Adequate coverage of water, sanitation, and hygiene (WaSH) infrastructure in Central America has been reported to be low in comparison with overall Latin American averages (Uytewaal, 2016). Previous research, however, has reported on the positive significance of WaSH interventions for the health of communities in these regions (Fewtrell et al., 2005; Moll, McElroy, Sabogal, Corrales, & Gelting, 2007). A primary objective for WaSH infrastructure in developing countries is to create barriers to transmission of bacterial contaminants from one person or animal to another person. These transmission pathways previously have been summarized as the five Fs: fingers, fluids, floors, foods, and flies (Centers for Disease Control and Prevention [CDC], 2013; Wagner, Lanoix, & World Health Organization [WHO], 1958).
Due to the variety of pathogen species, the differing severities of exposure, repeated exposures, and the impact on intestinal integrity of children, understanding of the relationships between WaSH infrastructure barriers and health outcomes is limited (Waddington, Snilstveit, White, & Fewtrell, 2009). Additionally, effectiveness of WaSH infrastructure on improving health outcomes has been shown to be geographically dependent because moving from one community or region to another can alter coverage rates, environmental realities, or cultural interactions (Botting et al., 2010).
The U.S. Agency for International Development (USAID) consistently collects household WaSH infrastructure data that include water sources, water treatment techniques, types of sanitation facilities, presence of soap at hand washing stations, and floor type or animal pen infrastructure. Furthermore, USAID collects specific child health data including child stunting, child wasting, child body mass index, and diarrheal occurrences (U.S. Agency for International Development [USAID], n.d.). Child stunting (or wasting) is defined as a child with a height-for-age (or weight-for-height) ratio 2 standard deviations below the World Health Organization (WHO) growth mean and is often used as a chronic (or acute) health indicator (WHO, 2010). Presence of diarrhea often is used as an acute measure of health and is defined by WHO as three or more loose stools in 24 hr (WHO, 2019). With regional WaSH infrastructure data coupled with health data, it is possible to assess trends over geographies and health outcomes to help identify significant infrastructure-based interventions that are likely to have the best return on investment for improving child health.
To assess differences in significant correlations between WaSH infrastructure and child health outcomes over both geography and type of health outcome in the Western Highlands of Guatemala, we assessed two datasets from USAID. We built structural equation models and tested them for five geographic regions and three types of health outcomes. We discuss the implications of these findings for governmental and nongovernmental organizations at international and local levels.
Data and Location
We assessed data from the USAID 2012 Food for Peace Baseline Survey (ICF International, 2014) and USAID 2013 Western Highlands Integrative Program Baseline Survey (Taylor, 2014) for five departments (states) including Huehuetenango, San Marcos, Quiche, Totonicapan, and Quetzaltenango. We describe the data collection methods elsewhere--but briefly, a clustered randomized survey was administered verbally to mothers in their local dialect while anthropometric measurements of the children were taken following the WHO protocol. Data were de-identified and provided to researchers for analysis upon approval by USAID. Table 1 reports environmental statistics on each department including mean elevation, mean temperatures, and mean rainfall. All five departments are in a set of mountain ranges collectively known as the Western Highlands. Commonalities among the population included 1) farming as the primary livelihood and 2) the level of socioeconomic status with over 51% of the population living below the poverty line (Prado Cordova, Wunder, Smith-Hall, & Borner, 2013; USAID, 2012). A majority of the population self-identified as a specific Mayan ethnicity, including Ixil, Quiche, Mam, and Popti, with each using their own distinct language (ICF International, 2014).
Table 2 shows the variables we selected to be analyzed in the models along with the associated questions and scales we used. Diarrhea and ZHAZ (height-for-age z-score) were selected as acute and chronic measures of health, respectively, while the latent variable EED (environmental enteric dysfunction)-a combination of ZHAZ, ZBMI (body-mass index z-score), ZWHZ (weight-for-height Z-score), and diarrhea--was created to represent medium-term measures of health. Additionally, each WaSH infrastructure variable was linked with the five-F transmission pathway in which it provided a barrier (CDC, 2013; Julian, 2016; Pruss, Kay, Fewtrell, & Bartram, 2002; Wagner et al., 1958).
Improved water source and water treatment infrastructure were associated with barriers of transmission via the fluid and food pathways. Having soap for hand washing was associated with barriers for the finger and food transmission pathways. An improved sanitation facility was associated with barriers for transmission for the floor and fly pathways. Finally, both having an animal pen and an improved household floor were associated with barriers for the floor transmission pathway. The 2013 dataset did not collect information on animal pens; therefore, type of flooring was selected as the substitute for the 2013 models (Berkman, Lescano, Gilman, Lopez, & Black, 2002; Zambrano, Levy, Menezes, & Freeman, 2014).
Three structural equation models (SEMs) were built and tested for five geographic regions and each model included five WaSH infrastructure variables (WaterSource, WaterTreat, SanitType, HygSoap, and AnimalPen/ FloorType) regressed on by a health variable (diarrhea, EED, or ZHAZ). SEM is a statistical modeling technique that combines path analysis and factor analysis to analyze multiple hypotheses simultaneously. Figure 1 depicts the basic graphical representation of the SEM where arrows are hypotheses, rectangles are observable variables, and ovals are latent variables. A latent variable (shown here as EED) is hypothesized to be an underlying factor that influences a set of indicator variables (shown here as ZHAZ, ZBMI, ZWHZ, and diarrhea).
As this factor is estimated, path analysis is used to compute and analyze...