Women ‘holding it’ in urban India: Toilet avoidance as an under-recognized health outcome of sanitation insecurity

ABSTRACT

Emerging research on sanitation challenges in the Global South increasingly uncovers health and social impacts by gender, particularly lack of sanitation safety. Women may employ strategies to avoid urination or defecation (‘holding it’) in the absence of safe sanitation, but the practice is not well understood. We quantitatively analyze survey data on women from urban slums across three cities in Maharashtra, India whose households constructed a toilet through an intervention programme. We assess relationships between household versus shared sanitation, perceptions of safety, and women’s toilet avoidance behaviours, including diet restriction. At baseline, women have more than three times the odds of reporting avoidance behaviours if they perceive a community toilet to be unsafe, even after controlling for other factors. Household water insecurity is also instrumental in the relationship between avoidance and lack of safety. Finally, avoidance exhibits a significant and major drop upon provision of a household toilet. This study provides substantial support for the prevalence of habitual toilet avoidance among vulnerable urban women without access to safe sanitation. We conclude with recommendations for policy approaches and call for more attention to the health repercussions of habitual toilet avoidance among women as a consequence of sanitation insecurity.

KEYWORDS:

• WASH
• gender inequality
• urban
• informal housing
• South Asia

Acknowledgements

The authors thank Shelter Associates staff, the University of South Florida College of Public Health Student Research Scholarship, and several anonymous reviewers and respected colleagues for their comments on this text.

Disclosure statement

No potential conflict of interest was reported by the author(s).

Notes

1 For the purposes of this research, we utilize the terms slum, informal settlement, and informal housing interchangeably to refer to housing that is unrecognized or semi-recognized by the state. We also draw from the UN-HABITAT’s definition: ‘A slum household [is] one in which the inhabitants suffer one or more of the following ‘household deprivations’: lack of access to improved water source, lack of access to improved sanitation facilities, lack of sufficient living area, lack of housing durability and lack of security of tenure’ (UN-HABITAT, 2016, p. 2).

2 Pune Mirror, 2019: https://punemirror.indiatimes.com/pune/civic/open-defecation-leads-to-plaints-of-harassment-at-slum-in-taljai/articleshow/69403323.cms.

3 Hindustan Times, 2017: http://www.hindustantimes.com/mumbai-news/mumbai-s-dirty-secret-out-in-the-open/story-tDSkqPwyeswBFgqS7TJiKI.html.

4 India Today, 2017: https://www.indiatoday.in/india/rajasthan/story/rajasthan-activist-lynched-by-civic-officials-983207-2017-06-17.

5 Infrastructure data and maps for settlements across several cities is available at https://shelter-associates.org/index.php#map-plugin.

6 Because some of these issues can change over time – for instance, sewer lines may be laid in that area by the municipality later on and the household may then construct a toilet – it is possible that some households lost to follow-up here may be captured in subsequent waves of data collection.

7 Due to SA’s policies, and the fact that data pertains to vulnerable human subjects, this dataset is not publicly accessible. Researchers may follow http://www.shelter-associates.org for access.

8 Subsets of the scale broken down thematically had much lower Cronbach’s alpha values, so a summed variable for the whole scale was used. While a principal components analysis (PCA) or multiple components analysis (MCA) for non-continuous data are commonly used to construct household possessions scales (Kaiser et al., 2017), an initial MCA suggested little variation (14%) being explained by the first principal component. Also, the way indices are coded may be more relevant than the actual weighting method Howe et al. (2008). In this case, all variables in the scale are binary, which may demonstrate higher agreement than scales with a mix of variable types. Finally, the utility of PCA/MCA tends to lie in cross-national or cross-regional studies with a large range in wealth, but the usage of the scale in a more specific population – urban slums in Maharashtra – across a narrow time span likely reduces this variability to some extent.

9 Initial bivariate models were constructed using a binary outcome of change from avoiding to not avoiding in comparison with everyone else, and the results were nearly identical to bivariate models predicting avoidance only at baseline.

10 Supplementary information available upon request.