Perceptions of water insecurity from urban and peri-urban Haiti: A quantitative analysis

Safe drinking water access has continued to be a growing issue in Haiti. Water accessibility, availability, and quality can have severe implications on health and safety, with those in urban areas often having more access. Key differences relating to water accessibility can be seen between the urban and peri-urban areas of Haiti. One major objective of this research is to examine the disparities between the two areas and determine limiting and enabling factors that are contributing to the perceived access to clean water. A cross-cultural household water insecurity experiences (HWISE) survey (n = 499) was distributed to determine barriers and accessibility to sufficient water quality and quantity at the household level. This paper explores the relationship between water insecurity between two urban and peri-urban communes in Haiti using this data. Fisher’s Exact and Kruskal-Wallis tests were used to identify significant differences between strata, and logistic regression was used to determine significant associations with water security outcomes. Results indicated there were differences in both the costs and the sources of drinking and non-drinking water between urban and peri-urban Haiti. Certain demographic and behavioral characteristics were associated with increased water insecurity, including a household size greater than five and experiencing injury during collection.


Introduction
Haiti is a small country in the Caribbean with limited natural resources and is currently ranked as one of the poorest nations in the Western Hemisphere [1]. Additionally, it has the lowest improved water and sanitation coverage in the region, which directly impacts human health, environmental health, and economic stability [2][3]. Access to safe drinking water continues to be a growing issue in Haiti, more so after the damaging earthquake in 2010 and the constant hurricane damages that the region suffers. Water and sanitation conditions following the 7.0 magnitude earthquake in 2010 that struck Léogâne worsened largely due to the destruction of property, such as community and individual water sources. At the time of the earthquake, Haiti was unprepared for natural disasters due to the lack of adequate infrastructure and was behavior in Gressier and Léogâne, Haiti. The survey was distributed by randomly selecting households within a grid that overlaid Gressier and Léogâne. For houses that were empty or did not want to participate, a dice app was used to determine how many houses should be skipped before continuing. The surveys were conducted by a team of local Haitians that spoke fluent English, Creole, and French, and had training in enumeration and REDCap survey software. Tablets were used to collect survey data and GPS points using REDCap software, and a voice recorder app was used to record each conversation. Prior to each survey, the enumerator read from the oral consent form outlining the project and the voluntary status of each participant. The surveys were recorded for full transcription and validation of the information captured.

Data analyses
Descriptive statistics were used to quantify the detailed distribution of sociodemographic and water characteristics; categories were then combined into most-similar groupings a priori. Fisher's Exact tests identified significant differences in characteristic between Gressier and Léogâne; significant two-by-two and three-by-two tests both indicate that the proportions of a variable differ between locations. Numeric measures were generally found to be non-normally distributed, therefore the Kruskall-Wallis test was used to identify significant differences in numeric measures between location and stratified water sources. Logistic regression was conducted to determine odds significantly associated with water characteristics. Variables were selected forward stepwise from a full model of socio-demographic and behavioral characteristics using Akaike's Information Criterion (AIC), a measurement that reduces fitting error while penalizing models with too many parameters [15]. Missing data were multiply imputed using the R package "mice" [16]; all analyses were conducted in R version 3.5.1.

Normative income, education, safe work, and safe water
A component of the HWISE survey asks respondents to place themselves on a rung of a "ladder" corresponding to their standing in the community in realms of income, education, safe work, and safe water. This rung is translated to a rating ten through one, with ten indicating low self-perception. These data were dichotomized into "low rung" and "high rung" if a respondent indicated above or below the community's 50 th percentile (median), which is identified in Table 1.

Participants
In total, there were 499 surveys that were analyzed, 295 were from the peri-urban commune and 204 were from a very concentrated area of the urban commune. The urban site, Léogâne, had fewer participants despite being more populated than Gressier largely because researchers wanted to focus on the more densely populated areas of the city. The average age of participants was approximately 36 years old in both communes, with roughly two adults and three children per household. Most participants in both Léogâne and Gressier lived in a house, although there are differences in whether the house is owned versus rented (Table 1). Other dwellings included apartments, farms, squatter community, refugee/internally displaced camp, or other. Occupations ranged from trading, education, and food service to being supported by family members within Haiti or abroad. Individual occupations are distributed differently between Gressier and Léogâne. About 15% of respondents from both locations work in the non-governmental organization sector, and about 10% work in the healthcare sector. Finally, 50% of respondents from Gressier report receiving education about health, compared to 39% of respondents in Léogâne, and about 20% of respondents from both locations have received health education from work. A potentially relevant difference between the communes (11%) is that respondents in Gressier received information about health at significantly higher rates than those in Léogâne. At the time the survey was given, respondents in Léogâne paid more for water in the past month than those in Gressier. Additionally, there is a significant difference in the distribution of water sources between Gressier and Léogâne. As shown in Table 1, respondents from Gressier report higher proportions of drinking water sourced from a pipe, pump, or well and lower proportions from rain, surface water, or other methods.

Water use characteristics
The mean Household Water Insecurity Security Experience score in Gressier is approximately 18, and in Léogâne roughly 20 ( Table 1). The cost of water (using the local currency, Haitian Gourdes-HTG) in Gressier and Léogâne is distributed differently, with the maximum cost in Gressier being 7000 HTG, compared to 3200 HTG in Léogâne, and median cost being 0 HTG and 100 HTG. Respondents in both Gressier and Léogâne source their drinking and nondrinking water from a range of sources, though Gressier has more diversity among non-drinking sources than Léogâne.
There are significant differences in the distribution of drinking and non-drinking water sources between Gressier and Léogâne. Specifically, a greater proportion of respondents in Gressier source their drinking water from rain, surface water, or other sources. Further, an overwhelming majority of respondents in Léogâne source their non-drinking water from a pipe, pump, or well, compared to a lower proportion of those in Gressier. There is also a significant difference in the proportion of individuals who get their drinking and non-drinking water sources from the same location.
When comparing numeric measures across drinking and non-drinking water sources in both locations, there are significant differences in the cost of water for protected or unprotected wells and rain, surface water, or other sources. Léogâne respondents who source their non-drinking water from a protected or unprotected well pay a median of 150 HTG. This is compared to drinking and non-drinking residents in Gressier and drinking residents in Léogâne, who pay a median of 0 HTG for protected or unprotected well water (Table 2).
When examining differences in numeric measures between water sources in the same location, there are significant differences in the cost of water by source. The cost of water significantly varies by source in Gressier and Léogâne drinking water and Gressier non-drinking water. The greatest median cost for a drinking source was 450 HTG for water from a small vendor or other person in Léogâne. The greatest median cost for non-drinking water was 600 HTG for water from a small vendor or other person (Table 3).
In Gressier, buying water was associated with sourcing non-drinking water from a pipe (OR = 3.09), and older age (OR = 1.65), but those who sourced drinking water from a pipe were at lower odds of buying water (OR = 0.38). In Léogâne, buying water was associated with experiencing past-year water shortage (OR = 1.27) and longer water collection times (OR = 1.17). However, respondents who reported high normative safe work, that they were responsible for their household's water, and who worked were at lower odds of buying their water (OR = 0.35, 0.46, 0.30, respectively). In Gressier, factors associated with sourcing drinking water from a pipe in Gressier was sourcing non-drinking water from a pipe, greater household size, working (OR = 50.25, 2.17, 1.72, respectively). In Léogâne, factors associated with sourcing drinking water from a pipe include longer water collection time (OR = 4.32) and greater household size (OR = 3.12), but receiving health education, buying water, and rating a high normative safe work were associated with decreased odds of sourcing drinking water from a pipe (OR = 0.23, 0.26, 0.19, respectively).
When considering past-year water shortage in Gressier, larger household size, longer water collection, water storage in the home, and rating high normative safe water were associated factors (OR = 2.38, 2.20, 2.12, 2.01, respectively), with high normative income being associated with lower odds (OR = 0.48). In Léogâne, past-year water shortage was associated with buying water (OR = 3.11) and experiencing injury while collecting water (OR = 3.31), and a high normative income was associated with decreased odds of past-year water shortage (OR = 0.39).
Finally, in Gressier, an HWISE score greater than the 50th percentile was associated with high rated normative safe water, experiencing injury while collecting water, longer water

Discussion
This study reinforced how there are often disparities between neighboring communities despite similarities in culture and demographics. The peri-urban commune in this study, Gressier, had a majority of its inhabitants using a form of piped water (31.6%) as their primary source of drinking water, whereas in Léogâne, the urban commune, respondents largely used a bottle or sachet (44.7%). While piped water is more accessible than pumped water or buying bottled water from vendors, the quality of the water is unknown at this time despite being classified an improved water source by WHO and UNICEF [10]. Piped water has been known to

Household Size
House or yard pipe 93 (31.  cause major health impacts and often need maintenance and repair, especially in low-income countries such as Cambodia; however, it is uncertain whether community water sources, such as tube wells or boreholes are a safer alternative [17][18][19][20][21]. For piped water specifically, DINEPA (Direction Nationale de l'Eau Potable et de l'Assainissement) of the Ministry of Public Works serves as the entity in Haiti that regulates service providers as well as implements policy [22]. However, due to the unstructured network of piping in Gressier, it is unclear if it falls under DINEPA's regulations and testing. The results of this study showed that nearly a quarter (22.1%) of those living in Léogâne received their drinking water from a hand-pumped well. This could potentially be due to the large number of NGOs and international organizations that dug wells in Léogâne after the 2010 earthquake [23]. Additionally, those that were more likely to have a higher HWISE score also experience injury upon water collection. While we did not specifically ask how they were injured, injury during water collection is not uncommon across low and middle income countries. The long term effects of water collection can result in fatigue, soft tissue damage, musculoskeletal damage, as well as adverse effects on the skeletal system [24][25]. Those living in rural areas are often more likely to experience injury collecting water due to their higher potential for malnutrition and poor health [24][25].
Comparisons of measured socioeconomic characteristics showed broad similarities between urban and peri-urban participants, which helps minimize potential confounding relationships. However, one difference identified is that Gressier participants received general health education in significantly higher proportions than those in Léogâne. The association between water and health is a present concern to residents of Gressier and Léogâne [26] and researchers have conducted health education campaigns about the association in the past, albeit in different communities [27]. Therefore, it is possible that receiving health information changed participant water practices, though the absolute difference of 11% indicates that this may only have minimal effects.
The distributions for sources of drinking and non-drinking water were significantly different between both settings. This indicates that water quality concern, routes for intervention, and health education needs may not be comparable, despite being in the same department. Additionally, the percentage of participants who received drinking and non-drinking water from the same source differed between both settings, again indicating that water practices between the two differed in a way significant to public health practice. Within both settings, stratified analyses showed there were no differences of median HWISE score between water sources (Table 4). This indicates that median HWISE score did not differ between water sources when adjusting for settings, drinking use, and non-drinking use. When comparing drinking and non-drinking water sources across both settings there were no differences in median HWISE scores identified (Table 3). That is, when adjusting for water source, the participant's drinking use, non-drinking use, urban, or peri-urban exposures were not associated with differences in HWISE score.
These results may indicate that major factors of water insecurity lie in water supply system before consumers seek it, and in ways that are insensitive to small geographic variations within departments; for example, a regional drought or the lack of commercial water presence. In order to address observed water insecurity, resources may need to be dedicated to supporting more robust water supply systems, rather than intervening on individual water sources. Instead, addressing individual water sources could be the target of interventions to infectious disease and other environmental health factors.
The United Nations has proposed a series of goals that aim to lessen various global health problems, which they call the Sustainable Development Goals. The underlying focus of these goals is directly related to water and sanitation. The sixth goal on the list aims to ensure availability and sustainable management of water and sanitation worldwide [3]. Globally, more than 2 billion people live in countries that have excess water stress, which is defined as the freshwater withdrawn to total renewable freshwater resources above a threshold of 25 percent [3]. This indicates that there is a large probability that these countries will experience water scarcity in the future [3]. Clean water and proper sanitation are crucial components when trying to improve health conditions within a population. In depth understanding of contributing factors and effective strategies in place in Haiti will allow for progress made towards the achievement of the 2030 Goal 6 put in place by the United Nations.
One of the strengths of this study was the demographic similarities that were shared between participants from Léogâne and Gressier. Characteristics such as age and household size were nearly identical, which reduces the confounding effects of these characteristics. However, limitations included a small sample size between both communes and potential bias within the Léogâne data due to UF students being present during data collection. Subsequently, future studies should build upon this research to target specific areas of water hygiene and insecurity.
Supporting information S1