Review is available at conformant HTML ... - US Forest … States. (Colford et al. 2006 ... Proxy...

Review of Epidemiological Studies of Drinking-Water Turbidity in Relation toAcute Gastrointestinal IllnessAnneclaire J. De Roos,1 Patrick L. Gurian,2 Lucy F. Robinson,3 Arjita Rai,1 Issa Zakeri,3 and Michelle C. Kondo41Department of Environmental and Occupational Health, Dornsife School of Public Health, Drexel University, Philadelphia, Pennsylvania, USA2Department of Civil, Architectural, and Environmental Engineering, College of Engineering, Drexel University, Philadelphia, Pennsylvania, USA3Department of Epidemiology and Biostatistics, Dornsife School of Public Health, Drexel University, Philadelphia, Pennsylvania, USA4Northern Research Station, Forest Service, U S. Department of Agriculture (USDA Forest Service), Philadelphia, Pennsylvania, USA

BACKGROUND: Turbidity has been used as an indicator of microbiological contamination of drinking water in time-series studies attempting to discernthe presence of waterborne gastrointestinal illness; however, the utility of turbidity as a proxy exposure measure has been questioned.

OBJECTIVES:We conducted a review of epidemiological studies of the association between turbidity of drinking-water supplies and incidence of acutegastrointestinal illness (AGI), including a synthesis of the overall weight of evidence. Our goal was to evaluate the potential for causal inference fromthe studies.METHODS:We identified 14 studies on the topic (distinct by region, time period and/or population). We evaluated each study with regard to modelingapproaches, potential biases, and the strength of evidence. We also considered consistencies and differences in the collective results.DISCUSSION: Positive associations between drinking-water turbidity and AGI incidence were found in different cities and time periods, and with bothunfiltered and filtered supplies. There was some evidence for a stronger association at higher turbidity levels. The studies appeared to adequatelyadjust for confounding. There was fair consistency in the notable lags between turbidity measurement and AGI identification, which fell between 6and 10 d in many studies.

CONCLUSIONS: The observed associations suggest a detectable incidence of waterborne AGI from drinking water in the systems and time periods stud-ied. However, some discrepant results indicate that the association may be context specific. Combining turbidity with seasonal and climatic factors,additional water quality measures, and treatment data may enhance predictive modeling in future studies. https://doi.org/10.1289/EHP1090

IntroductionTreatment of drinking water by utilities has remarkablyimproved public health by decreasing the risk of waterborneinfection in regions served by these water supplies (Ford2016). Nevertheless, acute gastrointestinal illness (AGI) inci-dence attributable to community water systems has been esti-mated to be as high as 12–16:4million cases annually in theUnited States. (Colford et al. 2006; Messner et al. 2006).There are substantial challenges to the quantification of water-borne illness from drinking water. Microbiological pathogensare diverse, including viruses, bacteria, and protozoa; there-fore, it is not realistic to routinely measure all of the patho-gens that contribute to AGI within any particular watersystem (e.g., for screening purposes). Proxy measures maygive a global indication of potential microbiological contami-nation, with the tradeoff of nonspecificity.

Turbidity, a measure of the cloudiness of water, has oftenbeen used as a proxy for microbiological contamination. Somestudies have found that turbidity was correlated with microbio-logical contamination in source water and filtered drinking water(LeChevallier et al. 1991a, 1991b). However, turbidity is a non-specific measure of the scattering of light by particles suspendedin water and is thus influenced by various types of particulates,including silt, clay, and organic matter, that can differ in

prevalence among water systems (Burlingame et al. 1998).Particulates are of concern in water systems not only becausemicrobes themselves are particulates, but more importantlybecause other, nonmicrobe particulates may serve as indicators ofpathogen presence (e.g., runoff may produce spikes in turbidityas both sediments and pathogens are washed into source waters),and furthermore, these other particulates may protect the patho-gens from disinfectants. Given that turbidity is a nonspecific indi-cator of particulates, turbidity itself is not directly linked tohealth concerns and is not expected to be a consistent indicator ofthe microbiological quality of water.

Despite its limitations, turbidity has been judged to be ofsufficient relevance to be included in the drinking-water regu-lations of many developed countries. The U.S. rules surround-ing turbidity (U.S. EPA 2015), which have been amended andrevised since the 1989 Surface Water Treatment Rule, cur-rently require filtered water supplies to conduct monitoring forturbidity at each individual filter at 15-min intervals; the rulesspecify that turbidity of combined filter effluent should be≤0:3 nephelometric turbidity units ðNTUÞ in at least 95% of themeasurements taken each month and no single measurementshould exceed 1NTU. The U.S. requirements are less strin-gent for unfiltered water supplies where the source water isconsidered sufficiently protected against microbiological contami-nation, allowing a maximum turbidity of 5NTU. Although regula-tions surrounding turbidity differ between developed countries,rules establishing maximum limits are in place in many regionsworldwide, including European countries, Canada, Australia,Japan, and South Africa (CWWA 2002). In several regions, waterutilities strive to optimize filtration performance to achieve turbid-ity levels below the regulatory limits (CWWA 2002), such as thevoluntary goal of 0:1NTU set by the American Water WorksAssociation Partnership for Safe Water (AWWA 2014).

Several epidemiological studies have investigated the associa-tion between turbidity of drinking-water supplies and incidenceof AGI. These studies have generally been analyzed as time-series, correlating regional turbidity levels to AGI counts overtime. Mann et al. reviewed six of the studies published before2007 (Mann et al. 2007), and concluded that an association

Address correspondence to A.J. De Roos, Nesbitt Hall, Room 658, 3215Market St., Philadelphia, PA 19104 USA. Telephone: (267) 359-6090. Email:[email protected] authors declare they have no actual or potential competing financial

interests.Received 12 September 2016; Revised 24 February 2017; Accepted 27

February 2017; Published 17 August 2017.Note to readers with disabilities: EHP strives to ensure that all journal

content is accessible to all readers. However, some figures and SupplementalMaterial published in EHP articles may not conform to 508 standards due tothe complexity of the information being presented. If you need assistanceaccessing journal content, please contact [email protected]. Our staffwill work with you to assess and meet your accessibility needs within3 working days.

Environmental Health Perspectives 086003-1

A Section 508–conformant HTML version of this articleis available at https://doi.org/10.1289/EHP1090.Review

https://doi.org/10.1289/EHP1090

mailto:[email protected]

http://ehp.niehs.nih.gov/accessibility/

mailto:[email protected]

https://doi.org/10.1289/EHP1090

between turbidity and AGI is likely in some settings or over acertain range of turbidity. Several additional studies have beenpublished on this topic since the Mann review, conducted inLe Havre and Nantes, France (Beaudeau et al. 2012, 2014b) andin Eastern Massachusetts, Atlanta, and New York City in theUnited States (Beaudeau et al. 2014a; Hsieh et al. 2015; Tinkeret al. 2010). Positive results have been interpreted by some tosuggest that there was a detectable excess risk of AGI from drink-ing water in the populations studied. However, there are issueswith the research that complicate causal inference, such as theuncertain utility of turbidity as a universal proxy for microbiolog-ical contamination, confounding by season and climatic varia-bles, and regional differences in source-water quality andtreatment practices.

Given a previously described association (Mann et al. 2007)and multiple new studies on the topic, we conducted a review tosynthesize the evidence for an association between drinking-water turbidity and AGI incidence. In addition to synthesis, weaimed to describe differences between the studies that mayexplain discrepant results. Our goal was to evaluate the potentialfor causal inference from the studies, in other words, the extent towhich the observed associations may be interpreted as indicatingthe presence of waterborne AGI from drinking water in theregions and time periods studied. Because our focus was oncausal inference, rather than the strength of associations, we usedqualitative, rather than quantitative, review methods to synthesizeand compare information from the studies. A secondary aim,based on our review, was to identify the seemingly most promis-ing approaches for future research. This body of work may pro-vide insight into the utility of turbidity measurements and thetime-series study design to detect increased risk of waterborneAGI from microbiological contamination of drinking-watersupplies.

MethodsWe conducted a literature search to identify and select epidemio-logical studies that have examined drinking-water turbidity inrelation to AGI. Our review was open to inclusion of studies writ-ten in English and published as articles in peer-reviewed journals,as government reports, or in gray literature including conferenceproceedings and dissertations. Our inclusion criteria did not spec-ify any particular time period or region. We identified studies byconducting searches in MEDLINE (PubMed), JSTOR, andProQuest databases, using search terms such as “turbidity ANDwater AND gastrointestinal.” In addition, we reviewed the citationlists of published reports to find additional studies, and conducteda Web of Science database linkage of identified peer-reviewedpapers to all subsequent studies appearing in the peer-reviewed lit-erature that cited one of the articles. We reviewed the abstract foreach retrieved study to evaluate its suitability for inclusion in theliterature review. The steps and results of our literature search aredetailed in Table 1. One additional study that was publishedshortly after we conducted our database search met our criteriaand was also included (Hsieh et al. 2015).

We extracted information from each study including design,region and time period, exposure (turbidity) measurement, out-come (AGI) definition and identification, statistical methods,assessment of confounding and other potential biases, effect mod-ification (interaction) by age or other factors, and the authors’overall conclusions. For each study region, we recorded theauthors’ description of source water, water treatment practices,and turbidity levels. We noted the few studies that were con-ducted during a recognized outbreak of waterborne AGI. Wespecified the water sample collection points for the turbiditymeasurements included in the research (e.g., finished effluent,

source water, or from within the distribution system) and thesummary turbidity metric used in regression models (e.g., dailyaverage, maximum). Studies evaluated multiple lags to testhypotheses regarding the length of time by which an increase inturbidity is followed by an increase in AGI incidence. We sum-marized each study’s approach for testing lags and recorded thelags with notable results. We also compared the notable lag timesacross the studies to evaluate consistency in the collective find-ings. For each study, we summarized the potential confoundersconsidered and discussed important omissions. We also noted theimpact of adjustment for individual covariates, where presentedby the study authors. We described where the association wasstronger in, or limited to a specific subgroup of the study popula-tion or the data, such as in analyses by age group, season, watersupply service area, or AGI definition. We noted subanalyses thatwere conducted in each study, such as sensitivity analyses toevaluate the influence of extreme turbidity levels, and describedwhere results differed from the overall analysis. Each of the co-authors reviewed the studies independently, before coming to-gether to discuss and critically evaluate the methods of the indi-vidual studies and overarching issues with the body of literature.

Results

Design and Methods of the StudiesWe identified 14 studies reporting on the association betweenmeasured levels of turbidity of drinking-water supplies and inci-dence of AGI (Aramini 2002; Bateson 2001; Beaudeau et al.2012, 2014a, 2014b; Egorov et al. 2003; Gilbert et al. 2006;Hsieh et al. 2015; Lim et al. 2002; Morris et al. 1998; Naumovaet al. 2003; Schwartz et al. 1997, 2000; Tinker et al. 2010). Thesewere considered 14 distinct studies as they were conducted in dif-ferent regions, time periods, and/or population subgroups.Several other reports were reviewed but were not considered dis-tinct studies becauses they reported results found in one of thereferences cited above (Aramini 2000; Bateson et al. 2000;Beaudeau et al. 1999; Tinker 2007) or covered the same region,time period, and population (Morris et al. 1996). The first studywas conducted following a documented outbreak of waterbornecryptosporidiosis in Milwaukee, Wisconsin in 1993 (Morris et al.1996, 1998). Investigators evaluated the association between tur-bidity levels and AGI both during and before the outbreak, using abroad definition of AGI, not specific to cryptosporidiosis.Additional studies have been conducted elsewhere in the UnitedStates (Philadelphia, Seattle, Atlanta, Eastern Massachusetts,and New York City), Canada (Vancouver, Edmonton, andQuebec City), France (Le Havre and Nantes), and Russia(Cherepovets). All of the studies were designed as time se-ries, in which turbidity measurements and AGI counts weresummarized within small time increments (usually daily) andthe summary measures were correlated over the study periodin order to estimate the relative incidence of AGI when com-paring across turbidity levels. Table 2 describes the sourcewater, drinking-water treatment methods including filtrationand type of disinfection, and turbidity levels of the water sup-plies studied. Table 3 shows details of the studies includingAGI definition, covariates, and findings.

Some of the studies were conducted under conditions thatwould not comply with current U.S. water quality regulations(Table 2). In the studies of the outbreak and pre-outbreak condi-tions in Milwaukee (Morris et al. 1996, 1998; Naumova et al.2003), the turbidity in the Linwood (North) Plant would not be incompliance with the current U.S. requirement to be ≤0:3NTU in95% of samples, even before the outbreak period. The water qual-ity for studies in Le Havre, France (Beaudeau et al. 2012) and


Cherepovets, Russia (Egorov et al. 2003) frequently had finishedwater turbidity exceeding 0:3NTU. There are a number of morerecent studies of water systems with turbidity levels generally<0:3NTU, including unfiltered water supplies from protectedsources, such as in Seattle (Bateson 2001), EasternMassachusetts (Beaudeau et al. 2014a), New York City (Hsiehet al. 2015), and Vancouver (Aramini 2002) and filtered suppliesderived from surface-water sources in Philadelphia (Schwartzet al. 1997, 2000), Atlanta (Tinker et al. 2010), Edmonton (Limet al. 2002), and Nantes, France (Beaudeau et al. 2014b). Itshould be noted that because of changing regulations and treat-ment practices, the turbidity levels reported within a particularregion are only directly representative of the time period studied.

The studies defined exposure using turbidity measurementsperformed as part of standard monitoring of drinking-water sup-plies. The specific instrumentation and methods used for mea-surement were not stated in most of the studies; however,regulatory bodies such as the U.S. EPA stipulate sampling meth-ods for compliance with regional rules (U.S. EPA 2015). The col-lection points for turbidity measurement in the studies wereprimarily finished (filtered) effluent from treatment plants andsource (raw) water of either unfiltered or filtered water supplies.Only one study included measurements taken within the distribu-tion system (Hsieh et al. 2015). Multiple daily measurementswere summarized for statistical analyses into daily mean, median,minimum, or maximum values or were summarized over longertime increments as a moving average.

Cases of AGI were defined in most of the studies by AGI-related diagnosis codes (e.g., ICD-9) for hospital admissions,emergency department visits, or physician office (outpatient) vis-its (Table 3). The sources of data for case identification wereMedicare records (reported by the Health Care FinancingAdministration) (Bateson 2001; Beaudeau et al. 2014a; Naumovaet al. 2003; Schwartz et al. 2000), specific regional health care

providers (Morris et al. 1996; Morris et al. 1998, Schwartz et al.1997; Tinker et al. 2010), or the government healthcare system inCanada (Aramini 2002, Lim et al. 2002). Other AGI case defini-tions were based on prescription drug sales in France (Beaudeauet al. 2012, Beaudeau et al. 2014b), calls to a health informationline in Quebec City, Canada (Gilbert et al. 2006), and self-reported symptoms during a short-term longitudinal data collec-tion in Russia (Egorov et al. 2003). The study conducted in NewYork City used syndromic surveillance of the chief complaintreported in emergency department visits to classify a diarrheasyndrome (Hsieh et al. 2015).

The studies typically applied generalized additive Poissonregression models to the time-series data. Early application ofgeneralized additive models (GAMs) to study the associationbetween water turbidity and AGI was given in Schwartz et al.(1997). Since then, other researchers of this topic have followeda similar method with some variations. The studies frequentlyapplied nonparametric smoothing methods, such as local polyno-mial regression (LOESS) or smoothing splines to adjust for pos-sible nonlinearity of the effects of seasonal cycles, long-termtrends, climatic factors, or other covariates. Studies also adjustedfor potential autocorrelation by inclusion of an autoregressiveterm representing lagged values of the AGI outcome (Aramini2002; Beaudeau et al. 2012, 2014b; Gilbert et al. 2006; Lim et al.2002; Morris et al. 1998; Naumova et al. 2003). Turbidity hasbeen modeled using either a linear term or nonparametricsmoothing, or with exploration of both types of relationships inseparate models. All of the studies evaluated multiple lag timesrepresenting the latency between turbidity measurement and AGIcase identification. The lag times were generally tested in multi-ple, separate models, and a few studies also considered multiplelags simultaneously in distributed lag models (Hsieh et al. 2015;Schwartz et al. 1997; Tinker et al. 2010). Significance testing,model fit, and various graphical methods such as temporal

Table 1. Database search for studies on turbidity of drinking-water supplies and risk of acute gastrointestinal illness.

Database and search terms Number of items retrievedArticles selected for review (not listed if

identified in previous search)

PubMed: turbidity AND water AND gastrointestinal 21 Beaudeau et al. 2014aSearch limited to peer-reviewed literature with all three of the searchterms appearing in any search field (title, key word, etc.)

Egorov et al. 2003Gilbert et al. 2006Morris et al. 1996Morris et al. 1998Schwartz et al. 1997Schwartz et al. 2000Tinker et al. 2010

PubMed: turbidity AND water AND gastroenteritis 19 Beaudeau et al. 2012Search limited to peer-reviewed literature with all three of the searchterms appearing in any search field (title, key word, etc.); Differs fromfirst search by use of word “gastroenteritis”

Beaudeau et al. 2014bNaumova et al. 2003

JSTOR: turbidity water gastrointestinal 526 No additional studies were identifiedSearch that casts a wide net to identify journal articles, conferenceabstracts, and books that include any combination of the search terms

JSTOR: turbidity AND gastro# AND “drinking water” 200 Bateson et al. 2000 (Abstract)Search includes a ‘wild card’ symbol (#), which will retrieve articleswith any word starting with “gastro,” including both gastrointestinaland gastroenteritis, and is also limited to items containing the words“drinking water” in any search field

Web of Science, cited reference search: for each previously selected peer-reviewed publication, search identifies subsequent articles that cited thepaper

All articles citing each selected article No additional studies were identified

Search of citations in each of the studies identified All studies cited in each study Aramini et al. 2000 (Government report)Lim et al. 2002Beaudeau et al. 1999

Proquest dissertations and theses: turbidity AND water AND (gastrointes-tinal OR gastroenteritis) in any search field except full text (includingfull text retrieved thousands of items)

14 Aramini 2002 (Dissertation)Tinker 2007 (Dissertation)

Contacted author directly to obtain full report 1 Bateson 2001 (Dissertation)


Tab

le2.Water

supplies,treatm

entm

ethods,and

turbidity

levelsin

thestudiesof

theassociationbetweenturbidity

ofdrinking-w

ater

suppliesandrisk

ofacutegastrointestinal

illness.

Region(study)

Tim

eperiod

Water

treatm

entp

lant

Source

water

(noted

ifprotected)

Treatmentd

uringthestudy

period

(described

inpaper)

Water

samples

evaluated

Turbidity

levelin

NTU,m

ean(range)

Turbidity

generally

<0:30

NTUin

water

supply?

Milw

aukee,USA

(Morrisetal.1

998;

similarregion

and

timeperiod

asMorrisetal.1

996

andNaumovaetal.

2003

)

Pre-outbreak:January

1992–M

arch

1993

How

ardAvenue

(South)plant,

Milw

aukeeWater

Works

Milw

aukeeRiver

Filtration,

chlorinatio

nFinished

effluent

(filtered)

Pre-outbreak:(

∼0:0

–0.25)a

Pre-outbreak:Y

es(<

0:2NTU98%

ofthetim

e)Outbreak:

March–

April1993

Outbreak:

(<0:05

–1.7)

aOutbreak:

No

Milw

aukee,USA

(Morrisetal.1

998;

similarregion

and

timeperiod

asMorrisetal.1

996

andNaumovaetal.

2003

)

Pre-outbreak:January

1992–M

arch

1993

LinwoodAvenue

(North)plant,

Milw

aukeeWater

Works

Milw

aukeeRiver

Filtration,

chlorinatio

nFinished

effluent

(filtered)

Pre-outbreak:(<0:1–

0.52)a

Datanotp

rovided

Outbreak:

March–

April1993

Outbreak:

(0.1–0

.33)

a

Philadelphia,USA

(Schwartzetal.

1997

;sim

ilarregion

andtim

eperiod

asSchw

artzetal.

2000

)

1992–1993

Belmontp

lant,

PhiladelphiaWater

Departm

ent

SchuylkillRiver

Flocculatio

n,sedimentatio

nfiltration,

chlorinatio

n,chloramineresidual

Finished

effluent

(filtered)

0.18

(10thto

90th

percentile:0.12

–0.24)

Yes

Philadelphia,USA

(Schwartzetal.

1997

;sim

ilarregion

andtim

eperiod

asSchw

artzetal.

2000

)

1992–1993

Queen

Laneplant,

PhiladelphiaWater

Departm

ent

SchuylkillRiver

Flocculatio

n,sedimentatio

nfiltration,

chlorinatio


Finished

effluent

(filtered)

0.20

(10thto

90th

percentile:0.14

–0.25)

Yes

Philadelphia,USA

(Schwartzetal.

1997

;sim

ilarregion

andtim

eperiod

asSchw

artzetal.

2000

)

1992–1993

Baxterplant,

PhiladelphiaWater

Departm

ent

Delaw

areRiver

Flocculatio

n,sedimentatio

nfiltration,

chlorinatio


Finished

effluent

(filtered)

0.17

(10thto

90th

percentile:0.13

–0.20)

Yes

Seattle,W

A,U

SA(B

ateson

2001

)1990–1994

Toltp

lant,S

eattle

PublicUtilities

ToltR

iver

watershed

(protected)

Chlorination(unfiltered

supply)

Source

water

(unfil-

teredsupply)

Daily

maxim

um:0

.64

(0.04–

2.52)

Datanotp

rovided

Atlanta,USA

(Tinker

etal.2

010)

1993–2004

Eight

treatm

entp

lants

Variety

ofsurface

water

sources:riv-

ers,stream

s,and

impoundm

ents

Mostp

lants:Coagulatio

n,sedimentatio

n,filtration,

chlorinatio

nOne

plant:Ozonatio

n,butn

osedimentatio

nThree

plants:U

Vand

chlorinatio

n

Finished

effluent

(filtered)

Source

water

Finished

effluent:

0.08

(10thto

90th

percentile:0.03

–0.15)

Source

water,d

aily

minim

um:7

.5(10th

to90th

percentile:

1.2–

16.0)

Yes

Massachusetts,U

SA(B

eaudeauetal.

2014a)

1998–2008

CarrollWater

TreatmentP

lant

(CWTP),

Massachusetts

Water

Resources

Authority

(MWRA)

Wachusettand

Quabbin

Reservoirs

(protected)

Ozonatio

n,chlorinatio

n,chloramination(unfiltered

supply)

CWTPchangedfrom

chlori-

natio

nto

ozonationduring

studyperiod

(in2005)

Source

water

(unfil-

teredsupply)

0.34

(0.17–0.68)

No(<

50%

ofsamples)


Tab

le2.

(Contin

ued.)

Region(study)

Tim

eperiod

Water

treatm

entp

lant

Source

water

(noted

ifprotected)

Treatmentd

uringthestudy

period

(described

inpaper)

Water

samples

evaluated

Turbidity

levelin

NTU,m

ean(range)

Turbidity

generally

<0:30

NTUin

water

supply?

New

YorkCity

,USA

(Hsieh

etal.2

015)

2002–2009

New

YorkCity

,New

YorkDepartm

ento

fEnvironmental

Protectio

n

Three

watersheds

locatednorthof

NYCthatdraininto

theCatskill,

Delaw

are,and

Crotonreservoirs

(protected)

Chlorination

Distributionsystem

(unfiltered

supply)

Source

water

(unfiltered

supply)

Distributionsystem

(unfiltered

supply):

0.97

(0.54–

2.38)

Source

water:0

.98

(0.50–

2.85)

No

Vancouver,C

anada

(Aramini2

002)

1992–1998

Greater

Vancouver

RegionalD

istrict

(GVRD)

Capilano,C

oquitlan,

andSeym

ourwater-

sheds(protected)

Chlorination

Source

water

(unfil-

teredsupply)

Capilano:1

.30(0.17–

19.0)

No(<

25%

ofsamples)

Seym

our:0.82

(0.15–

19.0)

Coquitlan:

0.54

(0.19–

8.30)

Edm

onton,

Canada

(Lim

etal.2

002)

1993–1997

Rossdaleplant,

EPC

ORWater

Services,Inc.

North

Saskatchew

anRiver

Clarificatio

n,filtration,

disinfectio

nFinished

effluent

(filtered)

Finished

effluent:

0.06

(0.059

–0.062)

Yes

(>95

%of

samples)

Source

water

Source

water:3

5(30.7,

39.3)

QuebecCity

,Canada

(Gilb

ertetal.2006

)2000–2002

QuebecCity

water

treatm

entp

lant

Saint-Charles

River

Pre-chlorinatio

n,coagulation,

decantation,

filtration,

ozo-

natio

n,post-chlorination

Finished

effluent

(filtered)

0.27

(0.11–0.75)

Datanotp

rovided

LeHavre,F

rance

(Beaudeauetal.

2012

)

Period

1:1994

–1996

Period

2:1997

–2001

SaintL

aurent

plant

(phi900andphi500

drains)

Karsticspringsfrom

chalkaquifers

Chlorination

Source

water

(unfil-

teredsupply)

phi900,P

eriod1:

0.13

(0.05–

1.22)

phi900,P

eriod2:

0.11

(0.05–

1.36)

phi500,P

eriod2:

0.10

(0.01–

0.90)

phi900:Y

esforboth

periods(about

98%

ofsamples)

phi500:D

atanot

provided

LeHavre,F

rance

(Beaudeauetal.

2012

)

Period

1:1994

–1996

Period

2:1997

–2001

Radicatelplant

Karsticspringsfrom

chalkaquifers

Directfiltratio

nandchlorina-

tion(exceptd

uringhigh

tur-

bidity

conditionswhen

coagulation-flocculatio

n-settlingwas

employed

before

filtration)

Finished

effluent

(filtered)

Source

water

Finished

effluent,

Period

1:0.34

(0.03–

1.46)

Finished

effluent,

Period

2:0.25

(0.01–

1.37)

Source

water,

Period

1:4.0(0.9–

70.2)

Source

water,

Period

2:4.0(0.5–

229)

No(<

75%

ofsamples

inPe

riod

1;<90%

ofsamples

inPe

riod

2)

Nantes,France

(Beaudeauetal.

2014b)

2002–2007

Nantes,France

Loire

River

Coagulatio

n,flocculatio

n,settling,

filtration,

ozona-

tion,

chlorinatio

n

Finished

effluent

(filtered)

0.05

(0.02–0.34)

Yes

Cherepovets,R

ussia

(Egorovetal.2

003)

1999

Cherepovets

Vodokanal

SheksnaRiver

Chlorination,

coagulation,

rapidsand

filtration,

second-

arychlorinatio

n(plant

rou-

tinelyneeded

toadjust

water

treatm

enttomeet

demand)

Finished

effluent

(filtered)

0.59

(0.21–1.33)b

No

a Estim

ated

from

figure.

b NTUconvertedfrom

mg/L,u

sing

form

ula:1NTU=0:576mg=

L.


Tab

le3.Designandresults

oftim

e-series

studiesof

theassociationbetweenturbidity

ofdrinking-w

ater

suppliesandrisk

ofacutegastrointestinal

illness.

Region(study)

Tim

eperiod

Populatio

nsubgroups

Turbidity

measure

AGIdefinitio

nsCovariatesconsidered

Lags

exam

ined

Associatio

nbetweenturbidity

andAGI

Additionalanalyses

Milw

aukee,USA

(Morrisetal.1

996;

Morrisetal.1

998)

Outbreak:

March–

April1993

Allresidents,by

water

servicearea,

agegroup(≤

18,

>18

y)

Finished

effluent

(filtered),2-wk

mean

Physicianoffice

(out-

patient)v

isits,emer-

gencydepartment

visits,and

hospital

admissions(ICD-9

codes007–009.3,

558.9)

atthe

MedicalCollege

ofWisconsin

Physicians

and

Clin

ics

Day-of-week

1wk

AGIoutpatient

visits,L

inwood

(North)plant,ages

≤18

y,1-wk

lag:

ER=53

%(95%

CI:−8%

,155%

)with

0:5-NTUincrease

of2-wkmeanturbidity

Morrisetal.1

998

presentsresults

for

daily

AGIwith

meanturbidity

bydaily

lags,inthe

form

ofcorrelation

coefficientsand

TERSplots

AGIem

ergencyandhospitalv

is-

its,L

inwood(N

orth)plant,ages

≤18

y,1-wklag:

ER=182%

(95%

CI:44%,4

52%)with

0:5-NTUincrease

of2-wk

meanturbidity

Linwood(N

orth)plant,ages

>18

y:Noassociation

AGIoutpatient

visits,H

oward

Avenue(South)plant,ages

≤18

y,1-wklag:

ER=36%

(95%

CI:−1%

,87%

)with

0:5-NTUincrease

of2-wk

meanturbidity

AGIem

ergencyandhospital,

How

ardAvenue(South)plant,

ages

≤18

y,1-wklag:

ER=73%

(95%

CI:19%,

150%

)with

0:5-NTUincrease

of2-wkmeanturbidity

AGIoutpatient

visits,H

oward

Avenue(South)plant,ages

>18

y,1-wklag:

ER=96%

(95%

CI:64%,1

34%)with

0:5-NTUincrease

of2-wk

meanturbidity

AGIem

ergencyandhospitalv

is-

its,H

owardAvenue(South)

plant,ages

>18

y,1-wklag:

ER=147%

(95%

CI:115%

,183%

)with

0:5-NTUincrease

of2-wkmeanturbidity

Milw

aukee,USA

(Morrisetal.1

996;

Morrisetal.1

998)

Pre-outbreak:

January1992–

March

1993

Allresidents,by

water

servicearea,

agegroup(≤

18,

>18

y)

Finished

effluent

(fil-

tered),2

-wkmean

Physicianoffice

(out-

patient)v

isits,emer-

gencydepartment

visits,and

hospital

admissions(ICD-9

codes007–009.3,

558.9)

atthe

MedicalCollege

ofWisconsin

Physicians

and

Clin

ics

Day-of-week

1wk

AGIoutpatient

visits,L

inwood

(North)plant,ages

≤18

y,1-wk

lag:

ER=60

%(95%

CI:−13

%,

192%

)with

0:5-NTUincrease

of2-wkmeanturbidity

Morrisetal.1

998

presentsresults

for

daily

AGIwith

meanturbidity

bydaily

lags,inthe

form

ofcorrelation

coefficientsand

TERSplots

AGIem

ergencyandhospitalv

is-

its,L

inwood(N

orth)plant,ages

≤18

y,1-weeklag:

ER=180%

(95%

CI:28%,5

12%)with

0:5-NTUincrease

of2-wk

meanturbidity

Noothersignificant

association

inotherw

ater

servicearea–age

groupcombinatio

ns


Tab

le3.

(Contin

ued.)

Region(study)

Tim

eperiod

Populatio

nsubgroups

Turbidity

measure

AGIdefinitio


Lags

exam

ined

Associatio

nbetweenturbidity

andAGI

Additionalanalyses

Milw

aukee,USA

(Naumovaetal.

2003

)

Outbreak:

March–

April1993

Elderly

(≥65

y)Finished

effluent

(fil-

tered)

from

How

ard

Avenue(South)

treatm

entp

lant,

daily

maxim

um

Hospitaladm

issions

(ICD-9

codes001–

009.9,

558.9,

276,

787,

789)

from

Medicarebilling

records

Seasonalcycles

Tim

etrend

Autoregressiveterm

0–18

dNonlin

earmodel,6

-dlag:

ER=310%

(p<0:05)w

ithdaily

maxim

umturbidity

of1–2NTUduring

outbreak

vs.

averagepre-outbreak

AGIrate

andturbidity

level

Linearmodel,6

–7dlag:

95%

CI

fortheERwas

(54%

,348%)

with

1:0-NTUincrease

ofdaily

maxim

umturbidity

Milw

aukee,USA

(Naumovaetal.

2003

)

Pre-outbreak:

January1992–

March

1993

Elderly

(≥65

y)Finished

effluent

(filtered)from

How

ardAvenue

(South)treatm

ent

plant,daily

maxim

um

Hospitaladm

issions

(ICD-9

codes001–

009.9,

558.9,

276,

787,

789)

from

Medicarebilling

records

Seasonalcycles

0–18

dNostatistically

significantasso-

ciationatanylagin

nonlinear

modela ,buta

visiblepeak

onTERSplot

at8-dlag

Tim

etrend

Autoregressiveterm

Philadelphia,USA

(Schwartzetal.

1997

)

1989–1

993

Children(age

range

notspecified)

Finished

effluent

(filtered)from

three

treatm

entp

lants,

daily

mean

Emergencydepart-

mentv

isits

(1992–

1993)andhospital

admissions(1989–

1993)(ICD-9

codes

001–009.9,

558.9)

atonechild

ren’s

hospital

Seasonalcycles

Tim

etrend

Tem

perature

Day-of-week

1–14

dAGIem

ergencyvisits,4

-dlag:

ER=7:2%

(95%

CI:2.8%

,11.7%)with

IQRincrease

(0:04NTU)of

daily

mean

turbidity

AGIem

ergencyvisits,1

0-dlag:

ER=6:7%

(95%

CI:2.2%

,11.4%)with

IQRincrease

(0:04NTU)of

daily

mean

turbidity

AGIhospitaladm

issions,8-d

lag:

ER=8:7%

(95%

CI:0%

,18%)with

IQRincrease

(0:04NTU)of

daily

mean

turbidity

Bywater

servicearea

Byagegroup(≤

2,>2y)

Includingsecondary

diagnoses

Two-lagmodels

Robustregression

Philadelphia,USA

(Schwartzetal.

2000

)

1992–1

993

Elderly

(≥65

y)Finished

effluent

(filtered)from

three

treatm

entp

lants,

daily

mean

Hospitaladm

issions

(ICD-9

codes001–

009.9,

558.9,

276,

787,

789)

from

Medicarebilling

records

Seasonalcycles

1–14

d9–11

dlag:

ER=9:0%

(95%

CI:

5.3%

,12.7%

)with

IQRincrease

(0:035

NTU)of

daily

mean

turbidity

Bywater

servicearea

Tim

etrend

Byagegroup(65–

74,

≥75

y)Tem

perature

Day-of-week

Seattle,U

SA(B

ateson

2001

)1990–1

994

Elderly

(≥65

y),b

yseason,

agegroup

(65–79,≥

80y)

Source

water

(unfiltered

supply),

daily

maxim

um

Hospitaladm

issions

(ICD-9

codes001–

009.9,

558.9,

276,

787,

789)

from

Medicarebilling

records

Seasonalcycles

Tim

etrend

Tem

perature

Precipitatio

nDay-of-week

1–21

dWinter,ages

≥80

y,excluding

days

with

>1NTUmaxim

umturbidity

,7-d

lag:

ER=11:6%

(95%

CI:3.3%

,20.4%

)with

0:1-NTUincrease

ofdaily

max-

imum

turbidity

Noconsistent

associations

found

foro

ther

agegroup-season

com-

binatio

nsor

with

inclusionof

maxim

umturbidity

>1NTUa


Tab

le3.

(Contin

ued.)

Region(study)

Tim

eperiod

Populatio

nsubgroups

Turbidity

measure

AGIdefinitio


Lags

exam

ined

Associatio

nbetweenturbidity

andAGI

Additionalanalyses

Atlanta,USA

(Tinker

etal.2

010)

1993–2

004

Allresidents

Finished

effluent

(fil-

tered)

from

eight

treatm

entp

lants,

daily

meanand

maxim

umSo

urce

water,d

aily

minim

umand

maxim

um

Emergencydepart-

mentv

isits

(ICD-9

codes001–004,

005.0,

005.4,

005.89,0

05.9,0

06–

007,

008.0,

008.42–

008.44,0

08.47,

008.49,0

08.5,

008.6,

008.8,

009,

558.9,

787.01–

787.03,7

87.91as

prim

aryor

second-

arydiagnosis)

Seasonalcycles

Tim

etrend

Tem

perature

Day-of-week

Federalh

olidays

Hospitalinclusion

Adjustm

entfor

pre-

cipitatio

nin

second-

aryanalysis

0–20

dFinished

effluent,0

–20ddistrib-

uted

lag:

ER=

−2%

(95%

CI:

−4%

,1%)w

ith0:1-NTU

increase

ofdaily

meanturbidity

Analysisof

3-dmovingaverage

finished

effluent

daily

meantur-

bidity

show

edhighestp

eakwith

6–8dlag(p>0:05)

Similarresults

with

maxim

umfinished

effluent

turbidity

Source

water,0

–20ddistributed

lag:

ER=6%

(95%

CI:4%

,8%)

with

10-N

TUincrease

ofdaily

minim

umturbidity

Analysisof

3-dmovingaverage

source

water

daily

minim

umturbidity

show

edhighestp

eak

with

7–9dlag(p

<0:05)

Similarresults

with

maxim

umsource-w

ater

turbidity

Bywater

servicearea

Byagegroup

Alternatediagnosis

categories

Exclusion

ofZIP

codesnotserved

100%

byonetreat-

mentp

lant

Controllin

gfor

rainfall

Alternatespecifica-

tionof

covariates

Massachusetts,U

SA(B

eaudeauetal.

2014a)

1998–2

008

Elderly

(≥65)

Source

water

(unfil-

teredsupply),daily

mean

Hospitaladm

issions,

emergencyonly

(ICD-9

codes001–

009.9,

558.9,

276,

787,

789)

from

Medicarebilling

records

Seasonalcycles

Tim

etrend

Monthly

cycles

Airtemperature

Water

temperature

Runoff

Day-of-week

Holidays

School

vacatio

ns

8–37

d(8–1

2-dlagselected

apriori)

8–12

dlag:

ER=3:4%

(95%

CI:

0.1%

,5.7%)with

0:1-NTU

increase

ofdaily

meanturbidity

Improved

fitw

ithnonlinear

modelandwith

algae-corrected

turbidity

a

8–12

dlag:

ER=5:8%

(95%

CI:

1.9%

,9,6%)with

0:1-NTU

increase

ofdaily

meanturbidity

atlevelsup

to0:33

NTU;n

oassociationathigher

levels

Algae-corrected

turbidity

,8–12d

lag:

ER=7:7%

(95%

CI:2.6%

,12.7%)with

0:1-NTUincrease

ofdaily

meanalgae-corrected

turbidity

atlevelsup

to0:25

NTU;n

oassociationat

higher

levels

Excluding

extrem

evalues

Excluding

tempera-

ture

covariates

Alternatespecifica-

tionof

covariates

Associatio

nsof

AGI

with

:fecalcoliform,

UV-absorbance,

algae,cyanobacter-

iae,chlorine

concen-

tration–

time

New

YorkCity

,USA

(Hsieh

etal.2

015)

2002–2

009

Allresidents,by

sea-

son,

agegroup(0–4,

5–17,1

8–64,≥

65y)

Distributionsystem

(unfiltered

supply),

daily

median

Source

water

(unfil-

teredsupply),daily

flow

-weightedmean

Syndromicsurveil-

lanceof

emergency

departmentv

isit

chiefc

omplaint

data,

groupedas

‘diarrhea

syndrome’

Seasonalcycles

Tim

etrend

Tem

perature

Precipitatio

nDay-of-week

Holidays

0–13

dDistributio

nsystem

turbidity

,sprin

g,6-dlag:

ER=5%

(95%

CI:3%

,6.5%)(CIe

sti-

mated

from

figure)

with

IQR

increase

ofdaily

medianturbidity

(1:0NTU,estim

ated

from

figure)

Associatio

nseen

inages

0–4and

5–17

y.Noassociationseen

inages

>17

ya

Noconsistent

associationin

other

seasons

Similarpattern

ofassociation

with

source-w

ater

turbidity

;and

resultfordistributio

nsystem

Negativebinomial

model

Distributed

lagmodel

Excluding

highesttur-

bidity

values

Evaluatingsubsetsof

years

Alternatespecifica-

tionof

covariates


Tab

le3.

(Contin

ued.)

Region(study)

Tim

eperiod

Populatio

nsubgroups

Turbidity

measure

AGIdefinitio


Lags

exam

ined

Associatio

nbetweenturbidity

andAGI

Additionalanalyses

turbidity

was

almostfully

explainedby

variability

insource-w

ater

turbidity

a

Vancouver,C

anada

(Aramini2

002)

1992–1

998

Allresidents,by

water

servicearea,

agegroup(<

2,2–

18,1

9–65,>

65y)

Source

water

(unfil-

teredsupply),daily

mean

Hospitaladm

issions

(ICD-9

codes001–

009,

5350,5

354,

5355,5

356,

558,

7870

oroneof

the

previous

codesas

secondarydiagnosis

with

prim

arycode

of276,

5781,6

910,

7806,7

830,

7832,

787,

or7890)from

natio

nalh

ealth

care

billing

Physicianoffice

(out-

patient)v

isits,emer-

gencyonly

(ICD-9

codes001–009,

558)

from

natio

nalh

ealth

-care

billing

Emergencydepart-

mentv

isits

with

AGI-relatedsymp-

toms(diarrhea,vom-

iting,g

astroenteritis,

enteritis,g

astritis)

from

onechild

ren’s

hospital

Seasonalcycles

Tim

etrend

Tem

perature

Precipitatio

nDay-of-week

Holidays

Fecalcoliform

Autoregressiveterm

0–39

dAssociatio

nsfrom

nonlinear

modelsweremostp

rominent

forages

2–18

yand19–6

5y,

with

lagtim

esof

3–6d,

7–9d,

12–1

6d,

and21–29da

Case–controlstudy

ofAGIcase

group

comparedto

anacuterespiratoryill-

ness

controlg

roup,

analyzed

usinglogis-

ticregression

Edm

onton,

Canada

(Lim

etal.2

002)

1993–1

997

Allresidents,by

age

group(<

2,2–18,

19–65,

>65

y)

Finished

effluent

(fil-

tered),d

aily

mean,

median,

and

maxim

umSo

urce

water,d

aily

mean,

median,

and

maxim

um

Hospitaladm

issions

(ICD-9

codes001–

009,

5350,5

354,

5355,5

356,

558,

7870

oroneof

the

previous

codesas

secondarydiagnosis

with

prim

arycode

of276,

5781,6

910,

7806,7

830,

7832,

787,

or7890)from

natio

nalh

ealth

care

billing

Physicianvisits(out-

patient),em

ergency

only

(ICD-9

codes

001–009,

558)

from

natio

nalh

ealth

care

billing

Seasonalcycles

Tim

etrend

Tem

perature

Precipitatio

nDay-of-week

Holidays

Finished

water

parti-

clecount

Source

water

fecal

coliform

Source

water

total

coliform

Source

water

temperature

Source

water

pHAutoregressiveterm

0–40

dNosignificantassociatio

nwith

finished

effluent

turbidity

atany

laga

Noassociationwith

source-w

ater

turbidity

a

Case–controlstudy

ofAGIcase

group

comparedto

anacuterespiratoryill-

ness

controlg

roup,

analyzed

usinglogis-

ticregression

QuebecCity

,Canada

(Gilb

ertetal.2006

)2000–2

002

Allresidents

Finished

effluent

(fil-

tered),d

aily

mean

Callsto

apublic

health

inform

ation

Seasonalcycles

Tim

etrend

1–40

dNonlin

earmodel,E

Rforpre-

dicted

AGIcountat0

:75NTU

Com

parisonof

mod-

elsfordaily

mean


Tab

le3.

(Contin

ued.)

Region(study)

Tim

eperiod

Populatio

nsubgroups

Turbidity

measure

AGIdefinitio


Lags

exam

ined

Associatio

nbetweenturbidity

andAGI

Additionalanalyses

linereportingAGI-

relatedsymptom

s(e.

g.,d

iarrhea,vomit-

ing,

nausea,abdom

i-nalcramps)

Season

(categorical)

Precipitatio

nDay-of-week

Holidays

Autoregressiveterm

daily

meanturbidity

relativ

eto

AGIcounto

ccurring

atthe

meandaily

turbidity

levelo

fentirestudyperiod

(0:27NTU):

11-d

lag:

ER=33

%(p

<0:05)

15–d

lag:

ER=53%

(p<0:05)

17-d

lag:

ER=76

%(p

<0:05)

turbidity

anddaily

maxim

umturbidity

LeHavre,F

rance

(Beaudeauetal.

2012

)

1994–2

001

Allresidents,by

water

servicearea,

drain,

andtim

epe-

riod

(modelfitted

foryears1997–2

001

first,then

appliedto

1994–1996)

Finished

effluent

(fil-

tered)

bydrain,

daily

mean

Source

water,d

aily

mean

Prescriptio

nsfor

antiemetics(A

TC

classificatio

nsA04A,A

03F)

probi-

oticantid

iarrheals

(A07F),intestin

alantip

ropulsives

(A07D,A

07X),in-

testinalabsorbents

(A07B,A

02X),in-

testinalanti-infec-

tious

agents(A

07A-

E),oralrehydration

salts

(noATCcode)

from

pharmacists’

network

Seasonalcycles

Tim

etrend

With

in-m

onth

cycling

Tem

perature

Day-of-week

Holidays

School

vacatio

nsAutoregressiveterm

3–15

dFinished

effluent,S

aint

Laurent

plant,drainphi5

00,1

997–

2001,6

–8dlag:

ER:2

7%(95%

CI:18%,3

6%)with

0:1-NTUincrease

ofdaily

mean

turbidity

Finished

effluent,S

aint

Laurent

plant,drainphi9

00,1

997–

2001,6

–8dlag:

ER:2

3%(95%

CI:17%,3

0%)with

0:1-NTUincrease

ofdaily

mean

turbidity

Finished

effluent,R

adicatal

plant,1997

–2001,

6–8dlag:

ER=2%

(95%

CI:−1%

,5%)

with

0:1-NTUincrease

ofdaily

meanturbidity

Source

water,R

adicatalplant,

1997–2

001,

7–9dlag:

ER=0:7%

(95%

CI:0.4%

,1%)

with

1-NTUincrease

ofdaily

meanturbidity

Nonlin

earmodelsdidnotshow

improved

fito

verlin

earmodels

Resultsonly

partially

reproduci-

blein

earliertim

eperiod;asso-

ciationseen

inboth

time

periodsin

Radicatalplantw

ithno

imposedsettling

Stratificationby

days

with

imposedset-

tling

vs.n

osettling

Nonlin

earfit

Nantes,France

(Beaudeauetal.

2014b)

2002–2

007

Allresidents,by

age

group(m

odelfitted

forages

<16

yfirst,

then

appliedto

ages

≥16

y)

Finished

effluent

(fil-

tered),d

aily

mean

Source

water,d

aily

mean

Prescriptio

nsfor

antiemetics(A

TC

classificatio

nsA04A,A

03F)

probi-

oticantid

iarrheals

(A07F),intestin

alantip

ropulsives

(A07D,A

07X),in-

testinalabsorbents

(A07B,A

02X),in-

testinalanti-infec-

tious

agents(A

07A),

oralrehydrationsalts

(noATCcode)from

Seasonalcycles

Tim

etrend

Tem

perature

Precipitatio

nWeekdays

Holidays

School

vacatio

nsRiver

flow

Finished

water

flow

Serviceinterventio

nsforbroken

pipe

repair

Hydrant

flushes

Chlorineresidual

Autoregressiveterm

7–9d

Finished

effluent,ages<16

y,7–

9dlag:

ER=3:0%

(95%

CI:

1.1%

,4.9%)with

0:1-NTU

increase

ofdaily

meanturbidity

Evidencefornonlinearity

ofeffect,w

ithno

cleara

ssociatio

n<0:45

NTUturbidity

,and

nearly

linearincreasesin

risk

athigher

levels

Finished

effluent,ages<16

y,7–

9dlag:

Evidenceof

nonlinearity

andsignificant

interactionwith

river

flow(strongerturbidity

associationathigh

river

flow)


Tab

le3.

(Contin

ued.)

Region(study)

Tim

eperiod

Populatio

nsubgroups

Turbidity

measure

AGIdefinitio


Lags

exam

ined

Associatio

nbetweenturbidity

andAGI

Additionalanalyses

natio

nalh

ealth

insur-

ance

database

Finished

effluent,ages<16

y,7–

9dlag:

ER=5:4%

(95%

CI:

0.5%

,10.5%

)with

daily

mean

turbidity

increase

from

0:06

to0:07

NTU,at9

5thper-

centile

riverflow

Source

water

turbidity

signifi-

cantly

associated

with

AGIin

ages

<16

y,butinconsistent

associationacross

lags

Finished

effluent,ages≥16

y,7–

9dlag:

ER=15:6%

(95%

CI:

9.7%

,21.9%

)with

daily

mean

turbidity

increase

from

0:06

to0:07

NTU,at9

5thper-

centile

riverflow

Cherepovets,R

ussia

(Egorovetal.2

003)

1999

Allages,byconsum

p-tio

nof

non-boiled

tapwater

Finished

effluent

(fil-

tered),d

aily

mean

Self-reported

Seasonalcycles

Tim

etrend

Day-of-week

Travel

Contactwith

farm

anim

als

Other

behavioral

factors

Consumptionof

non-

boiledwater

0–13

dPa

rticipantswith

consum

ptionof

non-boiledtapwater,7

-dlag:

ER=64%

(95%

CI:15%,

134%

)with

IQRincrease

(∼0:27

NTU)of

daily

mean

turbidity

Nostatistically

significantasso-

ciationatanylagin

participants

who

alwaysboiledtheirdrink-

ingwater

Nonlin

earmodelingresults

show

non

TERSplotssuggest

excess

risksof

AGIoccurred

only

athigher

levelsof

turbidity

(e.g.,>0:9NTU)

Note:AGI,acutegastrointestinalillness;C

I,confidence

interval;ICD,Interna

tiona

lClassificatio

nof

Diseases;IQ

R,interquartilerange;ER,excessrisk;N

TU,n

ephelometricturbidity

units;T

ERS,

temporalexposureresponse

surface.

a Statedby

author

(resultsnotshown).


exposure response surface (TERS) plots were used to determinethe importance of lagged effects. Multiple authors implementedmodeling approaches equipped to deal with possible overdisper-sion of the AGI count data, such as assuming a quasi-Poisson dis-tribution (Beaudeau et al. 2012, 2014b; Hsieh et al. 2015), oralternatively using negative binomial regression (Hsieh et al.2015) or robust regression (M-estimation) (Bateson 2001;Schwartz et al. 1997), or by scaling variance estimates to accountfor overdispersion (Tinker et al. 2010). Additionally, severalstudies assessed the influence of extreme turbidity values byexclusion of the highest values (Bateson 2001; Beaudeau et al.2012, 2014a, 2014b; Hsieh et al. 2015). Other approaches usedby investigators to assess the sensitivity of results to modelingchoices and model fit included comparison of a case–controlstudy design to the time-series analysis (Aramini 2002; Lim et al.2002), and following a split-sample approach to fit and test theirmodels in different subsets of the data (Beaudeau et al. 2012,2014b).

Study FindingsTable 3 shows the study findings. Positive associations betweenturbidity and AGI incidence were found in unfiltered water sup-plies from protected sources, such as in Seattle (Bateson 2001),Eastern Massachusetts (Beaudeau et al. 2014a), New York City(Hsieh et al. 2015), and Vancouver (Aramini 2002). Associationswere also observed in filtered supplies with relatively low turbid-ity levels, in Philadelphia (Schwartz et al. 1997; Schwartz et al.2000) and Nantes, France (Beaudeau et al. 2014b). Findings wererobust to alternate model specification (e.g., negative binomialmodels; M-estimation) (Bateson 2001; Hsieh et al. 2015;Schwartz et al. 1997) and exclusion of extreme turbidity values,above the 98th percentile (Beaudeau et al. 2012, 2014a, 2014b;Hsieh et al. 2015). The results of Tinker et al. (2010), fromAtlanta, are mixed. In this study, the multi-day association laggedover 0 to 20 d was significant for only one of eight filtered watersupplies; however, there was a positive association with source(raw) water turbidity. The one study that reported no relationshipbetween turbidity and AGI examined a filtered water supply inEdmonton with low levels of turbidity; the investigatorsobserved some statistically significant increases of AGI at spe-cific lags, but no overall “significant” relationships accordingto their criteria involving model fit and statistical significanceof lags over at least 2 consecutive days (Lim et al. 2002).

Comparison of estimated effect sizes was hindered by differ-ences between the studies, such as the turbidity contrasts used inrisk estimation, assumption of a linear or nonlinear associationbetween turbidity and AGI incidence, and the turbidity exposuremetric used in the time-series (e.g., daily or multi-day average,maximum). As shown in Table 3, there were some relativelylarge magnitude effects with large turbidity contrasts, such as 73–182% increases in AGI incidence observed per 0:5-NTU increasein Milwaukee water 2-wk average turbidity during the outbreak(Morris et al. 1996) and 33–76% AGI increases with a contrast of0:48NTU daily mean turbidity found in Quebec City (Gilbertet al. 2006). Smaller magnitude effect sizes were generally foundwith smaller turbidity unit contrasts, such as 7–9% AGI increasesper 0:04NTU daily mean turbidity in Philadelphia (Schwartzet al. 1997, 2000) and 3% increase per 0:01NTU daily mean tur-bidity inNantes, France (Beaudeau et al. 2014b).Highermagnitudeeffects were estimated in LeHavre, France (23–27%AGI increasesper 0:1NTU in the Saint Laurent plant) and Cherepovets, Russia(64% AGI increase per 0:27NTU), regions considered vulnerableto contamination during the time periods studied (Beaudeau et al.2012; Egorov et al. 2003). Other effects were disproportionallysmall, such as the 6% increase in AGI per 10-NTU increase in 3-d

maximum turbidity of AGI source (raw) water in Atlanta (Tinkeret al. 2010), and the 5% increase in AGI per approximately1:0-NTU increase of distribution system turbidity (turbidity inter-quartile range estimated from figure) found in New York City dur-ing the spring season (Hsieh et al. 2015). Interpretation of thesecomparisons is speculative, as the broad variation in effect sizesmay reflect differences in methods as well as true heterogeneity ofeffect among the studies.

In some studies, the association between turbidity and AGIincidence was limited to relatively high turbidity values. In thetwo studies that stratified data by season (Bateson 2001; Hsiehet al. 2015), positive findings were limited to the season with thehighest turbidity levels [winter, defined as October throughMarch in Bateson (2001); spring in Hsieh et al. ( 2015)]. In thespring season in New York City, when median distribution sys-tem turbidity levels were generally greater than 1:0NTU, Hsiehet al. (2015) found that this association was attenuated and loststatistical significance when excluding the top 20% of values.Further evidence of an association limited to higher turbidity lev-els comes from studies of the Milwaukee 1993 outbreak, in whichexclusion of the outbreak period, during which turbidity levelspeaked at 1:7NTU, resulted in loss of the association at theHoward Avenue (South) plant, which measured turbidity levelslower than 0:2NTU for most of the remaining study period(Morris et al. 1996, 1998; Naumova et al. 2003). Even during theoutbreak period, the association appeared strongest at the highestturbidity levels, with a sharp increase in AGI incidence with tur-bidity greater than 1:0NTU (estimated from TERS plots inMorris et al. 1998). The association observed in Cherepovets,Russia, a region with relatively high turbidity levels, was limitedto participants who consumed nonboiled tap water (Egorov et al.2003). Furthermore, this association only appeared at the higherturbidity levels observed in the study, with daily averages greaterthan approximately 0:9NTU (estimated from figure). Evidenceof nonlinearity was also found in Nantes, France, in a water sup-ply with very low overall turbidity, averaging 0:05NTU duringthe study period (Beaudeau et al. 2014b). In this study, there wasno clear association at turbidity levels lower than 0:045NTU,and a nearly linear positive association with AGI incidence at tur-bidity levels beyond this point. Complicating interpretation, somestudies found evidence that the association was restricted to lowerturbidity values. The association during winter season in Seattlewas observed only when the data were restricted to days withsource-water (unfiltered supply) turbidity levels <1NTU (80.5%of days) (Bateson 2001). In the study of an unfiltered water sup-ply in Eastern Massachusetts (Beaudeau et al. 2014a), the rela-tionship between turbidity and AGI showed increasing risk up toa turbidity level of 0:33NTU and no apparent additional increasein risk at higher levels.

Several studies examined source (raw) water turbidity in addi-tion to turbidity of finished water. In Nantes, France, source-water turbidity was significantly associated with AGI incidence,but without the consistency of effect across lags that was observedfor the association with finished (filtered) water (Beaudeau et al.2014b). In Atlanta, source-water turbidity was associated withdaily counts of AGI-related emergency department visits, but fin-ished (filtered) water turbidity was not (Tinker et al. 2010); theauthors reported that there was little correlation between turbiditymeasures from the source and finished water. Nevertheless,the pattern of association of 3-d moving average turbiditywith AGI in the Atlanta study (Tinker et al. 2010) was some-what similar between the minimum source-water turbidity andaverage finished water turbidity, with the effect rising from0-d lag to a peak at a 6- to 7-d lag, dipping to a low at an11- to 13-d lag, and rising again with a peak at 15-d lag


(estimated from figure), albeit with no statistically significanteffects for finished water. Similarly, the source-water turbidityof the Radicatal plant in Le Havre, France was strongly asso-ciated with AGI-related prescriptions, whereas the finished(filtered) effluent turbidity had a similar magnitude effect thatwas not statistically significant (Beaudeau et al. 2012). InEdmonton, there was consistency in the lack of an overallrelationship for both source-water and filtered-water turbiditywith AGI (Lim et al. 2002).The study by Hsieh et al. (2015),conducted in New York City, was the only study to includeturbidity measurements taken from within the distribution sys-tem. The authors found that an association between distribu-tion system water turbidity (unfiltered supply) and diarrhealevents, observed during the spring season, was almost com-pletely explained by the variation in source-water turbidity(Hsieh et al. 2015).

Season was an important confounder and every study adjustedfor it, by modeling covariates or by stratification. Hsieh et al.(2015) and Bateson (2001) presented only season-specific analy-ses. Other investigators demonstrated that they effectivelyadjusted for the nonlinear effects of season and other time trendsby showing plots of the residuals over the study period (Schwartzet al. 2000) or by describing noncorrelation of the residuals(Egorov et al. 2003; Tinker et al. 2010). Air temperature, day-of-the week, and holidays were also important confounders in moststudies. Several studies found that precipitation was not a con-founder of the association between turbidity and AGI (Aramini2002; Beaudeau et al. 2014b; Gilbert et al. 2006; Hsieh et al.2015; Lim et al. 2002) and other studies reported an associationbetween turbidity and AGI even with adjustment for precipitation(Bateson 2001; Tinker et al. 2010). The studies in Canada(Aramini 2002; Lim et al. 2002) evaluated residual confoundingfrom season and weather indirectly by conducting concurrenttime-series and case–control studies including the same AGIcases, with controls identified from healthcare visits for acute re-spiratory illness. The rationale behind control selection was thatturbidity (as a proxy for pathogens) was not expected to be acause of acute respiratory illness, but the seasonal pattern of thecontrol diagnoses was expected to be similar to AGI (peaking inthe winter months), thereby providing indirect adjustment forseasonal trends. Aramini (2002) found similar results in the case–control and time-series studies in Vancouver, with excess risk ofAGI peaking around the same turbidity levels and with similarlag times (as shown in TERS plots in the study), and Lim et al.(2002) reported null results with both study designs inEdmonton. The similarity of results with the two study designssuggests adequate adjustment of seasonal trends in the time-series analyses. Egorov et al. (2003) was the only study to collectindividual-level information on potential confounders, fromwhich they adjusted their time-series for behavioral factors suchas recreational water contact, attendance at summerhouses (withimplied lack of running water), and out-of-town trips. The associ-ation between drinking-water turbidity and AGI was strengthenedby adjustment for the individual behaviors (Egorov et al. 2003);however, potential biases from self-reported AGI in this studylimit interpretation.

There was limited and inconsistent evidence for effect mod-ification by age or AGI definition. Several studies that exam-ined multiple age groups found that associations were strongestin children (Beaudeau et al. 2014b; Hsieh et al. 2015; Morriset al. 1996; Tinker et al. 2010) or the oldest of the elderly(Bateson 2001; Schwartz et al. 2000), although another foundno association among the elderly (Hsieh et al. 2015). Somestudies conducted separate analyses for AGI counts based ondifferent case definitions. Morris et al. (1996) found generally

higher relative risks for the association of turbidity with AGIillness from emergency department visits and hospital admis-sions (combined) than from outpatient physicians’ office visitsin Milwaukee. In Philadelphia children, the association withturbidity did not differ remarkably between AGI defined fromemergency visits or hospitalizations (Schwartz et al. 1997). Theresults of the Vancouver study (Aramini 2002), analyzed byAGI definition and age group, do not show any clear patternsof effect modification; however, not all results were shown.

Several studies considered multiple variables for characteris-tics of source water, treatment, and water quality in their models,in addition to turbidity. The study in Eastern Massachusettsexamined confounding by other measurements taken in the watersystem, including fecal coliforms, UV-absorbance, algae, cyano-bacteriae, and water temperature, and found that adjustment forthese variables did not remarkably change the main associationof AGI incidence with turbidity of the unfiltered water supply(Beaudeau et al. 2014a). They did find, however, that algae-corrected turbidity was more strongly associated than overall tur-bidity with AGI-related hospital admissions, and resulted inimproved model fit. Source-water temperature was strongly andinversely associated with AGI, and replacement of the independ-ent model terms for algae-corrected turbidity and water tempera-ture with a bidimensional spline term (to model interactionbetween the two variables) also significantly improved the fit ofthe model (Beaudeau et al. 2014a). The study conducted inNantes, France found a strong association of AGI with finished(filtered) water turbidity on days with high river flow, and aweaker association at low flow, with a significant interactionbetween the two variables (Beaudeau et al. 2014b). In Le Havre,France, the model fit for finished effluent turbidity was signifi-cantly improved by inclusion of an interaction term indicatingdays with additional coagulation-flocculation-settling that wasimposed during times of high turbidity (Beaudeau et al. 2012).There was a statistically significant increased risk of AGI associ-ated with finished effluent turbidity on days without settling inboth children and adults, whereas the association was inconsis-tent between children and adults on days with the additional set-tling treatment.

Most studies evaluated lags between turbidity measurementand AGI outcome spanning from 1 to 13 or 14 d, and a few con-sidered lags extending beyond 20 d (Aramini 2002; Beaudeauet al. 2014a; Gilbert et al. 2006). Some investigators presentedresults for the entire span of lags examined, in tables or figures,such as three-dimensional TERS plots (Aramini 2002; Beaudeauet al. 2012; Egorov et al. 2003; Gilbert et al. 2006; Hsieh et al.2015; Lim et al. 2002; Morris et al. 1998). In other studies, lim-ited results were shown. Some studies took measures to avoidfalse positive results within the multiple testing framework, suchas by choosing an a priori lag structure for their main analysis(Beaudeau et al. 2014a; Tinker et al. 2010), only reporting onlags with statistical significance for multiple days in a row(Gilbert et al. 2006; Lim et al. 2002), or only reporting on thebest fitted model (Bateson 2001). Others addressed multiple com-parisons by estimating the likelihood of the total number of posi-tive associations observed, relative to the number expected underthe null hypothesis (Schwartz and Levin 1999; Schwartz et al.2000).

Specific lag times with notable associations are shown inFigure 1. Many of the studies found the most prominent or sig-nificant associations with turbidity exposure lagged by 6 to 10 d(Aramini 2002; Bateson 2001; Beaudeau et al. 2012, 2014a,2014b; Egorov et al. 2003; Hsieh et al. 2015; Morris et al.1998; Schwartz et al. 1997, 2000, Tinker et al. 2010). Somealso found associations with shorter (Egorov et al. 2003; Morris


Figure 1. This table provides information on lag days with associations reported in the epidemiological studies of drinking water turbidity in relation to acutegastrointestinal illness. Note: Dark gray shading indicates lag day with notable association (*indicates lagged multiday average); light gray shading indicateslag day not examined in the study. AGI, acute gastrointestinal illness; TERS, temporal exposure response surface.aStated by the author (comprehensive results not shown).bp<0.05.


et al. 1998; Schwartz et al. 1997) or longer lags (Aramini 2002;Beaudeau et al. 2012; Gilbert et al. 2006; Morris et al. 1998).

DiscussionA positive association between turbidity of drinking water andAGI incidence has been observed in different cities and time peri-ods, in regions with varying characteristics of source water, withboth unfiltered and filtered supplies, and with varying turbiditylevels. There is some consistency in the lag times with associa-tions (i.e., the time between the measured turbidity and identifica-tion of the AGI case), which fall between 6 and 10 d in manystudies. The studies appear to adequately adjust for possiblebiases of the time-series design, including confounding by sea-sonal cycles and other time trends, although it is acknowledgedthat unknown biases may still occur. There is evidence for nonli-nearity of the association from several studies, showing strongerassociations at higher levels of turbidity. The collective resultsreveal a broad association between turbidity and AGI that sug-gests low-level incidence of waterborne AGI from drinking waterwithin the systems and time periods studied.

The potential for causal interpretation was strongest fromstudies that presented results for all individual lags and found apattern of increased risk at several consecutive lags, rather than asingle peak (Aramini 2002; Beaudeau et al. 2012; Hsieh et al.2015). Causal inference was also strengthened when investigatorsfound similar results between their main analysis and with alter-nate modeling strategies (Aramini 2002; Hsieh et al. 2015;Schwartz et al. 1997; Schwartz et al. 2000) or with model valida-tion such as fitting and testing in a split sample (Beaudeau et al.2012, 2014b). The potential for causal inference was more lim-ited from studies that showed fewer, selected results (Bateson2001; Schwartz et al. 1997, 2000) or with results averaged acrossmultiple lag days, such as from distributed lag models (Tinkeret al. 2010). Findings of effects limited to low turbidity levels areinconsistent with the hypothesized causal relationship (Bateson2001; Beaudeau et al. 2014a), although not inconceivable. Somestudies had potential biases that rendered the results suggestive,at most, such as possible bias from self-reported AGI (Egorovet al. 2003) or diagnosis bias that could have followed newsreporting of the 1993 outbreak in Milwaukee, although this biaswould not have affected positive results found during the pre-outbreak period (Morris et al. 1996; Morris et al. 1998; Naumovaet al. 2003).

Given that the ratio of particulates to pathogens can varygreatly among water systems, turbidity alone may not beadequately correlated with microbiological contamination toserve as a useful proxy in every situation. The findings from NewYork City (Hsieh et al. 2015) and Seattle (Bateson 2001), inwhich an association was observed only in the season with thehighest turbidity, and those from Atlanta (Tinker et al. 2010) and

Le Havre, France (Beaudeau et al. 2012) in which there was asignificant association with source (raw)-water turbidity but notwith finished (filtered) water suggest that in some water sys-tems, the utility of turbidity as a proxy for microbiological con-tamination may be limited to high-turbidity scenarios.Nevertheless, as other studies found an association at low levelsof turbidity similar to those measured in Atlanta finished water(Beaudeau et al. 2014b; Schwartz et al. 1997, 2000), the dis-crepancies also suggest that the utility of turbidity as a proxymay be context specific—dictated by watershed characteristics,treatment system approaches and performance, and local envi-ronmental factors related to season. For this reason, future stud-ies would be most informative not by asking “Is there anassociation?” but, rather, “Under what conditions does an asso-ciation exist?”

The body of work demonstrates the efficacy of studies corre-lating turbidity with AGI counts in time-series for preliminaryinvestigation of the safety of water supplies. However, theapparent context-specific nature of turbidity in its associationwith AGI has implications for the optimal conduct of such stud-ies, including design issues and modeling strategies aimed to-ward elucidating specific- rather than generalized associations.Additionally, based on our review, we propose recommenda-tions for enhancing the potential for causal interpretation fromfuture studies.

Exposure MeasurementTurbidity measurements for the purpose of monitoring filtrationperformance are generally taken using continuously operating,automated turbidimeters at each individual filter [as per U.S.EPA guidance (U.S. EPA 2004)]. The measurement of turbidityusing automated turbidimeters is known to be imprecise, particu-larly at low levels (Burlingame et al. 1998). However, averagingof multiple measurements per day for use in time-series regres-sion would result in a summary measure with lower error than anindividual measurement. Some studies used longer averagingperiods than 1 d, such as Morris et al. (1996) who examinedMilwaukee water 2-wk average turbidity (lagged by 1 wk) inrelation to 2-wk AGI counts, or Tinker et al. (2010) who analyzedAtlanta water turbidity as a 3 d moving average across lags.Although the error of individual measurements is greatlydecreased with such an approach, the loss of variation by averag-ing over a longer time period may result in dilution of anytrue association. Two studies relied on a single turbidity measure-ment for their defined daily exposure—the daily maximum(Bateson 2001, Naumova et al. 2003)—and these single measure-ments are more subject to error than averages. The error and re-sultant misclassification of turbidity measurements would occurnondifferentially with respect to the AGI outcome (when consid-ering a linear relationship), but with left-truncation of low-level

Figure 1. (Continued.)


turbidity values at 0, it is unclear how such measurement errormight bias the results. Evaluation of both short-term averages(such as daily) and single measurements (such as maximum orminimum) of turbidity may be a reasonable approach to fittingthe most predictive model in a study, when coupled with evalua-tion of the influence of the averaging period and extreme outliers.

People consume water outside of their home, and potentiallyoutside their service area, and also consume bottled water.Individual use patterns may result in nonrepresentativeness ofone regional turbidity value for exposure of individual cases; thisis an inherent weakness given the ecological nature of data usedin the time-series design. The lack of individual-level informationon nontap water consumption and resultant exposure misclassifi-cation detracts from evaluation of the etiologic associationbetween microbiological quality of a drinking-water supply (asindicated by turbidity as a proxy measure) and AGI incidence.However, it should be noted that the lack of information does notdetract from the results of these studies for estimation of thepopulation-wide AGI risk associated with regional drinking-water quality, given real patterns of use, the interpretation mostrelevant for management of a specific water system.

Because the utility of turbidity as an exposure proxy formicrobiological contamination may be context specific, epidemi-ological research would be enhanced by combining turbidity withadditional seasonal and climatic variables, water quality, andtreatment measures to optimize predictive models of AGI inci-dence. Indeed, the association between turbidity and AGI differedby season in the only studies that reported this stratification(Bateson 2001; Hsieh et al. 2015), and studies that evaluated mul-tiple water indices found significant interactions of turbidity withfactors such as streamflow, water temperature, and systems oper-ations in their associations with AGI (Beaudeau et al. 2012,2014b). Climatic variables considered as confounders in the stud-ies, such as season and air temperature, could theoretically deter-mine the conditions in which turbidity is most correlated withmicrobiological contamination, and as such these variablesshould first be evaluated as effect modifiers, and secondly asconfounders.

Two studies suggest that the association between turbidityand AGI is more significant for source (raw) water than finished(filtered) water (Tinker et al. 2010; Beaudeau et al. 2012). Thismight seem counterintuitive as finished water would be expectedto be more representative of human exposure than raw water(except given exposure by recreational contact). The findingsmay be spurious or they may be confounded by recreational con-tact with surface waters, but they may alternatively be interpretedas indicating that turbidity of raw water may be more stronglycorrelated with microbiological contamination than the turbidityof filtered water. One would expect raw water pathogen loads tovary greatly, even over orders of magnitude, which should exceedthe variability following an engineered and tightly monitored pro-cess such as water filtration. Again, with the goal of elucidatingthe specific conditions under which an association exists, rawwater turbidity should be routinely evaluated, in addition to, andin combination with finished water turbidity.

Outcome AssessmentNondifferential misclassification of AGI is certainly present inthe studies. Many of the studies identified cases based on a broadgrouping of diagnosis codes for infectious gastroenteritis andalso included a catch-all code for noninfectious AGI, as well assymptom codes. The grouping of specific diagnoses into broadcategories (such as all infectious gastroenteritis including ICD-9codes 001–009) results in less misclassification for the group,overall, than for specific diagnoses. The most representative set

of diagnosis codes for waterborne AGI has not been determined;however, studies that evaluated the sensitivity of study results tovariations in case definition found little impact on their findings(Schwartz et al. 1997; Tinker et al. 2010). More extensive evalua-tion of the sensitivity of results to various case definitions wouldbe prudent, given the expected misclassification in diagnosiscodes.

The AGI outcome definitions are also subject to undercount-ing, resulting from lack of sensitivity of the data sources used.For example, although Medicare claims can be expected to befairly complete in capturing hospital admissions for persons≥65 y in the United States, hospitalizations are not a completerepresentation of all incident AGI cases. Other sources of data,such as prescriptions for AGI-related drugs or calls to a health in-formation line are also likely to represent only a fraction of allAGI cases on any given day. Undercounting was reflected in theSchwartz et al. (1997) study in Philadelphia, in which AGI casesin children were identified through emergency department visitsto one hospital, and rates differed considerably between the watertreatment service area located closest to the hospital (18.9 per1,000 person-years) and the service area located farthest away(1.5 per 1,000). Such undercounting of AGI, likely common toall the studies, reduces the study power but will not cause bias, aslong as the variation in counts over time is representative of thevariation in the underlying true counts (i.e., day-to-day increasesor decreases in the number of AGI-related hospitalizations arerepresentative of day-to-day changes in the true underlying num-ber of AGI cases). This assumption may be reasonable unless tur-bidity is related to the severity of AGI symptoms. In thissituation, AGI identified from data sources capturing more severecases (e.g., hospitalizations, emergency department visits) maybe disproportionately represented. This suggests that studiesshould explore effect modification by the sources of data used forAGI identification, in order to detect associations that may differby case severity.

ConfoundingIt is difficult to conceive of a variable not examined in the studiesthat could plausibly confound the relationship between turbidityand AGI incidence to the extent that it would produce a false pos-itive association in a consistent manner between the many studyregions and time periods examined. Season is a major factor ofconcern for confounding, as AGI exhibits seasonal cycles thathave nothing to do with microbiological water quality, such asfrom low air exchange rates in winter that make indoor transmis-sion of pathogens more common, and season (and related climaticfactors) is also highly correlated with source-water turbidity.Most studies used flexible nonlinear terms to estimate seasonaltrends and potential confounding effects of air temperature orother climatic variables, and several studies described adequateadjustment based on independence of the residuals over time(Egorov et al. 2003; Schwartz et al. 2000; Tinker et al. 2010). Inother studies there was little information reported on the model fitwith respect to the covariates. Because the studies in Vancouverand Edmonton, Canada (Aramini 2002; Lim et al. 2002) foundsimilar patterns of association of turbidity with AGI in both time-series and case–control analyses, this suggests that seasonal trendswere accounted for in the time-series analyses, and also stronglyargues against false positive findings in the Vancouver study. Inaddition to primary consideration of seasonal and climatic factorsas effect modifiers in future analyses (discussed above) and sec-ondarily as confounders, improved reporting of the adequacy ofseasonal trends adjustment would enhance causal interpretation ofthe results.


Multiple studies have reported an association between heavyprecipitation and increases in AGI incidence using the time-seriesstudy design (Guzman Herrador et al. 2015). Overadjustment is aconcern, as precipitation directly affects turbidity levels by caus-ing runoff of pathogens into source waters (thus, precipitationmay be a proxy for turbidity). Nevertheless, adjustment for pre-cipitation was inconsequential in the studies that examined it,possibly suggesting that other local factors influence turbidity toa greater extent than precipitation. It seems most prudent toexclude precipitation from models of the turbidity–AGI relation-ship as it may interfere with accurate estimation of the turbidityeffect.

An advantage of the time-series design is its inherent adjust-ment for potential confounders that do not greatly vary by thetime scale used in the analysis (e.g., such as socioeconomic statusand neighborhood factors, which do not vary by day); for thisreason, individual-level measurements are often not needed.Despite this practical advantage, targeted data collection onpotential confounders would be useful to inform the evidencebase for causal inference from studies of drinking-water turbidityand AGI incidence. The hybrid design employed by Egorovet al. (2003) in Russia illustrates collection and examinationof individual-level information (recreational water contact,consumption of nonboiled tap water, out-of-town trips) forinclusion in a time-series analysis; however, the small size ofthe study (367 participants), which allowed for data collec-tion, also limits study power and is likely not a feasible strat-egy in regions with lower rates of AGI and more pristinedrinking-water conditions. Rather than individual-level datacollection from each person in a study, efforts could insteadfocus on data collection from a representative sample of per-sons in the region to obtain information on the variation ofpotential confounders over time. For example, recreationalcontact with source waters is known to pose an increased riskof AGI (Sunger and Haas 2015). Such contact would vary byseason, and could thus confound the association betweendrinking-water turbidity and AGI. The sheer number of per-sons exposed through drinking water greatly overwhelms thenumber of persons exposed through recreational contact,which argues against the potential for unmeasured recrea-tional water contact to cause a spurious association betweenturbidity and AGI. Nevertheless, studies would benefit fromcollection of additional information on recreational watercontact by regional residents, such as the frequency and var-iability of contact over time—by season, day-of-the week,and holidays. This distributional information would allowfor application of probabilistic bias analysis methods toplace bounds on the possible bias from the (unmeasured)confounder on the risk estimates for the main effect of inter-est (Lash et al. 2009).

Evaluation of Multiple LagsAll of the studies evaluated multiple lag times to test multiplehypotheses regarding the latency between exposure and AGI.The lags represent the elapsed time from the point of sampling(usually as effluent from the water utility) until presentation of anAGI case to the health system, theoretically incorporating thetime before water reaches the consumer’s tap, incubation periodsfor common waterborne microbiological infections, and delay inseeking medical care for AGI. Given the ecological nature of thetime-series study design, the lag time represents the average lagamong AGI cases within the region studied. Testing multiple lagsis admittedly exploratory, but necessary given that differentpathogens have different incubation periods and an investigatortypically does not have knowledge at the outset about the

pathogens of concern within a particular water system. Many ofthe studies found the most significant or prominent associationswith lag times of 6–10 d. Lag times from 6 to 10 d may be con-sistent with latency periods for common waterborne AGI, such asviruses, giardiasis, and cryptosporidiosis, but this depends on theamount of time water spends in each of the distribution systemsstudied. The study of the Milwaukee outbreak, with known con-tamination by Cryptosporidium, found prominent lags between 6and 10 d, as well as longer lags of 13–16 d, which the authorsspeculated may reflect secondary transmission of the initialwaterborne outbreak (Morris et al. 1998; Naumova et al. 2003).Differing lag times between studies may suggest different water-borne pathogens or distribution system residence times amongthe study regions; however, inference regarding specific patho-gens will ultimately rely on identification of specific agents inwater samples.

In this type of analysis in which associations are exploredacross multiple lags, the pattern of association may be more im-portant than the statistical significance of individual associations.The most satisfying presentations of multiple lags tested showedthe entire range of results for the association between turbidityand AGI at all lag times examined, in tables or figures (Aramini2002; Beaudeau et al. 2012; Egorov et al. 2003; Gilbert et al.2006; Hsieh et al. 2015; Lim et al. 2002; Morris et al. 1998). Thiscomprehensive presentation allows the reader to assess whetherassociations represent generalized patterns of increasing riskacross multiple lags or appear as single associations showing nopattern with consecutive lags. A separate issue arises from sum-marizing results over multiple lags examined, as from a distrib-uted lag model, which may dilute any association limited to afew consecutive days. Comparison of lag times between differentstudies would be optimized by consistency in presentation of allresults, such as for single-day lags over a period of at least 2 wkspanning lags from 6 to 10 d.

Modeling StrategiesWith some exceptions (Beaudeau et al. 2014a; Hsieh et al. 2015;Schwartz et al. 2000), plots of the raw data and its relationship tothe fitted model, or residual diagnostic plots or other diagnosticinformation were not shown, making it difficult for the reader toascertain whether the model is a good fit. Authors rarelydescribed formal testing of the Poisson distributional assumptionthat the sample variance equals its mean, which may be violatedwhen spikes occur in AGI counts, although multiple authorsemployed alternate models equipped to handle overdispersedcount data. The linearity of the relationship between AGI andlagged turbidity is another assumption that was not explicitly jus-tified in most of the studies, and evidence for a nonlinear associa-tion from several studies underscores the need for evaluation ofthe suitability of a linear fit to the relationship within a particularwater supply. Additionally, in most studies, there was little infor-mation reported on the model fit with respect to the covariates.Imperfect model fit with respect to turbidity or the covariatesdoes not necessarily undermine basic conclusions about the exis-tence of a relationship between AGI and lagged turbidity, butmay impact estimates of this relationship, for example by intro-ducing some bias into parameter estimates or by invalidating esti-mated standard errors and thus the stated level of significance,which can lead to false conclusions either away from or towardthe null. Improved reporting of distributional assumptions andmodel fit would allow greater confidence in interpretation ofresults. Special attention should be paid in the studies to the fit ofmodels with respect to extreme values (spikes) of either turbidityor AGI.


Future DirectionsMoving forward, it may be useful to conduct additional studies ofturbidity in relation to AGI to screen drinking-water supplies forsource-water quality and treatment effectiveness. Time-seriesstudies are recommended as a reasonable first-step for epidemio-logical evaluation of a water supply; although the time-seriesdesign has weaknesses that limit causal inference, such studiesare relatively inexpensive to conduct, given ongoing generationof turbidity (and other water quality) measurements by water sys-tems and increasing availability of electronic health records.These studies may help identify regions, seasons, and source-water conditions of potential concern, which could then guidemore targeted research and data collection to help explain thosehigh-risk conditions. Targeted data collection would be useful toinform the evidence base for causal inference from time-seriesstudies of drinking-water turbidity and AGI incidence, such asdata to better understand water use or confounder distributionsover time, or through water sampling for pathogens. Samplingefforts to describe the relationship between turbidity and patho-gens across regions under various conditions (different source-waters, climatic conditions, and treatment approaches) may addvaluable information about the specific contexts in which turbid-ity is most useful as a proxy for microbiological contamination.Ultimately, the collective results from a variety of preliminarystudies may help to effectively allocate funding for more exten-sive studies (e.g., randomized trials) to regions most likely tobenefit from the information.

There are important gaps in the conditions that have beenstudied. Cities with pristine source-waters that employ filtrationhave not been well studied. This may reflect a tendency amongresearchers to undertake, and among funding agencies to sponsorstudies in areas in which positive associations are likely to befound. This approach is reasonable initially when the existence ofany sort of association would be unclear, but as a substantialnumber of studies show positive associations, the scope of futurestudies should be broadened so that the conditions under whichsuch associations exist can be better demarcated. Following thisline of reasoning, a priority for future study is the evaluation ofsystems employing enhanced disinfection, such as UV. An asso-ciation between turbidity and AGI has been found in chlorinatedand ozonated supplies; the association might not be present whenUV disinfection is used, given that this form of disinfectionis effective even against highly resistant pathogens such asCryptosporidium oocysts (Nasser 2016). Three of the treatmentplants studied in Atlanta (Tinker et al. 2010) employed UV disin-fection, but the paper did not list the type of treatment specific toeach water plant. New York City and Boston are particularlyattractive locations for studies of UV disinfection, as the methodhas been recently adopted in those regions, and previous studiesfound an association between turbidity and AGI (Beaudeau et al.2014a, Hsieh et al. 2015). Hence, a subsequent study could assesswhether the implementation of UV disinfection has removed theobserved association between turbidity and illness. Studies atsites using other advanced disinfection methods might be simi-larly valuable. Multi-region investigations using unified methodswould be an efficient approach to further our understanding ofthe relationship, particularly given contrasts of interest betweenstudy regions (such as in turbidity levels or treatment methods).

The studies we reviewed were generally designed to evaluateAGI risks in relation to drinking-water turbidity as a proxy formicrobiological contamination of source-water that is notfully eliminated by treatment. These studies (with the exceptionof Hsieh et al. 2015) did not evaluate AGI risk caused bymicrobiological contamination introduced within the distributionsystem (such as through main breaks, intrusions, and biofilms).

Distribution system contamination has been implicated in water-borne disease outbreaks (Ford 2016), and furthermore may be acause of endemic (nonoutbreak) waterborne AGI, as suggestedby an epidemiological study linking longer water residence timein the distribution system with increased AGI incidence inAtlanta (Tinker et al. 2009). In future research, it may be usefulto partition out health risks between various inputs to the watersystem (i.e., potential points of contamination), in order to under-stand more fully where to target interventions. Hsieh et al. (2015)addressed the question by simultaneously modeling both distribu-tion system turbidity and source-water turbidity in relation toAGI. The association with distribution turbidity was almost com-pletely explained by source-water turbidity (Hsieh et al. 2015),suggesting that the fraction of AGI associated with turbidity wasnot caused by contamination introduced within the distributionsystem. Likewise, Beaudeau et al. (2014b) found significantlyincreased risk associated with turbidity of finished effluent, withsimultaneous adjustment for pipe break service interventions(which itself had a small, but not statistically significant relation-ship with AGI). Additional studies with multivariable evaluationof source-water/effluent turbidity in addition to indicators ofpotential contamination within the distribution system may shedlight on the relative importance of contamination inputs from var-ious points in the water system.

Our understanding of the relationship between drinking-waterturbidity and AGI incidence would be improved through consis-tency of analysis and reporting in future studies. Greater consis-tency will more readily allow direct comparison of results acrossregions, water systems, and time periods, as well as enable meta-analyses to quantitatively summarize the effect and evaluate fac-tors contributing to heterogeneity of effect. To enhance consis-tency, we suggest evaluation of daily average turbidity (ratherthan longer averaging periods or alternate turbidity metrics) andevaluation of daily lags with presentation of results for all lagtimes examined. Although authors may determine that longer ex-posure or lag periods lead to the best-fitted model in their study,presentation of the results for daily average turbidity and dailylag times in a supplement or by request would allow comparisonof results, as stated. We also suggest fitting both linear and non-linear associations between turbidity and AGI incidence, withpresentation of the linear association, where appropriate, even iflimited to certain levels of turbidity or one season. The nonlinearassociation across multiple lags is effectively summarized usingTERS plots; however, providing risk estimates (and confidenceintervals) for the contrast between defined turbidity levels at keylag times is needed to allow accurate comparison of the magni-tude of effect across studies. Quantitative comparison (and sum-marization) of results across studies may add to the weight ofevidence for a causal interpretation of waterborne AGI fromdrinking water in the regions and time periods studied, and couldultimately lead to more accurate estimation of the total risk (at-tributable risk) through this exposure pathway.

ConclusionsIn summary, multiple time-series studies have observed an associa-tion between turbidity of drinking-water supplies and risk of AGI.Associations have been observed in unfiltered and filtered watersystems and at levels of relatively high and low turbidity. The posi-tive studies suggest an underlying risk of waterborne AGI duringthe time periods at which the systems were studied. Nevertheless,inconsistencies between the studies indicate that the utility of turbid-ity as a proxy for microbiological contamination may be contextspecific. The body of work demonstrates the efficacy of studies cor-relating turbidity with AGI counts in time-series for preliminaryinvestigation of the safety of water supplies. The context-specific


nature of the association between drinking-water turbidity and AGIsuggests that future research will be most effective if strategizedtowards elucidating specific- rather than generalized associations;for example, through exploration of effect modification by seasonalvariables, stream conditions, and water treatment. Time-series resultssupplemented by targeted data collection to help determine whetherthere is indeed a causal link may be useful as a means to assess theeffectiveness of utilities in managing various conditions posingincreased risk for exposure to microbiological agents of AGI.

ReferencesAmerican Water Works Association (AWWA). 2014. Partnership for Safe Water

Guidelines for Phase IV Application for the “Excellence in Water Treatment”Award. Denver, CO:AWWA.

Aramini JJ. 2002. The Use of Temporal-Spatial Analyses in the EpidemiologicalInvestigation of Endemic and Epidemic Waterborne Gastroenteritis [Dissertation].Ontario, Canada:University of Guelph.

Aramini J, McLean M, Wilson J, Holt J, Copes R, Allen B, et al. 2000. Drinkingwater quality and health-care utilization for gastrointestinal illness in GreaterVancouver. Can Commun Dis Rep 26(24):211–214, PMID: 11211600.

Bateson TF. 2001. Control of Time-Varying Confounding in Studies of EnvironmentalTime Series [Dissertation]. Boston, MA:Harvard University.

Bateson T, Levin R, Koenig J, Schwartz J. 2000. Unfiltered drinking water turbidityand gastrointestinal illness in Seattle's elderly [Abstract]. Presented at TwelfthConference of the International Society for Environmental Epidemiology. Buffalo,New York. 19–22 August 2000.

Beaudeau P, Le Tertre A, Zeghnoun A, Zanobetti A, Schwartz J. 2012. A time seriesstudy of drug sales and turbidity of tap water in Le Havre, France. J WaterHealth 10(2):221–235, https://doi.org/10.2166/wh.2012.157.

Beaudeau P, Payment P, Bourderont D, Mansotte F, Boudhabay O, Laubies B, et al.1999. A time series study of anti-diarrheal drug sales and tap-water quality. Int JEnviron Health Res 9(4):293–311, https://doi.org/10.1080/09603129973092.

Beaudeau P, Schwartz J, Levin R. 2014a. Drinking water quality and hospital admis-sions of elderly people for gastrointestinal illness in Eastern Massachusetts,1998–2008. Water Res 52:188–198, PMID: 24486855, https://doi.org/10.1016/j.watres.2014.01.005.

Beaudeau P, Zeghnoun A, Corso M, Lefranc A, Rambaud L. 2014b. A time seriesstudy of gastroenteritis and tap water quality in the Nantes area, France,2002–2007. J Expo Sci Environ Epidemiol 24(2):192–199, PMID: 23443240,https://doi.org/10.1038/jes.2013.5.

Burlingame GA, Pickel MJ, Roman JT. 1998. Practical applications of turbidity mon-itoring. J Am Water Works Assoc 90:57–69.

Colford JM, Roy S, Beach MJ, Hightower A, Shaw SE, Wade TJ. 2006. A review ofhousehold drinking water intervention trials and an approach to the estimation ofendemic waterborne gastroenteritis in the United States. J Water Health 4(suppl 2):71–88, PMID: 16895086, https://doi.org/10.2166/wh.2006.018.

CWWA (Canadian Water Works Association). 2002. Drinking Water Disinfection andTurbidity Requirements – A Global Perspective. Ottawa, Ontario, Canada:CWWA.

Egorov AI, Naumova EN, Tereschenko AA, Kislitsin VA, Ford TE. 2003. Daily variations ineffluent water turbidity and diarrhoeal illness in a Russian city. Int J Environ HealthRes 13(1):81–94, PMID: 12745350, https://doi.org/10.1080/0960312021000071567.

Ford T. 2016. Water and Health. In: Environmental Health: From Global to Local.Frumkin H, ed. San Francisco:Jossey-Bass, 413–450.

Gilbert M, Levallois P, Rodriguez M. 2006. Use of a health information telephoneline, Info-Santé CLSC, for the surveillance of waterborne gastroenteritis. JWater Health 4(2):225–232, PMID: 16813015.

Guzman Herrador BR, de Blasio BF, MacDonald E, Nichols G, Sudre B, Vold L, et al.2015. Analytical studies assessing the association between extreme precipita-tion or temperature and drinking water-related waterborne infections: a review.Environ Health 14:29, PMID: 25885050, https://doi.org/10.1186/s12940-015-0014-y.

Hsieh JL, Nguyen TQ, Matte T, Ito K. 2015. Drinking water turbidity and emer-gency department visits for gastrointestinal illness in New York City, 2002–2009. PLoS One 10:e0125071, PMID: 25919375, https://doi.org/10.1371/journal.pone.0125071.

Lash, TL, Fox, MP, Fink AK. 2009. Applying Quantitative Bias Analysis toEpidemiologic Data. New York:Springer.

LeChevallier MW, Norton WD, Lee RG. 1991a. Giardia and Cryptosporidium spp. infiltered drinking water supplies. Appl Environ Microbiol 57(9):2617–2621, PMID:1768135.

LeChevallier MW, Norton WD, Lee RG. 1991b. Occurrence of Giardia andCryptosporidium spp. in surface water supplies. Appl Environ Microbiol57(9):2610–2616, PMID: 1822675.

Lim G, Aramini J, Fleury M, Ibarra R, Meyers R. 2002. Investigating the Relationshipbetween Drinking Water and Gastro-enteritis in Edmonton, 1993–1998. Ottawa,Ontario, Canada:Health Canada.

Mann AG, Tam CC, Higgins CD, Rodrigues LC. 2007. The association betweendrinking water turbidity and gastrointestinal illness: a systematic review.BMC Public Health 7:256–263, PMID: 17888154, https://doi.org/10.1186/1471-2458-7-256.

Messner M, Shaw S, Regli S, Rotert K, Blank V, Soller J. 2006. An approach fordeveloping a national estimate of waterborne disease due to drinking waterand a national estimate model application. J Water Health 4(suppl 2):201–240,PMID: 16895092, https://doi.org/10.2166/wh.2006.024.

Morris RD, Naumova EN, Griffiths JK. 1998. Did Milwaukee experience waterbornecryptosporidiosis before the large documented outbreak in 1993? Epidemiology9(3):264–270, PMID: 9583417.

Morris RD, Naumova EN, Levin R, Munasinghe RL. 1996. Temporal variation indrinking water turbidity and diagnosed gastroenteritis in Milwaukee. Am JPublic Health 86(2):237–239, PMID: 8633742.

Nasser AM. 2016. Removal of Cryptosporidium by wastewater treatment proc-esses: a review. J Water Health 14(1):1–13, PMID: 26837825, https://doi.org/10.2166/wh.2015.131.

Naumova EN, Egorov AI, Morris RD, Griffiths JK. 2003. The elderly and waterborneCryptosporidium infection: gastroenteritis hospitalizations before and duringthe 1993 Milwaukee outbreak. Emerging Infect Dis 9(4):418–425, PMID:12702220, https://doi.org/10.3201/eid0904.020260.

Schwartz J, Levin R. 1999. Drinking water turbidity and health. Epidemiology 10(1):86–90, PMID: 9888287.

Schwartz J, Levin R, Goldstein R. 2000. Drinking water turbidity and gastrointestinalillness in the elderly of Philadelphia. J Epidemiol Community Health 54(1):45–51, PMID: 10692962.

Schwartz J, Levin R, Hodge K. 1997. Drinking water turbidity and pediatric hospitaluse for gastrointestinal illness in Philadelphia. Epidemiology 8(6):615–620,PMID: 9345659, https://doi.org/10.1097/00001648-199710000-00001.

Sunger N, Haas CN. 2015. Quantitative microbial risk assessment for recreationalexposure to water bodies in Philadelphia. Water Environ Res 87(3):211–222,PMID: 25842531.

Tinker SC. 2007. Drinking Water and Gastrointestinal Illness in Atlanta, 1993–2004[Dissertation]. Atlanta, GA:Emory University.

Tinker SC, Moe CL, Klein M, Flanders WD, Uber J, Amirtharajah A, et al. 2009.Drinking water residence time in distribution networks and emergency depart-ment visits for gastrointestinal illness in Metro Atlanta, Georgia. J WaterHealth 7(2):332–343, PMID: 19240359, https://doi.org/10.2166/wh.2009.022.

Tinker SC, Moe CL, Klein M, Flanders WD, Uber J, Amirtharajah A, et al. 2010.Drinking water turbidity and emergency department visits for gastrointestinalillness in Atlanta, 1993–2004. J Expo Sci Environ Epidemiol 20(1):19–28, PMID:18941478, https://doi.org/10.1038/jes.2008.68.

U.S. EPA (U.S. Environmental Protection Agency). 2004. Long Term 1 EnhancedSurface Water Treatment Rule Turbidity Provisions: Technical GuidanceManual. Office of Water (4606M) EPA 816-R-04-007. August 2004.

U.S. EPA (U.S. Environmental Protection Agency). 2015. National Primary DrinkingWater Regulations. 40 CFR Part 141. RIN 2040–ZA26. Fed Reg 82(7):3518–3552.


https://www.ncbi.nlm.nih.gov/pubmed/11211600

https://doi.org/10.2166/wh.2012.157

https://doi.org/10.1080/09603129973092


https://doi.org/10.1016/j.watres.2014.01.005

https://doi.org/10.1016/j.watres.2014.01.005


https://doi.org/10.1038/jes.2013.5


https://doi.org/10.2166/wh.2006.018


https://doi.org/10.1080/0960312021000071567



https://doi.org/10.1186/s12940-015-0014-y


https://doi.org/10.1371/journal.pone.0125071

https://doi.org/10.1371/journal.pone.0125071




https://doi.org/10.1186/1471-2458-7-256

https://doi.org/10.1186/1471-2458-7-256


https://doi.org/10.2166/wh.2006.024




https://doi.org/10.2166/wh.2015.131

https://doi.org/10.2166/wh.2015.131


https://doi.org/10.3201/eid0904.020260




https://doi.org/10.1097/00001648-199710000-00001



https://doi.org/10.2166/wh.2009.022


https://doi.org/10.1038/jes.2008.68

Review is available at conformant HTML ... - US Forest … States. (Colford et al. 2006 ... Proxy...

Documents

Transcript of Review is available at conformant HTML ... - US Forest … States. (Colford et al. 2006 ... Proxy...