Long-term impact of a community-led sanitation campaign in India, 2005–2016

Abstract Objective To evaluate the long-term impact of a community-led total sanitation campaign in rural India. Methods Local organizations in Odisha state, India worked with researchers to evaluate a community-led total sanitation campaign, which aimed to increase the demand for household latrines by raising awareness of the social costs of poor sanitation. The intervention ran from February to March 2006 in 20 randomly-selected villages and 20 control villages. Within sampled villages, we surveyed a random subset of households (around 28 households per village) at baseline in 2005 and over the subsequent 10-year period. We analysed changes in latrine ownership, latrine functionality and open defecation among approximately 1000 households. We estimated linear probability models that examined differences between households in intervention and control villages in 2006, 2010 and 2016. Findings In 2010, 4 years after the intervention, ownership of latrines was significantly higher (29.3 percentage points; 95% confidence interval, CI: 17.5 to 41.2) and open defecation was significantly lower (−6.8 percentage points; 95% CI: −13.1 to −1.0) among households in intervention villages, relative to controls. In 2016, intervention households continued to have higher rates of ever owning a latrine (26.3 percentage points; 95% CI: 20.9 to 31.8). However, latrine functionality and open defecation were no longer different across groups, due to both acquisition of latrines by control households and abandonment and deterioration of latrines in intervention homes. Conclusion Future research should investigate how to maintain and rehabilitate latrines and how to sustain long-term behaviour change.


Introduction
Over 800 million people defecate in the open rather than use a toilet. 1 Though global mortality from diarrhoeal diseases has recently declined, 1 poor water and sanitation are responsible for hundreds of thousands of child deaths, as well as high rates of diarrhoea, malnutrition and stunting. 2,3 Given the strong link between health, human development and productivity, 4 and the benefits of privacy, dignity and gender empowerment, 5 improving sanitation remains a key sustainable development goal. 6 Crucial in achieving objectives for improved sanitation is the need for long-term changes in behaviour that are deeply embedded in community cultures and norms. 7 Weak institutions, insufficient local capacity and dysfunctional markets for sanitation improvements make it especially challenging to achieve sustained behaviour change. 8 At least two major knowledge gaps have hampered prior efforts to overcome barriers to better sanitation. First, scalingup sanitation initiatives requires translating successful trials of their efficacy to achieve widespread use. 9,10 For example, community-led approaches recognize that effective promotion must change deep-seated attitudes and practices, rather than simply providing a latrine. Despite the appeal of communityled total sanitation, rigorous tests of its implementation are rare, and questions remain concerning the value of various promotion strategies. A second major gap is that few studies in the water, sanitation and hygiene sector have examined whether behaviour changes following an intervention are sustained in the long term. Several studies have rigorously evaluated the short-term behavioural and health impacts of sanitation interventions. [11][12][13][14][15][16] Yet recent systematic reviews [17][18][19][20][21] identified only one peer-reviewed sanitation evaluation with a control group that examined the impact beyond 3 years. A study in 1996 showed that while diarrhoea rates were lower 6 years after a water, sanitation and hygiene intervention, only 63.9% of the latrines (425 out of 665) remained functional. 22 A greater focus on sustainability is needed for several reasons. First, the behaviour change effects of health interventions have often been shown to decline over time; 17 shortterm evidence may therefore overestimate the true potential of interventions. Even households with access to improved sanitation may continue or revert to open defecation as time progresses. 23 Second, deterioration of latrines due to the cost of maintenance may outweigh the benefits over time. 24 Third, more effective programme design relies on understanding how policies improve sustainability of health interventions.
To begin to address these knowledge gaps, our study describes the evolution of latrine ownership, functionality, use and sanitation beliefs over 10 years following a community-led total sanitation intervention in Odisha state, India. We focused on these outcomes because they are key indicators of sustainability for this intervention. The study setting is particularly relevant for examining the long-term progression of sanitation behaviour; India has high rates of poor sanitation and roughly half of 1.2 billion Indians still practise open defecation. 25 The design of the Odisha intervention drew on insights from Objective To evaluate the long-term impact of a community-led total sanitation campaign in rural India. Methods Local organizations in Odisha state, India worked with researchers to evaluate a community-led total sanitation campaign, which aimed to increase the demand for household latrines by raising awareness of the social costs of poor sanitation. The intervention ran from February to March 2006 in 20 randomly-selected villages and 20 control villages. Within sampled villages, we surveyed a random subset of households (around 28 households per village) at baseline in 2005 and over the subsequent 10-year period. We analysed changes in latrine ownership, latrine functionality and open defecation among approximately 1000 households. We estimated linear probability models that examined differences between households in intervention and control villages in 2006, 2010 and 2016. Findings In 2010, 4 years after the intervention, ownership of latrines was significantly higher (29.3 percentage points; 95% confidence interval, CI: 17.5 to 41.2) and open defecation was significantly lower (−6.8 percentage points; 95% CI: −13.1 to −1.0) among households in intervention villages, relative to controls. In 2016, intervention households continued to have higher rates of ever owning a latrine (26.3 percentage points; 95% CI: 20.9 to 31.8). However, latrine functionality and open defecation were no longer different across groups, due to both acquisition of latrines by control households and abandonment and deterioration of latrines in intervention homes. Conclusion Future research should investigate how to maintain and rehabilitate latrines and how to sustain long-term behaviour change.
multiple domains, and was successful in addressing short-term demand and supply barriers to behaviour change, especially among the rural poor. 11 Our research also has policy relevance. In 2000, the government of India launched a total sanitation campaign aiming to achieve universal rural sanitation coverage by 2012, with the emphasis on community-level information, education and communication and subsidies for latrine building. However, the programme had minimal impacts on open defecation behaviour and no effect on child health outcomes. 15 In 2014, the Swachh Bharat Mission was launched with renewed efforts to end open defecation by 2019. Despite the success of that programme in delivering millions of latrines, the persistence of open defecation 7 highlights the need for more research that is focused on the barriers to long-term latrine use and related outcomes. 26 The aim of our study was to examine the long-term impacts of a communityled total sanitation campaign in Odisha, India on latrine adoption and open defecation, and to study barriers to continued latrine use.

Baseline study
The original cluster-randomized control trial tested the effects of a community-led total sanitation-type campaign in Odisha state in 2005-2006. 11 The study included 40 villages from two adjacent blocks in Bhadrak district (population around 1.5 million), where very little sanitation promotion had previously occurred. Twenty of these villages were randomized to receive a village-level intensive promotion campaign, while the other 20 villages served as controls. To obtain baseline data from sample villages, a random subset of households (around 28 households per village) were surveyed in August and September 2005. The baseline sample in each arm was generally representative of rural conditions in the region (Table 1 From February to March 2006, a specialized field team worked closely with community-based organizations and state and local government leaders to mobilize the target population and implement the campaign in the 20 intervention villages. The campaign included community education activities on the benefits of latrines, including health, safety and dignity, and privacy for women; providing local support for building pit latrines; and awareness-raising about government subsidies for latrine building (Box 1). A post-intervention survey conducted in August and September 2006, 6 months after the intervention, showed that latrine ownership had increased by 24.1% (from 40/534 to 168/521 households) in treatment villages relative to controls (from 71/552 to 70/529 households), and had increased more among belowthe-poverty-line households that were eligible for government subsidies. 11 Additionally, the intervention was linked to improvements in child health. 27

Follow-up study
We examined the sustainability of changes in latrine ownership, functionality, open defecation and sanitationrelated perceptions, using surveys administered to the original households about 4 and 10 years after the intervention, in July and August 2010 and January and February 2016, respectively. Surveys were conducted in Oriya by trained enumerators who were recent or current university students, although not the same as those at baseline. Additional questions were included in the later rounds to better understand latrine use and abandonment. Fig. 1 depicts the study timeline. We were able to resurvey most of the original households in 2010 (96.0%; 1043/1086 households) and 2016 (97.5%; 1059/1086 households). For households that we did not relocate in 2016, we replaced the household with a neighbouring household.
The 2010 and 2016 surveys were conducted with the approval of the Duke University intuitional review board (protocols 2799 and D0352). Local permission to conduct these surveys was obtained from village leaders and via informed consent from all participating households.

Data analysis
To quantify the effects of the promotion effort, we used linear probability models to examine the differences between households in intervention villages and controls in 2006, 2010 and 2016. We used the following equation: (1) where y ijt is the outcome y measured for household i in village j in year t, and T j is the treatment status of village j. The outcome variables that we measured were: (i) households ever owning and

Box 1. The community-led total sanitation campaign in Odisha, India, 2006
A specialized field team worked closely with community-based organizations and state and local government leaders to implement the different components of the campaign: 11 (i) A set of intensive information, education and communication activities that gathered villages together to evoke a collective response to widespread open defecation. Activities included: • defecation mapping: villagers drew maps of their village and noted the relative distances between common defecation sites and water sources, crops and schools; • walk of shame: villagers walked around their village identifying common defecation sites and noting the proximity of sites to other important community locations (e.g. water sources or health clinics); and • faecal mass calculations: a facilitator worked with villagers to collect and weigh the faeces gathered from these village defecation sites.
(ii) Establishing accessible latrine production centres in villages. The aim was to provide materials and support for to households and reduce the costs of pit latrine construction.
(iii) Raising of people's awareness about subsidies for latrine building. At the time, few households in the study site were taking advantage of the available subsidies from the Indian government to households below the poverty line. The information campaign took place in 2006 only. Although village production centres created an infrastructure that enabled continuing support to households, due to limited follow-up, it is unclear whether these supply-side activities continued. Finally, while the intervention only delivered information about the availability of subsidies in 2006, the government of India continues to offer subsidies to below-the-poverty-line households.
Impact of community led total sanitation campaign, India Jennifer Orgill-Meyer et al. We ran separate regressions for the outcomes in 2006, 2010 and 2016. Thus, β 1 represents the average difference between households in intervention and control villages in a given year. To allow for intra-village correlation, we clustered standard errors ε ij for all analyses at the village level. We also controlled for the few baseline variables X ij0 (i.e. baseline latrine use, television ownership and satisfaction with village sanitation) that were somewhat unbalanced between treatment and control villages in 2005. We also conducted logistic regressions and found that the results were consistent across both sets of models.
For the outcomes in categories (i), (ii) and (iv) above, we conducted the analyses using the full sample of households. In these cases, the coefficient β 1 represents the effect of being in an intervention village in 2005 on the outcome of interest. For category (iii), the regressions were only for households owning latrines at the time of the survey. β 1 in this case measures how latrine function and quality indicators vary across households owning latrines in the intervention versus control villages. For ownership and abandonment, we conducted separate analyses for the households below and above the poverty line, to assess how subsidy eligibility related to sustainability.
If nongovernmental organizations (NGOs) or local governments had become more involved in control villages after the intervention concluded, we might observe differential rates of latrine adoption and open defecation behaviour in the long term. We tested category (v) above in two ways. First, we used community surveys conducted with village leaders in 2016 that asked about the sanitation activities of NGOs and local government within the village. Second, we used the household surveys from 2010 and 2016 that asked household members to recall specific sanitation activities occurring in their village.

Latrine adoption and ownership
In intervention villages the proportion of households that had ever acquired a     (Fig. 2). By 2010, roughly half of in communityled total sanitation villages (271/517 households) had ever built a latrine, compared with only a quarter (131/527 households) in control villages. Controlling for baseline household variables, the increase in the percentage of households having a latrine in 2010 was 29.3 percentage points (95% confidence interval, CI: 17.5 to 41.2) higher than in control villages ( Table 2). The increase was higher among below-the-povertyline households (38.1 percentage points; 95% CI: 24.9 to 51.2), who could obtain subsidies to help defray the costs of constructing the latrine.
Despite the fact that intervention households were more likely to have ever owned a latrine in 2016 (Fig. 2), there was no significant difference in current ownership of a latrine between treatment and control villages in 2016 (Fig. 3), due to higher rates of adoption by new households in control villages after 2010 and latrine abandonment of intervention households. Nearly all intervention households who acquired latrines after 2005 were still using them in 2010, but by 2016 more households in intervention villages reported having ever abandoned a latrine (21.7 percentage points; 95% CI: 14.6 to 28.7; Table 2). Consequently, there were no significant differences in ownership rates in 2016 between control and intervention villages (1.3 percentage points; 95% CI: −11.1 to 13.7). Correlates of latrine adoption and the effects of the intervention on latrine use across different ages and gender can be found in the data repository. 28

Open defecation
Consistent with these differential patterns in ownership across villages, there was a decline in open defecation in the short-term between the 2005 and 2006 surveys among intervention villages (from 94.8% to 71.8%) relative to control villages (from 87.8% to 90.5%; Fig. 4). This small decline relative to latrine adoption suggests relatively low latrine usage.  (Table 2). However, open defecation in intervention villages increased relative to 2006 (from 374/521 to 446/517), even as latrine ownership increased in this group (Fig. 3). Since latrine abandonment was relatively low in 2010, these higher rates of open defecation suggest that households did not fully switch to using latrines 4 years after the intervention. In the longer-term   Research Impact of community led total sanitation campaign, India Jennifer Orgill-Meyer et al.
due to both a decline in open defecation in control villages (Fig. 4) and abandonment of latrines in intervention villages (Fig. 3). The effect of the intervention on open defecation by individuals of different ages and sexes can be found in the data repository. 28

Latrine abandonment
Given the initial success of the intervention, and increased adoption of latrines among intervention households in 2010, it is perhaps surprising that households would abandon latrines and continue open defecation in the long term. The lack of difference in current ownership rates reported in 2016 was a combination of catch-up in control villages and abandonment in intervention villages. The latrine pit being full did not appear to be a major reason for abandonment. Few households owning latrines in 2016 reported ever needing to empty their pit (31/135 treatment households and 29/153 control households), which points to relatively low use of latrines. Our data suggest, however, that construction quality was an important factor. By 2016, latrines were more likely to have observable structural problems in intervention relative to control households (effect size: 24.1%; SE: 5.8) and existing latrines were less likely to be functional in intervention households (−26.3%; SE: 5.6; Table 3). Latrines in intervention villages were generally 2.6 years older than those in control villages (P = 0.02) and were more likely to have problems in 2010 and 2016, whereas in 2006 these latrines were new and less likely to have problems ( Table 3).
Memory of the promotion effort appears to have faded quickly. In 2006 more households in intervention villages remembered sanitation promotion activities (effect size versus controls: 58.1%; SE: 4.6), whereas in 2010 there was no significant difference (effect size: 4.8%; SE: 3.7) and by 2016 households in treatment villages were less likely to remember promotion activities (effect size: −18.7%; SE: 6.2; Table 3). In addition, a high percentage of control households in 2010 reported remembering promotion activities (82.9%; 437/527) and it is therefore possible that recent sanitation activities in these villages were partially responsible for the increase in latrines in control villages. Compared with control households, more households in intervention villages reported pressure to build latrines from community members in 2010 (effect size: 11.8%; standard error of mean, SE: 5.9), but not in 2016 (effect size: −0.5%; SE: 4.0). Even so, more households in intervention villages aspired to rebuild latrines than those in control villages, even in 2016 (effect size: 12.5%; SE: 5.0; Table 3). The correlates of latrine abandonment can be found in the data repository. 28

Involvement of other organizations
Respondents in treatment and control villages reported similar levels of NGO involvement (effect size: 0.0%; SE: 9.7) and government sanitation assistance (effect size: 5.0%; SE: 13.5; Table 4). Households in control villages remembered sanitation activities taking place in their villages in 2010 and 2016 more   Impact of community led total sanitation campaign, India Jennifer Orgill-Meyer et al. Table 3. Research Impact of community led total sanitation campaign, India Jennifer Orgill-Meyer et al.

Outcomes of the community-led total sanitation-type intervention among control and intervention households in
than did households in treatment villages (Table 3 and Table 4). Thus, some of the increase in control latrine ownership over the course of the study was likely due to more sanitation promotion happening in control villages following the intervention.

Discussion
We returned to the site of an intensive behaviour change campaign and found that acquisition and use of latrines remained higher in treatment villages 4 years after the intervention. The medium-term success of this intervention can be attributed to several features. First, the campaign focused on a range of sanitation benefits, including health, privacy and dignity, and therefore helped to motivate the use of latrines. Second, the intervention recruited local builders and created an infrastructure for continued sales of latrine materials even after the intervention ended, which allowed households to make investments even after 2006. Third, the campaign targeted communities, creating the possibility of spillovers across peers and over time.
Finally, the intervention helped to connect below-the-poverty-line households with government-sponsored subsidies, rendering latrines more affordable. While we observed many new latrines in 2010 relative to 2006, those facilities could have been built long before 2010. Indeed, many treatment households reported their intention to build a latrine in 2006, 11 likely spurred on by the behaviour change campaign implemented earlier that year. The numbers of new latrines acquired up to 2010 suggest that the promotion campaign had effects on investment that persisted for at least 4 years, even as the conventional national campaign was progressing in the background. Despite this medium-term success, after 10 years there was no significant difference in latrine ownership and open defecation between treatment and control communities. Some of the decline in the effect of the original campaign was due to acquisition of latrines over time in control villages, which could be the result of increased sanitation activity in control villages following the intervention. Yet this decline was also driven by treatment households abandoning latrines or discontinuing use. There are several reasons why a household might abandon a latrine. Latrines might have filled up, and households lacked the know-how or financial resources to empty them; maintenance costs for latrines might have increased over time; and community-led total sanitation may not have sufficiently shifted local norms favouring open defecation.
Our results suggest three main challenges in sustaining improved sanitation behaviour, relating to both hard (infrastructure) and soft (information, education and communication) inputs. First, open defecation may persist even among households that built latrines; latrine use thus did not become a universal social norm. Second, campaign messages lose effect over time, which may contribute to the persistence of open defecation even in intervention villages. More regular messaging may be critical to overcome short memories and deep-rooted habits. 7,29 Third, the Notes: All estimates are from linear probability models that include control variables that were unbalanced at baseline. The results omitting controls, available upon request from the corresponding author, are not substantively different. Treatment effect (%) reports the coefficient on the treatment variable from these linear probability models, thus measuring the difference between the means in treatment and control villages. All estimates pertain to the effect of the 2006 community led total sanitation intervention, as conducted at the community level, and are thus intention-to-treat results. In 2016, we asked respondents if they remembered any of the sanitation activities within the past 5 years. In 2010, we asked respondents if they remembered any of the sanitation activities ever occurring, with a follow-up question about the timing of the activity. SEs of the treatment effects are clustered at the village level.
Impact of community led total sanitation campaign, India Jennifer Orgill-Meyer et al. quality of latrines deteriorates over time, creating a risk that households will abandon them. Given household budget constraints and thin supply chains for maintenance services in our study setting, ongoing financial support for latrine upkeep (parts and labour) may be necessary to sustain the impact of interventions. 30 In the context of quasi-public goods, such as sanitation and especially in resource-poor settings with affordability challenges, continued provision of information and maintenance of the infrastructure often requires governmental or other institutional support. Such support should not be assumed, however. While government and donors continue to work to increase sanitation coverage among the poor, they typically devote less attention to maintaining investments. Our findings are particularly important in the context of the current Swachh Bharat Mission programme in India, which primarily focuses on infrastructure provision. Given the high cost of the programme, it is important that policy-makers do not neglect the elements required to ensure persistent declines in open defecation and hence longer-term benefits.
Our findings also point to the need for more work that examines the success of interventions beyond a few years. More research on the sustainability of sanitation interventions would help to confirm whether the findings of this study are relevant to experiences in other contexts. Such long-term research is uncommon in public health and other applied fields, in part because it is so costly. Nonetheless, understanding how and why programmes affect outcomes in the long term is necessary for improving the sustainability of public health interventions. ■ Funding: Duke University and the National University of Singapore provided support for this research.