Context and intentions: practical associations for fecal sludge management in rural low-income Cambodia

Finite storage capacities of household pit latrines make safely managing fecal sludge a recurring challenge for 2.7 billion people globally. Frequently without guidance from standards or regulation, rural latrine owners choose how to manage their own fecal sludge. However, their intentions – what behavioral science says are the best predictors of future behaviors – when pits fill are poorly understood, inhibiting the development of safe fecal sludge management (FSM) solutions and deteriorating public and environmental health. Using survey data commonly measured by development practitioners, we analyze response frequencies and their associations with contextual factors, such as location, month that the survey was administered, and poverty level. We also use binomial logistic regression to determine if contextual factors can be used to predict the intentions of rural Cambodian latrine owners when pits fill. We found that four in ten rural latrine owners intend to manage their fecal sludge unsafely (41%), and one in six did not have a plan (16%). Desirable FSM intentions increased markedly after rice harvest and varied markedly across provinces. Many predictors of desirable FSM intentions, such as location and satisfaction with the household’s latrine, were also identified. Associations between FSM intentions and contextual factors can be used to help predict FSM behaviors and improve FSM service delivery, behavior change campaigns, and product design. However, future work should seek to characterize the complete decision-making processes of rural latrine owners when pits fill. This is an Open Access article distributed under the terms of the Creative Commons Attribution Licence (CC BY 4.0), which permits copying, adaptation and redistribution, provided the original work is properly cited (http://creativecommons.org/licenses/by/4.0/). doi: 10.2166/washdev.2020.103 ://iwaponline.com/washdev/article-pdf/10/2/191/713031/washdev0100191.pdf James Harper (corresponding author) Department of Civil, Environmental and Architectural Engineering, College of Engineering and Applied Science, University of Colorado at Boulder, 1111 Engineering Drive, 422 UCB, Boulder, CO 80309-0422, USA E-mail: james.harper@colorado.edu Angela Bielefeldt Amy Javernick-Will Department of Civil, Environmental and Architectural Engineering, College of Engineering and Applied Science, University of Colorado at Boulder, 1111 Engineering Drive, 428 UCB, Boulder, CO 80309-0428, USA Toeur Veasna Monitoring and Evaluation Manager, iDE Cambodia, 97A, St. Taphon (19BT), Boeung Tumpun 12351, Cambodia Chris Nicoletti Director of Measurement, Evaluation and Learning, iDE Global, 1031 33rd St #270, Denver, CO 80205, USA This article has been made Open Access thanks to the generous support of a global network of libraries as part of the Knowledge Unlatched Select initiative.

exposure to fecal sludge (Still & Foxon ). Unsafe methods include a latrine owner emptying their own pit (self-empty); releasing fecal sludge into floodwaters by opening a pit's lid during a flood (flooding out); or stopping latrine use.
It is important to note that while self-empty can be performed safely if personal protective equipment is worn, and proper techniques are used (Strande & Brdjanovic ), the literature describing rural fecal sludge management (FSM) in Cambodia shows that these safe procedures are rarely followed during self-empty (Kong & Bartell ; SNV ); thus, this study considers self-empty to be unsafe. Safe FSM, which includes the collection, transport, treatment, and disposal of fecal sludge, must be achieved to mitigate serious health and environmental consequences.
In rural communities, latrine owners typically choose how to manage their own fecal sludge. Research, however, has largely focused on technologies and economics (Strande & Brdjanovic ) while largely ignoring household-level decision-making (Jenkins Cumming & Cairncross ). Cambodia, a country with a large rural population that relies on on-site sanitation systems, a study in three provinces (n ¼ 39) and a study in two provinces (n ¼ 633) provide some information about household-level decisionmaking when pits fill (Kong & Bartell ; PSI ; SNV ). However, due to their limited geographic distributions, these studies could not draw conclusions about how intentions, the best predictor of future behaviors (Fishbein & Ajzen ), when pits fillhereafter called 'FSM intentions'vary by region, season, and other contextual factors across Cambodia, the focus of this study.
FSM intentions describe how willing a person is to perform certain behaviors related to FSM. Various models of decision-making describe how context affects the formation of intentions (Darnton ). Contextual factors can be categorized as personal (e.g., education, emotions, age, gender, income, religion), societal (e.g., social norms, culture, economy, politics), or physical (e.g., geography, season, weather, environment) (Fishbein & Ajzen ). The contextual factors that are relevant to one behavior may be irrelevant to another. Past studies of rural sanitation identified factors relevant to latrine adoption and maintenance behaviors, including availability of funds, physical space, service provision, the defecation practices of neighbors (a social norm), and a latrine owner being satisfied with their latrine (an attitude) ( Jenkins & Scott ; Coffey et al. ). While all components of the decision-making process are of interest, we analyze if contextual factors commonly measured by development practitioners and other agencies can be used to predict the FSM intentions of rural Cambodian latrine owners across seven provinces.

Research context and data collection
Cambodia is a lower-middle-income country of approximately 16 million people with approximately 80% living in rural areas (Asian Development Bank ). Much of Cambodia's economy (60%) is based on rice cultivation along its major waterways ( Figure A1 in Appendix A) (Food and Agriculture Organization of the United Nations ).
Since 2010, the international development organization iDE has studied how latrine owners in rural Cambodia perceive and interact with sanitation systems, which are primarily pour-flush single-pit latrines (iDE Cambodia ).
To improve sanitation marketing, iDE Cambodia developed a survey in 2014 to describe the experiences, intentions, and behaviors of rural latrine owners regarding the construction, use, maintenance, and improvement of their household's latrine. Sixty-seven survey questions asked about latrine owners' current and past defecation behaviors; intentions when pits fill; household demographics and locations; latrine construction details and costs; satisfaction with their latrine and its installer; and various personal, societal, and physical contextual factors. The questionnaire was developed and administered prior to iDE partnering with the first three authors to study FSM intentions.
The survey was administered to 3715 rural latrine owners who had purchased their latrine within 6 to 12 months before the survey. This sampling frame was selected to understand how latrine owners that do not currently have full pits are planning for their pit filling in the future, which usually occurs after 3-6 years (iDE Cambodia ). Each latrine owner was randomly selected from rural households that had purchased their latrine from an iDE-affiliated latrine installer in seven Cambodian provinces (Oddar Meanchey, Banteay Meanchey, Siem Reap, Kampong Thom, Kandal, Prey Veng, and Svay Rieng; Figure A1). Local iDE research assistants that speak native Khmer and live locally were trained both in classroom and field exercises to ensure accurate and repeatable survey administration. Surveys were administered in-person with rural latrine owners, and the data collected was reviewed by iDE's Monitoring and Evaluation managers and the authors for accuracy and completeness.
Responses were recorded by research assistants digitally in TaroWorks

Data analysis
We categorized responses to the question 'When your pit is getting full, what do you intend to do?' as desirable or undesirable FSM intentions. Desirable FSM intentions are held by latrine owners that intend to manage their FS safely by either paying for professional emptying or installing a new pit. Within this study, desirable FSM intentions provide an opportunity for safe FSM; for example, paying for professional emptying provides the opportunity for safe FSM because regulating FSM service operators is at least possible, while regulating how every household empties their own latrine is effectively impossible. All other intentions, including self-emptying, stopping latrine use, and being undecided, were defined in this study to be undesirable because they are unlikely to provide an opportunity for safe FSM. We analyzed the frequency of specific FSM intentions, their desirability, and associations with personal, societal and physical contextual factors. Personal contextual factors included poverty level (type 1, type 2 or not) (Ministry of Planning Government of Cambodia ); latrine usage (frequently, sometimes, rarely, unknown); defecation behaviors before constructing the household's latrine (practiced open defecation, used neighbor's latrine); satisfaction with the latrine and its installer (very unsatisfied through very satisfied in five steps, unknown); whether a household had recommended a latrine or its installer to a friend; and whether a household intended to improve their latrine within three years (improve shelter, flush-water reservoir, shower, new pit, sink, other). We investigated one societal contextual factor: how many members of their community households believed to practice open defecation (most, some, few). Physical contextual factors included challenges with the latrine (clogging, smells, flooding, insufficient flush water, full or overflowing); the province where the household was located; whether the household was located near a river or pond; the year and month the data was gathered; and the monthly average rainfall for the household's province.
See associations that are of low uncertainty and practical benefit to rural FSM are reported and discussed. We define 'of low uncertainty' in this study as 'having a statistical significance (p) that is less than 0.1' and 'of practical benefit' as 'capable of helping predict rural FSM behaviors or improve rural FSM services, behavior change campaigns, policies or product design'. In this study, we report associations between FSMintention desirability and factors that either (1) have been shown to associate with other aspects of rural sanitation systems (e.g., latrine access or use) in the literature; or (2) provide practical benefit to rural FSM development. Statistical significances (p-values) were calculated using the chi-squared test and are reported for all tests, and statistical importances are reported as either Cramer's v or a point-biserial correlation r pb .
A binomial logistic regression model was created to predict the desirability of FSM intentions (dependent variable).
Contextual factors measured in the survey were used as independent variables in the model. The model was developed iteratively, starting with all contextual factors measured in the survey as inputs. After the model was run once, we removed the factor that contributed the smallest reduction in residual deviance. This process was repeated until all factors included in the model provided practical benefit to predicting FSM-intention desirability with low uncertainty. A total of 245 responses were omitted from this analysis due to incomplete data. The statistical significance of each factor in the model was calculated using an analysis of variance (ANOVA) test that compared the full model to the reduced model. A 95%/5% training/testing dataset structure was used, and Nagelkerke's pseudo-R 2 is reported to describe model quality.

RESULTS AND DISCUSSION
Results of practical benefit with low uncertainty are discussed in the following sections, and a summary of these results is shown in Table 1 for clarity. Comprehensive results are available in the online appendices.

Frequency of undesirable FSM intentions
Responses of 3715 rural latrine owners to the question 'When your pit is getting full, what do you intend to do?' are shown in Table 2, and all contextual factors are shown in Table B1. Four in ten rural latrine owners reported undesirable FSM intentions (41%, Tables 1 and 2). However, few respondents intended to stop using their latrine when their pits fill (2%), and a majority intended to improve their  Table B2 for details. ***: p < 0.01, **: p < 0.05, *: p < 0.1 in association to desirability of FSM intentions.
-Not applicable or no important association found.
latrine within the next three years (54%), indicating a strong desire to continue using their latrine (Table B1). Relatedly, most respondents reported high satisfaction with their latrine (86%) and high latrine use (97%, Table B1).  These results all point to the institutionalization of practice when norms and beliefs influence future intentions.
While the underlying cause of these connections is unknown, it may be related to making a public commitment to sustainable sanitation ( Jenkins & Scott ).

Physical contextual factor: survey administration timing
FSM intentions varied with the year of data collection (p ¼ 0.000, v ¼ 0.14, Tables 1 and B2). The frequency of intending to pay for professional emptying remained relatively constant at ∼35% between 2015 and 2017, while the frequency of intending to install a new pit doubled, and the frequency of intending to self-empty halved (Table B3).
Indecision was common among latrine owners (11% to 19%), while intending to stop using a latrine or having other intentions remained low (Table B3). These results were unexpected because rainfall is known to strongly affect pit sludge levels (Strande & Brdjanovic ). However, results showed that unless a pit overflows, which was reported rarely (1% , Table B1

Predicting FSM intentions
A binomial logistic regression model was used to examine the relative strengths of the associations between contextual factors and FSM intentions described above, and identify any confounding in the above results. The final model used nine contextual factors as inputs and predicts FSM-intention desirability via Equation (1): The resulting equation is: indicates the range of estimates for the various responses for each factor (see Table B5 for confidence intervals and other details).
Characteristics of each factor in the model are shown in Table B4; factors are arranged by decreasing deviance, which describes the statistical importance of each factor.
Estimates, standard errors, confidence intervals, z-values and significances of the coefficients in the binomial logistic regression model are shown in Table B5.
All of the results described in the above sections are reproduced in the model. Additionally, the model shows that Satisfaction has a high deviance, and its estimates show that any satisfaction score other than 1 (very unsatisfied) is a good predictor of desirable FSM intentions.  Ideally, all components of the decision-making process should be characterized in detail to understand how intentions are formed and explain any differences between intentions and behaviors that may occur. This requires a targeted study design and enormous person-hour investment for study development and data collection and analysis.

Recommendations for development practitioners
We recommend that future work conduct a comprehensive study that characterizes each component of the decisionmaking process, including behavioral determinants and control-related topics, of rural Cambodian latrine owners when pits fill.
Future work should also randomize question order to reduce bias and survey households that purchased latrines outside of iDE's sales network to improve generalizability.
Future studies should continue to employ local research assistants of the same gender as respondents, and respondents should be encouraged to be wholly honest in their responses. Lastly, while the rural areas of the provinces sur- With recent rapid improvements in latrine access after decades of effort and enormous monetary investment, it is critical that the associated improvements in public health continue to push rural communities globally towards achieving the goals of SDG6 (UN ). This will require achieving safe FSM in rural communities globally and will rely on a thorough understanding of how households make decisions when their latrine pits fill.