International Journal of Hygiene and Environmental Health

International Journal of Hygiene and Environmental Health 220 (2017) 513–530
Contents lists available at ScienceDirect
International Journal of Hygiene and

Environmental Health
journal homepage: www.elsevier.com/locate/ijheh
Water quality, compliance, and health outcomes among utilities

implementing Water Safety Plans in France and Spain
Karen E. Setty a,∗ , Georgia L. Kayser a,b , Michael Bowling c , Jerome Enault d ,
Jean-Francois Loret d , Claudia Puigdomenech Serra e , Jordi Martin Alonso f ,
Arnau Pla Mateu e , Jamie Bartram a
a
The University of North Carolina, Chapel Hill, Department of Environmental Sciences and Engineering, 170 Rosenau Hall, CB #7400, Chapel Hill, NC, USA
b
The University of California, San Diego, Department of Family Medicine and Public Health, 9500 Gilman Dr, 0628, La Jolla, CA 92093-0628, USA
c
The University of North Carolina, Chapel Hill, Department of Health Behavior, 302 Rosenau Hall, CB #7440, Chapel Hill, NC, USA
d
Suez, Centre International de Recherche sur l’Eau et l’Environnement (CIRSEE), 38 rue du President Wilson, 78230 Le Pecq, France
e
Suez, Centre Tecnològic de l’Aigua (CETAQUA), Carretera d’Esplugues, 75, 08940 Cornellà de Llobregat, Barcelona, Spain
f
Aigües de Barcelona, Carrer General Batet 1-7, 08028, Barcelona, Spain
a r t i c l e i n f o a b s t r a c t
Article history: Water Safety Plans (WSPs), recommended by the World Health Organization since 2004, seek to proac-
Received 15 November 2016 tively identify potential risks to drinking water supplies and implement preventive barriers that improve
Received in revised form 31 January 2017 safety. To evaluate the outcomes of WSP application in large drinking water systems in France and Spain,
Accepted 19 February 2017
we undertook analysis of water quality and compliance indicators between 2003 and 2015, in conjunc-
tion with an observational retrospective cohort study of acute gastroenteritis incidence, before and after
Chemical compound studied in this article:
WSPs were implemented at five locations. Measured water quality indicators included bacteria (E. coli,
Aluminum (PubChem CID: 5359268)
fecal streptococci, total coliform, heterotrophic plate count), disinfectants (residual free and total chlo-
Bromate (PubChem CID: 84979)
Chlorine (PubChem CID: 24526)
rine), disinfection by-products (trihalomethanes, bromate), aluminum, pH, turbidity, and total organic
carbon, comprising about 240 K manual samples and 1.2 M automated sensor readings. We used multiple,
Keywords: Poisson, or Tobit regression models to evaluate water quality before and after the WSP intervention. The
Water safety plan compliance assessment analyzed exceedances of regulated, recommended, or operational water quality
HACCP thresholds using chi-squared or Fisher’s exact tests. Poisson regression was used to examine acute gas-
Drinking water quality
troenteritis incidence rates in WSP-affected drinking water service areas relative to a comparison area.
Gastrointestinal illness
Implementation of a WSP generally resulted in unchanged or improved water quality, while compliance
Water treatment
Regulatory compliance improved at most locations. Evidence for reduced acute gastroenteritis incidence following WSP imple-
mentation was found at only one of the three locations examined. Outcomes of WSPs should be expected
to vary across large water utilities in developed nations, as the intervention itself is adapted to the needs
of each location. The approach may translate to diverse water quality, compliance, and health outcomes.
© 2017 Elsevier GmbH. All rights reserved.
1. Introduction (EU) 2015/1787EU, 2015) and may be required as early as 2018.

In contrast to reactive approaches to water quality surveillance
In 2004, the World Health Organization Guidelines for Drink- and management, water purveyors who use WSPs seek to com-
ing Water Quality recommended that water suppliers develop and prehensively prevent problems from occurring. This management
implement Water Safety Plans (WSPs) to help proactively main- (“software”) intervention involves a continuous feedback loop of
tain safe public drinking water supplies and reduce health impacts risk identification, implementation of controls, and evaluation of
from water contamination events (Bartram et al., 2009). WSPs are whether risks are under control, stemming from the hazard analysis
now used in many world regions and required by national legisla- and critical control point (HACCP) approach used widely to ensure
tion in some countries. They were introduced into the European food safety. WSPs may or may not involve concurrent infrastruc-
Union Drinking Water Directive in 2015 (Commission Directive ture (“hardware”) upgrades or changes, depending on which risks
are identified and prioritized for each system. The WSP team, once
formed, conducts a thorough analysis of all potential risks to the
drinking water supply from source to tap, prioritizes these risks,
∗ Corresponding author.
E-mail address: [email protected] (K.E. Setty).
http://dx.doi.org/10.1016/j.ijheh.2017.02.004
1438-4639/© 2017 Elsevier GmbH. All rights reserved.
514 K.E. Setty et al. / International Journal of Hygiene and Environmental Health 220 (2017) 513–530
and establishes critical control points where ongoing monitoring and possible study locations for this project (Loret et al., 2016).
should take place (Bartram et al., 2009). It led to a ranking of reported WSP benefit categories among 21
More recently, evaluation frameworks and indicators have been drinking water utility managers as well as an average WSP labor
proposed to measure progress toward WSP goals and evaluate investment estimate of 10.5 person-months (full-time equivalent)
gains. Numerous indicators can relay the effectiveness of WSPs, for implementation and 4 person-months/year for ongoing WSP
broadly spanning inputs (e.g., funding and time commitment), maintenance.
activities/outputs (e.g., number of team meetings), outcomes (e.g.,
operational efficiency or cost savings), and impacts (e.g., water
quality or health improvements) (Gelting et al., 2012). Changes 2. Methods
related to the WSP process can take place across all categories,
although the former categories may show earlier and more measur- 2.1. Site selection
able change when compared to more distal outcomes and impacts.
Lockhart et al. (2014) recommends evaluating specific indicators To evaluate water quality, compliance, and health outcomes of
within four categories: institutional, operational, financial, and pol- WSP implementation, we undertook an observational retrospective
icy outcomes. A review of WSP evaluations to date (Kot et al., 2015) cohort study at five locations (locations 1–4 in France and loca-
found primary reported benefits of the WSP approach to include tion 5 in Spain) where WSPs were implemented between 2006 and
improvements in organizational structure or daily procedures, bet- 2013. Three (locations 1, 3, and 5) included a paired nearby com-
ter risk awareness among water operators, more efficient water parison area with no WSP implementation. Data availability was a
management practices, improved compliance with water regula- strongly limiting factor, so intervention and comparison areas were
tions, and a reduction in customer complaints. Another systematic not randomly selected (comparison area selection criteria are listed
review suggested financial outcomes of WSPs have the clearest evi- in Table A1). The five locations included in the study correspond to a
dence base, even though operational outcomes are more frequently total of 15 drinking water treatment plants and groundwater treat-
documented (String and Lantagne, 2016). The review concludes ment facilities (Table 1). Inclusion criteria specified either surface
that outcome and impact evaluation data demonstrating WSP value water or influenced groundwater sources, WSP implementation in
remain weak. the production and/or distribution system, and water quality data
Although a central goal of WSPs is to reduce the risk of available for at least two years before and after WSP implementa-
water contamination events, limited evaluation data is available tion. Each system had obtained ISO 22000 food safety management
to demonstrate WSP effectiveness at decreasing drinking water certification at the end of the WSP implementation period, one of
pathogen or chemical exposures, as well as corresponding health several existing WSP models (ISO, 2005). At locations 1, 2, and 4,
improvements. The impact of WSPs on human health has been only the production system was certified (the drinking water treat-
investigated in Iceland, one of the first countries to legislate their ment plants and/or groundwater treatment facilities). Location 5
use in 1995. Data collected before and after WSPs were intro- included two intervention areas: a “full WSP” where the produc-
duced showed measurably less contaminated water, significantly tion and distribution systems were certified and a “partial WSP”
fewer cases of diarrhea, and improved compliance with drinking certifying only the distribution system. In the partial WSP area,
water standards (Gunnarsdóttir et al., 2012a). Iceland is a unique water from another purveyor’s drinking water treatment plant is
developed country with a high quality groundwater supply, where delivered to the local service area via a main pipe. Most locations
chlorination is not used to disinfect drinking water supplies. We provided both production and distribution network water quality
sought to repeat this type of investigation at five locations in monitoring data; location 2 was limited to production samples only
France and Spain, with a focus on generalizing outcomes across and the partial WSP area of location 5 was limited to distribution
large population centers in developed nations served by chlori- samples only.
nated surface water and surface-influenced groundwater supplies. In some cases, the municipal boundaries where health data was
These regions have relatively low burdens of diarrheal disease com- reported did not fully coincide with the water service areas. The
pared to developing nations (WHO, 2010); still, the population location 1 intervention area, location 3 comparison area, and all
experiences a costly annual health burden from viral gastroen- areas for location 5 were considered to have virtually 100% corre-
teritis (especially norovirus) transmission, some of which stems spondence between the population served by health care providers
from water-related outbreaks (Kowalzik et al., 2015; Flahault and and water service providers. In contrast, the comparison area for
Hanslik, 2010; Beaudeau et al., 2008; Lopman et al., 2003). Surface location 1 was being supplemented at a rate of about 40% by
drinking water sources in France and Spain are affected by diverse drinking water from another source, although it was groundwa-
human and animal fecal influences, such as overland runoff and ter expected to be of higher quality than the 60% water supplied by
cross-contamination from wastewater pipes (Therre et al., 2008). the surface water treatment plant included in the study. At loca-
Cryptosporidiosis and giardiasis remain a concern for drinking tion 3, water supply coverage ranged from 21.5% to 60% within the
water managers, especially in spring and autumn periods of heavy four “intervention” municipalities where health data was collected.
rainfall. The two municipalities with 60% coverage were again receiving
The primary goals of this study were to characterize changes mixed water supplemented by another higher quality groundwa-
in water quality, compliance, and gastrointestinal disease inci- ter source. In the two municipalities with lower coverage rates,
dence following WSP implementation. We aimed to demonstrate 20–30% of inhabitants were receiving all of their water from the
the impacts of WSPs, as well as to note the presence of factors that WSP-affected source, while others were receiving only water from
might be used to improve WSP implementation and performance another source. 60% was considered the minimum coverage per-
in the future. A secondary goal was to evaluate the outcomes of full centage, so the main health results (Tables 5 and 6) exclude the
versus partial WSPs, where the scope is limited to only the produc- two municipalities with less than 30% exposure to the drinking
tion or distribution system. This project followed an earlier phase water intervention, although they were considered for sensitivity
of research into WSP inputs and outcomes within the Suez net- analysis.
work of utilities. Suez is a large multinational company based in Because this was a retrospective, observational study and gath-
France, named for their involvement in building the Suez Canal. A ering additional data was not possible, power calculations were
2014 study, which quantified costs and ranked perceived benefits not performed to designate minimum sample sizes. A minimum
of WSPs by surveying utility managers, helped to narrow the goals of two years of water quality data and one year of health data
K.E. Setty et al. / International Journal of Hygiene and Environmental Health 220 (2017) 513–530 515
Table 1
Characteristics of each study location, including the nature of the WSP intervention, comparison area characteristics (if available), population served (rounded to nearest
thousand), the number of matched municipalities included in health effects analysis for locations 1, 3, and 5, and the source water and treatment scheme of drinking water
treatment plants (DWTPs) or groundwater treatment facilities (GTFs).
Location (and nature of WSP intervention) Pop Served Municipalities Served DWTP or GTF DWTP or GTF Treatment Schemea
(and percent exposure Water Source
to water supply)
1 645,000
Intervention (production only) 43,000 1 (100%) Surface water Coagulation/sedimentation, rapid sand filtration,
ozonation, GAC filtration, utrafiltration, pH
stabilization, chlorination
Comparison 602,000 10 (60%) Surface water Coagulation/sedimentation, rapid sand filtration,
ozonation, GAC filtration, chlorination
2 1,000,000
Surface water Coagulation/sedimentation, GAC filtration,
ozonation, GAC filtration, UV, chlorination
Intervention (production only)
ozonation, ultrafiltration and chlorination
ozonation, GAC filtration, UV, chlorination
Groundwater GAC filtration, chlorination
3 77,000
Intervention (production/distribution) 43,000 4 (2 at 60%, 2 at Influenced Direct GAC filtration, UV, chlorination
20–30%) groundwater
Comparison 24,000 1 (100%) Protected Iron removal, chlorination
groundwater
4 73,000
Surface water Coagulation/sedimentation, rapid sand filtration,
Intervention (production only) ozonation, chlorination
Groundwater Iron removal, filtration, and chlorination
(summer only)
Groundwater Chlorination
(summer only)
5 325,000
Intervention (production/distribution) 148,000 2 (100%) Surface and Pre-oxidation (ClO2 ), coagulation/sedimentation,
groundwater rapid sand filtration, (50% to line 1) ozonation/GAC
filtration, (50% to line 2) ultrafiltration/reverse
osmosis, chlorination
Intervention (distribution only) 117,000 4 (100%) Pre-oxidation (Cl2 + ClO2 ),
Surface water
coagulation/sedimentation, GAC
filtration, chlorination
Comparison 60,000 1 (100%) ” ”
a
GAC = granular activated carbon.
was required in the before and after periods, and all possible data sensors were considered separately where available (locations 3
was requested. Specific pre- and post-WSP implementation time and 5). We selected twelve water quality parameters to evaluate
periods for each study location were then trimmed to sets of 12- water quality, treatment process effectiveness, and possible human
month intervals preceding the initiation of WSP team meetings health risk, including E. coli, fecal streptococci, total coliform, het-
(“before”) and following ISO 22000 certification (“after”) (Table 2). erotrophic plate count, trihalomethanes (THMs), bromate, free
Differences in climate exist among locations 1 and 2 (in northern residual chlorine, total residual chlorine, aluminum, total organic
France where rainfall peaks in May), locations 3 and 4 (south- carbon, turbidity, and pH. Data cleaning involved attribution of the
western France where rainfall peaks November to January), and study location, time period (before, during, or after the interven-
location 5 (northeastern Spain where rainfall peaks September to tion), and presence or absence of a WSP intervention, comprising
November). Because heavy seasonal rainfall could affect source about 240,500 manual samples and more than 1.24 million online
water quality parameters such as turbidity and might influence the sensor readings for a total of nearly 1.5 million water quality data
utilities’ performance, this approach served to maximize the period points.
of observation and sample size while controlling for seasonal influ- Detection limits for the equipment and/or test method used in
ences on health and water quality data. Data from “during” WSP water quality data collection are listed in Table A2. Left-censored
implementation (periods of 10–24 months from the initiation of data with detection limits of one or lower were set to zero (to
team meetings to certification) was excluded. match pre-processing of the French data), while left-censored data
with detection limits above one were set to half the detection limit.
Right-censored data were set to the detection limit. For microbial
2.2. Water quality analysis
water quality parameters, absence was set to zero and presence
was set to one. Data points that were blank or otherwise could not
Water quality data were provided by employees of the Suez-
be resolved were left as missing data. In some cases, water quality
affiliated drinking water supplier at each study location, including
data represented water that did not reach the consumer, (e.g., an
the parameter, unit, date, time (if applicable), and monitoring sta-
alarm or scheduled maintenance event might trigger containment
tion. Data sets were produced via either routine internal water
and disposal of a water batch and/or emergency cross-connection
quality monitoring or external quality control involving inde-
with alternate water supplies), but it was nevertheless included as
pendent sampling and analysis by health authorities. Data from
an event relevant to the WSP and performance history. Precise his-
manual sampling records were pooled while data sets from online
Table 2 the effect of suspicious extreme values in the data set that may
Time periods of water quality and health/population data availability at each loca-
have been affected by maintenance or other events, even though
tion, trimmed to 12-month intervals before and after WSP implementation.
imperfect historical records were available to justify data removal.
Location Water Quality Data Availability Health/Population Data Finally, variance ratios were computed between the pre- and post-
Availability
intervention period as another indicator of water quality control.
1 Before: 1 Jan 2008–31 Dec Before: 1 Jan 2010 – 31 Dec
2010 (3 years) 2010 (1 year)
WSP Implementation: 1 Jan 2011 – 31 Oct 2011 (10 months) Value or ln(Value) = ␤0 + ␤1 ∗Time + ␤2 ∗Proximity +
After: 1 Nov 2011 – 31 Oct After: 1 Nov 2011 – 31 Oct
2015 (4 years) 2015 (4 years)b ␤3 ∗Site + ␤4 ∗Time ∗ Site (1)
2 Before: 1 Jan 2003 – 31 Dec (data not available)
2005 (3 years)
WSP Implementation: 1 Jan Value or ln(Value) = ␤0 + ␤1 ∗Time + ␤2 ∗Proximity (2)
2006 – 31 Mar 2007 (15
months)
After: 1 Apr 2007 – 31 Mar
2015 (8 years) Value or ln(Value) = ␤0 + ␤1 ∗Time + ␤2 ∗Proximity +
3 Before: 13 Nov 2010 – 12 Nov Before: 13 Nov 2010 – 12 Nov ␤3 ∗FullSite + ␤4 ∗PartialSite + ␤5 ∗Time ∗ FullSite + ␤6
2012 (2 years) 2012 (2 years)
WSP Implementation: 13 Nov 2012 – 20 Dec 2013 (13 months) ∗Time ∗ PartialSite (3)
After: 21 Dec 2013 – 20 Dec After: 21 Dec 2013 – 20 Dec
2015 (2 years) 2015 (2 years)b
4 Before: 1 Jan 2003 – 31 Dec (data not available)

2.3. Compliance analysis
2006 (4 years)
WSP Implementation: 1 Jan The compliance analysis compared water quality data before
2007 – 31 Mar 2008 (15 and after WSP implementation to relevant thresholds, including:
months)
European Union (EU) Drinking Water Directive quality limits and
After: 1 Apr 2008 – 31 Mar
2015 (7 years) quality references, national quality limits and quality references,
Suez internal recommended practices (for France), location-
5 Before: 1 Jan 2005a – 31 Dec Before: 1 Jan 2005 – 31 Dec
2007 (3 years) 2007 (3 years)
specific WSP critical limits, and location-specific WSP operational
WSP Implementation: 1 Jan 2008 – 31 Dec 2009 (24 months) limits. National quality limits and references for France and Spain
After: 1 Jan 2010 – 31 Dec 2015 After: 1 Jan 2010 – 31 Dec 2015 often closely matched the EU Drinking Water Directive (98/83/CE)
(6 years) (6 years) (Ministère de La Santé et des Solidarités, 2007; Ministerio de
a
Some datasets (online sensors datasets for turbidity, total organic carbon, and la Presidencia, 2003; EU, 1998, 2015). Some regional legislation
free chlorine; and critical control parameters used for compliance analysis including was applied to location 5 only (Generalitat de Catalunya, 2005).
trichloroethylene/tetrachloroethylene, nickel, chromium VI, and iodine absorption) Within French and Spanish national regulations, “quality limits” are
begin 1 Jan 2006 for a total of two years in the before period.
b
For locations 1 and 3, extrapolated population data was used during the after
intended as upper limits while “quality references” are intended as
period to enable comparison with case numbers for 2014 and 2015. indicators of good practice. Internally defined thresholds were gen-
erally the most stringent, since compliance had operational but not
regulatory implications.
torical records of maintenance, flushing, spiking, equipment failure We evaluated compliance for the same set of 12 parameters used
and other events/activities that potentially affected individual sam- for water quality modeling, as well as any additional parameters
ples were not available; therefore, suspected outliers remained in relevant to critical control points (at location 5 only). Between 9 and
the dataset. 22 relevant thresholds existed and were investigated at each loca-
All statistical analyses were performed using SAS 9.4 software. tion. A 2 × 2 table was constructed using the number of passes and
For chemical parameters, multiple regression was applied to detect fails in each time period. Significant differences in cell sizes were
significant differences between the pre-intervention (“before”) and then evaluated using either a chi-squared or Fisher’s exact test for
post-intervention (“after”) time periods. The model controlled for unequal proportions (Fisher’s exact test was used when any cell size
a clustered sampling design (samples clustered by monitoring was five or less). A p-value correction was again applied within each
station), proximity to the treatment facility (production versus location’s family of compliance statistical tests using the adaptive
distribution samples), and, if available, non-WSP comparison site Holm procedure. Data analysis definitions of “compliance” were
conditions over the same time period (at locations 1, 3, and 5). matched as closely as possible to the utility managers actual use
Chemical parameters censored by a high detection limit, especially of the data (e.g., reporting of manual monitoring data to regula-
bromate, occasionally achieved better fit with a Tobit regression tory bodies and use of real-time online sensor data for internally
model of the same form. Microbial parameter data sets were fit with maintained critical control points), but needed to be simplified in
a corresponding Poisson regression model (based on the natural some cases to permit analysis of incomplete historical data records.
log of the dependent variable, value), owing to the non-continuous These analyses are denoted as “simplified” in Tables 4 and B9 . For
count nature of the data. example, if the time resolution of historic sampling was twice daily,
For the locations with comparison areas, beta coefficients the data could not be analyzed for hourly changes. Therefore, anal-
are reported for the interaction term consisting of time period ysis results may differ from the utilities’ self-reported compliance
(before/after) and WSP presence (no/yes) (Eq. (1)). For locations records, since more complex verifications involving time persis-
2 and 4, the beta value represents the effect of time period alone tence, equipment substitution, and repeat sampling apply to some
(Eq. (2)). For location 5, two sets of dummy and interaction vari- thresholds in practice.
ables were used to represent the full (production and distribution)
WSP versus the partial (distribution only) WSP intervention (Eq. 2.4. Health analysis
(3)). To determine significance, a p-value correction was applied
within each location’s family of water quality statistical tests To evaluate health before and after WSP intervention, acute gas-
using the adaptive Holm procedure. Sensitivity testing examined troenteritis incidence data was provided by the national or regional
Table 3
Summary of statistically significant water quality outcomes reported in Tables B4 through B8, grouped by improvement, degradation, and neutral changes. “Neutral” changes
may or may not be considered desirable depending on the individual needs of the drinking water utility.
Location Improvements in water quality Degradation in water quality Neutral changes in water quality
post-intervention post-intervention post-intervention
1 Aluminumb Bromate pH
2 Trihalomethanes – –
3 – Total coliform, heterotrophic plate Free chlorine (sensors)
count, turbidity (sensors)
4 Heterotrophic plate count, aluminum, turbidity – Free chlorine, total chlorine, pH
5 (full WSPa ) Trihalomethanes, total organic carbon, Bromate Free chlorine (sensors), pH
turbidity (manual and sensors)
5 (partial WSPa ) Turbidity – Free chlorine, pH
a
The full WSP applied to both the production and distribution system; the partial WSP applied to the distribution system only.
b
Remained constant relative to comparison area.
Table 4
Summary of statistically significant changes in compliance after WSP implementation at each intervention location. Detailed noncompliance rates and other test results can
be found in Table B9.
Location Parameter Change post-intervention Threshold definitionb Source
1 Total coliform Increase in compliance >0 MPN/100 ml EU and French quality limit
Turbidity Increase in compliance >2 NTU French quality reference for tap
2 None
3 Free chlorine (sensors) Increase in compliance <0.05 mg/l WSP critical limit for chlorination
4 Bromate Increase in compliance >10 ␮g/l EU and French quality limit

Free chlorine (surface water plant only) Increase in compliance <0.2 mg/l *simplified WSP operational limit (surface water plant)
Aluminum Increase in compliance >100 ␮g/l Suez internal recommended practice
5 (full WSPa ) Total coliform Increase in compliance >0 MPN/100 ml EU and Spanish quality limit
THMs Increase in compliance ≥50 ␮g/l Suez recommendation for plant outlet (in France)
Increase in compliance >100 ␮g/l EU and Spanish quality limit for network
Free chlorine (sensors) Increase in compliance <0.2 mg/l *simplified Catalunya/WSP critical limit for chlorination
Aluminum Increase in compliance >200 ␮g/l EU and Spanish and regional quality limit
Turbidity (sensors after Decrease in compliance ≥0.75 NTU WSP operational limit
sand filters) Decrease in compliance ≥1 NTU *simplified WSP critical limit
Turbidity (sensors) Increase in compliance >0.5 NTU Spanish quality reference for plant outlet
Nickel Increase in compliance >20 ␮g/l EU and Spanish quality limit (for tap)
5 (partial WSPa ) None

a
b
Direction indicates when a sample does not attain the recommended range of values. Some thresholds were simplified to enable comparison with historical data.
Self-reported records might differ based on the application of time duration, repeat sampling, or equipment validation procedures.
Table 5
Incidence of acute gastroenteritis before and after WSP implementation at both intervention and comparison (“comp”) areas and model interaction term between time and
site. Negative beta values correspond to a reduction in acute gastroenteritis in the intervention area.
Location Incidence rate Incidence rate Incidence rate ratio Ratio of incidence rate Model beta and Wald Model p-value
(cases per 1000 (cases per 1000 and 95% confidence ratios and 95% 95% confidence Chi-Square *significant
person-years) person-years) interval (after: confidence interval interval (time* site)
(before) (after) before) (interv: comp)
1 93.6 100.2 1.07 (1.03, 1.11) 0.96 (0.93, 1.00) −0.037 (−0.072, −0.001) 4.11 0.043*
(comp) 78.0 86.5 1.11 (1.10, 1.12)
3 126.5 111.8 0.88 (0.85, 0.92) 1.01 (0.96, 1.07) 0.013 (−0.042, 0.069) 0.22 0.640
(comp) 102.6 89.5 0.87 (0.84, 0.91)
5 (full WSPa ) 0.304 0.418 1.37 (1.13, 1.68) 1.32 (0.96, 1.82) 0.276 (−0.044, 0.597) 2.86 0.091
5 (partial WSPa ) 0.397 0.564 1.42 (1.17, 1.73) 1.37 (0.99, 1.87) 0.310 (−0.008, 0.628) 3.66 0.056
(compa ) 0.513 0.534 1.04 (0.81, 1.34)
a
The full WSP applied to both the production and distribution system, the partial WSP applied to the distribution system only, and the comparison area had no WSP. All
areas of location 5 have much lower case numbers due to the differences in public health surveillance methods between France and Spain.
public health authorities: Santé Publique France (formerly Institut locations 1, 3, and 5 by matching the geographical boundaries
de Veille Sanitaire) at locations 1 and 3 and Servei Català de la Salut of intervention (WSP) and comparison (non-WSP) water service
(CatSalut) at location 5. The data collection mechanism in Spain areas with between one and ten corresponding municipalities
relies on the Spanish Minimum Basic Data Set (MBDS) hospital reg- (“municipalities served” in Table 1). Population data at the level of
istry, whereas the French data collection method was revised in municipality was provided by the National Institute of Statistics and
2010 to capture a greater percentage of acute gastroenteritis cases Economic Studies (INSEE) in France and the Statistical Institute of
(estimated at around 32% of all cases) by relying on state-provided Catalonia (IDESCAT) in Spain. Because the French population data is
prescription drug reimbursements (Bounoure et al., 2011). Case released after a two-and-a-half-year time lag, population estimates
numbers reported by municipality of residence were pooled for needed to be extrapolated for the years 2014–2015 to enhance or
Table 6
Incidence of acute gastroenteritis by age group at location 1, 3, and 5 intervention and comparison (“comp”) areas. Negative beta values correspond to a reduction in acute
gastroenteritis in the intervention area.
Location Age group Incidence rate Incidence rate Incidence rate Ratio of incidence rate Model beta and 95% Wald Model p-value
(cases per 1000 (cases per 1000 ratio and 95% ratios and 95% confidence interval Chi-Square *signif.
person-years) person-years) confidence confidence interval (time* site)
(before) (after) interval (after: (interv: comp)
before)
1 1–4/0–4b 236.9 301.2 1.27 (1.18, 1.37) 1.02 (0.94, 1.10) 0.017 (−0.057, 0.091) 0.21 0.644
(comp) 235.5 294.3 1.25 (1.22, 1.28)
1 5–14 168.5 174.9 1.04 (0.97, 1.11) 0.95 (0.88, 1.02) −0.051 (−0.123, 0.020) 1.97 0.160
(comp) 141.1 154.2 1.09 (1.07, 1.12)
1 15–64 70.8 70.8 1.00 (0.95, 1.05) 0.94 (0.89, 0.99) −0.062 (−0.112, −0.011) 5.78 0.016*
(comp) 57.1 60.7 1.06 (1.05, 1.08)
1 65+ 24.9 26.4 1.06 (0.87, 1.29) 0.90 (0.73, 1.10) −0.109 (−0.313, 0.095) 1.09 0.296
(comp) 21.5 25.4 1.18 (1.12, 1.25)
3 1–4/0–4b 398.4 353.7 0.89 (0.81, 0.97) 0.96 (0.84, 1.09) −0.043 (−0.169, 0.083) 0.45 0.505
(comp) 388.2 359.7 0.93 (0.85, 1.01)
3 5–14 238.8 198.1 0.90 (0.83, 0.97) 1.03 (0.92, 1.16) 0.032 (−0.083, 0.148) 0.30 0.583
(comp) 198.1 172.6 0.87 (0.80, 0.95)
3 15–64 97.3 84.7 0.87 (0.83, 0.92) 1.03 (0.95, 1.12) 0.032 (−0.047, 0.110) 0.63 0.429
(comp) 79.7 67.2 0.84 (0.79, 0.89)
3 65+ 55.4 55.0 0.99 (0.86, 1.15) 1.12 (0.91, 1.37) 0.109 (−0.094, 0.313) 1.11 0.293
(comp) 43.8 39.1 0.89 (0.78, 1.02)
a
5 (full WSP ) 0–14 1.025 1.128 1.10 (0.83, 1.46) 0.96 (0.55, 1.69) −0.039 (−0.602, 0.523) 0.02 0.891
5 (partial WSPa ) 1.535 1.734 1.13 (0.87, 1.47) 0.99 (0.57, 1.71) −0.014 (−0.565, 0.537) 0.00 0.961
(compa ) 0.871 0.997 1.14 (0.70, 1.86)
5 (full WSPa ) 15+ 0.174 0.276 1.58 (1.19, 2.10) 1.59 (1.06, 2.38) 0.462 (0.055, 0.869) 4.95 0.026*
5 (partial WSPa ) 0.192 0.335 1.74 (1.29, 2.35) 1.74 (1.15, 2.65) 0.557 (0.138, 0.975) 6.80 0.009*
(compa ) 0.449 0.448 1.00 (0.75, 1.34)
a
The full WSP applied to both the production and distribution system, the partial WSP applied to the distribution system only, and the comparison area had no WSP. All
areas of location 5 have much lower case numbers due to the differences in public health surveillance methods between France and Spain.
b
For French data (locations 1 and 3), the number of acute gastroenteritis cases exclude infants under age 1, because they are not expected to be exposed to tap water
consumption. The population data includes infants.
enable comparison with cases reported at locations 1 and 3, respec- 2.5. Informal audits
tively. Extrapolations were based on linear estimates of growth for
each municipality using 2010–2013 data. Lastly, qualitative questionnaires were developed to better
Because the number of years in each time period varied, acute understand the nature of the data and the WSP intervention at each
gastroenteritis was characterized as a rate: the incidence (number site, as well as to gather information about perceptions and expec-
of new cases) per 1000 person-years. Population data could further tations of undertaking the WSP. Informal WSP audits were carried
be matched to existing age divisions in the health data to strat- out in June/July 2016 at locations 1, 3, and 5, incorporating a semi-
ify the analysis by ages under 5, 5–14, 15–64, and over 65 years at structured group discussion and a guided tour of a drinking water
locations 1 and 3, and by ages 0–14 and over 15 at location 5. To sta- treatment plant. Questions were developed using the World Health
tistically compare acute gastroenteritis incidence before and after Organization and International Water Association’s Practical Guide
WSP implementation, a Poisson regression model was applied with to Auditing Water Safety Plans (WHO and IWA, 2015) as well as an
population as the offset (person-years) and controlling for the com- interview guide used in Iceland by Gunnarsdóttir (2012; Appendix
parison area conditions (the base level of the site variable; Eq. (4)). 1). Because the questionnaires were intended as a starting point
Comparison area values were included to isolate the effect of the for discussion, sessions were structured loosely and answers were
drinking water intervention, as opposed to data reporting, overall not forced on all questions; however, information was specifically
health, or other changes that may have affected the whole region. requested on significant events that took place during the study
The location 5 model separated the full WSP (production and distri- period, and expectations of change in water quality or health data
bution) and partial WSP (distribution only) intervention areas (Eq. as a result of the WSP (reported in Tables B2 and B3, respectively).
(5)). Cases reported by municipality were pooled within each area Shorter questionnaires adapted for electronic rather than in-person
and clustering was not considered in the model. Finally, sensitiv- delivery were then developed and completed by a member of the
ity of the model was tested relative to assumptions of population Suez research team for locations 2 and 4 in August 2016.
extrapolation and service area coverage.
3. Results
ln(Cases/Person-Year) = ␤0 + ␤1 ∗Time + ␤2 ∗Site +
3.1. Water quality
␤3 ∗Time ∗ Site (4)
Several significant water quality differences were observed
between the pre-implementation and post-implementation peri-
ods at the intervention area. Mean values and model results by
ln(Cases/Person-Year) = ␤0 + ␤1 ∗Time + ␤2 ∗FullSite + ␤3 location, parameter, and time period are detailed in Tables B4
through B8, and summarized as improvements, degradation, and
∗PartialSite + ␤4 ∗Time ∗ FullSite + ␤5 ∗Time ∗ PartialSite (5)
neutral outcomes in Table 3. Changes listed in the “improvements”
column are considered desirable, since drinking water quality man-
agers seek to reduce the concentration of these constituents for
Table 7
Sensitivity test of acute gastroenteritis outcomes in the overall population and by age group at location 3 intervention and comparison (“comp”) areas, adding health data
from two municipalities with less than 30% water coverage to the intervention area. Values from the comparison area remain unchanged.
Location Age group Incidence rate Incidence rate Incidence rate ratio Ratio of incidence rate Model beta and Wald Model
(cases per 1000 (cases per 1000 and 95% confidence ratios and 95% 95% confidence Chi-Square p-value
person-years) person-years) interval (after: confidence interval interval (time* site) *signif.
(before) (after) before) (interv: comp)
3 All 127.0 115.4 0.91 (0.88, 0.93) 1.04 (0.99, 1.10) 0.041 (−0.008, 0.091) 2.71 0.100
(comp) 102.6 89.5 0.87 (0.84, 0.91)
3 0–4/1–4a 389.1 373.1 0.96 (0.90, 1.02) 1.03 (0.93, 1.16) 0.034 (−0.076, 0.145) 0.37 0.542
(comp) 388.2 359.7 0.93 (0.85, 1.01)
3 5–14 237.5 211.4 0.89 (0.84, 0.94) 1.02 (0.92, 1.13) 0.022 (−0.081, 0.125) 0.17 0.678
(comp) 198.1 172.6 0.87 (0.80, 0.95)
3 15–64 98.8 89.2 0.90 (0.87, 0.94) 1.07 (1.00, 1.15) 0.068 (−0.002, 0.139) 3.61 0.058
(comp) 79.7 67.2 0.84 (0.79, 0.89)
3 65+ 54.9 52.4 0.96 (0.86, 1.07) 1.07 (0.90, 1.28) 0.070 (−0.106, 0.246) 0.60 0.438
(comp) 43.8 39.1 0.89 (0.78, 1.02)
a
For French data (locations 1 and 3), the number of acute gastroenteritis cases exclude infants under age 1, because they are not expected to be exposed to tap water
consumption. The population data includes infants.
operational or health reasons. In contrast, “degradation” refers to an reduction where it was not found to change originally (␤ = −0.044,
increase in a parameter that managers seek to minimize. Changes in raw p-value < 0.001). Likewise, dropping two extreme values for
“neutral” parameters, while statistically significant, might or might free chlorine at the location 1 comparison area improved the model
not be considered desirable (and therefore relevant to operational fit and made the reduction statistically significant (␤ = −0.047, raw
or health outcomes) depending on the goals of the particular drink- p-value < 0.001). Additional sensitivity testing at locations 2 and 4
ing water utility at any given time. examined the effects of seasonality, since several months of addi-
The water quality parameters that changed significantly var- tional partial-year data was available in the after period. It showed
ied from one location to the next, and few patterns were observed no changes in the significance of test results when this data was
(Table 3). Several microbial water quality parameters demon- included.
strated little variation from zero and models could not be fit
(Tables B4 through B8). Heterotrophic plate count and total col- 3.2. Compliance
iform worsened at location 3, while heterotrophic plate count
improved at location 4. Likewise, turbidity increased at location Like water quality, changes in compliance following WSP imple-
3, but decreased at locations 4 and 5 (full and partial WSPs). Bro- mentation also varied, as measured by rates of noncompliance with
mate (a byproduct of using ozonation to treat water) increased at between nine and twenty-two relevant internal or external water
locations 1 and 5 while trihalomethanes (a byproduct of chlori- quality benchmarks per location. Significant outcomes ranged from
nation) decreased at locations 2 and 5 (full WSP). In the neutral zero to ten per location (summarized in Table 4). Nearly all repre-
category, free residual chlorine increased at locations 3 and 4 while sented improvements in water quality, with a couple of exceptions.
it decreased at location 5 (full and partial WSPs), although free While turbidity compliance increased at locations 1 and 5 (full WSP,
chlorine levels at location 5 were intentionally kept higher than outlet), it decreased at location 5 (full WSP, operational and critical
locations 3 and 4 due to a difference in local regulatory standards limits after sand filters). Aluminum compliance increased at both
(see typical median values in Table B1). locations 4 and 5 (full WSP). Location 2 and the location 5 partial
Increased numbers of samples, especially with regards to online WSP area showed no changes in compliance. Some parameters dis-
sensor data, increased statistical power to detect small differences. played movement around a relevant threshold in this component of
Of potential relevance to water utility managers, some changes the study, even though the change in the values themselves from
in sample means were observed with descriptive statistics but before to after WSP implementation was not statistically signifi-
not found to be statistically significant, possibly due to relatively cant as reported in Table 3 (e.g., turbidity at location 1; bromate
low power or low numbers of non-zero data points. Examples at location 4). The opposite also held true, in that some statistically
included a decrease in heterotrophic plate counts, turbidity, and tri- significant changes in water quality levels as reported in Table 3 did
halomethanes at location 1 and a decrease in heterotrophic plate not correspond to increased or decreased compliance with applica-
count and aluminum at location 5 (full WSP) (detailed in Tables ble thresholds. Full results of the compliance analysis can be found
B4 and B8). As another indicator of water quality control, the WSP in Appendix B (Table B9).
intervention sometimes resulted in changes in the distribution of
data around the mean, reported as the variance and variance ratio
(F-value) in Tables B4 through B8. Both increases and reductions 3.3. Health
in variance were observed; variance declined notably (with a vari-
ance ratio equal to or exceeding 4:1) for turbidity at location 1, Health outcomes for the overall population also varied across
heterotrophic plate count at locations 4 and 5 (full WSP), and tri- locations (Table 5). Location 1, where total coliform and turbidity
halomethanes at location 5 (full WSP). compliance improved, showed a statistically significant decrease
With regards to sensitivity testing, exclusion of extreme values in the incidence of acute gastroenteritis relative to a comparison
did not alter significance of the statistical test for free chlorine or area, following the implementation of a WSP (p = 0.043, ␣ = 0.05).
turbidity at location 2, free chlorine and free chlorine sensors at This corresponds to about a 4% reduction in acute gastroenteritis
location 3, total chlorine at location 4, or free chlorine and turbid- incidence, when comparing the incidence rate ratios between the
ity sensors at location 5. Testing did indicate a potential reduction intervention and comparison areas. At location 3, where total col-
in free and total chlorine levels (neutral management outcomes) iform and turbidity levels increased, the WSP intervention area did
at location 1. Exclusion of four extreme values for total chlorine at not experience a significant change in acute gastroenteritis relative
the location 1 intervention area showed a statistically significant to the comparison area (p = 0.640). Looking at the overall popula-
tion for location 5, where water quality and compliance generally
improved, initially showed no significant change in acute gastroen- across parameters and locations; therefore, statistical power to
teritis for the full or partial WSP intervention areas (p = 0.091 and detect changes varied among data sets. Individual parameter mod-
0.056, respectively). The larger case numbers at locations 1 and els at each site had a sample size ranging from 12 (for bromate
3 enhanced the statistical power of the overall population test to at location 3, where a model could not be fit) to 16,203, with an
detect smaller differences in incidence rates. Location 1 had the average of about 4296 samples (Tables B4 through B8). Significant
largest sample size on the order of 278,000 total cases and 3.2 mil- changes were more easily detected in online sensor datasets with
lion person-years, while location 3 (excluding two municipalities sample numbers up to 531,603. Some undesirable changes were
with low water service coverage) included roughly 20,000 total observed at some locations (Tables 3 and 4), but a widespread
cases and 189,000 person-years. Both case numbers and incidence harmful effect was not indicated over the period of the interven-
rates for location 5 (which had a total of about 1,300 cases and 2.9 tion.
million person-years) were much lower than those for locations In particular, microbial indicator levels at location 3 appeared to
1 and 3 due to differences in public health surveillance methods worsen following the WSP intervention. This might be explained by
between Spain and France. the utility managers’ desire to keep chlorine levels as low as pos-
Stratification of health data by age group offered additional sible in response to concerns about disinfection by-products and
information (Table 6). The pooled results appeared to be driven their potential adverse health effects. Due to local socio-political
by the adult population (ages 15+), which contributed about four pressure to eliminate chlorine usage, this location is looking to
to five times the number of person-years to the analysis. When pilot an unchlorinated water supply in the near future. The WSP
data were stratified by the four possible age groups (under 5, at this location did not emphasize control of microbial contamina-
5–14, 15–64, and 65+) at location 1, only the 15–64 age group tion due to the perceived high quality of the groundwater source.
demonstrated a statistically significant reduction in acute gastroen- Interim turbidity compliance (after the sand filters) at location 5
teritis incidence (p = 0.016). Likewise, stratification by children also worsened, although the finished water saw the opposite effect,
(ages 0–14) and adults (ages 15+) at location 5 showed a statis- with an improvement (decrease) in turbidity levels and improved
tically significant increase in acute gastroenteritis incidence for compliance based on manually sampled and sensor water qual-
adults only at both the full WSP and partial WSP intervention areas ity data. Finally, bromate formation increased at locations 1 and 5,
(p = 0.026 and 0.009, respectively). At all three locations, incidence potentially due to changes in ozonation practices.
rates for children were higher than those reported among adults. The microbial water quality data, especially for E. coli and fecal
Normally populations with poorer immunity (young children and streptococci, overwhelmingly consisted of values below the detec-
the elderly) might be expected to exhibit greater changes in health tion limit (i.e., absence or less than one colony-forming unit per
outcomes, but this held true only for the 65+ age group. It should 100 ml). This scenario is common across developed nations with
be noted that for French data (locations 1 and 3), the reported chlorinated drinking water supplies. Although a sign of low health
case numbers of acute gastroenteritis (numerators of the rates) risk, such a data distribution hinders the ability to characterize
in the under 5 age group exclude infants (under age 1), because changes in baseline water quality. By reducing both pathogen and
they are not expected to consume tap water. The population data indicator bacteria levels, the practice of chlorine disinfection may
(denominators of the rates) does include infants, serving to artifi- even preclude detection of potential health risks if the signal from
cially reduce incidence rates for this age group at locations 1 and 3 the indicator organism is eliminated but the pathogen remains
across both intervention and comparison areas. viable. A quantifiable value might be elicited by (a) sampling water
Regarding sensitivity analyses, exclusion of extrapolated just prior to disinfection, (b) using larger sample volumes, or (c)
2014–2015 population data from the location 1 analysis did not detecting the presence of pathogens themselves rather than indica-
substantially affect the outcome. Assessing only data reported tor organisms. Under current regulatory scenarios, these measures
through 2013 resulted in a similar approximately 5% reduction in would add to (not replace) ongoing compliance monitoring efforts.
acute gastroenteritis incidence at the intervention area relative to Owing to the potential advantages, molecular methods for direct
the comparison area (␤ = −0.055, p = 0.005). Further, where health pathogen detection were being developed and validated during the
data and water supply coverage did not completely coincide, the study’s site visits. Drinking water utilities may adopt rapid testing
effects of WSP implementation on acute gastroenteritis incidence methods in the future as the technology becomes more refined and
may have been diluted (in the case of no change in the additional widely available.
water source over time) or confounded (in the case of beneficial or Among the four measured microbial water quality indicators,
detrimental changes in the additional water source over time) by total coliform and heterotrophic plate counts showed greater vari-
exposure to other drinking water sources. At location 3, addition of ability than E. coli and fecal streptococci, demonstrating significant
the two municipalities with less than 30% water supply coverage changes in some cases. Total coliform compliance improved sig-
did not alter significance of results, which showed no effect of the nificantly at locations 1 and 5 (full WSP), while levels worsened
WSP implementation on acute gastroenteritis (Table 7). at location 3. Heterotrophic plate counts decreased significantly
at only one of the five locations, and were not examined in the
compliance portion of our study (owing to the lack of appli-
4. Discussion cable compliance thresholds for piped water supplies in France
and Spain). This finding corresponds somewhat with results from
4.1. Water quality and compliance outcomes Iceland, where two out of the five locations examined indi-
vidually showed significant drops in mean heterotrophic plate
Many beneficial water quality and compliance changes, corre- counts (Gunnarsdóttir et al., 2012a). When pooled across locations
sponding to an expectation of reduced risk to consumers, were included in the Iceland study, heterotrophic plate counts exceed-
identified in this study. Changes in specific water quality param- ing 10 colony-forming units were significantly less likely following
eters between the pre- and post-implementation periods varied the WSP intervention (Gunnarsdóttir et al., 2012a). The Iceland
from one location to the next. This finding is compatible with study examined unchlorinated drinking water suppliers and also
the nature of the intervention, since the general WSP approach included smaller water suppliers (<5000 inhabitants) that typically
should be adapted to each location based on the specific pri- have higher initial non-compliance rates.
oritized risks. Owing to the observational nature of the study,
sample sizes were not assigned and were not always consistent
4.2. Health outcomes pathogens, in contrast to the unchlorinated treatment schemes in

Iceland.
Changes in the incidence of acute gastroenteritis are generally
difficult to discern due to public health surveillance data limita- 4.3. Relationship between water quality and health outcomes
tions and the existence of multiple exposure routes (CDC, 2011).
Bottled water consumption and self-treatment of gastrointestinal Stability or decreases in bacteria levels might be expected as
disease apply to large swaths of the population in France and Spain, a precursor to declining acute gastroenteritis incidence, demon-
inhibiting the ability to associate WSPs and health outcomes. Only strative of the mode of disease transmission. Fecal indicator
about two-thirds of the population is expected to consume tap bacteria, including E. coli and fecal streptococci are interpreted
water (Therre et al., 2008), and the majority of acute gastroen- as signs of fecal contamination, whereas total coliform and het-
teritis cases are self-treated or resolve without treatment (Lopman erotrophic plate count are more indicative of general sanitary
et al., 2003). Large background fluctuations in acute gastroenteritis conditions and potential pathogen presence. Higher turbidity may
stem from the dominance of other pathogen transmission routes, also be correlated with poor pathogen removal and increased
especially the annual winter peak in person-to-person norovirus rates of acute gastroenteritis (Beaudeau et al., 2012). At loca-
transmission (Chikhi-Brachet et al., 2002; Arena et al., 2014), mak- tion 1, where a significant health effect was found, turbidity and
ing drinking water exposure a relatively minor contributor to the total coliform compliance correspondingly improved and turbid-
burden of disease (Lopman et al., 2003). ity variance decreased significantly post-intervention (F = 12.85,
Further, the acute hospital records used in Spain are expected p < 0.0001; Table B4). Location 3, in contrast, where no health effect
to capture only a small percentage (perhaps 1–2%) of actual cases, was found, exhibited a worsening of total coliform, heterotrophic
making trend extrapolation fairly difficult. Representatives of Santé plate count, and turbidity levels (Table B6).
Publique France indicated the prescription-based reporting used At location 5 for the full WSP intervention only, the vari-
since 2010 in France may have strengthened the location 1 and 3 ance of heterotrophic plate count decreased significantly (F = 153.8,
analyses by capturing about 32–33% of total cases (Bounoure et al., p < 0.0001), and total coliform compliance improved (Tables 3 and
2011). Our estimates of acute gastroenteritis incidence were gen- 4). Further, turbidity levels in the treated water improved, although
erally on the same order of magnitude as others in the literature turbidity compliance after the sand filters (an interim measure-
(Van Cauteren et al., 2012; Kowalzik et al., 2015; Chikhi-Brachet ment within the treatment plant) was noted to worsen after the
et al., 2002). The overall incidence rates found in France were some- intervention (Table 4). Finally, compliance with free chlorine stan-
what higher than other sources, while the incidence rates found in dards improved at the full intervention area, while free chlorine
Spain were somewhat lower, illustrating differences in surveillance levels dropped slightly at both the full WSP and partial WSP
methods. Performing pooled analysis across locations (e.g., via a intervention areas. Thus, the overall water quality evidence does
multi-level model) would be useful, but is precluded by the small not necessarily support increase pathogen exposure as a causal
number of case studies and differences in data collection methods. precursor for increased acute gastroenteritis incidence. Possible
A prospective, randomized controlled study design might be rec- hypotheses to explain the outcome might be: (1) other expo-
ommended as the gold standard for overcoming confounding and sures related to acute gastroenteritis, such as foodborne pathogen
data consistency constraints. outbreaks (as mentioned above), or (2) presence of pathogenic
One of the three case studies did demonstrate a reduction in organisms that are not well-correlated with fecal indicator bacteria
acute gastroenteritis incidence following WSP implementation, and/or resistant to current treatment schemes.
corresponding to about a 4% decrease in acute gastroenteritis inci-
dence in the overall population, or 6% in the 15–64 age group 4.4. Comparison among locations
(Tables 5 and 6). This occurred at location 1, which had the
largest health dataset. Location 5 full (production and distribu- Although the five case studies were not directly compared,
tion) and partial (distribution only) WSPs showed the opposite of the full WSP at location 5 resulted in the most dramatic water
the expected effect, with significantly higher post-implementation quality improvements, probably because (a) some baseline values
rates of acute gastroenteritis among adults in both intervention (e.g., trihalomethanes, total organic carbon) were initially higher
areas, relative to the comparison area. Of the three locations exam- (Table B1) due to the nature of the primary water source (a small,
ined, location 5 had the lowest reported case numbers, leading seasonally dry river with several upstream influences) and (b) the
to low statistical power and wide confidence intervals. Unfortu- WSP implementation and post-implementation periods involved
nately, the low number of cases per municipality could have been several major upgrades to drinking water treatment processes
driven by outbreak events, whether stemming from drinking water, (Table B2). This suggests that ongoing, iterative improvement (a
foodborne, or other exposures. CatSalut was unable to share any core characteristic of the WSP approach) may help produce lasting
additional information about possible outbreaks during the after- effects on water quality. Location 4 also reported several “hard-
WSP study period. ware” upgrades and showed a number of significant improvements
The health outcome at location 1 (an overall 4% reduction in water quality and compliance, although no comparison area was
in acute gastroenteritis) corresponds fairly well with the magni- available to confirm these stemmed from the WSP.
tude of the 14% reduction in diarrhea found by pooling results The study design did evaluate differences among WSPs applied
across locations studied in Iceland (Gunnarsdóttir et al., 2012a). to (a) the production system (drinking water treatment plants or
At an individual level, diarrheal incidence declined significantly at groundwater treatment facilities) only, (b) the distribution system
five of seven observed locations in Iceland. Observation periods only, or (c) both the production and distribution systems. The par-
were longer, averaging nearly 12 years (compared to an average tial WSP at location 5 did not seem to have as strong an effect on
of six years for the locations in France and Spain), which may water quality, compliance, and health outcomes as the full WSP
have enabled enhanced observation of health impacts. The Iceland covering both the drinking water treatment plant(s) and the dis-
study was also an observational retrospective cohort study and, tribution system. This indicates that most WSP outcomes found
like our study, was limited by data nonconformity and lack of con- in our study stemmed from changes related to the drinking water
trol for confounding factors. Differences in our findings could also treatment plants. Locations where only the treatment plant was
stem from the practice of chlorination in France and Spain, which certified (locations 1, 2, and 4) did not show a discernable trend
may provide a residual protective effect against some waterborne when compared to the locations with a WSP covering both pro-
duction and distribution (location 3 and the location 5 full WSP or are neutral to WSP implementation, including the perception of
area). This is understandable because each location had its own aesthetics as a surrogate for water safety (enabling) and belief that
distinctive attributes. water should be free (limiting) (Omar et al., 2017). Factors found
Based on these case studies, a piecemeal approach to WSP to correlate with higher performing WSPs in Iceland included: fre-
implementation would not be recommended. Locations that pur- quent internal and external audits; a working WSP steering group;
chase water from or sell water to other suppliers might be good understanding of the WSP among staff; cooperation among
constrained to managing and improving only partial components senior management, health authorities, and the local government;
of the drinking water supply system. Further, household-level pip- and a training plan, especially for field workers (Gunnarsdóttir et al.,
ing and delivery systems, where some distribution monitoring 2012b). Factors inherent to WSP effectiveness across twelve Asian
samples are taken, can only be partially manipulated by a water Pacific countries included external financial support, formal poli-
utility-focused intervention (e.g., via residual chlorine dosing), and cies/regulations, and WSP-related record keeping, especially for
this lack of control over privately owned delivery systems could often-overlooked qualitative measures (Kumpel et al., In prep).
reduce WSP effectiveness. Where possible, the World Health Orga-
nization’s WSP guidance recommends considering all risks from 5. Conclusions
the source to the tap (Bartram et al., 2009). If prioritizing lim-
ited resources is necessary, and in the absence of other indications, We selected five case studies of WSP implementation outcomes
the evidence from this study suggests concentrating on the water that were likely to be generalizable to other chlorinated drinking
treatment facilities. water treatment systems in developed nations. WSP implementa-
tion improved water quality and compliance with relevant water
4.5. Limitations and future recommendations quality thresholds at a majority of locations (Tables 3 and 4).
Identified adverse effects were fewer, adding to the weight of evi-
Further study would help to elicit which particular attributes dence that WSPs offer operational performance benefits (String and
of the locations and/or WSPs most strongly enable beneficial out- Lantagne, 2016; Kot et al., 2015). Epidemiological analysis at one
comes. In this study, changes in water quality, compliance, and of three locations suggested that WSPs among large chlorinated
health were tied to the time period of the WSP intervention (from drinking water treatment systems in developed nations may reduce
the initiation of WSP team meetings to the ISO 22000 certifica- acute gastroenteritis incidence (Table 5), although validity of this
tion date). Specific causal investigations of identified changes were finding is limited by differences among the three observed case
not undertaken. Reported significant events that coincided with studies and potential sources of confounding. In particular, location
the study period can be found in Table B2. Measures taken during 5 showed an anomalous increase in acute gastroenteritis that was
the WSP implementation process typically included team forma- not clearly explained by water quality trends. Outcomes of WSPs
tion and meetings, documentation and posting of operating and should be expected to vary across locations, since the intervention
emergency procedures, initiation of special staff training sessions itself is adapted to the needs of each site. As such, the WSP approach
on risk management, designation of critical control points, instal- may translate to diverse water quality, compliance, and health out-
lation of online sensors, and occasionally equipment or treatment comes. Scenarios of stability or beneficial change identified in this
technology upgrades. study might be considered desirable among drinking water util-
Utility managers’ expectations of the degree of change matched ity managers. Future research should focus on eliciting the causal
fairly well with actual outcomes (Table B2), suggesting that aware- factors that enhance successful application of the WSP approach,
ness and deliberate intent or action to address specific water quality and on identifying best practices. Such information can be used
issues may have played a key role in creating that change. Expec- to improve individual utilities’ WSP implementation practices and
tations were gathered before data analysis results were shared, refine global WSP guidance.
although managers were likely able to make qualitative judgments
based on pre-existing familiarity and knowledge of the drinking
Declaration of interest
water treatment system. For example, managers at location 5 added
a reverse osmosis step to the drinking water treatment plant about
Suez, AESN, and the Royster Society of Fellows provided finan-
the same time the ISO 22000 was certified, fully intending to reduce
cial support for the research and article preparation. Suez-affiliated
levels of trihalomethanes to meet new European Union regulatory
staff (including Loret, Enault, Puigdomenech Serra, Pla Mateu, and
requirements.
Martin Alonso) were involved in study design, data collection, data
Some prominent transformations attributed to the WSP mech-
analysis and interpretation, report writing, and the decision to sub-
anism, as cited by utility managers during site visits, included
mit the article for publication. Dr. Bartram has served on Suez
formalization/documentation of risk management procedures and
committees as an unremunerated adviser.
the recognition of water as a food product among all levels of staff,
including those with primarily construction-oriented tasks, which
resulted in greater awareness of potential health risks during daily Acknowledgements
operations. Additional mechanisms suspected of affecting change
come from 2014 cost/benefit questionnaires administered among The authors greatly appreciate the assistance of all Suez
WSP-adopting Suez utilities, which cite both changes in human staff members who facilitated site visits and data collection for
behavior and improved reaction time to alarms for critical control this study. Special thanks extend to Catherine Galey and Pascal
points, especially chlorination (Loret et al., 2016). Beaudeau from Santé Publique France for providing access to public
In addition to those already mentioned, factors of interest for health surveillance data in France and advice regarding its inter-
future study might include sensitivity of outcomes to time since pretation. Suez, the Seine River Basin Agency (AESN; specifically
certification, age/condition of the water treatment and distribution Véronique Lahoussine), and the Royster Society of Fellows at UNC
system, and diversity/cohesiveness of the WSP team. The political Chapel Hill generously provided financial support for this study.
economy, including local community and organizational readiness We also are grateful to Maria Gunnarsdóttir and Olivier Schlosser
is expected to influence WSP outcomes (Kot et al., 2015). One study for reviewing and commenting on the draft manuscript. Statisti-
observing cultural influences on WSP implementation in India, cal analysis support was provided by the Odum Institute at the
Uganda, and Jamaica identified twelve themes that enable, limit, University of North Carolina at Chapel Hill.
Appendix A. Methodological detail
Table A1
Criteria for selection of paired comparison areas at locations 1, 3, and 5.
Study location Comparison area selection criteria
Locations 1 and 3 • Geographical proximity (same metropolitan area)

• Accessibility of water quality and health data over the study period
• Sufficient population (minimum of the same order of magnitude as the intervention area)
• Status of WSP implementation (municipality not served by an ISO22000-certified drinking water treatment facility or distribution network)
• Same (preferred) or similar water source
• Similar water treatment scheme
Location 5 • Geographical proximity (same metropolitan area)

• Accessibility of water quality and health data over the study period (required consent of one non-Suez utility to participate)
• Sufficient population (minimum of 60,000 inhabitants)
• Status of WSP implementation (municipality not served by an ISO22000-certified drinking water treatment facility or distribution network)
Table A2
Detection limits for left- and right-censored data applicable to monitoring equipment/test procedures used during the study period and data cleaning procedures.
Parameter Locations 1–4 (manual Location 1 Location 5 (manual monitoring) Location 5 (automated
monitoring) (automated sensors) sensors)
E. coli <1/100 ml set to 0 – Pres/100 ml set to 1; Abs/100 ml set to 0; –

Blanks set to missing (2 values)
Fecal Streptococci <1/100 ml set to 0 – – –
Total coliform <1/100 ml set to 0 – Pres/100 ml set to 1; Abs/100 ml set to 0; –
>2400 set to 2400; blanks set to missing (2
values); 4 values with decimal places rounded
to nearest whole colony
Heterotrophic plate <1/ml set to 0 – <1/ml set to 0; >300/ml set to 300 (one value) –
count (22 ◦ C)
Trihalomethanes <1 ␮g/l set to 0 – <2, <3.5, <4 ␮g/l set to 1, 1.75, 2 –
Bromate < 1 ␮g/l set to 0 – <0.5, <1.5, <2, <7.5, <10 ␮g/l set to 0, 0.75, 1, –
3.75, 5
Free chlorine <0.01 mg/l set to 0 <0.01 mg/l set to 0 <0.1, <0.15, <0.2 mg/l set to 0 <0.10 mg/l set to 0
Total chlorine <0.01 mg/l set to 0 – <0.05, <0.10, <0.2 mg/l set to 0 –
Aluminum <0.005 mg/l set to 0 – <20, <25 ␮g/l set to 10, 12.5 –
Total organic carbon <0.2 mg/l set to 0 – <1 mg/l set to 0 <0.2 mg/l set to 0
Turbidity <0.1 NTU set to 0 <0.02 NTU set to 0 <0.10, <0.20 NTU set to 0 <0.015, <0.10, <0.20
NTU set to 0
pH All values within range – All values within range –
Trichloroethylene + – – <0.5, <0.6, <1 ␮g/l set to 0 –
tetracholorethylene
Nickel – – <1, <4, <5, <8 ␮g/l set to 0, 2, 2.5, 4 –
Chromium VI – – – <10 ␮g/l set to 5
Iodine Index – – Blanks set to missing (176 values) –
Appendix B. Results detail
Table B1
A comparison of median water quality values (based on manually collected treated water samples) for each parameter across intervention locations prior to WSP implemen-
tation shows differing baseline water quality, attributable to variation in source water and treatment schema.
Parameter Unit Median water quality value prior to WSP implementation
Location 1 Location 2 Location 3 Location 4 Location 5 (full) Location 5 (partial)
E. coli MPN/100 ml 0 0 0 0 0 0
Fecal Streptococci MPN/100 ml 0 0 0 0 – –
Total coliform MPN/100 ml 0 0 0 0 0 0
Heterotrophic plate count (22 ◦ C) MPN/ml 0 0 1 1 1 0
Trihalomethanes ␮g/l 12.70 11.25 3.25 – 106.30 50.55
Bromate ␮g/l 0 0 0 0 2.40 –
Free chlorine mg/l 0.17 0.34 0.07 0.05 0.63 0.53
Total chlorine mg/l 0.23 0.40 0.11 0.05 0.71 0.80
Aluminum mg/l 0.011 0.026 0.007 0.069 0.045 0.070
Total organic carbon mg/l 1.103 1.10 0.58 0.80 1.80 2.10
Turbidity NTU 0 0.06 0.10 0.10 0.31 0.21
pH pH units 7.36 7.58 7.40 8.185 7.445 7.69
Table B2
All reported significant events occurring during the study period (prior to, during, and after WSP implementation) at each intervention location.
Location Date Event
1 2010 Merger with company on opposite side of river

July 2010 Ultrafiltration installed
1 Jan 2011 WSP implementation begins
31 Oct 2011 WSP implementation ends
2011–13 Interim plant manager
Dec 2015 – Jan 2016 Replaced ultrafiltration membranes to reduce breakage/cut down on bench testing
2 1 Jan 2006 WSP implementation begins

31 Mar 2007 WSP implementation ends
2007 Replacement of sand filtration with GAC filtration at one surface water plant
2010 UV treatment installed at one of three surface water plants
2011 UV treatment installed at another of the three surface water plants
3 2008 Renovation of distribution system (including replacement of main pipe)

2007–09 HACCP planning on production sites
2009 Gasoline spill at supermarket resulted in legal suits
2010 HACCP planning on the distribution network
2011 Perchlorate contamination event (one water source discontinued)
2011 Several upgrades implemented, including:
• Chlorine dioxide and chlorine and UV disinfection added before chlorine and GAC filtration to reinforce disinfection
capacity); previously discarded water when turbidity was high
• Online sensors added to network
• Water batch isolation
13 Nov 2012 WSP implementation begins
20 Dec 2013 WSP implementation ends
2014 Flood event spikes total organic carbon (treatment stopped; water did not reach consumers)

2007–08 Several upgrades implemented, including:
• Adaptation of ozonation to limit bromate formation
• Monitoring station installation upstream of the water intake and batch reservoir
• Emergency interconnection with other treatment plants
31 Mar 2008 WSP implementation ends

2008 Severe drought
Sept–Oct 2009 Gradual commissioning of reverse osmosis membranes
31 Dec 2009 WSP implementation ends
2010–2011 Improvements in sand filtration (post coagulation with ferric chloride) to address aluminum
Early 2013 Low water availability
May 2013 Replacement of all the sand filter beds
2013 Replacement of ultrafiltration membranes
Oct 2013 – Feb 2014 Switch to groundwater sources only due to dioxin contamination of surface water by wastewater treatment plant
2013–14 Raw water pH adjustment to optimize coagulation with aluminum sulfate
2014 Replacement of reverse osmosis membranes began
2014 Improvements to water mixing step before division into two treatment lines (increased amount treated by
ultrafiltration/reverse osmosis)
2013–2015 Adjustment of the ozone treatment to minimize formation of bromides
2015 Decrease use of chlorine dioxide in the pretreatment step
Oct–Dec 2015 Switch to groundwater sources only due to dioxin contamination of surface water by wastewater treatment plant
Table B3
Utility managers’ expectations of change at each location (“Should the water quality parameter increase, decrease, or stay the same following WSP implementation?”).
Parameter Location 1 Location 2 Location 3 Location 4 Location 5 (full/partial)
E. coli same same same same same

Fecal Streptococci same same same same –
Total coliform same same same same same
Heterotrophic plate count (22 ◦ C) same same same same same
Trihalomethanes decrease same same – decrease
Bromate same same same decrease increase
Free chlorine same same same same same
Total chlorine decrease same same same same
Aluminum same same same same decrease
Total organic carbon decrease same same same decrease
Turbidity decrease same same same decrease
pH same same same same same
Table B4
Minimum, maximum, and mean water quality values before and after WSP implementation, along with the number of samples, direction/magnitude of change, variance
ratio, and raw and adjusted p-value for regression model interaction term at location 1 intervention and comparison (“comp”) sites.
Parameter (units) N % non-zero Mean/variance Mean/variance Change in Variance ratio Model beta Model Model p-value
(before) (after) mean (F) *reduced by (time*site) p-value (adjusted)
≥4:1 (raw) *significant
E. coli (MPN/100 ml) 1390 0% 0 (0) 0 (0) 0 – – – –

(comp) 5744 0.03% 0 (0) 0.002 (0.005) 0.002 –
Fecal streptococci 1018 0% 0 (0) 0 (0) 0 – −1.40 1.000 1.000
(MPN/100 ml)
(comp) 5493 0.05% 0 (0) 0.002 (0.006) 0.002 –
Total coliform 1390 0.65% 0.019 (0.039) 0.001 (0.001) −0.018 39.00* −2.30 0.094 0.658
(MPN/100 ml)
(comp) 5737 0.64% 0.044 (1.157) 0.033 (0.342) −0.011 3.38
Heterotrophic plate 1122 35.7% 10.032 (1434) 2.949 (405) −7.083 3.54 −1.01 0.085 0.595
count (MPN/ml)
(comp) 5517 29.1% 7.716 (2277) 6.745 (1094) −0.971 2.08
THMs (␮g/l) 78 98.7% 12.298 (32.7) 11.153 (23.3) −1.145 1.40 −4.81 0.100 0.658
(comp) 96 72.9% 7.939 (103.2) 10.850 (110.4) 2.911 1.07
Bromate (␮g/l) 101 42.6% 0 (0) 1.244 (2.62) 1.244 – 14.81a <0.001 <0.001*
(comp) 210 40% 0.684 (3.67) 1.331 (2.38) 0.647 1.54
Free chlorine (mg/l) 1393 96.7% 0.190 (0.016) 0.190 (0.012) 0 1.33 −0.047 0.810 1.000
(comp) 5818 96.4% 0.377 (44.99) 0.423 (67.70) 0.046 1.50
Total chlorine (mg/l) 1392 97.9% 0.243 (0.019) 0.275 (0.438) 0.032 23.05 −0.218 0.259 1.000
(comp) 5181 98.2% 0.315 (0.280) 0.560 (111.6) 0.245 398.57
Aluminum (mg/l) 909 68.0% 0.013 (0) 0.014 (0) 0.001 – −0.016 <0.001 <0.001*
(comp) 3928 72.9% 0.010 (0) 0.026 (0) 0.016 –
Total organic carbon 407 100% 1.181 (0.076) 0.966 (0.032) −0.215 2.38 −0.014 0.893 1.000
(mg/l)
(comp) 205 100% 1.280 (0.172) 1.053 (0.072) −0.227 2.40
Turbidity (NTU) 1003 31.3% 0.243 (1.684) 0.148 (0.131) −0.095 12.85* −0.106 0.280 1.000
(comp) 4910 92.6% 0.220 (0.046) 0.225 (0.078) 0.005 1.70
pH (pH units) 936 100% 7.371 (0.020) 7.582 (0.010) 0.211 2.00 0.222 <0.001 <0.001*
(comp) 5753 100% 7.568 (0.020) 7.565 (0.024) −0.003 1.20
a
Tobit regression was used with highly censored data sets.
Table B5
Minimum, maximum, and mean water quality values before and after WSP implementation, along with the number of samples, direction/magnitude of change, variance
ratio, and raw and adjusted p-value for regression model time parameter among location 2 intervention site production samples.
Parameter (units) N % non-zero Mean/ variance Mean/ variance Change in Variance ratio Model beta Model Model p-value
(before) (after) mean (F) *reduced by (time) p-value (adjusted)
≥4:1 (raw) *signif.
E. coli (MPN/100 ml) 9069 0% 0 (0) 0 (0) 0 – – – –

Fecal streptococci 11183 0.02% 0 (0) 0 (0) 0 – – – –
(MPN/100 ml)
Total coliform 11170 0.06% 0 (0) 0.001 (0.001) 0.001 – – – –
(MPN/100 ml)
Heterotrophic plate 4610 22.5% 1.151 (110.6) 1.545 (146.2) 0.004 1.32 0.294 0.082 0.328
count (MPN/ml)
THMs (␮g/l) 535 94.8% 10.956 (44.91) 7.492 (42.435) −3.464 1.06 −3.464 0.006 0.024*
Bromate (␮g/l) 843 15.3% 0.975 (7.014) 0.467 (2.063) −0.508 3.40 −0.739a 0.307 1.000
Free chlorine (mg/l) 11713 100% 0.344 (0.003) 0.390 (16.076) 0.046 5358.67 0.046 0.311 1.000
Total chlorine (mg/l) 10008 100% 0.410 (0.003) 0.415 (0.003) 0.005 1.00 0.005 0.429 1.000
Aluminum (mg/l) 3615 87.6% 0.028 (0.001) 0.029 (0) 0.001 – 0.000 0.965 1.000
Total organic carbon 4134 99.5% 1.124 (0.068) 1.066 (0.076) −0.058 1.12 −0.057 0.257 1.000
(mg/l)
Turbidity (NTU) 10846 78.9% 0.064 (0.004) 0.061 (0.032) −0.003 8.00 −0.003 0.065 0.260
pH (pH units) 8126 100% 7.538 (0.035) 7.624 (0.012) 0.086 2.92 0.086 0.126 0.504
a
Tobit regression was used with highly censored data sets.
Table B6
Minimum, maximum, and mean water quality values before and after WSP implementation, along with the number of samples, direction/magnitude of change, and raw and
adjusted p-value for model interaction term at location 3 intervention and comparison (“comp”) sites. Data sets are from manual water samples unless otherwise noted as
coming from online sensors.
(before) (after) mean (F) *reduced by (time* site) p-value (adjusted)
E. coli (MPN/100 ml) 499 0.40% 0 (0) 0.793 (146.5) 0.793 – – – –

(comp) 301 0% 0 (0) 0 (0) 0 –
Fecal streptococci 500 0.20% 0 (0) 0.182 (9.09) 0.182 – – – –
(MPN/100 ml)
(comp) 303 0% 0 (0) 0 (0) 0 –
Total coliform 499 1.40% 0.004 (0.004) 1.128 (170.2) 1.124 42550 7.07 <0.001 <0.001*
(MPN/100 ml)
(comp) 302 1.66% 0.062 (0.334) 0.014 (0.014) −0.048 23.86*
Table B6 (Continued)
(before) (after) mean (F) *reduced by (time* site) p-value (adjusted)
Heterotrophic plate 523 56.4% 14.689 (3302) 60.456 45.767 33.25 2.09 0.019 0.019*
count (MPN/ml) (109796)
(comp) 303 38.0% 11.230 (2368) 5.366 (873.1) −5.864 2.71
THMs (␮g/l) 41 82.9% 4.620 (23.32) 6.245 (14.10) 1.625 1.65 1.39a 0.275 0.275
Bromate (␮g/l) 12 0% 0 (0) 0 (0) 0 – – – –
Free chlorine (mg/l) 361 80.6% 0.091 (0.011) 0.193 (3.312) 0.102 301.1 0.109 0.422 0.422
(comp) 158 84.8% 0.088 (0.006) 0.073 (0.004) −0.015 1.50
Free chlorine (sensors) 419982 97.6% 0.125 (0.018) 0.140 (0.006) 0.015 3.00 0.015a <0.001 <0.001*
Total chlorine (mg/l) 189 86.2% 0.114 (0.010) 0.115 (0.014) 0.001 1.40 0.029 0.289 0.289
(comp) 14 57.1% 0.027 (0.003) 0.020 (0.001) −0.007 3.00
Aluminum (mg/l) 92 93.5% 0.008 (0) 0.006 (0) −0.002 – −0.002 0.305 0.305
(comp) 17 47.1% 0.001 (0) 0.002 (0) 0.001 –
Total organic carbon 130 100% 0.734 (0.194) 0.986 (0.228) 0.243 1.18 0.199 0.069 0.069
(mg/l)
(comp) 51 100% 0.222 (0.004) 0.343 (0.022) 0.121 5.5
Turbidity (NTU) 411 65.2% 0.169 (0.142) 0.266 (1.419) 0.097 9.99 0.185 0.058 0.058
(comp) 224 45.5% 0.182 (0.208) 0.094 (0.032) −0.088 6.50*
Turbidity (sensors) 436293 99.4% 0.282 (0.361) 0.336 (0.341) 0.054 1.06 0.055a <0.001 <0.001*
pH (pH units) 83 100% 7.419 (0.017) 7.406 (0.014) −0.013 1.21 0.031 0.256 0.256
(comp) 246 100% 7.889 (0.009) 7.853 (0.007) −0.036 1.29
a
Where comparison data sets were not available, the beta reported is for time only.
Table B7
adjusted p-value for model time parameter at the location 4 intervention site.
(before) (after) mean (F) *reduced by (time) p-value (adjusted)
E. coli (MPN/100 ml) 3165 0.19% 0.001 (0.001) 0.005 (0.036) 0.004 36.00 – – –
Fecal streptococci 3529 0.11% 0.003 (0.003) 0 (0) −0.003 – – – –
(MPN/100 ml)
Total coliform 3515 0.94% 0.010 (0.017) 0.030 (0.304) 0.020 17.88 1.059 0.086 0.258
(MPN/100 ml)
Heterotrophic plate 3307 39.13% 10.375 (9170.4) 3.367 (327.3) −7.008 28.02* −1.120 <0.001 <0.001*
count (MPN/ml)
Bromate (␮g/l) 742 19.27% 2.446 (30.44) 1.643 (22.27) −0.803 1.37 −0.257 0.619 1.000
Free chlorine (mg/l) 3440 75.15% 0.134 (0.023) 0.147 (0.020) 0.013 1.15 0.022 0.002 0.006*
Total chlorine (mg/l) 3476 73.16% 0.130 (0.027) 0.166 (0.409) 0.036 15.15 0.042 0.013 0.039*
Aluminum (mg/l) 2589 91.35% 0.069 (0.002) 0.041 (0.000) 0.028 – −0.030 <0.001 <0.001*
Total organic carbon 562 99.11% 0.828 (0.070) 0.812 (0.185) −0.016 2.64 −0.017 0.808 1.000
(mg/l)
Turbidity (NTU) 3401 82.95% 0.167 (0.063) 0.130 (0.072) −0.037 1.14 −0.036 0.003 0.009*
pH (pH units) 893 100% 8.133 (0.086) 8.103 (0.105) −0.030 1.22 −0.118 <0.001 <0.001*
Table B8
adjusted p-value for model interaction term at location 5 intervention and comparison (“comp”) sites. Data sets are from manual water samples unless otherwise noted as
coming from online sensors.
(before) (after) mean (F) *reduced by (time*site) p-value (adjusted)
≥4:1 (raw) *signif.
E. coli (MPN/100 ml) 1641 0.1% 0.002 (0.002) 0 (0) −0.002 – – – –

(partial)a 1992 0.05% 0 (0) 0.001 (0.001) 0.001 – – – –
(comp) 360 0% 0 (0) 0 (0) 0 –
Total coliform 1641 2.6% 0.181 (4.69) 2.942 (7050) 2.761 1502.9 – – –
(MPN/100 ml)
(partial) 1992 0.8% 0.006 (0.008) 0.157 (22.08) 0.151 2760.4 – – –
(comp) 329 1.2% 0 (0) 0.208 (5.56) 0.208 –
Heterotrophic plate 64 45.3% 18.545 (3121) 1.619 (20.29) −16.926 153.8* −3.083 0.031 0.062
count (MPN/ml)
(partial) 144 38.2% 3.708 (350.1) 3.896 (937.8) 0.188 2.68 −0.596 0.707 1.000
(comp) 156 40.4% 620.2 (7.42 M) 1335.9 (237 M) 715.6 32.00
THMs (␮g/l) 4683 100% 112.508 11.817 (51.84) −100.691 46.71* −96.168 <0.001 <0.001*
(2421.5)
(partial) 138 100% 51.177 (157.8) 52.095 (275.1) 0.918 1.74 5.124 0.253 0.759
(comp) 35 100% 57.133 (313.6) 48.012 (219.3) −9.121 1.43
Bromate (␮g/l) 2150 99.5% 3.127 (7.76) 5.384 (26.60) 2.257 3.43 2.263 <0.001 <0.001*
(partial) 71 100% – 4.940 (0.254) – – – – –
(comp) 6 100% 5 (0) 5 (0) 0 –
Free chlorine (mg/l) 5034 98.1% 0.529 (0.002) 0.520 (0) −0.009 – −0.072 0.079 0.158
(partial) 11354 99.7% 0.620 (0.030) 0.588 (0.037) −0.032 1.23 −0.096 0.004 0.012*
(comp) 325 97.9% 0.462 (0.042) 0.545 (0.038) 0.083 1.11
Free chlorine (sensors) 283678 100% 1.068 (0.036) 0.893 (0.025) −0.175 1.44 −0.175b <0.001 <0.001*
Total chlorine (mg/l) 4970 99.6% 0.722 (0.062) 0.643 (0.047) −0.079 1.32 −0.022 0.681 1.000
(partial) 11220 100% 0.799 (0.032) 0.771 (0.043) −0.028 1.34 0.028 0.552 1.000
(comp) 13 92.3% 0.214 (0.001) 0.157 (0.011) −0.057 11.00
Aluminum (␮g/l) 6699 100% 54.010 (1638) 36.317 (950) −17.693 1.72 −21.070 0.121 0.242
(partial) 141 100% 72.011 (1355) 60.609 (706) −11.402 1.92 −14.863 0.319 0.957
(comp) 79 100% 47.650 (1658) 50.169 (714.8) 2.519 2.32
Total organic carbon 67 76.1% 1.924 (0.585) 0.945 (0.623) −0.979 1.06 −1.067 <0.001 <0.001*
(mg/l)
(partial) 143 99.3% 2.085 (0.220) 2.330 (0.383) 0.245 1.74 0.157 0.428 1.000
(comp) 35 100% 2.467 (0.200) 2.492 (0.588) 0.025 2.94
Total organic carbon 18925 93.1% 2.760 (0.588) 0.918 (0.187) −1.842 3.14 −1.842 b
<0.001 <0.001*
(sensors)
Turbidity (NTU) 1592 65.1% 0.369 (0.201) 0.162 (0.092) −0.207 2.18 −0.234 <0.001 <0.001*
(partial) 1990 64.6% 0.208 (0.165) 0.093 (0.140) −0.115 1.18 −0.143 0.011 0.033*
(comp) 327 93% 0.312 (0.053) 0.344 (0.573) 0.032 10.81
Turbidity (sensors) 87054 77.5% 0.257 (0.008) 0.120 (6.510) −0.137 813.75 −0.138b <0.001 <0.001*
pH (pH units) 503 100% 7.475 (0.044) 7.440 (0.056) −0.035 1.27 −0.219 <0.001 <0.001*
(partial) 797 100% 7.683 (0.031) 7.577 (0.056) −0.106 1.81 −0.289 <0.001 <0.001*
(comp) 325 100% 7.639 (0.019) 7.820 (0.035) 0.181 1.84
a
b
Where comparison data sets were not available, the beta reported is for time only.
528
Table B9
Detailed results of compliance analysis involving relevant internal and external thresholds. Both raw and adjusted p-values are shown.
Location Parameter Data Source Thresholdb Threshold Source % % Chi-SquareC or P-value

(Definition of Non-compliance Non-compliance Fisher’s Exact (adjusted) *signif.
exceedance) (before) (after) TestF
1 E. coli Prod/Dist >0 MPN/100 ml EU and French quality limit 0% 0% – –

Total coliform Prod/Dist >0 MPN/100 ml EU and French quality limit 1.17% 0.14% 673F 0.019(0.038)*
THMs Production >30 ␮g/l more than WSP operational limit 0% 0% – –
10% of the time
Production ≥50 ␮g/l Suez internal recommended practice 0% 0% – –
for plant outlet
Distribution >100 ␮g/l EU and French quality limit for 0% 0% – –
K.E. Setty et al. / International Journal of Hygiene and Environmental Health 220 (2017) 513–530
network
Bromate Prod/Dist >10 ␮g/l EU and French quality limit 0% 0% – –
Aluminum Prod/Dist >200 ␮g/l EU and French quality reference 0% 0% – –
Production >100 ␮g/l Suez internal recommended practice 0% 0% – –
and WSP operational limit
Turbidity Production >0.5 NTU French quality reference for plant 1.03% 0.55% 192F 1.000(1.000)
outlet
Production >1 NTU French quality limit for plant outlet 0% 0% – –
Distribution >2 NTU French quality reference for tap 4.04% 0.98% 309F 0.020(0.040)*
pH Prod/Dist <6.5 or >9 pH units EU and French quality limit 0% 0% – –
2 E. coli Production >0 MPN/100 ml EU and French quality limit 0% 0% – –

Total coliform Production >0 MPN/100 ml EU and French quality limit 0% 0.09% 3188F 0.203(1.000)
THMs Production ≥50 ␮g/l Suez internal recommended practice 0% 0.21% 56F 1.000(1.000)
for plant outlet
Bromate Production >10 ␮g/l EU and French quality limit 0.68% 0.29% 145F 0.435(1.000)
Free chlorine Production ≤0.05 or ≥0.7 WSP critical limit 0.03% 0.02% 3202F 1.000(1.000)
Aluminum Production >200 ␮g/l EU and French quality reference 0% 0.03% 674F 1.000(1.000)
Production >100 ␮g/l Suez internal recommended practice 0.45% 0.37% 671F 0.734(1.000)
Total organic carbon Production >2 mg/l French quality reference 0.08% 0.03% 1224F 0.505(1.000)
Turbidity Production >0.5 NTU French quality reference for plant 0.04% 0.26% 2776F 0.025(0.225)
outlet
Production >1 NTU French quality limit for plant outlet 0% 0.07% 2777F 0.349(1.000)
pH Production <6.5 or >9 pH units EU and French quality limit 0% 0% – –
3 E. coli Prod/Dist >0 MPN/100 ml EU and French quality limit 0% 0.73% 224F 0.504(1.000)
Total coliform Prod/Dist >0 MPN/100 ml EU and French quality limit 0.44% 2.19% 224F 0.135(0.810)
THMs Production ≥50 ␮g/l Suez internal recommended practice 0% 0% – –
for plant outlet
Distribution >100 ␮g/l EU and French quality limit for 0% 0% – –
network
Free chlorine Production(sensors) <0.05 mg/l WSP critical limit for chlorination 7.53% 2.81% 4783C <0.001(<0.001)*
Aluminum Prod/Dist >200 ␮g/l EU and French quality reference 0% 0% – –
Prod/Dist >100 ␮g/l Suez internal recommended practice 0% 0% – –
Total organic carbon Prod/Dist >2 mg/l French quality reference 2.08% 1.41% 47F 1.000(1.000)
Turbidity Production(sensors) >0.5 NTU *simplified Suez internal recommended practice 4.72% 4.73% 0.0004C 0.983(1.000)
Production >0.5 NTU French quality reference for plant 2.38% 0% 41F 0.452(1.000)
outlet
Production >1 NTU French quality limit for plant outlet 0% 0% – –
Distribution >2 NTU French quality reference for tap 0.65% 1.23% 154F 1.000(1.000)
pH Prod/Dist <6.5 or >9 pH units EU and French quality limit 0% 0% – –
4 E. coli Prod/Dist >0 MPN/100 ml EU and French quality limit 0.13% 0.21% 797F 1.000(1.000)
Total coliform Prod/Dist >0 MPN/100 ml EU and French quality limit 0.78% 1.02% 0.47C 0.495(1.000)
Bromate Prod/Dist >10 ␮g/l EU and French quality limit 13.11% 4.55% 8.24C 0.004(0.040)*
Free chlorine Production <0.2 mg/l *simplified WSP operational limit (surface water) 14.79% 7.65% 11C 0.001(0.010)*
Production <0.15 mg/l WSP operational limit (groundwater) 44.83% 55.70% 1.01C 0.316(1.000)
Production <0.1 mg/l WSP critical limit (surface water) 2.57% 2.19% 0.13C 0.718(1.000)
Production <0.05 mg/l *simplified WSP critical limit (groundwater) 6.90% 18.99% 27F 0.149(1.000)
Aluminum Prod/Dist >200 ␮g/l EU and French quality reference 0.28% 0% 721F 0.078(0.780)
Prod/Dist >100 ␮g/l Suez internal recommended practice 23.79% 1.18% 384C <0.001(0.001)*
Total organic carbon Production >2 mg/l French quality reference 0% 1.79% 172F 0.107(1.000)
Turbidity Production >0.5 NTU French quality reference for plant 2.84% 4.12% 1.03C 0.309(1.000)
outlet
Production >1 NTU French quality limit for plant outlet 0.28% 1.15% 351F 0.270(1.000)
Distribution >2 NTU French quality reference for tap 0.12% 0.12% 814F 1.000(1.000)
pH Prod/Dist <6.5 or >9 pH units EU and French quality limit 0% 0.13% 110F 1.000(1.000)
5 (full WSP )a E. coli Prod/Dist >0 MPN/100 ml EU and Spanish quality limit 0.24% 0% 822F 0.500(1.000)
K.E. Setty et al. / International Journal of Hygiene and Environmental Health 220 (2017) 513–530
Total coliform Prod/Dist >0 MPN/100 ml EU and Spanish quality limit 4.49% 0.61% 787F <0.001(<0.001)*
THMs Production ≥50 ␮g/l Suez recommendation for plant outlet 94.92% 0.32% 4217C <0.001(<0.001)*
(in France)
Distribution >100 ␮g/l EU and Spanish quality limit for 84.21% 0% 6F <0.001(<0.001)*
network
Bromate Prod/Dist >10 ␮g/l EU and Spanish quality limit 2.82% 4.14% 69F 1.000(1.000)
Production ≥7.5 ␮g/l *simplified Regional internal recommended 7.04% 6.59% 66F 0.808(1.000)
practice/WSP critical limit
Free chlorine Distribution >1 mg/l Spanish quality limit in network 1.50% 0.85% 3.63C 0.057(0.399)
Production(sensors) <0.5 mg/l Catalunya/WSP operational limit 0.50% 0.45% 3.85C 0.050(0.398)
Production(sensors) <0.2 mg/l *simplified Catalunya/WSP critical limit for 0.13% 0.06% 35.9C <0.001(<0.001)*
chlorination
Total chlorine Distribution >2 mg/l Spanish quality limit in network 0.06% 0% 3372F 1.000(1.000)
Aluminum Prod/Dist >200 ␮g/l EU and Spanish and regional quality 0.55% 0.06% 1450F <0.001(0.003)*
limit
Turbidity (after sand Production (sensors) ≥0.75 NTU WSP operational limit 0.12% 1.27% 3473F <0.001(<0.001)*
filters) Production (sensors) ≥1 NTU*simplified WSP critical limit 0.03% 0.38% 3476F <0.001(<0.001)*
Turbidity (at outlet) Production (sensors) >0.5 NTU Spanish quality reference 1.44% 0.24% 402C <0.001(<0.001)*
Production (sensors) >1 NTU EU and Spanish quality limit for plant 0.01% 0.04% 15071F 0.156(0.936)
outlet
Turbidity (in network) Distribution >5 NTU Spanish quality limit 0.13% 0% 771F 0.485(1.000)
pH Prod/Dist <6.5 or >9.5 pH units EU and Spanish quality limit 0% 0% – –
Iodine adsorption Production <550 mg I2/g WSP operational level for granular – 20.56% – –
activated carbon
Production <400 mg I2/g WSP critical level for granular – 9.25% – –
activated carbon
Nickel Production >20 ␮g/l EU and Spanish quality limit (for tap) 0.43% 0% 1400F <0.001(0.002)*
Chromium VI Production >50 ␮g/l EU and Spanish quality limit 0% 0% – –
Trichloro- Production >10 ␮g/l EU and Spanish quality limit 0% 0% – –
ethylene + tetrachloro-
ethylene
5(partial WSP1 ) E. coli Distribution >0 MPN/100 ml EU and Spanish quality limit 0% 0.09% 826F 1.000(1.000)
Total coliform Distribution >0 MPN/100 ml EU and Spanish quality limit 0.48% 1.11% 822F 0.147(0.882)
THMs Distribution >100 ␮g/l EU and Spanish quality limit for 0% 1.02% 40F 1.000(1.000)
network
Bromate Distribution >10 ␮g/l EU and Spanish quality limit 0% 0% – –
Free chlorine Distribution >1 mg/l Spanish quality limit in network 0.35% 0.33% 0.05C 0.827(1.000)
Total chlorine Distribution >2 mg/l Spanish quality limit in network 0% 0% – –
Aluminum Distribution >200 ␮g/l EU and Spanish and regional quality 0% 1.04% 45F 1.000(1.000)
limit
Turbidity Distribution >5 NTU Spanish quality limit 0.12% 0.09% 825F 1.000(1.000)
pH Distribution <6.5 or >9.5 pH units EU and Spanish quality limit 0% 0% – –
a
b
Some thresholds were simplified to enable comparison with historical data. Self-reported records might differ based on the application of time duration, repeat sampling, or equipment validation procedures.
529
References International Organization for Standardization (ISO), 2005. Food safety

management systems – requirements for any organization in the food chain,
Arena, C., Amoros, J.P., Vaillant, V., Ambert-Balay, K., Chikhi-Brachet, R., Jourdan-Da Retrieved from http://www.iso.org/iso/home/standards/management-
Silva, N., Varesi, L., Arrighi, J., Souty, C., Blanchon, T., Falchi, A., Hanslik, T., 2014. standards/iso22000.htm.
Acute diarrhea in adults consulting a general practitioner in France during Kumpel, E., Delaire, C., Peletz, R., Kisiangani, J., Rinehold, A., DeFrance, J.,
winter: incidence, clinical characteristics, management and risk factors. BMC Sutherland, D., Khush, R. (In prep). Evaluating Water Safety Plans: Lessons
Infect. Dis. 14, 574, http://dx.doi.org/10.1186/s12879-014-0574-4. Learned from Assessments in the Asia-Pacific Region.
Bartram, J., Corrales, L., Davidson, A., Deere, D., Drury, D., Gordon, B., Howard, G., Kot, M., Castleden, H., Gagnon, G.A., 2015. The human dimension of water safety
Rinehold, A., Stevens, M., 2009. Water Safety Plan Manual: Step-by-step Risk plans: a critical review of literature and information gaps. Environ. Rev. 23,
Management for Drinking-water Suppliers. World Health Organization, 24–29, http://dx.doi.org/10.1139/er-2014-0030.
Geneva, Switzerland. Kowalzik, F., Riera-Montes, M., Verstraeten, T., Zepp, F., 2015. The burden of
Beaudeau, P., de Valk, H., Vaillant, V., Mannschott, C., Tillier, C., Mouly, D., Ledrans, norovirus disease in children in the European Union. Pediatric Infect. Dis. J. 34
M., 2008. Lessons learned from ten investigations of waterborne (3), 229–234, http://dx.doi.org/10.1097/INF.0000000000000546.
gastroenteritis outbreaks, France 1998–2006. J. Water Health 6 (4), 491–503, Lockhart, G., Oswald, W.E., Hubbard, B., Medlin, E., Gelting, R.J., 2014. Development
http://dx.doi.org/10.2166/wh.2008.051. of indicators for measuring outcomes of water safety plans. J. Water Sanit. Hyg.
Beaudeau, P., Le Tertre, A., Zeghnoun, A., Zanobetti, A., Schwartz, J., 2012. A time Dev. 4 (1), 171–181.
series study of drug sales and turbidity of tap water in Le Havre, France. J. Lopman, B., Reacher, M., van Duijnhoven, Y., Hanon, F.X., Brown, D., Koopmans, M.,
Water Health 10 (2), 221–235, http://dx.doi.org/10.2166/wh.2012.157. 2003. Viral gastroenteritis outbreaks in Europe, 1995–2000. Emerg. Infect. Dis.
Bounoure, F., Beaudeau, P., Mouly, D., Skiba, M., Lahiani-Skiba, M., 2011. Syndromic 9 (1), 90–96, http://dx.doi.org/10.3201/eid0901.020184.
surveillance of acute gastroenteritis based on drug consumption. Epidemiol. Loret, J.F., Blaudin de Thé, C., Martin Alonso, J., Puigdomenech Serra, C., Kayser, G.,
Infect. 139, 1388–1395, http://dx.doi.org/10.1017/S095026881000261X. Bartram, J., 2016. Assessing the costs and benefits of water safety plans. In:
Centers for Disease Control and Prevention (CDC), 2011. A Conceptual Framework Paper Presented at the IWA World Water Congress, Brisbane, Australia.
to Evaluate the Impacts of Water Safety Plans. Atlanta, Retrieved from http:// Ministère de La Santé et des Solidarités, 2007. Arrêté du 11 janvier 2007 relatif aux
www.cdc.gov/nceh/ehs/gwash/publications.htm. limites et références de qualité des eaux brutes et des eaux destinées à la
Chikhi-Brachet, R., Bon, F., Toubiana, L., Pothier, P., Nicolas, J.C., Flahault, A., Kohli, consommation humaine mentionnées aux articles R. 1321-2, R. 1321-3, R.
E., 2002. Virus diversity in a winter epidemic of acute diarrhea in France? J. 1321-7 et R. 1321-38 du code de la santé publique. Journal Officiel De La
Clin. Microbiol. 40 (11), 4266–4272. République Française, NOR: SANP0720201A.
European Union (EU), 1998. Council Directive 98/83/EC of 3 November 1998 on the Ministerio de la Presidencia, 2003. Real decreto 140/2003, de 7 de febrero, por el
quality of water intended for human consumption. Off. J. Eur. Commun., L que se establecen los criterios sanitarios de la calidad del agua de consumo
330/32. humano. BOE Núm. 45.
European Union (EU), 2015. Commission Directive (EU) 2015/1787 of 6 October Omar, Y.Y., Parker, A., Smith, J.A., Pollard, S.J.T., 2017. Risk management for
2015 amending Annexes II and III to Council Directive 98/83/EC on the quality drinking water safety in low and middle income countries – cultural influences
of water intended for human consumption. Off. J. Eur. Union, L 260/6. on water safety plan (WSP) implementation in urban water utilities. Sci. Total
Flahault, A., Hanslik, T., 2010. Epidemiology of viral gastroenteritis in France and Environ. 576, 895–906.
Europe. Bulletin De l’Academie Nationale De Medecine 194 (8), 1415–1425, String, G., Lantagne, D., 2016. A systematic review of outcomes and lessons learned
discussion 1424–1425. from general, rural, and country-specific Water Safety Plan implementations.
Gelting, R.J., Delea, K., Medlin, E., 2012. A conceptual framework to evaluate the Water Sci. Technol.: Water Supply 16 (6), 1580–1594, http://dx.doi.org/10.
outcomes and impacts of water safety plans. J. Water Sanit. Hyg. Dev. 2 (2), 2166/ws.2016.073.
103–111. Therre, H., de Valk, H., Vaillant, V., Beaudeau, P., Mouly, D., 2008. Assessment of
Generalitat de Catalunya, Departament de Salut Direcció General de Salut Pública Waterborne Outbreaks Investigated in France Since 1998 and Key
(2005). Vigilància i Control Sanitaris de les Aigües de Consum Humà de Recommendations. French Institute for Public Health Surveillance (InVS),
Catalunya. Saint-Maurice, France, Retrieved from: http://opac.invs.sante.fr/doc num.
Gunnarsdóttir, M.J., Gardarsson, S.M., Elliott, M., Sigmundsdottir, G., Bartram, J., php?explnum id=8081.
2012a. Benefits of water safety plans: microbiology, compliance, and public Van Cauteren, D., De Valk, H., Vaux, S., Le Strat, Y., Vaillant, V., 2012. Burden of
health. Environ. Sci. Technol. 46 (14), 7782–7789, http://dx.doi.org/10.1021/ acute gastroenteritis and healthcare-seeking behaviour in France: a
es300372h. population-based study. Epidemiol. Infect. 140 (4), 697–705, http://dx.doi.org/
Gunnarsdóttir, M.J., Gardarsson, S.M., Bartram, J., 2012b. Icelandic experience with 10.1017/S0950268811000999.
water safety plans. Water Sci. Technol. 65 (2), 277–288. World Health Organization (WHO), International Water Association (IWA), 2015. A
Gunnarsdóttir, M.J., 2012. Safe Drinking Water: Experience with Water Safety Practical Guide to Auditing Water Safety Plans. WHO and IWA, Geneva,
Plans and Assessment of Risk Factors in Water Supply (Doctoral Dissertation). Switzerland.
Faculty of Civil and Environmental Engineering, University of Iceland, World Health Organization (WHO), 2010. Burden of Disease (in DALYs)
Reykjavik. Attributable to Water, Sanitation and Hygiene (’000), by Disease and WHO
Member State, 2004. Global Health Observatory (GHO) Data, Retrieved from:
http://www.who.int/gho/phe/water sanitation/burden/en/index2.html.

International Journal of Hygiene and Environmental Health

Uploaded by

Copyright:

Available Formats

International Journal of Hygiene and Environmental Health

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

International Journal of Hygiene and Environmental Health

Uploaded by

Copyright:

Available Formats

International Journal of Hygiene and Environmental Health 220 (2017) 513–530

Contents lists available at ScienceDirect

International Journal of Hygiene and

Water quality, compliance, and health outcomes among utilities

1. Introduction (EU) 2015/1787EU, 2015) and may be required as early as 2018.

4 Before: 1 Jan 2003 – 31 Dec (data not available)

Location Parameter Change post-intervention Threshold deﬁnitionb Source

4 Bromate Increase in compliance >10 ␮g/l EU and French quality limit

5 (partial WSPa ) None

4.2. Health outcomes pathogens, in contrast to the unchlorinated treatment schemes in

Appendix A. Methodological detail

Study location Comparison area selection criteria

Locations 1 and 3 • Geographical proximity (same metropolitan area)

Location 5 • Geographical proximity (same metropolitan area)

E. coli <1/100 ml set to 0 – Pres/100 ml set to 1; Abs/100 ml set to 0; –

Appendix B. Results detail

Parameter Unit Median water quality value prior to WSP implementation

Location 1 Location 2 Location 3 Location 4 Location 5 (full) Location 5 (partial)

Location Date Event

1 2010 Merger with company on opposite side of river

2 1 Jan 2006 WSP implementation begins

3 2008 Renovation of distribution system (including replacement of main pipe)

4 01 Jan 2007 WSP implementation begins

5 1 Jan 2008 WSP implementation begins

Parameter Location 1 Location 2 Location 3 Location 4 Location 5 (full/partial)

E. coli same same same same same

E. coli (MPN/100 ml) 1390 0% 0 (0) 0 (0) 0 – – – –

E. coli (MPN/100 ml) 9069 0% 0 (0) 0 (0) 0 – – – –

E. coli (MPN/100 ml) 499 0.40% 0 (0) 0.793 (146.5) 0.793 – – – –

E. coli (MPN/100 ml) 1641 0.1% 0.002 (0.002) 0 (0) −0.002 – – – –

Location Parameter Data Source Thresholdb Threshold Source % % Chi-SquareC or P-value

1 E. coli Prod/Dist >0 MPN/100 ml EU and French quality limit 0% 0% – –

2 E. coli Production >0 MPN/100 ml EU and French quality limit 0% 0% – –

References International Organization for Standardization (ISO), 2005. Food safety

You might also like