these reductions in reported OD behavior were smaller than the gains in improved toilet coverage. For example, although treatment groups reports lower OD levels, overall levels remain high (73% for adult men). Among the 630 households in intervention villages that had a JMP improved toilet at follow-up, 41% reported that adult men still practiced daily OD and 64% reported that children < 5 years practiced daily OD. Among 339 control village households with a JMP improved toilet at follow-up, 32% reported that adult men practiced daily OD and 54% reported that children < 5 years practiced daily OD. Although OD behavior remained prevalent among young children in both groups, households in intervention villages were more likely to report proper disposal of child feces based on the JMP definition (27% v. 18%). We observed no difference between groups in the proportion of households with no human or animal feces observed in their living area around the structure (40% in both groups). Drinking Water Quality (Table 7, Panel 3) Drinking water quality was poor in the study population. In control villages, 82% (331/403) of household drinking water samples tested positive for E. coli; water quality was slightly improved in intervention households where 77% (310/404) of samples tested positive for E. coli. 15 / 41 Intervention households also had slightly lower concentrations of E. coli than control households based on quantitative measures (1.25 vs 1.40 log10 E. coli CFU per 100 ml). Of 514 village source samples tested, 74% (210/282) of the sources in control villages and 70% (163/232) in intervention villages tested positive for E. coli. The results are statistically not significant in the “adjusted” model. Intestinal Parasite Infections (Table 7, Panel 4) In the subsample of 1,150 children who provided stool specimens, 5.7% (66) had some helminth infection and the majority of the infections (50/66) were Ascaris. Besides Ascaris, all remaining infections were tapeworms; no children were infected with Trichuris or hookworm. We observed no difference in helminth prevalence between intervention and control groups. Giardia infection was common, and consistent with slightly improved water quality in the intervention group, we observed lower Giardia prevalence among children in intervention villages (18%) compared to children in control villages (23%). However, in the adjusted model, while the direction of effect is the same the difference between intervention and control villages is not statistically significant. We observed no Cryptosporidium infections in the study children, and a low prevalence of E. histolytica (33 out of 1150; 2.9%). All specimens that tested positive for a protozoan infection using ELISA were confirmed as positive by the PCR assays (there were no false positives). Caregiver Reported Illness (Table 7, Panel 5) Diarrhea prevalence did not differ between intervention and control groups (8%), and acute respiratory illness was more prevalent in the intervention group than in the control group (17% v. 13%). We observed no significant differences between groups in negative control caregiver- reported outcomes including bruising/abrasions (1.4% v. 1.3%) and itchy skin/scalp (2.5% v. 2.2%) suggesting that a positive response bias was less likely. Anemia and Anthropometry (Table 7, Panel 6) Anemia was prevalent (54%) in the study children, and children were also small based on international growth standards. For example, mean weight-for-age Z-scores were –1.83 (control) and – 1.92 (intervention), and mean length-for-age Z-scores were –2.16 (control) and –2.19 (intervention). We observed no differences between groups in anemia prevalence or most anthropometric measurements. 4.4 Subgroup Results Table 8 presents the results of subgroup analyses of the TSC program impacts. We found that overall the program impacts on toilet construction, use of toilet, and reduction in open defecation were concentrated in households that did not have a toilet at baseline and those that were Below Poverty Line (BPL). For example, among BPL households, the intervention increased JMP improved toilet coverage by 32 percentage points compared to control (49% v. 17%) and it reduced open defecation among women by 21 percentage points (71% v. 92%). 16 / 41 Despite larger improvements in toilet construction and open defecation practices among BPL households or those with no toilet at baseline, we did not observe consistently larger improvements in child health in those subgroups. Instead, we observed larger reductions in diarrhea and Giardia among households who had a toilet at baseline, yet in this subgroup the intervention did not increase toilet construction, toilet use, or reduce open defecation. This suggests that the differences in child health could be due to chance given the small size of the subgroup (<100 individuals per arm) and the large number of subgroup comparisons made. Discussion Summary of Main Results This 80-village study conducted in rural Madhya Pradesh represents the one of the first large- scale, randomized evaluations of India’s Total Sanitation Campaign (TSC) to include child health measurement. Based on the latest reported CLTS triggering activity in each village, the program was not implemented to its full extent at the time of the follow-up survey (only 14 of 40 intervention villages had > 6 months of exposure). However, the information from other sources and especially the exposure to TSC related activities measured at the village level shows clear evidence of more social mobilization and behavior change activities as well as toilet construction in intervention villages compared to control villages, We observed as well some evidence for contamination in 10 of 40 control villages close to our follow-up survey. The incomplete program delivery and slight contamination of the control group would in theory bias our findings towards an underestimate of program impact. Despite the absence of full exposure to the program at the time of our follow-up measurement, we found that the TSC increased toilet coverage by 19 percentage points in intervention villages compared to control (41% v. 22%). Improvements in sanitation-related behavior were more modest: reported open defecation among adults decreased by 10 percentage points (e.g., 73% v. 83% among women) and correct child feces disposal increased by 9 percentage points (27% v. 18%). Among households in the intervention arm with a JMP improved toilet, 41% reported that adults still practiced daily OD. We found that program impacts on toilet coverage and sanitation-related behavior were concentrated almost entirely among households without a toilet at baseline and households that were below the poverty line, which is consistent with the program design that provided higher amounts of subsidies to such households. Reduced levels of E. coli in intervention household drinking water and lower Giardia prevalence in intervention children are consistent with the possibility that the TSC program reduced waterborne transmission of this pathogen in intervention villages. However, we found no evidence for reductions in helminth infection or diarrheal disease, and no evidence for more distal impacts on anemia or growth during the study period. 17 / 41 Interpretation Many child health studies of water, sanitation, and hygiene interventions have been small- scale, short-duration efficacy trials [41]; these trials are an essential step in documenting the potential for interventions to impact child health, but ultimately the external validity of such studies remains open to debate. This study was designed to measure the impacts of one of the largest public health interventions in human history: India’s TSC has touched the majority of rural Indian households since its inception in 1999. Effectiveness studies that measure the impact of large-scale government programs face numerous challenges compared to tightly controlled efficacy trials. A central challenge is ensuring good compliance and high fidelity of intervention implementation. In this study, we found evidence of TSC activities in 10 of 40 control villages, and in no case did intervention villages achieve “Open Defecation Free” status. Although a finding of incomplete implementation is an important finding in itself, ITT estimates of program impacts with imperfect compliance will tend to underestimate the effect possible under perfect compliance. If the effects of rural toilet promotion campaigns are heterogeneous with village- level toilet coverage, then this study provides information only about conditions where toilet coverage is < 80% (with most villages closer to 40%). In theory, the effects of similar interventions on child health could be larger if intervention campaigns achieve higher levels of toilet coverage, but the question can only be answered empirically in future studies that achieve higher levels of toilet coverage in intervention villages and allow for a longer duration of program exposure. A second challenge in studying large-scale programs that are not implemented by evaluation investigators is that implementation timelines may differ greatly from the original evaluation timeline. The follow-up measurement in this study (21 months following baseline) was the latest possible point we could measure outcomes under the grant timeline and our expectation that the program would soon expand into control villages. Although it is possible that impacts on diarrheal disease could begin relatively soon after intervention, as documented in short-duration efficacy trials [41], we would expect impacts on intestinal parasite infection and growth to potentially accrue more slowly. Ascaris ova are viable in the environment for many months (possibly years) [42]; for this reason, the hygienic disposal of feces may require many months to measurably reduce the quantity of viable helminth eggs in the village environment and subsequent parasite infection. Similarly, we would expect any benefit of sanitation interventions to child height to require many months of exposure to improved conditions since the causal mechanism is thought to work through a cumulative reduction of symptomatic and asymptomatic infections over the first years of life [13,43]. Given the large improvements in toilet construction documented in this study, an additional child health follow-up measurement in the study villages would yield valuable information over a longer period of exposure to improved sanitation conditions. To provide some context to our results, we reviewed findings from recent studies of similar sanitation programs. Our findings contrast with the preliminary results reported in two working 18 / 41 papers that study India’s TSC, both of which have attributed large improvements in child height- for-age to the TSC, and a third working paper that evaluated a large-scale sanitation program in Indonesia. Spears [44] estimated that children living in districts with an average level of TSC implementation had improved height of +0.2 Z-scores and reduced infant mortality by 4 per 1000. This study used administrative monitoring data collected by the TSC to measure program exposure combined with national surveys for child outcome measurements. However, official TSC monitoring data likely overstates actual implementation conditions. For example, in this study the TSC block coordinators reported that village level toilet coverage ranged from 47% to 97% in intervention villages, whereas village level toilet coverage in our study population ranged from 5% to 79% (median 42%). Correlated measurement errors in exposures or confounders usually lead to unpredictable biases that are often non-conservative [44–48]. In a second study from Maharashstra, Hammer and Spears [45] report results from a cluster- randomized trial that evaluated the TSC and was similar in design to the present study. Sixty villages were randomized to control or intervention in February 2004, and outcomes measured in August of 2004 and 2005. At the final measurement, the study reported a modest improvement of 8 percentage points in toilet construction but a strong increase of +0.4 Z in height-for-age among children < 5 years. The effects of the TSC on height-for-age are greater in magnitude than most complementary feeding efficacy trials in infants (mean +0.28 Z, range -0.04, 0.69) and far greater than effects observed in complementary feeding effectiveness trials [46,47]. While the study design is robust and the findings are important, the large effects on height-for-age just 6 months after enrollment when most children are beyond the primary window for growth faltering (0-24 months) [48] will need further research to determine whether the results are biologically plausible. A companion study to the current evaluation was conducted in East Java, Indonesia [21]. Consistent with our results, the intervention in Indonesia significantly increased toilet installation and reduced open defecation, albeit at rates lower than in this study. Among households with no toilet at baseline (N=941), the Indonesia study documented a small but significant improvement in health outcomes including diarrhea, helminth infection, and anthropometry. However, these results while promising should be viewed with caution as this subsample is small and the experiment was not designed nor powered to examine health impacts in subsamples. Our findings are broadly consistent with several recent studies of sanitation promotion interventions in South Asia. An earlier randomized evaluation in Orissa, India, between 2005 and 2006 randomly allocated the TSC program to 20 villages (20 reserved as control), and documented a 19-percentage point increase in household toilet ownership 5-7 months after implementation; the investigators found that a significant share of households were in the process of building their toilets, and suggested that later measurement would be helpful to document the full impact of the program. The findings from the present study are also consistent with results from an evaluation of the Sanitation Hygiene Education and Water supply in Bangladesh (SHEWA-B) [49]. The SHEWA-B project had the goal of improving sanitation, 19 / 41 hygiene, and water supply for more than 20 million rural Bangladeshis; a matched cohort study in 100 villages found modest improvements in implementation and behavioral outcomes, but no impact on child diarrhea after 2 years of intervention [49]. Arnold et al. [50] documented large improvements in toilet coverage, smaller improvements in open defecation, no improvements in diarrheal disease or height-for-age Z-scores in a matched cohort study of combined sanitation, water, and hygiene interventions in Tamil Nadu, India (a pilot implementation of TSC). A similar percentage of households with toilets reported adult daily OD in the Tamil Nadu study (39%) as in this study (41%), despite implementation being complete for more than 23 months in most intervention villages in Tamil Nadu. The toilet coverage and open defecation levels in the Tamil Nadu study were similar to those we observed in Madhya Pradesh, however diarrhea prevalence was much lower in Tamil Nadu (1.8%) and the authors inferred that open defecation was unlikely to be an important source of diarrheal pathogen transmission for children during their study. Although the findings from the present study are consistent with the hypothesis that open defecation may not be important source of diarrheal pathogen transmission for young children, we did not collect extensive information to confirm or rule out all transmission pathways. We would expect the TSC program to reduce child exposure to fecal pathogens through two main environmentally mediated pathways (Figure 1): (i) reduced open defecation of sick adults and older siblings or other children in the household compounds where young children may encounter feces, and (ii) reduced contamination of village drinking water supplies by fecal transport through the environment. Intervention households were more likely to report hygienic methods of child feces disposal than control villages (27% v. 18%), but there was no difference in the absence of human or animal feces observed during household visits (40% in both groups) and we did not observe lower prevalence of Ascaris infections (Table 7). This suggests that the program did not successfully reduce pathogen transmission to children through soil. However, lower levels of E. coli in drinking water and lower Giardia prevalence in intervention villages compared to control villages provides some objective evidence to suggest that the TSC program may have reduced waterborne pathogen transmission. In the study population, 43% of households used bore wells that draw water from shallow groundwater aquifers that could be contaminated with feces, and 33% used piped water that can be cross-contaminated with sewage and run off due to negative pressure in the pipes during non-supply hours. Giardia may have proven to be a more sensitive measure of child health impact compared to diarrhea because it was more prevalent in this study population (Table 7) and it is more specific to a transmission pathway the intervention is thought to block (waterborne transmission). These findings certainly merit future research, but we recommend caution in over-interpreting the result by itself because it was the only measure of health improvement among many measured in this study. The majority of intervention studies in the water, sanitation, and hygiene sector have used caregiver reported diarrhea as their primary outcome [41], which theoretically could be subject to differential, biased reporting in unblinded trials [27,28]. We included numerous objective child 20 / 41 health outcomes in this study (parasite infection, anemia, anthropometry) to complement caregiver-reported diarrhea. However, we did not include objective measures of sanitation behavior (disposal of child feces, toilet use and open defecation), which relied on reports from study participants. To the extent that our measurements of reported outcomes were subject to courtesy bias, we may have over-estimated toilet use and under-estimated open defecation prevalence in the study population. Furthermore, if the bias was differential by treatment group, then we would expect the study to have over-estimated the improvements due to intervention because we would expect the intervention households to be more sensitized to the stigma of open defecation. Measures of toilet use could be improved in future sanitation studies through the use of passive sensors mounted in the latrine [51]. Conclusions Despite encouraging increases in toilet construction and reductions in open defecation, we found no consistent evidence for improvements in multiple health outcomes among children < 24 months at enrollment. The lack of impact on downstream measures such as anemia and height- for-age would be expected given the delay in program implementation and relatively short exposure period by the follow-up survey. Intervention impacts on more proximate infectious outcomes, including diarrhea or intestinal parasite infections, were plausible under the timeline of the current study, but were not observed. The present study cannot rule out multiple competing hypotheses for lack of impact on infectious outcomes, which include: inadequate exposure time to the intervention, insufficient reductions in open defecation, and other dominant transmission pathways (e.g., person-to-person). However, the present study documents that India’s TSC was effective at convincing rural Indian households to build improved toilets but the corresponding reduction in open defecation has been less. Even among intervention households with a JMP improved toilet, 41% report that adults still practice daily open defecation. Toilet use has clearly lagged behind toilet construction, and even following intervention there remains room for large improvements in both toilet construction and sanitation related behavior change. The majority of adults (74%) in intervention villages report that they still practice daily open defecation, and open defecation among children < 5 remains even higher (84%). The study also found that subsidies provide a strong incentive for rural Indian households to build toilets and reduce open defecation, particularly among poor households who received the largest subsidy. Households below the poverty line (BPL) were far more likely to build toilets (+31%) than non BPL households (+15%). Nevertheless, the use of constructed toilets appears to lag behind their construction: although BPL households were less likely to report daily adult open defecation (-19%) compared to non BPL households (-6%), these reductions were not commensurate with the increase in toilet construction. These results suggest that subsidies likely accelerate toilet construction, but are insufficient to guarantee use. An assessment of 56 districts across India by WSP identified processes related to improved performance under TSC [52]. 21 / 41 WSP assessment identified that in high performing districts (>90% toilet use in a representative survey of GPs that won NGP award) the subsidy for toilet construction was given as an incentive whereas in low performing districts (<35% toilet use) subsidy was used to push toilet construction without without commensurate efforts at demand creation. This impact evaluation is not designed to answer how the subsidy affects sustainability of the use of toilets or to study different subsidy mechanisms. However, this study provides a basis for future research to answer these questions. Taken together, the study suggests that the TSC program shows promise for the successful expansion of access to improved sanitation in rural India. However, the challenge faced by rural India is increasing the use of newly constructed toilets to achieve commensurate reductions in open defecation. Additional measures of toilet construction, use, and maintenance over longer time periods in this cohort would provide additional evidence to support or refute whether the improvements observed in the present study are sustained (or even expanded) over a longer period. To the extent that such improvements reduce pathogen transmission, we would expect larger health improvements than those observed in the present study. 22 / 41 References 1. Mara D, Lane J, Scott B, Trouba D (2010) Sanitation and health. PLoS Med 7: e1000363. doi:10.1371/journal.pmed.1000363. 2. UNICEF, WHO (2012) World Health Organization and United Nations Children’s Fund Joint Monitoring Programme for Water Supply and Sanitation (JMP). Progress on Drinking Water and Sanitation: 2012 Update. New York and Geneva: UNICEF and WHO. 3. Unicef, WHO (2009) Diarrhoea: Why children are still dying and what can be done. Unicef / WHO. 4. Clasen TF, Bostoen K, Schmidt W-P, Boisson S, Fung IC-H, et al. (2010) Interventions to improve disposal of human excreta for preventing diarrhoea. Cochrane Database Syst Rev 6. doi:http://dx.doi.org/10.1002/14651858.CD007180.pub2. 5. Norman G, Pedley S, Takkouche B (2010) Effects of sewerage on diarrhoea and enteric infections: a systematic review and meta-analysis. Lancet Infect Dis 10: 536–544. doi:10.1016/S1473- 3099(10)70123-7. 6. Barreto ML, Genser B, Strina A, Teixeira MG, Assis AMO, et al. (2010) Impact of a City-Wide Sanitation Programme in Northeast Brazil on Intestinal Parasites Infection in Young Children. Environ Health Perspect. Available: http://www.ncbi.nlm.nih.gov/pubmed/20705544. Accessed 20 November 2010. 7. Mascarini-Serra LM, Telles CA, Prado MS, Mattos SA, Strina A, et al. (2010) Reductions in the prevalence and incidence of geohelminth infections following a city-wide sanitation program in a Brazilian Urban Centre. PLoS Negl Trop Dis 4: e588. doi:10.1371/journal.pntd.0000588. 8. Ziegelbauer K, Speich B, Mäusezahl D, Bos R, Keiser J, et al. (2012) Effect of sanitation on soil- transmitted helminth infection: systematic review and meta-analysis. PLoS Med 9: e1001162. doi:10.1371/journal.pmed.1001162. 9. Kar K, Chambers R (2008) Handbook on Community-Led Total Sanitation. Plan International (UK). Available: http://www.communityledtotalsanitation.org/resource/handbook-community-led-total- sanitation. Accessed 12 February 2013. 10. Chavasse DC, Shier RP, Murphy OA, Huttly SR, Cousens SN, et al. (1999) Impact of fly control on childhood diarrhoea in Pakistan: community-randomised trial. Lancet 353: 22–25. 11. Emerson PM, Lindsay SW, Alexander N, Bah M, Dibba S-M, et al. (2004) Role of flies and provision of latrines in trachoma control: cluster-randomised controlled trial. Lancet 363: 1093–1098. doi:10.1016/S0140-6736(04)15891-1. 12. Emerson PM, Simms VM, Makalo P, Bailey RL (2005) Household pit latrines as a potential source of the fly Musca sorbens–a one year longitudinal study from The Gambia. Trop Med Int Health 10: 706–709. doi:10.1111/j.1365-3156.2005.01432.x. 13. Checkley W, Buckley G, Gilman RH, Assis AM, Guerrant RL, et al. (2008) Multi-country analysis of the effects of diarrhoea on childhood stunting. Int J Epidemiol 37: 816–830. doi:10.1093/ije/dyn099. 23 / 41 14. Schmidt W-P, Boisson S, Genser B, Barreto ML, Baisley K, et al. (2010) Weight-for-age z-score as a proxy marker for diarrhoea in epidemiological studies. J Epidemiol Community Health 64: 1074– 1079. doi:10.1136/jech.2009.099721. 15. Humphrey JH (2009) Child undernutrition, tropical enteropathy, toilets, and handwashing. Lancet 374: 1032–1035. 16. Prendergast A, Kelly P (2012) Enteropathies in the developing world: neglected effects on global health. Am J Trop Med Hyg 86: 756–763. doi:10.4269/ajtmh.2012.11-0743. 17. Checkley W, Gilman RH, Black RE, Epstein LD, Cabrera L, et al. (2004) Effect of water and sanitation on childhood health in a poor Peruvian peri-urban community. Lancet 363: 112–118. 18. Pattanayak SK, Yang JC, Dickinson KL, Poulos C, Patil SR, et al. (2009) Shame or subsidy revisited: social mobilization for sanitation in Orissa, India. Bull World Health Organ 8: 580–587. 19. Waterkeyn J, Cairncross S (2005) Creating demand for sanitation and hygiene through Community Health Clubs: a cost-effective intervention in two districts in Zimbabwe. Soc Sci Med 61: 1958–1970. 20. O’Loughlin R, Fentie G, Flannery B, Emerson PM (2006) Follow-up of a low cost latrine promotion programme in one district of Amhara, Ethiopia: characteristics of early adopters and non-adopters. Trop Med Int Health 11: 1406–1415. 21. Cameron L, Shah M, Olivia S (2013) Impact Evaluation of a Rural Sanitation Behavior Change Project in Indonesia. World Bank. 22. Kar K (2003) Subsidy or self-respect? Participatory total community sanitation in Bangladesh. 23. Goldman N, Vaughan B, Pebley AR (1998) The use of calendars to measure child illness in health interview surveys. Int J Epidemiol 27: 505–512. 24. Arnold BF, Galiani S, Ram PK, Hubbard AE, Briceño B, et al. (2013) Optimal Recall Period for Caregiver-reported Illness in Risk Factor and Intervention Studies: A Multicountry Study. Am J Epidemiol 177: 361–370. doi:10.1093/aje/kws281. 25. Baqui AH, Black RE, Yunus M, Hoque AR, Chowdhury HR, et al. (1991) Methodological issues in diarrhoeal diseases epidemiology: definition of diarrhoeal episodes. Int J Epidemiol 20: 1057–1063. 26. Lipsitch M, Tchetgen ET, Cohen T (2010) Negative controls: a tool for detecting confounding and bias in observational studies. Epidemiology 21: 383–388. doi:10.1097/EDE.0b013e3181d61eeb. 27. Wood L, Egger M, Gluud LL, Schulz KF, Jüni P, et al. (2008) Empirical evidence of bias in treatment effect estimates in controlled trials with different interventions and outcomes: meta- epidemiological study. BMJ 336: 601–605. doi:http://dx.doi.org/10.1136/bmj.39465.451748.AD. 28. Schmidt W-P, Cairncross S (2009) Household Water Treatment in Poor Populations: Is There Enough Evidence for Scaling up Now? Environ Sci Technol 43: 986–992. 24 / 41 29. Cogill B (2003) Anthropometric indicators measurement guide. Washington, D.C.: Food and Nutrition Technical Assistance Project, Academy for Educational Development. 30. de Onis M, Onyango AW, Broeck JV den, Chumlea WC, Martorell R (2004) Measurement and standardization protocols for anthropometry used in the construction of a new international growth reference. Food Nutr Bull 25: S27–S36. 31. WHO (2006) WHO Child Growth Standards: Length/height-for-age, weight-for-age, weight-for- length, weight-for-height and body mass index-for-age: Methods and developments. 32. WHO (2011) Haemoglobin concentrations for the diagnosis of anaemia and assessment of severity. Vitamin and Mineral Nutrition Information System. Geneva, World Health Organization. 33. Helsel DR (1990) Less Than Obvious - Statistical Treatment of Data Below the Detection Limit. Environ Sci & Technol 24: 1766–1774. 34. Katz N, Chaves A, Pellegrino J (1972) A simple device for quantitative stool thick-smear technique in Schistosomiasis mansoni. Rev Inst Med Trop Sao Paulo 14: 397–400. 35. Garcia LS, Shimizu RY (1997) Evaluation of nine immunoassay kits (enzyme immunoassay and direct fluorescence) for detection of Giardia lamblia and Cryptosporidium parvum in human fecal specimens. J Clin Microbiol 35: 1526–1529. 36. den Hartog J, Rosenbaum L, Wood Z, Burt D, Petri, Jr WA (2013) Diagnosis of multiple enteric protozoan infections by enzyme-linked immunosorbent assay in the guatemalan highlands. Am J Trop Med Hyg 88: 167–171. doi:10.4269/ajtmh.2012.12-0142. 37. Cacciò SM, De Giacomo M, Pozio E (2002) Sequence analysis of the beta-giardin gene and development of a polymerase chain reaction-restriction fragment length polymorphism assay to genotype Giardia duodenalis cysts from human faecal samples. Int J Parasitol 32: 1023–1030. 38. Xiao L, Escalante L, Yang C, Sulaiman I, Escalante AA, et al. (1999) Phylogenetic analysis of Cryptosporidium parasites based on the small-subunit rRNA gene locus. Appl Environ Microbiol 65: 1578–1583. 39. Gonin P, Trudel L (2003) Detection and differentiation of Entamoeba histolytica and Entamoeba dispar isolates in clinical samples by PCR and enzyme-linked immunosorbent assay. J Clin Microbiol 41: 237–241. 40. Freedman DA (2006) On the so-called “Huber sandwich estimator” and “robust standard errors.” Am Stat 60: 299–302. 41. Waddington H, Snilstveit B (2009) Effectiveness and sustainability of water, sanitation, and hygiene interventions in combating diarrhoea. J Dev Eff 1: 295–335. 42. Roberts LS, Janovy J (2009) Foundations of Parasitology. 8th ed. United States: McGraw-Hill. 43. Petri WA, Guerrant RL, Miller M, Binder HJ, Levine MM, et al. (2008) Enteric infections, diarrhea, and their impact on function and development. J Clin Invest 118: 1277–1290. 25 / 41 44. Spears D (2012) Effects of Rural Sanitation on Infant Mortality and Human Capital: Evidence from India’s Total Sanitation Campaign (July 3, 2012 draft). Princeton University. 45. Hammer J, Spears D (2013) Village sanitation externalities and children’s human capital: Evidence from a randomized experiment by the Maharashtra government (Feb 3, 2013 draft). Princeton University. 46. Dewey KG, Adu-Afarwuah S (2008) Systematic review of the efficacy and effectiveness of complementary feeding interventions in developing countries. Matern Child Nutr 4 Suppl 1: 24–85. doi:http://dx.doi.org/10.1111/j.1740-8709.2007.00124.x. 47. Bhutta ZA, Ahmed T, Black RE, Cousens S, Dewey K, et al. (2008) What works? Interventions for maternal and child undernutrition and survival. Lancet 371: 417–440. 48. Victora CG, de Onis M, Hallal PC, Blossner M, Shrimpton R (2010) Worldwide timing of growth faltering: revisiting implications for interventions. Pediatrics 125: e473–e480. doi:http://dx.doi.org/10.1542/peds.2009-1519. 49. Huda TMN, Unicomb L, Johnston RB, Halder AK, Sharker MAY, et al. (2012) Interim evaluation of a large scale sanitation, hygiene and water improvement programme on childhood diarrhea and respiratory disease in rural Bangladesh. Soc Sci Med 75: 604–611. doi:10.1016/j.socscimed.2011.10.042. 50. Arnold BF, Khush RS, Ramaswamy P, London AG, Rajkumar P, et al. (2010) Causal inference methods to study nonrandomized, preexisting development interventions. Proc Natl Acad Sci U S A 107: 22605 – 22610. doi:10.1073/pnas.1008944107. 51. Clasen T, Fabini D, Boisson S, Taneja J, Song J, et al. (2012) Making sanitation count: Developing and testing a device for assessing latrine use in low-income settings. Environ Sci Technol. Available: http://dx.doi.org/10.1021/es2036702. 52. Water and Sanitation Program (WSP) (2010). Linking Service Delivery Processes and Outcomes in Rural Sanitation: Findings from 56 Districts in India. Global Scaling Up Sanitation Project Report. 26 / 41 Figure 1. Hypothesized causal mechanisms for intervention impact, and measurements in the evaluation. Figure 1. Timing of latest Community-Led Total Sanitation triggering in intervention villages relative to follow-up survey according to Block TSC Coordinators Figure 2. Expected time for the village to become ODF relative to follow-up survey based on the opinion of Block TSC Coordinators 27 / 41 Table 1. Number of Household Questionnaires Completed in Baseline and Follow-up Surveys and Attrition of Samples Follow-up Baseline Intact Panel Attrition [Number (%)] Region Treat Control Total Treat Control Total Treat Control Total Treat Control Total DHAR 762 756 1,518 487 482 969 404 394 798 83 (17%) 88 (18%) 171 (18%) Bagh 77 71 148 40 49 89 35 24 59 5 (13%) 25 (51%) 30 (34%) Dahi 38 38 76 26 25 51 25 23 48 1 (4%) 2 (8%) 3 (6%) Dhar 78 76 154 50 50 100 40 41 81 10 (20%) 9 (18%) 19 (19%) Gandhwani 73 77 150 50 50 100 39 45 84 11 (22%) 5 (10%) 16 (16%) Kukshi 38 38 76 25 25 50 17 24 41 8 (32%) 1 (4%) 9 (18%) Manawar 78 77 155 49 49 98 43 45 88 6 (12%) 4 (8%) 10 (10%) Nalchha 76 76 152 49 47 96 35 33 68 14 (29%) 14 (30%) 28 (29%) Nisarpur 76 76 152 49 38 87 38 26 64 11 (22%) 12 (32%) 23 (26%) Sardarpur 114 113 227 75 74 149 64 67 131 11 (15%) 7 (9%) 18 (12%) Tirla 38 38 76 24 25 49 21 22 43 3 (13%) 3 (12%) 6 (12%) Umarban 76 76 152 50 50 100 47 44 91 3 (6%) 6 (12%) 9 (9%) KHARGONE 763 758 1,521 489 496 985 418 439 857 71 (15%) 57 (11%) 128 (13%) Barwah 152 151 303 99 97 196 82 88 170 17 (17%) 9 (9%) 26 (13%) Bhagvanpura 76 76 152 50 50 100 38 45 83 12 (24%) 5 (10%) 17 (17%) Bhikangaon 77 76 153 49 49 98 40 39 79 9 (18%) 10 (20%) 19 (19%) Gogawan 77 76 153 49 50 99 41 44 85 8 (16%) 6 (12%) 14 (14%) Kasrawad 114 114 228 70 75 145 65 69 134 5 (7%) 6 (8%) 11 (8%) Khargone 76 75 151 48 50 98 44 44 88 4 (8%) 6 (12%) 10 (10%) Maheshwar 114 114 228 74 75 149 65 67 132 9 (12%) 8 (11%) 17 (11%) Ziranya 77 76 153 50 50 100 43 43 86 7 (14%) 7 (14%) 14 (14%) TOTAL 1,525 1,514 3,039 976 978 1,954 822 833 1,655 154 (16%) 145 (15%) 299 (15%) 28 / 41 Table 2. Test for Baseline Balance in Covariates after randomization. Intervention (I) Control (C) I-C N Mean N Mean Mean SE Household Characteristics Number of members in a Household 976 6.95 978 6.82 0.13 0.14 U5 Children Age in months 1683 21.89 1707 22.12 -0.23 0.45 Age of HH Head 976 45.34 978 43.18 2.16 1.12* Whether HH Head went to school 954 0.50 952 0.53 -0.03 0.05 SCST HHs 935 0.70 905 0.71 -0.02 0.05 Household construction type 976 0.57 978 0.60 -0.03 0.05 Monthly HH income (module 3 & 4) 976 11293 978 11022 271 1072 PC based wealth index 962 -0.14 964 -0.15 0.01 0.24 WASH Behaviors JMP defined improved water source 976 0.89 978 0.80 0.10 0.05** Self-reported drinking water treatment 976 0.68 978 0.66 0.02 0.04 Food is completely covered 968 0.64 969 0.60 0.04 0.04 Soap & water present, HW place used post 969 0.44 972 0.54 -0.10 .06^ defecation PCG of u5 child reports HW w/ soap after fecal 978 0.62 985 0.64 -0.02 0.04 contact in last 24 Hrs Child Nutrition Child ever breastfed& 1026 0.99 1037 0.99 0.00 0.01 & Child Still breastfeeding 1013 0.91 1021 0.90 0.02 0.01 # Iron pills, syrup given 1019 0.07 1033 0.06 0.01 0.02 Drugs for intestinal worms given past 6# 1025 0.19 1033 0.16 0.03 0.02 Did receive VitA dose last 6 months# 1013 0.30 1029 0.32 -0.01 0.03 Program Outputs/Outcomes JMP Improved Sanitation 975 0.14 978 0.12 0.01 0.03 JMP correct child feces disposal 976 0.16 978 0.13 0.03 0.03 29 / 41 Intervention (I) Control (C) I-C N Mean N Mean Mean SE Garbage obs. in kitchen or house 974 0.39 972 0.43 -0.04 0.03 Feces are not observed in living area around HH 973 0.41 976 0.38 0.03 0.05 Health Status Diarrhea 7-day prevalence# 1683 0.13 1707 0.12 0.01 0.02 # Diarrhea 14-day prevalence 1683 0.15 1707 0.15 0.01 0.02 # ALRI 7-day prevalence 1683 0.11 1707 0.10 0.01 0.03 # ALRI 14-day prevalence 1683 0.13 1707 0.11 0.02 0.03 Child weight (to 0.1 kg) # 957 7.04 943 6.99 0.05 0.09 # Child height (to 0.1 cm) 932 69.43 933 68.11 1.33 0.65** # Child arm circ. (to 0.1 cm ) 921 13.01 895 12.99 0.02 0.13 # Weight-for-age z-score 957 -2.20 943 -2.18 -0.02 0.09 # Length/height-for-age z-score 932 -1.38 933 -1.81 0.42 0.22* # Arm Circ-for-age z-score 921 -1.31 895 -1.33 0.02 0.12 # Weight-for-height z-score 895 -1.68 879 -1.43 -0.25 0.16 # Anemic: Hb < 110 g/l 293 0.88 329 0.93 -0.05 0.02* Water Microbiology Log10 E. coli in household drinking water (CFU/100 ml) 172 2.01 174 2.06 -0.05 0.09 Houshold drinking water is contaminated with E. coli 172 0.96 174 0.98 -0.02 0.02 & For children less than 24 months of age # For children less than 60 months of age z-test is used for the group mean by clustering standard error at GP level. The significance is noted as: ^ Statistically significant at  = 0.15 * Statistically significant at  = 0.1 ** Statistically significant at  = 0.05 *** Statistically significant at  = 0.01 30 / 41 Table 3. Analysis of Attrition in the baseline panel Attrition/Lost Households Intact Households Indicators T C T-C T C T-C N Mean N Mean Mean SE N Mean N Mean Mean SE Household Characteristics Number of members in a Household 154 6.45 145 6.68 -0.23 0.31 822 7.04 833 6.84 0.20 0.15 U5 Children Age in months 263 22.26 247 22.05 0.20 1.24 1420 21.82 1460 22.13 -0.31 0.48 Age of HH Head 154 41.26 145 42.74 -1.48 2.18 822 46.1 833 43.26 2.84 1.19** Whether HH Head went to school 150 0.39 143 0.38 0.01 0.08 804 0.52 809 0.55 -0.03 0.05 SCST HHs 150 0.85 130 0.77 0.08 0.07 785 0.67 775 0.7 -0.04 0.06 Household construction type 154 0.48 145 0.6 -0.12 0.07 822 0.59 833 0.61 -0.02 0.05 monthly HH income (module 3 & 4) 154 4601 145 5790 -1189 1364 822 12547 833 11933 614 1140 PC based wealth index 151 -0.93 142 -0.29 -0.64 0.36* 811 0.01 822 -0.13 0.14 0.25 WaSH Behaviors JMP defined improved water source 154 0.86 145 0.86 0.00 0.05 822 0.9 833 0.79 0.11 0.05** Self-reported drinking water 154 0.6 145 0.64 -0.04 0.06 822 0.7 833 0.67 0.03 0.04 treatment Food is completely covered 154 0.56 145 0.51 0.05 0.06 814 0.66 824 0.62 0.04 0.04 Soap & water present, HW place 153 0.33 144 0.4 -0.07 0.07 816 0.47 828 0.57 -0.10 0.06 used post PCG of u5 child reports HW w/ soap 151 0.56 145 0.59 -0.04 0.06 827 0.63 840 0.65 -0.02 0.04 after fecal contact in last 24 Hrs Child Nutrition Child ever breastfed& 160 0.97 153 0.99 -0.02 0.02 866 0.99 884 0.98 0.01 0.01* & Child Still breastfeeding 153 0.92 152 0.87 0.05 0.04 860 0.91 869 0.9 0.01 0.01 # Iron pills, syrup given 158 0.09 153 0.07 0.02 0.03 861 0.07 880 0.06 0.01 0.02 Drugs for intestinal worms given 160 0.18 153 0.16 0.02 0.05 865 0.19 880 0.16 0.03 0.02 past 6# Did receive VitA dose last 6 months# 156 0.26 152 0.33 -0.07 0.05 857 0.31 877 0.31 0.00 0.03 Program Outputs/Outcomes JMP Improved Sanitation 154 0.05 145 0.08 -0.03 0.05 821 0.15 833 0.13 0.02 0.04 31 / 41 JMP correct child feces disposal 154 0.1 145 0.12 -0.02 0.04 822 0.17 833 0.14 0.03 0.03 Garbage obs. in kitchen or house 154 0.45 145 0.42 0.03 0.06 820 0.38 827 0.43 -0.05 0.03 Feces are not observed in living area 152 0.41 144 0.34 0.07 0.06 821 0.41 832 0.39 0.02 0.05 around HH Health Status Diarrhea 7-day prevalence# 263 0.13 247 0.13 0.00 0.04 1420 0.13 1460 0.12 0.01 0.02 # Diarrhea 14-day prevalence 263 0.14 247 0.15 -0.01 0.04 1420 0.16 1460 0.15 0.01 0.02 # ALRI 7-day prevalence 263 0.14 247 0.11 0.02 0.05 1420 0.11 1460 0.1 0.01 0.03 # ALRI 14-day prevalence 263 0.15 247 0.12 0.03 0.05 1420 0.13 1460 0.11 0.01 0.03 # Child weight (to 0.1 kg) 141 6.86 131 6.89 -0.03 0.21 816 7.07 812 7.01 0.07 0.09 # Child height (to 0.1 cm) 141 69.3 126 68.3 1.02 1.23 791 69.46 807 68.08 1.38 0.65** # Child arm circ. (to 0.1 cm ) 139 12.8 122 12.9 -0.10 0.26 782 13.05 773 13.00 0.05 0.13 # Weight-for-age z-score 141 -2.41 131 -2.14 -0.27 0.16* 816 -2.16 812 -2.186 0.03 0.09 # Length/height-for-age z-score 141 -1.41 126 -1.54 0.13 0.34 791 -1.38 807 -1.85 0.47 0.22** # Arm Circ-for-age z-score 139 -1.52 122 -1.37 -0.15 0.23 782 -1.27 773 -1.32 0.05 0.12 # Weight-for-height z-score 135 -1.85 118 -1.57 -0.28 0.24 760 -1.65 761 -1.41 -0.24 0.16 # Anemic: Hb < 110 g/l 43 0.88 36 0.94 -0.06 0.08 250 0.88 293 0.92 -0.04 0.02* Water Microbiology Log10 E. coli in household drinking 24 2.08 22 1.99 0.10 0.20 148 2 152 2.07 -0.07 0.09 water (CFU/100 ml) Houshold drinking water is 24 0.96 22 0.95 0.00 0.06 148 0.96 152 0.98 -0.02 0.02 contaminated with E. coli & For children less than 24 months of age # For children less than 60 months of age z-test is used for the group mean by clustering standard error at GP level. The significance is noted as: * P < 0.1 ** P < 0.05 *** P < 0.01 32 / 41 Table 4. Program Implementation as per Block Coordinator Interview and Records for Intervention Villages (Gram Panchayats; GPs) Timing of latest TSC No of trt No of Months latest No of GPs % of % of related activities wrt EL GPs triggering done where Post % of Total No HHs IHL in No of (Feb 2011) where before EL triggering HHs Block of HH in with use – GPs 1-3 4-6 CLTS Follow up with GPs U/C opinion After mths mths triggering Mean Min Max activities are IHL toilets of BC before before is done done Bagh 2 0 0 2 2 1 1 1 0 732 64% 35% 50% Dahi 1 0 1 0 1 8 8 8 0 339 89% 11% 10% Dhar 2 0 2 0 2 1 1 1 2 966 61% 39% 10% Gandhwani 2 0 0 2 2 1 1 1 0 1071 49% 51% 10% Dhar District Kukshi 1 0 1 0 1 -1 -1 -1 1 568 91% 10% 50% Manawar 2 0 2 0 2 2 2 2 0 770 90% 10% 75% Nalchha 1 0 1 0 1 0 0 0 1 333 83% 0% 10% Nisarpur 2 0 0 2 2 8 8 8 0 953 88% 1% 10% Sardarpur 3 1 0 0 3 -1 -1 -1 0 1415 95% 5% 63% Tirla 1 1 0 0 1 16 16 16 1 243 51% 16% 10% Umarban 2 0 2 0 2 7 0 14 1 851 97% 3% 85% Barwah 4 0 4 0 4 1 1 1 0 1964 62% 20% 70% Bhagvanpura 2 0 0 2 2 8 8 8 2 853 91% 6% 60% Khargone District Bhikangaon 2 0 0 2 2 10 10 10 0 824 80% 25% 18% Gogawan 2 0 0 2 2 1 1 1 0 979 47% 53% 50% Kasrawad 3 0 3 0 3 1.33 1 2 0 2060 66% 68% 68% Khargone 2 0 2 0 2 3 3 3 0 735 67% 33% 70% Maheshwar 3 3 0 0 3 8.67 8 9 3 2382 59% 9% 58% Ziranya 2 0 0 2 2 7 7 7 0 1267 82% 17% 90% 33 / 41 Table 5. Implementation of TSC program at GP level and program exposure of the households DHAR KHARGONE COMBINED Variable Description T C T-C T C T-C T C T-C N Mean N Mean Mean SE N Mean N Mean Mean SE N Mean N Mean Mean SE Implementation of and Exposure to Program in GP as per GP Secretary Interview Implemented TSC activities or 0.14 0.11 20 0.60 20 0.45 0.15 0.16 18 0.83 20 0.30 0.53 38 0.71 40 0.38 0.34 received TSC funds *** *** Implemented TSC activities or 0.15 0.13 0.1* received TSC funds or Nirmal 20 0.50 20 0.20 0.30 18 0.83 20 0.25 0.58 38 0.66 40 0.23 0.43 ** *** ** Vatika funds Report some CLTS triggering 0.12 0.13 0.09 20 0.95 20 0.50 0.45 18 0.89 20 0.65 0.24 38 0.92 40 0.58 0.35 activity in GP *** * *** % of GP secys personally exposed to - 18 1.00 18 0.89 0.11 0.07 16 0.94 19 0.95 0.08 34 0.97 37 0.92 0.05 0.05 any CLTS tool 0.01 Avg No of CLTS tools GP secy 0.81 0.65 18 6.50 18 4.22 2.28 16 6.63 19 5.16 1.47 1.01 34 6.56 37 4.70 1.86 exposed to *** *** Implementation of and Exposure to Program as per School teachers interview Avg No of schools in GP 19 4.68 20 3.85 0.83 0.59 20 4.05 18 3.33 0.72 0.51 39 4.36 38 3.61 0.75 0.4* % of schools in GP exposed to any 19 0.87 20 0.81 0.06 0.10 20 0.96 18 0.89 0.07 0.07 39 0.92 38 0.85 0.07 0.06 CLTS Avg No of CLTS tools school were 0.61 0.53 19 5.45 20 4.52 0.93 0.85 20 5.57 18 3.15 2.42 39 5.51 38 3.87 1.64 exposed *** *** 0.11 0.08 % of Schools in GP with toilets 19 0.71 20 0.45 0.26 20 0.62 18 0.49 0.14 0.13 39 0.67 38 0.47 0.20 ** ** % of schools reporting toilet 19 0.13 20 0.08 0.05 0.06 20 0.31 18 0.18 0.14 0.14 39 0.22 38 0.13 0.10 0.08 construction post baseline Implementation of and Exposure to Program as per Anganwadi workers Interview Avg No of Anganwadis in GP 20 3.30 20 3.05 0.25 0.29 20 3.05 20 2.90 0.15 0.47 40 3.18 40 2.98 0.20 0.27 % of AW in GP exposed to any 0.09 0.05 20 0.94 20 0.74 0.20 20 0.98 20 0.91 0.06 0.04 40 0.96 40 0.83 0.13 CLTS tool ** ** Avg No of CLTS tools AWs were 0.87 0.65 0.54 20 5.61 20 4.08 1.52 20 6.22 20 4.43 1.79 40 5.91 40 4.26 1.66 exposed to * *** *** % of AW observed to have toilet 20 0.26 20 0.18 0.08 0.09 20 0.21 20 0.19 0.02 0.11 40 0.24 40 0.18 0.05 0.07 % of AW reporting toilet - - - 20 0.65 20 0.78 0.10 20 0.67 20 0.71 0.11 40 0.66 40 0.75 0.07 construction post baseline 0.14 0.04 0.09 Implementation of and Exposure to Program as per Transect Walk (rapid HH survey) 100 - 198 190 - % of HHs aware about TSC 0.37 920 0.43 0.11 980 0.32 980 0.33 0.00 0.11 0.35 0.38 0.08 0 0.06 0 0 0.03 34 / 41 DHAR KHARGONE COMBINED Variable Description T C T-C T C T-C T C T-C N Mean N Mean Mean SE N Mean N Mean Mean SE N Mean N Mean Mean SE % of HH with usable toilets 100 0.09 198 190 0.07 0.40 920 0.22 0.18 0.1* 980 0.32 980 0.17 0.15 0.36 0.19 0.17 (Observed) 0 * 0 0 ** % of HH with under construction 100 0.04 198 190 0.07 0.14 920 0.06 0.07 980 0.24 980 0.08 0.16 0.13 0.19 0.07 0.11 IHL (observed) 0 * 0 0 * % of HH with toilet superstructure 100 - 198 190 - 0.13 920 0.13 0.01 0.05 980 0.14 980 0.15 0.06 0.14 0.14 0.04 (door, roof, walls - observed) 0 0.02 0 0 0.01 100 198 190 % of HH with water near IHL 0.17 920 0.16 0.00 0.07 980 0.15 980 0.12 0.03 0.04 0.16 0.14 0.02 0.04 0 0 0 % of HH where possibility of use is 100 198 190 0.33 920 0.20 0.12 0.09 980 0.23 980 0.17 0.06 0.09 0.28 0.19 0.10 0.06 confirmed by enumerators 0 0 0 Exposure of Individual households to program related activities (main HH survey) Did HH receive WaSH msg from 152 151 765 0.27 759 0.26 0.01 0.05 763 0.32 757 0.29 0.03 0.05 0.30 0.27 0.02 0.04 mass media 8 6 Did HH change Wash behavior due 152 151 765 0.19 759 0.16 0.03 0.05 763 0.23 757 0.20 0.03 0.05 0.21 0.18 0.03 0.03 to mass media 8 6 Did HH receive WaSH msg from 0.04 0.04 148 147 0.03 744 0.21 733 0.13 0.09 740 0.26 744 0.07 0.19 0.24 0.10 0.14 personal Visits ** *** 4 7 *** Did HH change Wash behavior due 0.03 0.03 148 147 0.02 744 0.13 733 0.06 0.07 740 0.17 744 0.05 0.12 0.15 0.05 0.09 to Personal visits *** *** 4 7 *** Did HH participate or is aware of 0.05 0.04 153 151 0.03 766 0.26 760 0.18 0.08 764 0.32 759 0.14 0.18 0.29 0.16 0.13 any of CLTS triggering activities * *** 0 9 *** Does HH know of masons to build 153 151 766 0.65 760 0.64 0.01 0.05 764 0.79 758 0.77 0.02 0.04 0.72 0.70 0.01 0.04 toilets 0 8 Does HH know of shop to buy toilet - - 153 151 - 766 0.74 760 0.81 0.05 764 0.87 759 0.88 0.04 0.80 0.85 0.03 mater 0.07 0.01 0 9 0.04 Does HH know of about TSC/NGP 0.05 153 151 0.04 766 0.25 760 0.22 0.03 0.05 764 0.30 759 0.21 0.09 0.27 0.21 0.06 program ** 0 9 * z-test is used for the group mean by clustering standard error at GP level. The significance is noted as: * P < 0.1 ** P < 0.05 *** P < 0.01 35 / 41 Table 6. Trend between baseline and follow-up surveys in the intervention and control groups. CONTROL INTERVENTION Follow-up (F) Baseline (B) F-B Follow-up (F) Baseline (B) F-B Description N Mean N N Mean N N Mean N Mean Mean SE Household Characteristics Number of members in a Household 1514 7.04 996 6.80 0.24 0.13* 1525 7.11 996 6.96 0.16 0.13 U5 Children Age in months 2609 29.40 1737 22.13 7.27 0.45*** 2600 28.79 1719 21.93 6.86 0.42*** Age of HH Head 1514 43.17 996 43.15 0.02 1.01 1525 43.42 996 45.42 -2.00 1.01** Whether HH Head went to school 1506 0.52 969 0.53 -0.01 0.04 1513 0.57 972 0.49 0.07 0.04* SCST HHs 1489 0.71 922 0.71 0.00 0.05 1508 0.65 951 0.70 -0.05 0.06 Household construction type 1514 0.61 996 0.61 0.00 0.05 1525 0.64 996 0.57 0.06 0.05 monthly HH income (module 3 & 4) 1514 9172 996 11024 -1853 906** 1525 10780 996 11241 -461 1411 PC based wealth index 1509 0.02 982 -0.14 0.16 0.26 1523 0.16 982 -0.14 0.30 0.24 WaSH Behaviors JMP defined improved water source 1514 0.95 996 0.80 0.15 0.04*** 1525 0.97 996 0.89 0.08 0.03*** Self-reported drinking water treatment 1514 0.68 996 0.67 0.01 0.04 1525 0.67 996 0.69 -0.02 0.04 Food is completely covered 1156 0.78 987 0.60 0.18 0.03*** 1190 0.77 988 0.64 0.13 0.03*** Soap & water present, HW place used post 1269 0.44 990 0.54 -0.11 0.05** 1334 0.49 989 0.44 0.05 0.05 HW w/ soap after fecal contact in last 24 Hrs 1498 0.59 1003 0.64 -0.05 0.04 1512 0.61 998 0.61 0.00 0.04 Child Nutrition Child ever breastfed 771 0.98 1055 0.99 -0.01 0.01 812 0.98 1046 0.99 -0.01 0.01 Child Still breastfeeding 753 0.90 1039 0.90 0.01 0.01 795 0.88 1033 0.91 -0.03 0.02** Iron pills, syrup given 2341 0.01 1051 0.06 -0.05 0.01*** 2296 0.02 1039 0.07 -0.06 0.01*** Drugs for intestinal worms given past 6 2350 0.10 1051 0.16 -0.06 0.03** 2326 0.14 1045 0.19 -0.05 0.03* Did receive VitA dose last 6 months 798 0.19 1047 0.31 -0.12 0.04*** 956 0.21 1033 0.30 -0.09 0.04** Program Outputs/Outcomes JMP Improved Sanitation 1512 0.22 996 0.13 0.10 0.04** 1522 0.41 995 0.13 0.28 0.04*** JMP correct child feces disposal 1514 0.18 996 0.14 0.05 0.03* 1525 0.27 996 0.16 0.11 0.03*** Garbage obs. in kitchen or house 1476 0.37 990 0.43 -0.06 0.04* 1491 0.36 994 0.39 -0.03 0.03 No human/animal feces observed in HH 1500 0.40 994 0.38 0.02 0.05 1512 0.40 993 0.41 0.00 0.04 living area Health Status diar period prev, last 07 days 2609 0.08 1737 0.12 -0.04 0.02** 2600 0.07 1719 0.13 -0.06 0.01*** diar period prev, last 14 days 2609 0.08 1737 0.15 -0.06 0.02*** 2600 0.08 1719 0.15 -0.07 0.02*** alri period prev, last 07 days 2609 0.13 1737 0.10 0.03 0.03 2600 0.16 1719 0.11 0.05 0.03* alri period prev, last 14 days 2609 0.13 1737 0.11 0.02 0.03 2600 0.17 1719 0.13 0.04 0.03 Child weight (to 0.1 kg) # 2161 10.28 943 6.99 3.29 0.16*** 2154 10.07 957 7.04 3.03 0.1*** Child height (to 0.1 cm) # 2185 82.31 933 68.11 14.21 0.49*** 2175 81.68 932 69.43 12.25 0.62*** 36 / 41 CONTROL INTERVENTION Follow-up (F) Baseline (B) F-B Follow-up (F) Baseline (B) F-B Description N Mean N N Mean N N Mean N Mean Mean SE Child arm circ. (to 0.1 cm#) 2191 13.80 895 12.99 0.82 0.11*** 2197 13.78 921 13.01 0.77 0.12*** Weight-for-age z-score# 2161 -1.83 943 -2.18 0.35 0.1*** 2154 -1.92 957 -2.20 0.28 0.09*** Length/height-for-age z-score# 2185 -2.16 933 -1.81 -0.35 0.13*** 2175 -2.19 932 -1.38 -0.81 0.20*** Arm Circ-for-age z-score# 2191 -1.34 895 -1.33 -0.01 0.11 2197 -1.34 921 -1.31 -0.03 0.11 Weight-for-height z-score# 2054 -0.83 879 -1.43 0.60 0.12*** 2054 -0.85 895 -1.68 0.83 0.16*** Anemic: Hb < 110 g/l 1922 0.51 331 0.93 -0.42 0.04*** 1919 0.56 297 0.88 -0.32 0.04*** Microbiology Log10(ecoli) in household DW (ecoli in 403 1.40 176 2.06 -0.66 0.08*** 404 1.26 178 2.02 -0.76 0.11*** CFU/100 ml) Whether HH DW is contaminated with ecoli 403 0.82 176 0.98 -0.16 0.03*** 404 0.77 178 0.96 -0.19 0.03*** Any protozoa infection prevalence& 569 0.26 88 0.13 0.13 0.04*** 581 0.22 93 0.13 0.09 0.04** Entamoeba histolytica infection prevalence& 569 0.02 88 0.08 -0.05 0.03* 581 0.03 93 0.09 -0.05 0.03* Giardia lambia infection prevalence 569 0.23 88 0.05 0.19 0.03*** 581 0.18 93 0.04 0.14 0.03*** Soil transmitted helminth infection 569 0.06 88 0.06 0.00 0.02 581 0.06 93 0.04 0.02 0.03 prevalence# Any helminth or protozoa infection 569 0.31 88 0.18 0.13 0.05** 581 0.27 93 0.17 0.10 0.04** prevalence Z-test was used for the group mean by clustering standard error at village level. The significance is noted as: * P < 0.1 ** P < 0.05 *** P < 0.01 37 / 41 Table 7. Differences at follow-up between TSC intervention and control groups for main program outputs, outcomes, and impacts Difference due Intervention Difference due to Intervention Control Group Outputs, Outcomes, Impacts ITT unadjusted ITT adjusted1 N Mean SE N Coeff SE R2 N Coeff SE R2 Panel 1: Program Exposure Did HH receive WASH mass media message 1511 0.27 0.01 3034 0.023 0.028 0.05 2,972 0.000 0.024 0.14 Did HH change WASH behavior due to mass media 1511 0.18 0.01 3034 0.03 0.026 0.05 2,972 0.019 0.023 0.11 Did HH receive WASH message from personal visits 1472 0.10 0.01 2951 0.140*** 0.021 0.07 2,892 0.127*** 0.023 0.08 Did HH change WASH behavior due to personal visits 1472 0.06 0.01 2951 0.095*** 0.018 0.04 2,892 0.080*** 0.018 0.05 Did HH participate or is aware of CLTS Triggering 1514 0.16 0.01 3039 0.135*** 0.026 0.07 2,977 0.140*** 0.026 0.08 Does HH know of masons to build toilets 1513 0.70 0.01 3038 0.015 0.023 0.08 2,976 0.023 0.026 0.09 Does HH know of shop to buy toilet mater 1514 0.85 0.01 3039 -0.043** 0.019 0.11 2,977 -0.017 0.021 0.13 Does HH know of TSC/NGP 1514 0.21 0.01 3039 0.062** 0.026 0.07 2,977 0.053** 0.025 0.12 Panel 2: Toilet Coverage and Sanitation-related Behavior HH with JMP Improved Sanitation 1512 0.22 0.01 3034 0.189*** 0.035 0.09 2,972 0.178*** 0.035 0.13 Observe that HH is using toilet 1504 0.17 0.01 3024 0.104*** 0.029 0.08 2,962 0.093*** 0.025 0.15 HH Reported OD by men 1514 0.84 0.01 3039 -0.095*** 0.029 0.07 2,977 -0.087*** 0.024 0.14 HH Reported OD by women 1514 0.83 0.01 3039 -0.102*** 0.028 0.07 2,977 -0.091*** 0.025 0.15 HH Reported OD by children 1514 0.89 0.01 3039 -0.053** 0.021 0.04 2,977 -0.054*** 0.017 0.11 HH reports JMP correct child feces disposal 1514 0.18 0.01 3039 0.087*** 0.021 0.06 2,977 0.075*** 0.019 0.10 No human/animal feces observed in HH living area 1500 0.40 0.03 3012 0.006 0.026 0.11 2950 0.019 0.023 0.13 Panel 3: Drinking Water Quality Log10 E. coli in household drinking water (CFU/100 ml) 403 1.40 0.05 807 -0.153* 0.086 0.08 791 -0.046 0.09 0.09 E. coli present in household drinking water, % 403 0.82 0.02 807 -0.055* 0.028 0.05 791 -0.032 0.034 0.05 E. coli present in the source from where household 280 0.74 0.03 511 -0.115 0.077 0.14 504 -0.016 0.083 0.20 collected drinking water, % Panel 4: Intestinal Parasite Infections Any protozoan present 569 0.26 0.02 1150 -0.04 0.024 0.03 1,128 -0.027 0.028 0.04 Entamoeba histolytica present 569 0.02 0.01 1150 0.008 0.008 0.02 1,128 0.009 0.009 0.03 38 / 41 Difference due Intervention Difference due to Intervention Control Group Outputs, Outcomes, Impacts ITT unadjusted ITT adjusted1 N Mean SE N Coeff SE R2 N Coeff SE R2 Giardia lambia present 569 0.23 0.02 1150 -0.048** 0.024 0.03 1,128 -0.036 0.026 0.04 Any helminth present 569 0.06 0.01 1150 0.001 0.011 0.02 1,128 -0.005 0.012 0.03 Ascaris lumbricoides present 569 0.04 0.01 1150 -0.002 0.01 0.02 1,128 -0.011 0.01 0.03 Any intestinal parasite present 569 0.31 0.02 1150 -0.040* 0.023 0.03 1,128 -0.032 0.026 0.04 Panel 5: Caregiver reported illness in the last 7 days Diarrhea 2609 0.08 0.01 5209 -0.003 0.008 0.02 5,096 -0.002 0.008 0.03 Acute respiratory infection 2609 0.13 0.01 5209 0.038** 0.017 0.06 5,096 0.049** 0.02 0.07 Panel 6: Anemia and Anthropometry Anemic: Hb < 110 g/L 1922 0.51 0.01 3841 0.05 0.03 0.10 3,761 0.033 0.032 0.13 # Child weight (to 0.1 kg) 2161 10.28 0.06 4315 -0.229* 0.132 0.02 4,239 -0.13 0.108 0.59 # Child height (to 0.1 cm) 2185 82.31 0.25 4360 -0.678* 0.344 0.01 4,275 -0.242 0.275 0.75 # Child arm circ. (to 0.1 cm ) 2191 13.80 0.03 4388 -0.004 0.071 0.09 4,294 -0.022 0.073 0.24 Weight-for-age z-score# 2161 -1.83 0.03 4315 -0.095 0.079 0.02 4,239 -0.094 0.076 0.05 # Length/height-for-age z-score 2185 -2.16 0.04 4360 -0.034 0.081 0.02 4,275 -0.04 0.092 0.08 # MUAC-for-age z-score 2191 -1.34 0.02 4388 0.02 0.068 0.10 4,294 -0.022 0.065 0.14 # Weight-for-height z-score 2054 -0.83 0.04 4108 -0.018 0.089 0.04 4,035 0.029 0.086 0.05 # BMI z-score 2052 -0.60 0.04 4104 -0.062 0.09 0.03 4,030 -0.019 0.087 0.04 Abbreviations. BMI: Body Mass Index; CLTS: Community-Led Total Sanitation; Hb: Hemoglobin; HH: household; ITT: Intention-To-Treat; JMP: WHO/Unicef Joint Monitoring Programme for Water Supply and Sanitation; MUAC: Mid-Upper-Arm Circumference; OD: Open Defecation; WASH: Water, Sanitation, and Hygiene 1 Adjusted models included: whether household head attended school, whether household belongs to Scheduled Caste or Tribe, percentage of surveyed households in the GP with improved water source at baseline, and percentage of surveyed households in the GP with soap, water present at hand washing place used after defecation, and average height-for-age Z-score in the GP at the baseline. For child level outcomes, age of child in months and sex were also included. * P < 0.1 ** P < 0.05 *** P < 0.01 39 / 41 Table 8. Differences at follow-up between TSC intervention and control groups stratified by population subgroups Control Intervention ITT - Unadjusted ITT – Adjusted 1 N Mean SE N Mean SE N Coeff SE R2 N Coeff SE R2 HH with JMP Improved Sanitation Total Population 1512 0.22 0.01 1522 0.41 0.01 3034 0.189*** 0.035 0.09 2972 0.178*** 0.035 0.13 No toilet at Baseline 1319 0.11 0.01 1297 0.32 0.01 2616 0.202*** 0.031 0.11 2564 0.193*** 0.033 0.11 Toilet at Baseline 190 0.98 0.01 212 0.97 0.01 402 -0.018 0.019 0.05 392 0.008 0.014 0.07 BPL 375 0.17 0.02 300 0.48 0.03 675 0.317*** 0.046 0.16 660 0.330*** 0.054 0.18 Non-BPL 1137 0.24 0.01 1222 0.4 0.01 2359 0.152*** 0.037 0.09 2312 0.137*** 0.035 0.13 Observe that HH is using toilet Total Population 1504 0.17 0.01 1520 0.27 0.01 3024 0.104*** 0.029 0.08 2962 0.093*** 0.025 0.15 No toilet at Baseline 1311 0.06 0.01 1294 0.17 0.01 2605 0.108*** 0.021 0.07 2553 0.097*** 0.02 0.08 Toilet at Baseline 190 0.91 0.02 212 0.89 0.02 402 -0.026 0.029 0.1 392 -0.01 0.027 0.14 BPL 373 0.08 0.01 300 0.29 0.03 673 0.228*** 0.034 0.15 658 0.228*** 0.037 0.18 Non-BPL 1131 0.2 0.01 1220 0.27 0.01 2351 0.065** 0.031 0.08 2304 0.053* 0.027 0.15 HH Reported OD by men Total Population 1514 0.84 0.01 1525 0.75 0.01 3039 -0.095*** 0.029 0.07 2977 -0.087*** 0.024 0.14 No toilet at Baseline 1320 0.95 0.01 1297 0.85 0.01 2617 -0.096*** 0.019 0.06 2565 -0.089*** 0.018 0.07 Toilet at Baseline 191 0.12 0.02 214 0.13 0.02 405 0.008 0.038 0.1 395 0.006 0.035 0.18 BPL 375 0.93 0.01 300 0.74 0.03 675 -0.197*** 0.032 0.12 660 -0.198*** 0.035 0.16 Non-BPL 1139 0.81 0.01 1225 0.75 0.01 2364 -0.062* 0.032 0.07 2317 -0.054** 0.026 0.15 HH Reported OD by women Total Population 1514 0.83 0.01 1525 0.73 0.01 3039 -0.102*** 0.028 0.07 2977 -0.091*** 0.025 0.15 No toilet at Baseline 1320 0.94 0.01 1297 0.84 0.01 2617 -0.101*** 0.02 0.07 2565 -0.092*** 0.019 0.08 Toilet at Baseline 191 0.1 0.02 214 0.1 0.02 405 0 0.039 0.12 395 0.001 0.034 0.20 BPL 375 0.92 0.01 300 0.73 0.03 675 -0.206*** 0.033 0.13 660 -0.206*** 0.035 0.16 Non-BPL 1139 0.81 0.01 1225 0.73 0.01 2364 -0.068** 0.032 0.07 2317 -0.057** 0.027 0.16 HH Reported OD by children < 5 y Total Population 1514 0.89 0.01 1525 0.84 0.01 3039 -0.053** 0.021 0.04 2977 -0.054*** 0.017 0.11 No toilet at Baseline 1320 0.96 0.01 1297 0.91 0.01 2617 -0.055*** 0.012 0.03 2565 -0.054*** 0.011 0.04 Toilet at Baseline 191 0.41 0.04 214 0.42 0.03 405 0.044 0.064 0.11 395 -0.043 0.077 0.17 BPL 375 0.97 0.01 300 0.83 0.02 675 -0.139*** 0.024 0.1 660 -0.141*** 0.025 0.11 Non-BPL 1139 0.87 0.01 1225 0.84 0.01 2364 -0.025 0.025 0.04 2317 -0.028 0.019 0.12 Diarrhea in the past 7 days Total Population 2609 0.08 0.01 2600 0.07 0.01 5209 -0.003 0.008 0.02 5096 -0.002 0.008 0.03 No toilet at Baseline 2302 0.08 0.01 2231 0.08 0.01 4533 0.001 0.009 0.02 4435 0.003 0.009 0.03 40 / 41 Control Intervention ITT - Unadjusted ITT – Adjusted 1 N Mean SE N Mean SE N Coeff SE R2 N Coeff SE R2 Toilet at Baseline 302 0.06 0.01 343 0.03 0.01 645 -0.034* 0.019 0.04 630 -0.037 0.023 0.06 BPL 638 0.09 0.01 513 0.08 0.01 1151 -0.007 0.015 0.03 1120 -0.002 0.017 0.05 Non-BPL 1971 0.07 0.01 2087 0.07 0.01 4058 0 0.009 0.02 3976 0 0.011 0.03 Ascaris lumbricoides infection Total Population 569 0.04 0.01 581 0.04 0.01 1150 -0.002 0.01 0.02 1128 -0.011 0.01 0.03 No toilet at Baseline 487 0.05 0.01 482 0.04 0.01 969 -0.004 0.01 0.01 950 -0.013 0.01 0.03 Toilet at Baseline 82 0.04 0.02 92 0.04 0.02 174 -0.004 0.02 0.09 171 -0.005 0.041 0.14 BPL 147 0.07 0.02 109 0.04 0.02 256 -0.019 0.03 0.08 250 0.007 0.042 0.11 Non-BPL 422 0.04 0.01 472 0.04 0.01 894 0.007 0.01 0.02 878 -0.009 0.01 0.03 Giardia lambia infection Total Population 569 0.232 0.022 581 0.184 0.018 1150 -0.048** 0.024 0.03 1128 -0.036 0.026 0.04 No toilet at Baseline 487 0.232 0.023 482 0.185 0.02 969 -0.041 0.026 0.03 950 -0.036 0.029 0.04 Toilet at Baseline 82 0.232 0.04 92 0.185 0.054 174 -0.115** 0.053 0.18 171 -0.06 0.073 0.23 BPL 147 0.224 0.037 109 0.165 0.036 256 -0.045 0.048 0.08 250 -0.062 0.057 0.12 Non-BPL 422 0.235 0.023 472 0.189 0.021 894 -0.050* 0.027 0.03 878 -0.037 0.029 0.05 Weight-for-age z-score# Total Population 2161 -1.83 0.03 2154 -1.92 0.03 4315 -0.095 0.079 0.02 4239 -0.094 0.076 0.05 No toilet at Baseline 1909 -1.87 0.03 1854 -1.95 0.03 3763 -0.104 0.085 0.02 3699 -0.103 0.08 0.05 Toilet at Baseline 247 -1.56 0.09 275 -1.68 0.07 522 -0.132 0.134 0.05 510 -0.086 0.134 0.07 BPL 551 -2 0.06 441 -1.95 0.07 992 0.034 0.121 0.05 972 0.033 0.121 0.08 Non-BPL 1610 -1.78 0.04 1713 -1.91 0.03 3323 -0.143* 0.081 0.02 3267 -0.132* 0.079 0.05 Length/height-for-age z-score# Total Population 2185 -2.16 0.04 2175 -2.19 0.04 4360 -0.034 0.081 0.02 4275 -0.04 0.092 0.08 No toilet at Baseline 1928 -2.19 0.04 1873 -2.23 0.04 3801 -0.054 0.08 0.03 3727 -0.05 0.084 0.09 Toilet at Baseline 253 -1.86 0.11 281 -1.88 0.1 534 0.068 0.184 0.03 523 0.004 0.27 0.05 BPL 549 -2.22 0.08 439 -2.25 0.08 988 -0.01 0.143 0.05 963 0.057 0.16 0.12 Non-BPL 1636 -2.13 0.04 1736 -2.17 0.04 3372 -0.034 0.079 0.02 3312 -0.046 0.09 0.08 Abbreviations. BPL: Below Poverty Line; HH: household; ITT: Intention-To-Treat; JMP: WHO/Unicef Joint Monitoring Programme for Water Supply and Sanitation; OD: Open Defecation 1 Adjusted models included: whether household head attended school, whether household belongs to Scheduled Caste or Tribe, percentage of surveyed households in the GP with improved water source at baseline, percentage of surveyed households in the GP with soap, water present at hand washing place used after defecation, and average height-for-age Z-score in the GP at the baseline. For child level outcomes, age of child in months and sex were also included. * P < 0.1 ** P < 0.05 *** P < 0.01 41 / 41