Abstract
Exposure to fecal contamination is a leading cause of childhood infectious diseases in low- and middle-income countries. Low-quality sanitation infrastructure and inadequate maintenance can make on-site solutions prone to spillage, exposing children to sewage. This paper uses a unique dataset with independent verification of sewage in and around the parcels of more than 20,000 houses with on-site sanitation in peri-urban Bolivia. We analyze the relationship between exposure to sewage from overflowed sanitation infrastructure and the incidence of diarrhea in children under age five. The presence of sewage is associated with a 4 percentage point increase in the probability of diarrhea incidence – a relative increase of 22%. This relationship is driven by sewage within the boundaries of the property where the child resides, which is associated with a relative increase of 30% in the probability of the incidence of diarrhea. Our spatial analysis of sewage density shows that the probability of the incidence of diarrhea increases with the concentration of sewage in the immediate vicinity of the child's residence, suggesting negative spillovers from neighbors with overflowed on-site sanitation facilities. These potential negative health externalities provide a persuasive argument in favor of public interventions that adequately remove and treat fecal sludge.
INTRODUCTION
Lack of access to hygienic sanitation facilities has been associated with increased prevalence of diarrhea, which is the second leading cause of death in children under the age of five worldwide, killing around 760,000 children per year (WHO 2013). Policy solutions to low sanitation coverage have stressed the construction of on-site facilities (i.e., to deal with excreta where it is deposited), with the aim of reducing exposure to fecal pathogens from open defecation. However, while epidemiological studies have identified a clear link between sanitation and the prevalence of diarrhea (White et al. 1972; Esrey & Habicht 1986; Fewtrell et al. 2005; Kremer 2007), recent studies of the impact of expanding on-site sanitation coverage have found modest or no effects on child diarrhea (e.g., Clasen et al. 2014; Patil et al. 2014).
A possible explanation for the lack of a health impact from increased sanitation coverage may be related to poorly constructed and maintained on-site facilities and inadequate fecal sludge management that result in sewage runoff (referred to as ‘blackwater’). For example, Berendes et al. (2017a) found no effect on diarrhea prevalence from the adoption of poorly constructed toilets that discharge to open drains. Besides appropriate construction for the effective underground drainage of effluent, on-site sanitation systems require routine removal of sludge to avoid spillage (Carter 2013; Peal et al. 2014). Therefore, even if on-site sanitation coverage is high, construction is appropriate, and sanitation facilities are used consistently, septic tanks and cesspools overflow and release sewage if households fail to properly maintain their systems (Reed 1994; Holden 2008; Peal et al. 2014). In fact, a recent systematic review found little to no effect from on-site sanitation coverage on the presence of feces and flies and on pathogenic transmission pathways such as food and soil (Sclar et al. 2016).
Poor maintenance of sanitation facilities may occur for a host of reasons, including resource constraints to fix or clean the system, inadequate information regarding the importance of hygiene and sanitation, and an insufficient supply of sludge removal services. Particularly in densely populated peri-urban neighborhoods, sewage runoff from one household can produce negative externalities by contaminating public spaces such as roads and sidewalks and contributing to the spread of infectious diseases in neighboring children. Therefore, households do not assume the full costs and benefits of maintaining their sanitation systems, further reducing the incentives for proper sanitation maintenance.
The aim of this study is to estimate the association between sewage from overflowed on-site sanitation systems and the occurrence of diarrhea among children under five, and to document potential negative health externalities. To address this question, we analyze data from over 20,000 households in a low-income, peri-urban area of Santa Cruz de la Sierra, Bolivia. As with many such areas throughout the country, households in the study area have no access to piped sewerage, but near universal access to piped water and on-site sanitation. As such, the primary source of sewage contamination above ground is from septic tanks and cesspools that, when not properly maintained, leak runoff in the form of blackwater onto properties, streets, and sidewalks.
METHODS
Data
This study uses the Baseline Survey of Sanitation Systems in Peri-Urban Areas conducted by the Ministry of Water and Environment in the ‘Plan 3000’ area of Santa Cruz de la Sierra, Bolivia from October to December 2013. Information was collected from all households (census) in 26 neighborhoods, covering a total of 20,637 households, out of which, 7,158 had children under five years of age. As a result, health outcome information on diarrhea incidence was collected for 9,008 children. A unique feature of the survey was an independent visual inspection by the interviewer of the presence of sewage runoff inside the dwelling and in the immediate surrounding area. The survey also recorded geographic coordinates for each dwelling, household socio-demographic characteristics, the education and labor status of household members, and dwelling characteristics. The target respondent was the female head of household.
The health outcome of interest is diarrhea incidence, measured as a binary variable equal to 1 if the caretaker reported that a child in the household had diarrhea during the 2 weeks preceding the interview. The diarrhea prevalence of the study population is 0.18 (standard deviation of 0.38, as shown in Table 1). The explanatory variable of interest is exposure to sewage, which is obtained from the interviewer's independent visual inspection of the presence of blackwater in and around the property. Reporting of the visual inspection was standardized across interviewers as part of a module completed by the interviewer immediately following the survey based on observations conducted in and around the property. We define indoor sewage as the observation of the presence of blackwater in the property or within the dwelling, which was recorded as the interviewer walked through to inspect the toilet facilities typically located at the back of the property. We define outdoor sewage as runoff or pooling of blackwater observed on the sidewalk or half of the street closest to the property's front entry, anywhere between the start and endpoints of the parcel limits. Observation to determine if there was sewage outdoors was done after completing the survey in each household.
Descriptive statistics
Variable . | Mean . | Standard deviation . | Minimum . | Maximum . | Observations . |
---|---|---|---|---|---|
All households | |||||
Access to piped water | 0.998 | 0.040 | 0 | 1 | 20,447 |
Access to sanitation facility | 0.992 | 0.088 | 0 | 1 | 20,509 |
Toilet location | |||||
Inside dwelling | 0.161 | 0.368 | 0 | 1 | 20,350 |
Outside dwelling, but within the property | 0.839 | 0.368 | 0 | 1 | 20,350 |
Discharge | |||||
Sewerage | 0.005 | 0.074 | 0 | 1 | 20,350 |
Septic tank | 0.369 | 0.482 | 0 | 1 | 20,350 |
Pit latrine to cesspools | 0.584 | 0.493 | 0 | 1 | 20,350 |
Open ditch | 0.032 | 0.177 | 0 | 1 | 20,350 |
Other | 0.009 | 0.095 | 0 | 1 | 20,350 |
Cleaning frequency of septic tank/cesspool | |||||
Never | 0.583 | 0.493 | 0 | 1 | 18,642 |
Once per year | 0.209 | 0.407 | 0 | 1 | 18,642 |
Every 2 years | 0.081 | 0.273 | 0 | 1 | 18,642 |
Every 3 years | 0.108 | 0.310 | 0 | 1 | 18,642 |
More than every 3 years | 0.019 | 0.136 | 0 | 1 | 18,642 |
Households with children under five years old and reported indoor and/or outdoor sewage | |||||
Sewage exposure | |||||
Indoor and outdoor sewage | 0.134 | 0.340 | 0 | 1 | 6,387 |
Indoor sewage | 0.049 | 0.216 | 0 | 1 | 6,387 |
Outdoor sewage | 0.319 | 0.466 | 0 | 1 | 6,387 |
Children | |||||
Prevalence of diarrhea | 0.181 | 0.385 | 0 | 1 | 8,171 |
Age in months | 29.11 | 16.913 | 0 | 60 | 8,171 |
Female | 0.482 | 0.500 | 0 | 1 | 8,171 |
Variable . | Mean . | Standard deviation . | Minimum . | Maximum . | Observations . |
---|---|---|---|---|---|
All households | |||||
Access to piped water | 0.998 | 0.040 | 0 | 1 | 20,447 |
Access to sanitation facility | 0.992 | 0.088 | 0 | 1 | 20,509 |
Toilet location | |||||
Inside dwelling | 0.161 | 0.368 | 0 | 1 | 20,350 |
Outside dwelling, but within the property | 0.839 | 0.368 | 0 | 1 | 20,350 |
Discharge | |||||
Sewerage | 0.005 | 0.074 | 0 | 1 | 20,350 |
Septic tank | 0.369 | 0.482 | 0 | 1 | 20,350 |
Pit latrine to cesspools | 0.584 | 0.493 | 0 | 1 | 20,350 |
Open ditch | 0.032 | 0.177 | 0 | 1 | 20,350 |
Other | 0.009 | 0.095 | 0 | 1 | 20,350 |
Cleaning frequency of septic tank/cesspool | |||||
Never | 0.583 | 0.493 | 0 | 1 | 18,642 |
Once per year | 0.209 | 0.407 | 0 | 1 | 18,642 |
Every 2 years | 0.081 | 0.273 | 0 | 1 | 18,642 |
Every 3 years | 0.108 | 0.310 | 0 | 1 | 18,642 |
More than every 3 years | 0.019 | 0.136 | 0 | 1 | 18,642 |
Households with children under five years old and reported indoor and/or outdoor sewage | |||||
Sewage exposure | |||||
Indoor and outdoor sewage | 0.134 | 0.340 | 0 | 1 | 6,387 |
Indoor sewage | 0.049 | 0.216 | 0 | 1 | 6,387 |
Outdoor sewage | 0.319 | 0.466 | 0 | 1 | 6,387 |
Children | |||||
Prevalence of diarrhea | 0.181 | 0.385 | 0 | 1 | 8,171 |
Age in months | 29.11 | 16.913 | 0 | 60 | 8,171 |
Female | 0.482 | 0.500 | 0 | 1 | 8,171 |
Source: Authors' calculations based on Plan 3000 baseline survey.
While outdoor inspection was conducted in a public space and as such is reported for the full sample, the indoor visual inspection was conditional on the household's willingness to let the interviewer onto the property and/or in the house, raising the potential for nonrandom item nonresponse. The nonresponse rate for indoor sewage for the subsample of households with children below five years of age was 9%. However, observed characteristics including outdoor sewage and household demographic characteristics are balanced across respondents and nonrespondents (see Appendix A, available with the online version of this paper), alleviating concerns of nonresponse bias. Our final sample includes 6,387 households where indoor and outdoor sewage was reported, corresponding to 8,171 children.
Table 1 presents descriptive statistics of the sanitation infrastructure: 99% of households have piped water and on-site sanitation facilities, and 16% of those facilities are inside the dwelling and 84% outside but within the property. Sewage is discharged primarily to septic tanks (37%) and cesspools (58%), but 58% of these households reported never cleaning their septic tanks or cesspools. Only 1% of households have no facility and 3% of households discharge effluent into an open ditch. As such, the observed sewage inside and/or outside half of all households is not mainly a result of open discharge, but rather of overflowed on-site facilities that leak blackwater. Variation in the presence of indoor and outdoor sewage results in four binary variables that indicate: (1) indoor and outdoor sewage (13%); (2) indoor sewage only (5%); (3) outdoor sewage only (32%); and (4) no indoor or outdoor sewage (50%). (See Appendix B for a cross-tabulation of sewage exposure and types of sanitation discharge, available with the online version of this paper.) The correlation between indoor and outdoor sewage is just 15%, suggesting a fairly high level of independence between these two variables.
Analysis











Even if a household has strong preferences for maintaining a sanitary environment, exposure to fecal matter in public spaces such as streets and sidewalks depends on the behavior of neighbors with regard to the upkeep of their own on-site sanitation systems. We exploit the rich spatial variation in the concentration of outdoor sewage to estimate the association between density of neighboring houses with outdoor sewage (henceforth, sewage density) and the incidence of diarrhea in children under five. Figure 1 shows a map of the 26 neighborhoods where the survey was conducted. Each point on the map represents a dwelling's geographic coordinate. Shaded areas represent households where outdoor sewage from overflowed on-site sanitation facilities was identified, while lightly shaded areas represent households without outdoor sewage.
Plan 3000 map of the location of houses with outdoor sewage. The map covers the 14 square kilometers where the 26 neighborhoods included in the survey are located. Houses with outdoor sewage are denoted with a black circle and those without outdoor sewage with a white circle. Source: Authors' calculations using ArcGIS™ software with data from the Plan 3000 baseline survey.
Plan 3000 map of the location of houses with outdoor sewage. The map covers the 14 square kilometers where the 26 neighborhoods included in the survey are located. Houses with outdoor sewage are denoted with a black circle and those without outdoor sewage with a white circle. Source: Authors' calculations using ArcGIS™ software with data from the Plan 3000 baseline survey.
We measure sewage density as the number of neighboring houses with outdoor sewage within buffer circular arcs with a radius of 100 meters around each household (constructed with ArcGIS™ software). We then estimate model (1) with sewage density as the explanatory variable of interest, instead of the binary variables of sewage exposure. We additionally control for population density, which is constructed in the same way as sewage density.
RESULTS AND DISCUSSION
Household sewage exposure and child diarrhea
Child exposure to contaminated environments is likely to vary with age, as children change their behavioral patterns (i.e., crawl, walk with assistance, walk alone, play in the ground, etc.), increase their awareness of risks (i.e., putting dirty fingers inside the mouth), and are more able to make choices to protect their health as their immune system strengthens (WHO 2016). Child diarrheal incidence is expected to follow a similar pattern (Kattula et al. 2015).
Figure 2 depicts the evolution of the prevalence of diarrhea by age in months for children exposed to sewage (indoor and/or outdoor) and not exposed to sewage. For both groups we observe that the prevalence of diarrhea increases with age during the first 12 months, remains more or less stable through 24 months, and then declines steadily through 60 months. Furthermore, children exposed to sewage have a systematically higher diarrhea prevalence than children of the same age not exposed to sewage over much of the age distribution, particularly in the 10–60 month range. This observation is consistent with child development patterns that can lead to increased contact with contaminated environments, for example, as children become more mobile, typically crawling around 13 months and walking without assistance by 17 months (WHO 2006).
Sewage exposure and prevalence of diarrhea, by age in months. The x axis measures children's age in months and the y axis measures the mean of diarrhea incidence for all children within each age. ‘Sewage’ indicates households with any sewage exposure (indoors and/or outdoors) and ‘No sewage’ the opposite. The dark line represents children from households exposed to sewage and the dotted line represents children not exposed to sewage. Source: Authors' calculations based on Plan 3000 baseline survey.
Sewage exposure and prevalence of diarrhea, by age in months. The x axis measures children's age in months and the y axis measures the mean of diarrhea incidence for all children within each age. ‘Sewage’ indicates households with any sewage exposure (indoors and/or outdoors) and ‘No sewage’ the opposite. The dark line represents children from households exposed to sewage and the dotted line represents children not exposed to sewage. Source: Authors' calculations based on Plan 3000 baseline survey.
We estimate the association between exposure to sewage and the incidence of child diarrhea using model (1) and maximum-likelihood estimation. Table 2 shows the marginal effect at the mean of the observable covariates included in model (1). Table 2, column 1 shows that, for the whole sample of children, indoor and outdoor sewage exposure is significantly associated with an increased probability of diarrhea incidence in children under five. On average, any exposure to sewage is associated with a 4 percentage point increase in the probability of diarrhea incidence, a 22% increase relative to children not exposed to sewage. The presence of only indoor sewage is significantly associated with a 5.4 percentage point increase in the probability of diarrhea incidence, a relative increase of 30% compared to children in households with no sewage exposure. Although the coefficient on indoor sewage is larger than indoor and outdoor sewage, the difference is not statistically significant. For exposure to outdoor sewage, the estimate is small and insignificant. These results suggest that sewage from improperly maintained on-site sanitation facilities is a likely contributor to the prevalence of diarrhea, and that the primary vector is from exposure to sources inside the dwelling.
Association between child diarrhea incidence among children under five and exposure to sewage
. | . | (2) . | (3) . | (4) . | (5) . | (6) . |
---|---|---|---|---|---|---|
(1) Full sample . | Age in months: 0–12 . | Age in months: 12–23 . | Age in months: 24–36 . | Age in months: 37–48 . | Age in months: 49–60 . | |
Indoor and outdoor sewage exposure | 0.040*** | 0.019 | 0.043 | 0.008 | 0.084*** | 0.017 |
(0.014) | (0.025) | (0.033) | (0.030) | (0.022) | (0.000) | |
Indoor sewage exposure | 0.054*** | 0.046 | 0.136*** | −0.001 | 0.061* | 0.012 |
(0.020) | (0.036) | (0.050) | (0.046) | (0.035) | (0.000) | |
Outdoor sewage exposure | 0.010 | −0.012 | 0.020 | 0.022 | 0.010 | 0.002 |
(0.011) | (0.021) | (0.026) | (0.022) | (0.020) | (0.000) | |
Age in months | −0.002*** | 0.019*** | −0.001 | −0.002 | −0.001 | −0.000 |
(0.000) | (0.002) | (0.003) | (0.003) | (0.002) | (0.000) | |
Female | −0.006 | −0.009 | −0.009 | 0.032 | −0.030 | −0.015 |
(0.012) | (0.027) | (0.035) | (0.030) | (0.025) | (0.000) | |
Access to piped water | −0.070* | −0.028 | −0.149 | −0.061 | −0.084 | −0.049 |
(0.042) | (0.071) | (0.101) | (0.082) | (0.058) | (0.000) | |
Domestic animals (observation) | 0.003 | 0.015 | −0.016 | 0.010 | 0.009 | −0.011 |
(0.010) | (0.019) | (0.024) | (0.021) | (0.017) | (0.000) | |
Household covariates | Yes | Yes | Yes | Yes | Yes | Yes |
p value b1 = b2 | 0.515 | 0.511 | 0.088 | 0.855 | 0.535 | 0.873 |
p value b1 = b3 | 0.042 | 0.258 | 0.526 | 0.660 | 0.002 | 0.512 |
Prevalence of diarrhea when not exposed to sewage | 0.181 | 0.195 | 0.259 | 0.195 | 0.137 | 0.107 |
Observations | 8,148 | 1,779 | 1,636 | 1,690 | 1,674 | 1,334 |
Households | 6,387 | 1,760 | 1,617 | 1,660 | 1,653 | 1,333 |
. | . | (2) . | (3) . | (4) . | (5) . | (6) . |
---|---|---|---|---|---|---|
(1) Full sample . | Age in months: 0–12 . | Age in months: 12–23 . | Age in months: 24–36 . | Age in months: 37–48 . | Age in months: 49–60 . | |
Indoor and outdoor sewage exposure | 0.040*** | 0.019 | 0.043 | 0.008 | 0.084*** | 0.017 |
(0.014) | (0.025) | (0.033) | (0.030) | (0.022) | (0.000) | |
Indoor sewage exposure | 0.054*** | 0.046 | 0.136*** | −0.001 | 0.061* | 0.012 |
(0.020) | (0.036) | (0.050) | (0.046) | (0.035) | (0.000) | |
Outdoor sewage exposure | 0.010 | −0.012 | 0.020 | 0.022 | 0.010 | 0.002 |
(0.011) | (0.021) | (0.026) | (0.022) | (0.020) | (0.000) | |
Age in months | −0.002*** | 0.019*** | −0.001 | −0.002 | −0.001 | −0.000 |
(0.000) | (0.002) | (0.003) | (0.003) | (0.002) | (0.000) | |
Female | −0.006 | −0.009 | −0.009 | 0.032 | −0.030 | −0.015 |
(0.012) | (0.027) | (0.035) | (0.030) | (0.025) | (0.000) | |
Access to piped water | −0.070* | −0.028 | −0.149 | −0.061 | −0.084 | −0.049 |
(0.042) | (0.071) | (0.101) | (0.082) | (0.058) | (0.000) | |
Domestic animals (observation) | 0.003 | 0.015 | −0.016 | 0.010 | 0.009 | −0.011 |
(0.010) | (0.019) | (0.024) | (0.021) | (0.017) | (0.000) | |
Household covariates | Yes | Yes | Yes | Yes | Yes | Yes |
p value b1 = b2 | 0.515 | 0.511 | 0.088 | 0.855 | 0.535 | 0.873 |
p value b1 = b3 | 0.042 | 0.258 | 0.526 | 0.660 | 0.002 | 0.512 |
Prevalence of diarrhea when not exposed to sewage | 0.181 | 0.195 | 0.259 | 0.195 | 0.137 | 0.107 |
Observations | 8,148 | 1,779 | 1,636 | 1,690 | 1,674 | 1,334 |
Households | 6,387 | 1,760 | 1,617 | 1,660 | 1,653 | 1,333 |
Note: The outcome variable is diarrhea, measured as a binary variable that indicates whether a child had diarrhea during the 2 weeks preceding the interview. The three explanatory variables of interest are measured as binary variables that indicate whether a household has each type of sewage exposure. All models include neighborhood fixed effects and control for the following household covariates: number of males and females by age group, household head's gender, age, marital status, educational attainment, and employment status, dwelling type and ownership, asset ownership, and an indicator of whether the household can rely on somebody for emergencies and on somebody to lend money to improve the dwelling. Clustered standard errors at the household level are reported in parentheses. Sample size varies from the original sample size due to missing values in household covariates.
*Statistically different from 0 at the 10% level.
**Statistically different from 0 at the 5% level (none found).
***Statistically different from 0 at the 1% level.
Source: Authors' calculations based on Plan 3000 baseline survey restricted to households with children under five years old and with verification of indoor sewage.
Next, we disaggregate estimates by age, following the hypothesis that children will interact with sewage differently depending on their stage of development. Children younger than 12 months old who have more limited ability to move independently are expected to be less prone to direct contact with sewage, but more susceptible to environmental conditions given their developing immune systems. Children aged 12–23 months have increasing capacity for independent mobility around the home and land parcel, but are less prone to follow instructions or understand the risk of interacting with and ingesting soil or objects contaminated with fecal matter. Finally, children older than 48 months represent the age group with the most developed cognitive abilities, and are better able to self-regulate their interaction with sewage.
Results in Table 2 (columns 2–6) indicate that exposure to indoor and outdoor sewage is most strongly associated with increased probability of diarrhea incidence in the 37–48-month-old group. In this group, exposure to sewage increases the probability of diarrhea incidence by 8.4 percentage points, an increase of 61% relative to children of the same age group who are not exposed to sewage. Similarly, we find that indoor sewage is significantly associated with increased probability of diarrhea incidence in the 12–23-month-old group. Among this group, exposure only to indoor sewage is associated with a 13.6 percentage point increase in the probability of diarrhea incidence, an increase of 52% relative to children of the same age group who are not exposed to sewage. No significant effects of sewage are found in the 24–36 or 49–60-month-old groups. We speculate that the latter result may be attributed to children's improved ability for self-regulation as they grow older, although the absence of a significant relationship in the 24–36-month-old range remains a puzzle.
As robustness checks to our main specification in model (1) we estimate the relationship between sewage and diarrhea using a linear probability model (LPM), a mixed model, and matching methods. We also test sensitivity to alternative sets of covariates in model (1). Results discussed in Appendix D confirm that the estimates of Table 2 are robust to these alternative specifications. Furthermore, although the presence of sewage may be seasonal and linked to the incidence of flooding, we found no association between the amount of precipitation and the extent of sewage observed on a given date during the span of the survey (see Appendix E). (Appendices D and E are available with the online version of this paper.)
Externalities
We next turn to the analysis of externalities using sewage density as the explanatory variable of interest. Figure 3 plots the marginal association of child diarrhea with sewage density conditional on observed covariates included in regression model (1) and neighborhood fixed effects. The plot provides the estimated association for each additional neighboring house with outdoor sewage. The results are consistent with the presence of negative health externalities from outdoor sewage, as the likelihood of the occurrence of diarrhea in children under five increases with the number of neighbors with outdoor sewage.
Marginal association between child diarrhea and sewage density. This figure shows the marginal association between each level of sewage density and the probability of diarrhea incidence. Standard errors are clustered at the household level and confidence intervals are shown by the dashed line. The prevalence of diarrhea in children under five years old increases with the number of neighbors with outdoor sewage. Source: Authors' calculations based on Plan 3000 baseline survey restricted to households with children under five years old. The full sample of the Plan 3000 baseline survey is used for computing sewage density.
Marginal association between child diarrhea and sewage density. This figure shows the marginal association between each level of sewage density and the probability of diarrhea incidence. Standard errors are clustered at the household level and confidence intervals are shown by the dashed line. The prevalence of diarrhea in children under five years old increases with the number of neighbors with outdoor sewage. Source: Authors' calculations based on Plan 3000 baseline survey restricted to households with children under five years old. The full sample of the Plan 3000 baseline survey is used for computing sewage density.
An increase of 20 neighboring houses with outdoor sewage is associated with an increase in the probability of diarrhea incidence of about 3.8%. Compared to houses in areas with a low density of sewage (ten or less neighboring houses with outdoor sewage), living in a high-sewage-concentration area (70 or more neighboring houses with outdoor sewage) is associated with an approximate 14% increase in the probability of the incidence of diarrhea.
CONCLUSIONS
Although epidemiological studies have identified a clear link between exposure to fecal pathogens and the incidence of diarrhea, existing studies on the impact of on-site sanitation construction have found, at best, modest effects on the prevalence of diarrhea in children under the age of five. Less attention has been given to the role of sewage released into the environment from poorly constructed and maintained on-site sanitation facilities and improper fecal sludge management, which could attenuate the potential health gains from investments in on-site sanitation infrastructure.
The aim of this observational study has been to explore the association between sewage from overflowed on-site sanitation systems and the prevalence of diarrhea among children under five and to document the potential negative health externalities. Based on the independently verified presence of sewage within and just outside the properties of more than 6,000 households with children in a low-income peri-urban area of Santa Cruz de la Sierra, Bolivia, this study has found significant increases in the likelihood of the prevalence of diarrhea in children under five associated with the presence of sewage from on-site sanitation facilities. The study has also found evidence of negative externalities emerging from neighboring households with outdoor sewage. The estimates are within the range of other studies of the association of in-house exposure to fecal contamination and diarrhea incidence, such as Bartlett et al. (1992) (2.28, 95% confidence interval 1.19–4.39) and Traore et al. (1994) (1.38, 95% confidence interval 0.98–1.95), and enteric infections, such as Berendes et al. (2017b) (3.78, no confidence interval reported).
This study contributes to further understanding the absence of detectable health effects of on-site sanitation construction (e.g., Clasen et al. 2014; Patil et al. 2014) and the importance of externalities from sanitation behavior (Geruso & Spears 2015; Hathi et al. 2017; Jung et al. 2017). While much of the literature focuses on the adverse health effects of open defecation and the absence of sanitation facilities, our results highlight a potentially important but largely overlooked source of fecal contamination, that is, sewage overflow into the environment from inadequate construction, sanitation maintenance, and fecal sludge management. Furthermore, while much of the epidemiological literature includes more precisely measured indicators from stool, water, and soil samples tested for microbial presence (Yajima & Koottatep 2010; Pickering et al. 2012; Berendes et al. 2017a, 2017b), an important contribution of our study is the use of a large, high-powered sample and direct observations of sewage contamination to analyze self-reported diarrhea outcomes.
This study has notable strengths and limitations in measurement and analysis. We use a rich dataset on a sample of more than 20,000 households. Independently verifying sewage within and just outside the parcel corresponding to a household provides a unique assessment of the relationship between sewage exposure and diarrhea incidence. Rather than capturing sewage resulting from improper construction or open defecation (99% of households have a sanitation facility, and only 3% of households release fecal sludge into open ditches), most of the sewage observed in this context is the result of overflowed on-site sanitation facilities. We thus attribute our results to sewage leakage resulting from poor maintenance and inadequate fecal sludge management. Furthermore, rich spatial variation in the geographic concentration of outdoor sewage provided an ideal dataset to test for the presence of negative health externalities from sewage exposure in neighboring properties. A limitation of our study is that we do not observe directly the types of septic tanks and cesspools in place and the availability and functionality of sludge removal solutions and drains, the mechanisms that we presume link on-site sanitation with sewage runoff.
Measurement of diarrhea incidence relies on the response from the household caretaker, which is prone to misreporting, and the use of a 2-week recall period may underestimate diarrhea incidence (Zafar et al. 2010; Arnold et al. 2013). This misreporting, however, is unlikely to be correlated with our independently observed measure of sewage. Another potential concern is seasonal bias in diarrhea incidence (Luby et al. 2011). The survey used for this analysis was conducted over a period of less than three months, canvassing entire neighborhoods within a few days. As such, the neighborhood fixed effects should account for any differences in the timing of the interviews across neighborhoods. Lastly, because self-reported diarrhea reflects mixed etiologic agents that include viruses, intestinal parasites, and bacteria linked to enteric infections (Berendes et al. 2017a), more research is needed to better characterize the association with different groups of pathogens and further understand the implications for malnutrition, growth, and cognitive development (Humphrey 2009; Ngure et al. 2014).
Our estimates of the association between diarrhea incidence and exposure to sewage are robust to alternative specifications, including linear probability models, mixed models, matching methods, and alternative sets of covariates, lending additional credibility to our results. The primary risk to a causal interpretation of this relationship is the presence of unobserved confounders across households within a given neighborhood, including household-level preferences and health-related behaviors. Nevertheless, even when controlling for proxies of sanitary preferences (the presence of animals, water treatment, and trash in the yard), the results are qualitatively similar. An additional risk is linked to our externality analysis, since sewage density is likely not random, but rather affected by residential choices that also affect diarrhea prevalence. Yet, we argue that identifying the association between different levels of exposure to sewage and child diarrhea is an important first step to document the health effects and potential externalities from overflowed on-site sanitation facilities.
The evidence suggesting that sewage exposure is negatively associated with child health and the documentation of potential negative externalities across neighboring houses provides a strong argument in favor of public expenditure to maintain sanitation facilities and improve fecal sludge management. Because on-site sanitation facilities may be prone to spillage, our findings highlight the need for policy alternatives that prevent fecal contamination and ensure that children are fully protected in their neighborhood. Further research should explore whether sewer lines connected to treatment plants that incorporate proper technologies for removing and treating sludge and do not rely on routine maintenance by households are a cost-effective alternative, particularly in more densely populated urban areas where the marginal improvements in child health may well outweigh the costs of large infrastructure investments.
ACKNOWLEDGEMENTS
The authors thank Stephen Jenkins for numerous helpful discussions and suggestions and Gastón Gertner, Raúl Burgos, Teresa Reinaga, Vanessa Carrera, Maria Rodrigues, Gabriela Smarrelli, and Charis Arvanitis for helpful assistance. Data used for the analysis are from the baseline survey of the Bolivian Water and Sewerage Program (Phase I and Phase II) in Plan 3000 in Santa Cruz de la Sierra, collected by the Bolivian Ministry of Water and Environment with support from the Inter-American Development Bank and the Spanish Agency for International Development Cooperation (AECID). All opinions in this paper are those of the authors and do not necessarily represent the views of Bolivia's Ministry of Water and Environment, AECID, or the Inter-American Development Bank, its Executive Directors or the governments they represent.