1. Epidemiology and Global Health
  2. Microbiology and Infectious Disease
Download icon

Impact of the scale-up of piped water on urogenital schistosomiasis infection in rural South Africa

  1. Frank Tanser  Is a corresponding author
  2. Daniel K Azongo
  3. Alain Vandormael
  4. Till Bärnighausen
  5. Christopher Appleton
  1. Africa Health Research Institute, South Africa
  2. University of KwaZulu-Natal, South Africa
  3. University College London, United Kingdom
  4. Navrongo Health Research Centre, Ghana Health Service, Ghana
  5. Harvard T.H. Chan School of Public Health, United States
  6. University of Heidelberg, Germany
Research Article
Cite this article as: eLife 2018;7:e33065 doi: 10.7554/eLife.33065
8 figures, 4 tables and 3 additional files

Figures

Prevalence (95% CI) of Schistosoma haematobium infection by age and sex among children taking part in the parasitological survey (N = 2,105).

Darker colours represent heavy infections (≥50 eggs per 10 ml urine).

https://doi.org/10.7554/eLife.33065.002
Geographical variations in Schistosoma haematobium prevalence across the surveillance area obtained by a Gaussian kernel applied to participants’ precise household locations.

Approximate locations of participants’ households are shown (incorporating an intentional random error) with white dots representing an infected child. Superimposed on the map are the clusters of infection independently identified by the Kulldorff spatial scan statistic (cluster 5, low relative-risk; cluster 1–4, high relative-risk). The National Road can be seen running along the Eastern boundary of the surveillance area towards Mozambique.

https://doi.org/10.7554/eLife.33065.004
Time series of maps showing the coverage of piped water (%) between 2001 and 2007 in the study area (as measured using a Gaussian kernel approach) as well as mean piped water coverage over the full study period (bottom right).

Main roads are superimposed.

https://doi.org/10.7554/eLife.33065.006
Figure 4 with 1 supplement
Adjusted odds ratio of Schistosoma haematobium infection (95% CI) by piped water coverage in the surrounding local community - coverage quintile (Left) and continuous piped water coverage (Right).

The piped water coverage measure (2001–2007) is derived using a Gaussian kernel to calculate the proportion of all households in the unique local community surrounding each participant having access to piped water (Figure 3). Odds ratios are adjusted for age, sex, household assets, toilet in household, landcover class, distance to water body, altitude, slope, treatment in the last 12 months and school grade. Standard errors are adjusted for clustering by school and grade.

https://doi.org/10.7554/eLife.33065.007
Figure 4—figure supplement 1
Results of a parallel analysis using a Poisson regression.

The graph shows the adjusted prevalence ratio (95% CI) by piped water coverage in the surrounding local community. The piped water coverage measure (2001–2007) is derived using a Gaussian kernel to calculate the proportion of all households in the unique local community surrounding each participant having access to piped water (Figure 3). The resulting risk estimates are adjusted for age, sex, household assets, toilet in household, landcover class, distance to water body, altitude, slope, treatment in the last 12 months and school grade. Standard errors are adjusted for clustering by school and grade.

https://doi.org/10.7554/eLife.33065.008
Comparison of adjusted odds of Schistosoma haematobium infection (95% CI) in participants living in households with access to piped water (relative to participants without household-access to piped water) by sex.

The resulting risk estimates are adjusted for age, household assets, toilet in household, landcover class, distance to water body, altitude, slope, treatment in the last 12 months and school grade. Standard errors are adjusted for clustering by school and grade.

https://doi.org/10.7554/eLife.33065.011
Forest plot of Schistosoma haematobium infection according to household/individual level access to a safe water.

The data are taken from Grimes et al. systematic review (Grimes et al., 2014), based on 17 data-points (Abou-Zeid et al., 2012; Al-Waleedi et al., 2013; Awoke et al., 2013; Dame et al., 2006; Dawet, 2012; Farooq et al., 1966; Howarth et al., 1988; Knopp et al., 2013; Nworie et al., 2012; Reuben et al., 2013; Sady et al., 2013) to which we have added our study results. The sizes of the squares represent the weight given to each study, the rhombus is the effect-size with the black lines representing the 95% confidence intervals. The overall rhombus represents the combined effect-size, with the results of this study shown in red.

https://doi.org/10.7554/eLife.33065.012
Location of the study area in South Africa.
https://doi.org/10.7554/eLife.33065.013
Environmental control variables used in the statistical analysis with locations of 33 schools superimposed (yellow diamonds).

(Top left) Altitude in metres above sea level (MASL) (Top right) Distance to nearest water body (km) (Bottom left) Slope (degrees) (Bottom right) Satellite-derived landcover classification.

https://doi.org/10.7554/eLife.33065.014

Tables

Table 1
Characteristics of the primary school children within the survey who were linked to the population-based cohort (N=1976).
https://doi.org/10.7554/eLife.33065.003
CovariateTotalInfected (%)(95% CI)
Gender
 Female1016117 (11.5)(7.7–16.9)
 Male960217 (22.6)(17.5–28.7)
Age group
 911814 (11.9)(7.1–19.2)
 1036644 (12.0)(6.8–20.3)
 1159287 (14.7)(11.3–18.9)
 1246992 (19.6)(13.6–27.4)
 1322249 (22.1)(15.8–30.0)
 ≥1420948 (23.0)(15.1–33.3)
Community piped water (quintiles)*
 (Lowest) 1399102 (25.6)(13.0–44.1)
 241587 (21.0)(15.7–27.4)
 339765 (16.4)(10.5–24.6)
 438633 (8.5)(5.4–13.2)
 537947 (12.4)(8.9–16.9)
Household access to water
 No piped water30453 (17.4)(9.9–28.8)
 Piped water1672281 (16.8)(12.9–21.6)
Household assets quintiles
 (Poorest) 139883 (20.9)(13.3–31.1)
 238763 (16.3)(11.7–22.1)
 338963 (16.2)(11.7–22.0)
 436556 (15.3)(11.5–20.1)
 535957 (15.9)(10.4–23.6)
 Missing7812 (15.4)(9.5–23.9)
School grade
 Grade 51039186 (17.9)(12.2–25.5)
 Grade 6937148 (15.8)(10.4–23.2)
Praziquantel in the last 12 months
 No1933321 (16.6)(12.5–21.8)
 Yes4313 (30.2)(17.3–47.3)
Altitude (meters above sea level)
 <507629 (38.2)(19.3–61.4)
 50–100641168 (26.2)(19.2–34.7)
 100–150875108 (12.3)(9.5–16.0)
 150–20029622 (7.4)(4.4–12.4)
 >200887 (8.0)(3.5–17.2)
Distance to water body
 <1 km606112 (18.5)(14.2–23.7)
 1–2 km618122 (19.7)(14.0–27.0)
 2–3 km37668 (18.1)(10.8–28.7)
 >3 km37632 (8.5)(5.4–13.2)
Toilet in household
 No Toilet43864 (14.6)(9.9–21.1)
 Toilet1538270 (17.6)(13.3–22.8)
Land cover classification
 Closed Shrubland787184 (23.3)(17.6–30.3)
 Open Shrubland69683 (11.9)(8.8–15.9)
 Sparse Shrubland43752 (11.9)(8.7–16.0)
 Thickett5615 (26.8)(14.8–43.4)
Slope (quintiles)
 (Lowest) 139059 (15.1)(11.0–20.5)
 238672 (18.7)(12.9–26.3)
 340270 (17.4)(11.9–24.8)
 440470 (17.3)(12.4–23.7)
 539463 (16.0)(11.6–21.6)
  1. *Computes the proportion of households having access to piped water in the unique community surrounding each participant in the study (Figure 3). The Quintile (Q) ranges (min–max) are: Q1: 0–36; Q2: 37–59; Q3: 60–75; Q4: 76–92; Q5: 93–100.

Table 2
Significant spatial clusters (either unexpectedly high or low numbers) of Schistosoma haematobium infections identified by the Kulldorff spatial scan statistic (p<0.05) across the study area (see Figure 2)
https://doi.org/10.7554/eLife.33065.005
Cluster NumberRadius (km)Log-LikelihoodP-valuePrevalence (%)Relative Risk
10.9516.54<0.00184.65.63
22.6865.5<0.001504.66
31.1930.15<0.00149.24.53
41.9614.89 0.00241.92.99
52.9317.2<0.0010.760.05
Table 3
Logistic regression analysis of the risk factors of Schistosoma haematobium infection.

Model 0 gives the univariate results and Model 1 includes all variables in the model. In Model 2, piped water coverage in the immediate community surrounding each participant has been substituted with household-level piped water covariate.

https://doi.org/10.7554/eLife.33065.009
Model 0: UnivariateModel 1: Community coverageModel 2: Household access
CovariateaOR(95% CI)P-valueaOR(95% CI)P-valueaOR(95% CI)P-value
Community piped water quintiles (vs Lowest)†
 20.77(0.34, 1.75)0.5290.39‡(0.23, 0.66)<0.001
 30.57(0.22, 1.50)0.2500.30(0.15, 0.59)<0.001
 40.27(0.10, 0.71)0.0090.16(0.08, 0.33)<0.001
 50.41(0.17, 0.99)0.0480.12(0.06, 0.26)<0.001
Household access to water (vs No)
 Yes0.96(0.56, 1.64)0.8700.54(0.33, 0.89)0.017
Gender (vs Female)
 Male2.24(1.64, 3.08)<0.0012.62(1.92, 3.59)<0.0012.41(1.77, 3.28)<0.001
Age testing
 Per unit1.19(1.07, 1.31)0.0011.21(1.08, 1.36)0.0011.18(1.06, 1.31)0.002
Grade (vs Grade 5)
 Grade 60.86(0.45, 1.66)0.6480.76(0.51, 1.11)0.1490.77(0.47, 1.28)0.310
Praziquantel in last 12 months (vs No)
 Yes2.18(1.05, 4.52)0.0381.27(0.60, 2.71)0.5291.48(0.69, 3.16)0.307
Altitude Class (vs < 50)
 50–1000.58(0.27, 1.22)0.1470.47(0.23, 0.96)0.0390.50(0.23, 1.09)0.081
 100–1500.23(0.09, 0.60)0.0030.20(0.09, 0.43)<0.0010.20(0.08, 0.51)0.001
 150–2000.13(0.04, 0.39)<0.0010.09(0.03, 0.25)<0.0010.11(0.04, 0.33)<0.001
 >2000.14(0.04, 0.51)0.0040.08(0.03, 0.29)<0.0010.12(0.03, 0.44)0.002
Landcover class (vs Sparse Shrubland)
 Closed Shrubland2.26(1.55, 3.28)<0.0011.56(1.05, 2.31)0.0302.41(1.63, 3.58)<0.001
 Open Shrubland1.00(0.70, 1.44)0.9891.03(0.72, 1.47)0.8631.44(0.96, 2.17)0.079
 Thickett2.71(1.28, 5.75)0.0101.75(0.82, 3.73)0.1452.52(1.23, 5.18)0.012
Slope (square root)
 per unit0.98(0.83, 1.16)0.8181.02(0.89, 1.16)0.7940.91(0.79, 1.04)0.159
Distance to water body (vs < 1 km)
 1–2 km1.08(0.74, 1.58)0.6670.78(0.56, 1.08)0.1310.99(0.69, 1.41)0.946
 2–3 km0.97(0.54, 1.75)0.9280.72(0.47, 1.12)0.1431.04(0.62, 1.77)0.874
 >3 km0.41(0.23, 0.74)0.0030.25(0.12, 0.49)<0.0010.44(0.24, 0.79)0.007
Toilet in household (vs No)
 Yes1.24(0.90, 1.72)0.1861.24(0.87, 1.76)0.2291.20(0.84, 1.72)0.319
Household assets quintile (vs Poorest)
 20.74(0.50, 1.09)0.1230.88(0.60, 1.27)0.4800.87(0.61, 1.25)0.459
 30.73(0.49, 1.09)0.1270.78(0.51, 1.18)0.2350.75(0.48, 1.16)0.186
 40.69(0.42, 1.14)0.1430.81(0.47, 1.40)0.4500.67(0.40, 1.11)0.118
 50.72(0.41, 1.24)0.2280.80(0.48, 1.34)0.3890.62(0.36, 1.05)0.075
 Missing0.69(0.39, 1.21)0.1940.84(0.40, 1.80)0.6580.64(0.29, 1.40)0.258
  1. † Computes the proportion of households having access to piped-water in the unique community surrounding each participant in the study (Figure 3). The Quintile (Q) ranges (min–max) are: Q1: 0–36; Q2: 37–59; Q3: 60–75; Q4: 76–92; Q5: 93–100, ‡ Corresponding values for a model in which community-level piped-water coverage is used as a continuous variable: a 1% increase in the coverage of piped-water in the surrounding community, was independently associated with a 2.5% decrease in the odds of a Schistosoma haematobium infection (aHR=0.975; 95% CI: 0.966, 0.985; p-value<0.001).

Table 4
Multivariable model examining the socio-demographic predictors of Schistosoma haematobium infection stratified by gender.

Model 1 includes all variables in the model. In Model 2, piped water coverage in the immediate community surrounding each participant has been substituted with household-level piped water covariate.

https://doi.org/10.7554/eLife.33065.010
Model 1: Community-level coverage of piped waterModel 2: Household level access to piped water
FemalesMalesFemalesMales
CovariateaOR§ (95% CI)P-valueaOR§ (95% CI)P-valueaOR§ (95% CI)P-valueaOR§ (95% CI)P-value
Community piped water quintiles (vs Lowest)†
 20.24 (0.10–0.57)‡0.0020.56 (0.35–0.90)‡0.017
 30.21 (0.08–0.55)0.0020.37 (0.17–0.78)0.011
 40.16 (0.06–0.45)0.0010.16 (0.08–0.35)<0.001
 50.07 (0.02–0.20)<0.0010.17 (0.08–0.36)<0.001
Household access to water (vs No)
 Yes0.38 (0.19–0.75)0.0050.76 (0.41–1.42)0.379
Age at Testing
 Per unit9.62 (1.62–57.22)0.0141.24 (1.08–1.42)0.0036.85 (1.56–30.16)0.0111.20 (1.06–1.37)0.005
Age2
 Per unit0.92 (0.85–0.99)0.0240.93 (0.88–0.99)0.017
Toilet in household (vs No)
 Yes1.59 (0.88–2.88)0.1211.12 (0.70–1.86)0.6671.26 (0.75–2.14)0.3711.11 (0.66–1.86)0.687
Household Assets quintile (vs Poorest)
 20.46 (0.24–0.90)0.0261.27 (0.75–2.08)0.3440.55 (0.30–0.99)0.0451.24 (0.76–2.04)0.376
 30.35 (0.13–0.89)0.0321.21 (0.72–1.99)0.4530.35 (0.15–0.84)0.0191.19 (0.71–1.98)0.502
 40.42 (0.18–1.00)0.0611.18 (0.55–2.28)0.6460.40 (0.18–0.91)0.0290.94 (0.46–1.94)0.885
 50.66 (0.32–1.37)0.3070.90 (0.43–1.75)0.7690.54 (0.25–1.17)0.1150.71 (0.34–1.44)0.337
 Missing0.57 (0.20–1.66)0.2881.10 (0.38–2.94)0.8150.40 (0.12–1.35)0.1350.91 (0.34–2.42)0.845
  1. §All estimates simultaneously adjusted for landcover class, distance to water, altitude, slope, treatment in the last 12 months and school grade.

    † Computes the proportion of households having access to piped-water in the unique community surrounding each participant in the study (Figure 3). The Quintile (Q) ranges (min–max) are: Q1: 0–36; Q2: 37–59; Q3: 60–75; Q4: 76–92; Q5: 93–100

Additional files

Supplementary file 1

Shows the adjusted prevalence ratios (aPR) for the risk factors of Schistosoma haematobium infection, which are parallel to the results presented in Table 3.

Model 0 gives the univariate results and Model 1 includes all variables in the model. In Model 2, piped water coverage in the immediate community surrounding each participant has been substituted with the household-level piped water covariate.

https://doi.org/10.7554/eLife.33065.015
Supplementary file 2

Linear probability regression models showing the impact of piped water on schistosomiasis infection in primary school children across the study area.

Model 0 gives the univariate results and Model 1 gives the multivariate results for the availability of piped water in the community. Model 2 shows the instrumental variable estimation (IVE) results corresponding to Model 1, where the instrumental variable is the year that piped water was introduced into the community. Model 3 gives the multivariate results for piped water in the household and Model 4 shows the corresponding IVE results.

https://doi.org/10.7554/eLife.33065.016
Transparent reporting form
https://doi.org/10.7554/eLife.33065.017

Download links

A two-part list of links to download the article, or parts of the article, in various formats.

Downloads (link to download the article as PDF)

Download citations (links to download the citations from this article in formats compatible with various reference manager tools)

Open citations (links to open the citations from this article in various online reference manager services)