68 © IWA Publishing 2011 Journal of Water, Sanitation and Hygiene for Development | 01.

1 | 2011

Comparison and verification of four field-based

microbiological tests: H2S test, Easygel®, Colilert®,
PetrifilmTM
Patty Chuang, Stephanie Trottier and Susan Murcott

ABSTRACT

The UN defines water supplies as ‘improved’ or ‘unimproved.’ These indicators are easy to measure, Patty Chuang
301 Heidinger Drive,
but do not reflect water quality, which requires laboratory or field tests. Laboratory and test Cary, North Carolina, 27511,
USA
availability, expense and technical capacity are obstacles for developing countries.
This research compares and verifies four low-cost, field-based microbiological tests: the EC-Kit Stephanie Trottier
4110 Marlowe,
(Colilert® and Petrifilm™ tests), the H2S bacteria test, and Easygel®, against a standard method Montreal, Quebec, H4A 3M2
Canada
(Quanti-Tray®). The objectives are to: (1) verify the accuracy of the four field-based tests, (2) study the
Susan Murcott (corresponding author)
accuracy of these tests as a function of improved and unimproved sources; (3) recommend a single
Mass. Institute of Technology,
microbiological test, if appropriate, based on accuracy and cost, and/or (4) recommend a testing Civil and Environmental Engineering Dept.,
77 Massachusetts Ave.,
combination, if appropriate, based on accuracy and cost. 1-138 Cambridge, MA 02139,
USA
The tests of 500þ total water samples from Capiz Province, Philippines and Cambridge, MA E-mail: murcott@mit.edu

indicate that two-tests systems gave better results than a single test. Both the 100-mL H2S test þ
Petrifilm™ and the 20-mL H2S test þ Easygel® combinations yield promising results, in addition to
being inexpensive. None of the field-based tests should be used on their own. We recommend
further verification of a larger sample size and scale be undertaken before these testing
combinations are recommended for wider use.
Key words | Colilert®, drinking water, Easygel®, H2S, Petrifilm™, Quanti-Tray®

INTRODUCTION proportion of the population that uses an improved drinking

water source, and the proportion that uses an improved sani-
The United Nations (UN) describes water as ‘indispensable for tation facility (United Nations ). The ideal solution to
leading a life of human dignity’ and ‘a prerequisite for the this problem is to provide a drinking water supply that is
realization of other human rights’. Still, 13% of the world’s both reliable and safe. According to the UN, access to drinking
population, or 884 million people, lack access to an improved water means that the source is less than one kilometre away
drinking water source (WHO/UNICEF ) and, every year, from its place of use, and that it is possible to reliably obtain
at least 1.8 million people – the majority of them children under at least 20 litres of water per member per household per day.
the age of five (WHO/UNICEF ) – die from diarrhoeal The UN also defines safe drinking water as water with
diseases related to unsafe water, sanitation and hygiene. microbial, chemical and physical characteristics that meet
The UN has therefore developed Millennium Develop- World Health Organization (WHO) guidelines or national
ment Goal Target 7C, which aims to halve the proportion standards on drinking water quality (United Nations ).
of people without sustainable access to adequate water Although the accessibility and reliability of a drinking
and sanitation by the year 2015 (United Nations ). water source are relatively easy to ascertain, determining the
The indicators used for these numbers include the safety of a drinking water source requires the performance of
doi: 10.2166/washdev.2011.026
 69 P. Chuang et al. | Comparison of four field-based microbiological tests Journal of Water, Sanitation and Hygiene for Development | 01.1 | 2011

drinking water quality tests. The UN currently defines water and piped supplies) used throughout the province,
sources as ‘improved’ or ‘unimproved’, which does not with the exception that the Roxas City municipal water
necessarily reflect the actual quality of the drinking water. treatment plant sent treated water samples to a laboratory
However, laboratory testing is especially difficult in develop- on another island for regular testing. During Fall 2008, Dr
ing countries where funds, technology, laboratory facilities Jarvis Punsalan, MD, MPH, Director of Public Health
and trained personnel are in short supply. Also, water qual- (DPH) head of the Capiz Province Provincial Health
ity testing raises even more questions: What contaminants Office (PHO), received funding from the European Commis-
should the drinking water be tested for and what test(s) sion, the Philippines’ government’s Department of Health
should be used? Which sources should be sampled, how and United Nations Children’s Fund (UNICEF) to set up a
many of them should be sampled, and how often? water quality testing laboratory at Roxas Memorial Hospital,
Although there are many types of drinking water contami- in Roxas City, which would test for microbiological contami-
nation, this study will solely focus on microbiological nation. He contacted Susan Murcott, Senior Lecturer at the
contamination of drinking water, which occurs when drinking Massachusetts Institute of Technology (MIT), for advice on
water is contaminated at the source by human or animal faeces, the types of microbiological drinking water quality tests to
or through inappropriate transportation, handling or storage in conduct, and she recommended two types of tests: IDEXX
vessels in the household. This research aims to verify and com- Quanti-Tray and the EC-Kit. These tests will be described
pare the accuracy of four field-based tests (10-mL pre-dispensed more fully in the following sections.
Colilert®, 3M™ Petrifilm™, hydrogen sulfide (H2S) bacteria During 2009, Capiz’s PHO purchased EC-Kits and
® ®
test and Easygel ) against a standard method (Quanti-Tray ). Quanti-Tray tests. An incubator, ultraviolet (UV) light and
The drinking water sources tested were located in Capiz Quanti-Tray sealer were also purchased in order to conduct
Province, Philippines, and in Cambridge, Massachusetts. the Quanti-Tray tests. In May 2009, a Filipino non-profit
organization, ‘A Single Drop’, trained the Capiz PHO staff,
Project area municipal health officers and sanitary inspectors (SIs) on
how to sample water sources, use the EC-Kit and interpret
The Philippines is an archipelago made up of more than the sample results. The Quanti-Tray equipment finally
7000 islands, located in Southeast Asia. The population of arrived in November 2009 and, as part of that purchase,
the Philippines is estimated at almost 98 million as of July the laboratory staff of the PHO’s Roxas City office received
2009, making it the 12th most populated country in the training from the suppliers in the set-up and use of the
world. The country has a high rate of food- and water-related Quanti-Tray system. During October to December 2009, in
diseases such as bacterial diarrhoea, hepatitis A, typhoid collaboration with the MIT team, the PHO developed a
fever, dengue, malaria and Japanese encephalitis, which is water quality assessment survey designed to test 1000 differ-
worsened by the tropical marine climate (CIA ). ent water supplies from all 16 municipalities and Roxas City,
Capiz Province is situated on the northeastern part of which took place from December 2009 to March 2010.
Panay Island, located in the Western Visayas. It has a land In addition, the H2S and Easygel tests were suggested as
surface area of approximately 2600 km2 and has roughly potential complementary tests to the EC-Kit, to be verified
80 km of coastline. In 2007, the population of Capiz was during the Capiz Province water quality testing program.
estimated at approximately 701,000 with 148,000 people This water quality assessment survey would be the first-
living in the capital, Roxas City. The province is divided ever comprehensive drinking water quality testing in the
into 17 areas: 16 municipalities and Roxas City. province.
The main PHO participants in this project included Dr
Project initiation Jarvis Punsalan, MD, DPH, head of the Capiz PHO; Jane
Delos Reyes, Engineer, coordinator of the water quality test-
Until 2009, Capiz had never performed any testing of the ing program; Leo Biclar, medical technician responsible for
various drinking water sources (wells, springs, surface processing and interpreting the Quanti-Tray tests; and SIs at
 70 P. Chuang et al. | Comparison of four field-based microbiological tests Journal of Water, Sanitation and Hygiene for Development | 01.1 | 2011

the provincial and municipal levels who were in charge of Table 1 | WHO drinking water source classification

collecting the water samples and processing and interpreting

Improved sources of
the EC-Kit tests. drinking water Unimproved sources of drinking water

Piped water into dwelling, Unprotected dug well

Objectives of study yard or plot
Public tap/standpipe Unprotected spring
The objectives of this study are to: (1) verify the accuracy of Tubewell/borehole Vendor-provided water
the four field-based tests: 10-mL pre-dispensed Colilert, Protected dug well Tanker truck water
Petrifilm, H2S bacteria test (laboratory-made reagent in 10-, Protected spring Surface water (river, stream, dam,
20- and 100-mL sample volume; and the industry-made lake, pond, canal, irrigation
channel)
HACH reagent with 20-mL sample volume) and Easygel
Rainwater collection Bottled water*
against a standard method: Quanti-Tray, using STATA® stat-
* Bottled water is considered an ‘improved’ source of drinking water only where there is a
istical software; (2) study the accuracy of the four field-based
secondary source that is ‘improved’.
tests as a function of improved and unimproved sources; (3)
if it is determined that field-based microbiological tests are
to a UN-designated ‘improved’ drinking water supply (NSO
accurate, provide recommendations for the use of a single
). A summary of the Capiz Province and corresponding
test based on accuracy and cost; and/or (4) provide rec-
UN designation are presented in Table 2.
ommendations for the use of a testing combination, if
appropriate, based on accuracy and cost.

Drinking water sources and types

METHODS

The WHO Joint Monitoring Programme (JMP) for Water Research design

Supply and Sanitation has been assembling statistics on

drinking water and sanitation coverage since 1990. Since In December 2009, the Capiz PHO began the 1000 test

2000, the JMP has classified water sources as ‘improved’ program covering 16 municipalities and Roxas City. As col-

or ‘unimproved’ (WHO/UNICEF Joint Monitoring laborators to this project, the authors travelled to and

Programme for Water Supply and Sanitation 2005). The worked in Capiz for approximately 22 working days, begin-

overall assumption behind the improved/unimproved drink- ning on January 7, 2010.

ing water source categories is that improved sources are The research plan and methodology for this study were

more likely to provide safe water than unimproved sources. comprised of the following steps.

The Drinking Water Source Classification is presented in 1. Elaboration of a study design concerning the villages/
Table 1. sampling zones to be tested by the MIT team and PHO SIs
According to the Philippines’ government’s regulatory using the Quanti-Tray and EC-Kit tests. This study design
definitions, ‘improved’ water sources include three levels: was prepared by Punsalan and Reyes, in collaboration
Level 1, which consists of point sources, Level 2, which with Susan Murcott of MIT, Tom Mahin of the Massachu-
consists of communal faucet systems and Level 3, which setts Department of Environmental Protection and the
consists of piped systems with individual household connec- four-person Masters of Engineering student team from MIT.
tions. The unimproved water sources, referred to as 2. Elaboration of a study design concerning the villages/
‘doubtful sources’ in the Philippines, consist of open dug sampling zones to be tested using the H2S bacteria tests
wells, unimproved springs, rainwater and surface water (laboratory-made reagent: 10-, 20- and 100-mL sample
sources. According to the National Statistical Coordination volume and industry-made reagent: HACH with 20-mL
Board (NSCB) of the Philippines, as of 2000, 119,000 house- sample volume) and Easygel test. This study design was
holds in Capiz (or 92% of the Capiz population) had access prepared by Trottier, in collaboration with Murcott.
 71 P. Chuang et al. | Comparison of four field-based microbiological tests Journal of Water, Sanitation and Hygiene for Development | 01.1 | 2011

Table 2 | Levels of drinking water sources in the Philippines Water samples were then randomly selected: the names of
qualified villages or zones per town were put in a box and
Capiz
UN Capiz PHO Province drawn randomly with 25% additional names drawn as reserve
designation designation Source types population
in case of inaccessibility of those initially selected. Water
Unimproved Doubtful Unimproved 8% source selection was based on accessibility and their use by
springs, open
at least ten nearby households in the sampling zone:
dugwells or wells
that need priming, • For each selected zone having doubtful sources, five of
surface water, or
these sources were randomly selected and tested.
rainwater
collectors • For each village randomly selected for Level 1 supply test-
Improved Level 1 Stand-alone point 92% ing, five Level 1 water sources were randomly selected for
sources, including testing.
shallow wells,
tubewells with
• For each village randomly selected for Level 2 supply test-
handpumps ing, one reservoir was randomly selected and five of its
Level 2 Piped water supply outlets were tested. Water sources tested were the reservoir
with communal outlets. A maximum of five outlets per reservoir were tested.
water points, from
boreholes,
• For each village randomly selected for Level 3 supply test-
tubewells ing, five households accessing water from these sources
Level 3 (piped Piped water supply were randomly selected and tested per zone. Water sources
connection with private water tested were every tenth household within the zone until the
on premises) points, such as a
needed number of samples (five) was attained.
household
connection The only exception to the aforementioned study design was
the Level 3 water supply for Roxas City. Since all of Roxas
City has a piped, chlorinated water supply, this was tested
3. Field testing in Capiz Province, Philippines, using the
separately using chlorine residual testing instead of the cost-
Quanti-Tray, EC-Kit (Colilert and Petrifilm), H2S bacteria
lier bacteriological testing. The Hach® Free Chlorine Pocket
and Easygel tests. The field testing was undertaken by the
Colorimeter II Test Kit was used to test randomly selected
Capiz Province PHO and municipal SIs in collaboration
sample locations. The 85 samples tested for chlorine
with the MIT team.
residual showed poor compliance with the DOH standard
4. Field testing of the Charles River in Cambridge,
of 0.2 to 0.5 mg/L of free chlorine in the distribution system.
Massachusetts, using the Quanti-Tray, H2S bacteria and
The study design plan for H2S and Easygel testing con-
Easygel tests. The field testing was undertaken in April
sisted of analysing a subset of all the water samples tested
2010 by Chuang and Trottier.
(by the Capiz PHO and municipal SIs) during the month of
5. Statistical comparison of the four field-based tests (Colilert,
January 2010, by biasing the sample selection toward Doubt-
Petrifilm, H2S bacteria and Easygel tests) in order to verify
ful or Level 1 sources and known contaminated sources.
their accuracy and provide recommendations concerning
the suitability of these tests for determining microbiological
contamination of drinking water. Sample collection and testing

The study design plan established by Punsalan and Reyes Drinking water samples were collected directly from the
consisted of selecting one sampling zone for every 5000 source or from the point-of-use by the authors or municipal
population (e.g. a municipality with a population of 30,000 SIs. In some circumstances where the point-of-use location
would have 6 sampling zones selected), where zones were was not sufficient for sampling (e.g. a storage tank that
distributed according to the ratio of water sources accessed was no longer filled with water), then samples were col-
by the residents of the particular municipality. lected from the household storage containers. To ensure
 72 P. Chuang et al. | Comparison of four field-based microbiological tests Journal of Water, Sanitation and Hygiene for Development | 01.1 | 2011

sterile conditions, water samples were collected into three total coliform and E. coli) as low as 1 MPN/100 mL and
separate containers: sterile 100-mL polystyrene vessels for up to 200.5 MPN/100 mL of sample, whereas the Quanti-
Quanti-Tray laboratory testing, 100-mL sterile sampling Tray/2000 provides bacterial counts up to 2419 MPN/
bags for EC-Kit field testing and 300-mL sterile sampling 100 mL.
bags for H2S bacteria testing and Easygel testing. The Quanti-Tray can test both chlorinated and
Figure 1 summarizes the general water sampling and un-chlorinated samples. However, chlorinated samples
testing methodology used by the PHO and the authors in should first be treated with sodium thiosulfate before the
the field research. The water samples were identified using Quanti-Tray reagent is added.
a labelling system created by members of the Capiz PHO. The Quanti-Tray is easy-to-use, rapid and accurate, and
has been approved by the United States Environmental
Microbiological test methods Protection Agency (EPA) for drinking, source/surface,
ground and wastewaters (IDEXX ). However, one of
The two microbiological drinking water quality tests the main drawbacks of the Quanti-Tray is its cost, since
used for the PHO’s water quality assessment program Quanti-Tray requires the use of an expensive sealer, and
were Quanti-Tray and EC-Kit. In addition, the H2S the trays and reagents were particularly expensive ($21/
bacteria and Easygel tests were suggested as test in Capiz), which makes this method generally unafford-
potential complementary tests to the EC-Kit, to be able in developing countries.
verified during the Capiz Province water quality testing Inadvertently, the Quanti-Tray tests purchased by the
program. Capiz PHO and used during the Capiz laboratory analyses
were the regular 50-well Quanti-Tray, whereas the Quanti-
Quanti-Tray Tray tests purchased at MIT and used during the laboratory
studies were the Quanti-Tray/2000. In retrospect, it would
The IDEXX Quanti-Tray and Quanti-Tray/2000 are enzyme- have been more useful for the Capiz PHO to purchase the
substrate coliform tests (Standard Methods 9223) that use Quanti-Tray/2000 since many of the water samples tested
semi-automated quantification methods based on Most using Quanti-Tray had results that were higher than the
Probable Number (MPN). Quanti-Tray detection limit (200.5 MPN/100 mL). How-
The MPN method uses multiple qualitative (presence/ ever, since the Capiz PHO was going to use the Quanti-
absence) data points to generate a maximum probability Tray to test drinking water samples, there was no reason
coliform count per 100-mL value, given by a standard to suspect that so many water samples would go above the
MPN table. The Quanti-Tray provides bacterial counts (of Quanti-Tray detection limit.

Colilert and Petrifilm (EC-Kit)

A portable microbiology testing kit was initially developed

by Robert Metcalf, Professor of Microbiology at California
State University at Sacramento and one of the original
founders of Solar Cookers International. He introduced
this method to Susan Murcott in Kenya in 2005 and con-
ducted training at MIT in June 2008. Susan Murcott then
modified the testing kit to also include a waist-belt incubator
and other materials needed in order to perform and inter-
pret the tests. The waist-belt incubator, which incubates
water samples using body temperature alone, serves as a
Figure 1 | Flow diagram representing sampling/testing methodology. cheaper, portable and more convenient alternative to
 73 P. Chuang et al. | Comparison of four field-based microbiological tests Journal of Water, Sanitation and Hygiene for Development | 01.1 | 2011

traditional incubators that are often costly and usually presence of faecal contamination. Venkobachar et al.
require electricity. She also created several different () later developed a second test medium, M2, which
model sizes of the product and branded it with the name consisted of the original M1 medium with the addition of
‘EC-Kit’. Susan Murcott introduced the technology to L-cystine, which was shown to increase the sensitivity and
the Non-Governmental Organization (NGO) ‘A Single reliability of the H2S test (Pillai et al. ).
Drop’, and introduced the director of that NGO to The M2 medium was used throughout the water quality
Robert Metcalf, after which they brought the technology testing program in Capiz Province in January 2010 for all
to the Philippines. sample sources and for all sample volumes (10-, 20- and
The EC-Kit contains two complementary tests for 100-mL). Indeed, since the H2S test reagent includes a
E. coli: the Colilert 10-mL presence/absence (P/A) test chlorine-neutralizing compound (sodium thiosulfate), the
and Petrifilm test. The Colilert test contains the same formu- H2S test is a suitable microbiological test for chlorinated
lation as in the Quanti-Tray tests, only it is reduced to its water supplies.
simplest form: a single P/A test of a 10-mL sample. How- Another H2S test used in this study was the industry-
ever, the Colilert test has a detection limit equivalent to made HACH Pathoscreen. This test uses a powder-form,
10 MPN/100 mL. In the Colilert test, the substrate is hydro- dehydrated H2S test reagent, suitable for a 20-mL (or
lyzed by the total coliform by-products, and reacts with a 100-mL) sample volume.
specific enzyme found in E. coli. A positive result is given
by a yellow sample (presence of total coliforms), or a Easygel
sample that fluoresces under long-wave UV illumination in
the dark (presence of E. coli) after 24-hour incubation The Easygel test is a quantitative water quality test that uses
(Gerba ). The Petrifilm test provides a quantitative an enzyme substrate method to provide a total coliform and
count of total coliform bacteria colonies (red colonies with E. coli bacterial count present in a 0.5-mL to 5-mL sample
gas bubbles after 24-hour incubation) and E. coli colonies volume, depending on the microbiological quality of the
(blue colonies with gas bubbles after 24-hour incubation) water sample tested.
with a 1-mL sample volume. The Easygel medium contains a sugar linked to different
Colilert can test both chlorinated and un-chlorinated dyes that react with coliform and E. coli products to produce
samples. However, chlorinated samples should first be trea- pink colonies (total coliform) and purple colonies (presence
ted with sodium thiosulfate before the Colilert reagent is of E. coli) (Micrology Laboratories ).
added. On the other hand, the Petrifilm cannot be used to Easygel can be used to test both chlorinated and
test chlorinated samples. un-chlorinated samples, without the addition of sodium
The EC-Kit is simple, low cost and easy to use. The thiosulfate.
most promising features of the EC-Kit are that it can be One of the main advantages of the Easygel test is that it
used by virtually anyone who receives the brief 15- to serves as an agar replacement. Agar is difficult to prepare,
30-minute training, and bacterial incubation is requires specific reagents and equipment, and preparation
performed using the waist-belt incubator, so it is comple- is both labour and time consuming. However, the Easygel
tely portable. sample processing procedure is very simple: 0.5 mL to 5 mL
of the water sample is pipetted into the Easygel reagent
H2S test bottle and the resulting mixture is poured into the pre-trea-
ted petri dish and allowed to set for 30 minutes. A sample
The H2S test using the original M1 medium is a well-known, volume of 5 mL was used for all Easygel samples in Capiz
simple and low-cost P/A test, developed by Manja et al. and in Cambridge, Massachusetts.
(). The test identifies the presence of H2S-producing An added benefit of the Easygel test is that if an electric
bacteria, associated with faecal contamination in a volume incubator is not available, samples can be incubated at ambi-
of water, which has been shown to correlate with the ent temperature (Micrology Laboratories ). So Easygel is
 74 P. Chuang et al. | Comparison of four field-based microbiological tests Journal of Water, Sanitation and Hygiene for Development | 01.1 | 2011

an economical, hassle-free and portable alternative, which The four new, field-based tests (and combinations) were
makes it convenient for field use, in developing countries. analysed using contingency tables, general statistical ana-
lyses (True Results (TR), False Positives (FP) and False
Negatives (FN)) and by calculating the error and pro-
portional reduction in error. More information on each of
RESULTS AND DISCUSSION
these tests is provided below.

A total of 521 water samples were collected in Capiz

Province (the overall test program was of 1000 sources. Frequency distribution table
We report here on a total of 521 water samples, a subset
of the overall work, which was not yet completed at the A 2  2 frequency distribution table (Table 3) is a table used
time of this report). Each water sample was tested in the in bi-variate analyses and is composed of two rows, cross-
field using at least two tests: one field-based test (Colilert, classified by two columns. It is often used to display data
Petrifilm, H2S bacteria test and Easygel) and the standard that can be classified by two different variables (e.g. New
method (Quanti-Tray). In Capiz, 521 samples were tested Test and Standard Method), each of which has two possible
using the EC-Kit tests (Colilert and Petrifilm), 163 samples outcomes, in this case Presence or Absence. Each of the four
were tested using the H2S bacteria test and 43 samples cells (a, b, c and d ) represents the number of times the out-
were tested using the Easygel test. Furthermore, 40 water come falls within that cell.
samples were also collected in Cambridge, Massachusetts, Similarly, a 3  3 frequency distribution table is a table
in order to provide additional comparative data for the used in tri-variate analyses and is composed of three rows,
H2S bacteria tests (10-, 20- and 100-mL sample volume) cross-classified by three columns. They serve the same
and Easygel test. purpose as 2  2 frequency distribution tables; however,
In total, 521 samples were tested using the EC-Kit tests; since enumerative tests give more information as to the
204 samples were tested using the H2S bacteria tests; and 83 degree of E. coli contamination, then a higher degree
samples were tested using the Easygel test. It is important to frequency distribution table can be set up with two different
note that although both total coliform and E. coli were variables (e.g. New Test and Standard Method), each of
tested, we report here on E. coli contamination only because which has multiple outcomes (WHO risk levels: Confor-
total coliform is not necessarily from the faeces of humans mity, Low, Intermediate, High, Very High).
and mammals and some coliform can occur naturally in
the environment. Therefore, total coliform is not the ideal General statistical analyses
indicator of faecal contamination (in contrast, E. coli and
most thermotolerant coliform are). Therefore, E. coli is the When a New Test is being compared against a Standard
most widely used test to determine whether there is Method, the percentage of true results (TRs ¼ a þ d), false
human or animal faeces in drinking water. positives (FPs ¼ b) and false negatives (FNs ¼ c) is calculated.
These results provide information as to the ‘correctness’
of the given test (TR), and also specify the tendency of a test
Statistical analyses
to incorrectly flag a positive result when it should be

The data presented here is a compilation of the data col-

lected in Capiz Province from December 2009 to March Table 3 | Frequency distribution table

2010. The statistical analysis for Capiz Province results

Standard method
compares the field-based microbiological tests (Colilert, Pet-
New test presence Absence
rifilm, H2S test and Easygel) to Quanti-Tray. All statistical
Presence a b
results presented here were analysed using the STATA
Absence c d
Release II software.
 75 P. Chuang et al. | Comparison of four field-based microbiological tests Journal of Water, Sanitation and Hygiene for Development | 01.1 | 2011

negative, or to incorrectly flag a negative result when it frequency distribution tables presented in Tables 4 and 5,
should be positive. respectively. In other words the black-tone areas in these
two tables indicate an exact match and the hatch-marked
Error and proportional reduction in error areas represent an error. This is useful information for veri-
fication of the new field-based tests, but as a water quality
The error associated with a given test is the sum of FP and test an overestimate of the risk level still yields useful infor-
FN results, divided by the total number of tests. The pro- mation. So, the error values of interest actually follow the
portional reduction in error, λ, is a measure of ‘how good hatch-marked areas presented in Table 6.
one becomes at making predictions’ starting from an initial
test result prediction (with corresponding Error1) and then
adding another piece of information (in this case, a new, Minimum detection limit and FP/FN

field-based test) to obtain a test result that will hopefully

yield a better prediction (with corresponding Error2). The Throughout the following section, minimum detection limit

formula for λ is provided below. and FP/FN results are used to help determine the accuracy
and validity of the new, field-based tests presented here.

Error1  Error2 Therefore, it is important to understand the inherent differ-

l¼ ences between these two terms.
Error1
First, ‘detection limits’ refers to the lowest concentration
In this case, the initial assumption was that UN-designated of a substance (in this case, E. coli or H2S-producing bac-
unimproved water sources (or Doubtful sources in the Phi- teria) that can be distinguished from the absence of that
lippines) were all contaminated (High/Very High Risk substance. Therefore, ideally, a microbiological test would
Level or Presence of contaminant), and that UN-desig- have a low minimum detection limit (i.e. the test is said to
nated improved water sources (or Levels 1 through 3
Table 5 | Calculations for error – exact match (EC-Kit only)
in the Philippines) were all safe (Conformity/Low Risk
Level or Absence of contaminant). The error associated Quanti-Tray

with the initial assumption was calculated: it was deter-

High/
mined that for unimproved sources, the initial error was Conformity/ very
low Intermediate high
15%; and for improved sources, the initial error was
63%. In other words, the assumption that an unimproved Water source Conformity/
type þ EC-Kit low
water source is contaminated would be incorrect 15% of
Intermediate
the time, and the assumption that an improved water
source is safe would be incorrect 63% of the time. High/very
high
The calculations for error (Error1 and Error2) for tests
and test combinations were done in two ways. The ‘exact
match’ method reports the percentage frequency of tests Table 6 | Calculations for error for single, new, field-based tests and testing combinations

results in the black-tone areas in the 2  2 and 3  3

Quanti-Tray

Table 4 | Calculations for error for new, field-based tests High/

Conformity/ very
low Intermediate high
Quanti-Tray

Water source Conformity/

Presence Absence
type þ low
Water source type þ additional Presence EC-Kit Intermediate
test High/very
Absence high
 76 P. Chuang et al. | Comparison of four field-based microbiological tests Journal of Water, Sanitation and Hygiene for Development | 01.1 | 2011

be ‘sensitive’) so that it could detect the presence of a sub- Table 8 | TR, FP and FN results for Colilert test, Capiz samples

stance, even at low concentrations. Second, FP/FN

True results (%) False positives (%) False negatives (%)
represents the percentage of sample results that were incor-
Colilert test 75 6 19
rect. More specifically, FPs represent results that were
positive by the new, field-based test but which did not con-
tain any of the looked-for substance. Conversely, FNs are source. However, it should be noted that the error following
results that were negative by the new, field-based test, but the addition of the Colilert test is still at 12% for unimproved
which contained the looked-for substance. and 27% for improved sources.
Detection limits and FP/FN results are both necessary
in order to understand the test results given by the new, Petrifilm results
field-based microbiological tests. All in all, these comparative,
statistical tools measure different things: FP/FN results The frequency distribution table for the Petrifilm test results
tell us how good a given test is at quantifying the results; compared to Quanti-Tray are presented in Table 10. The cor-
whereas, the minimum detection limit tells us how good responding TR, FP and FN values are presented in Table 11
the test is at receiving a (low) signal. for Capiz samples.
The Petrifilm test yielded highly accurate results (88%).
Colilert results The FNs had Petrifilm indicating a higher risk level category
than Quanti-Tray. While ideally one would hope for perfect
The frequency distribution table for the Colilert test results congruence of the two test methods, it is preferable to have
compared to Quanti-Tray are presented in Table 7. The cor- over-reporting of risk (false positives) than under-reporting
responding TR, FP and FN values are presented in Table 8 of risk (false negatives). Table 12 presents the error and pro-
for Capiz samples. portional reduction in error values for the Petrifilm test
The Colilert test yielded relatively accurate results results.
(75%). However, it is of concern that the FN (19%) was
much higher than the FP (6%). This may be because the Table 9 | Error and proportional reduction in error for Colilert test, Capiz samples
detection limit for the Colilert test is 10 colony forming
Unimproved sources Improved sources
units (CFU) per 100 mL, or 1 CFU per 10 mL of sample.
Initial error ¼ 15% Initial error ¼ 64%
For the purposes of distinguishing between unimproved Test Error λ n Error λ N
and improved water sources, these false results do not err
Colilert 12% 25% 521 27% 58% 521
on the side of caution. Table 9 presents the error and pro-
portional reduction in error values for the Colilert test
results. It is interesting to note that the addition of the
Table 10 | Frequency distribution table for Petrifilm test, Capiz samples
Colilert test greatly improves the initial predictions based
on UN-designated improved/unimproved categories alone. Quanti-Tray

In fact, for unimproved sources, it decreased the error by Presence Absence

25%, and for improved sources, it decreased the error by Petrifilm test Presence 353 19
58%, resulting in a more accurate description of the water Absence 43 106

Table 7 | Frequency distribution table for Colilert, Capiz samples

Table 11 | TR, FP and FN results for Petrifilm test

Quanti-Tray
Presence Absence
True results (%) False positives (%) False negatives (%)

Colilert test Presence 242 32 Petrifilm 88 4 8

Absence 101 146 test
 77 P. Chuang et al. | Comparison of four field-based microbiological tests Journal of Water, Sanitation and Hygiene for Development | 01.1 | 2011

Table 12 | Error and proportional reduction in error for Petrifilm, Capiz samples Table 14 | TR, FP and FN Results for H2S bacteria tests

Unimproved sources Improved sourcess True False False

Initial error ¼ 15% Initial error ¼ 64% results positives negatives
(%) (%) (%)
Test Error λ n Error λ n

Petrifilm 37% –138% 521 39% 39% 521 Laboratory-made 10-mL H2S test 80 9 11
reagents (n ¼ 203)
20-mL H2S test 84 10 6
(n ¼ 203)
For unimproved sources, the addition of Petrifilm did
100-mL H2S test 80 16 4
not reduce our error, but in fact increased it (λ ¼ –138%). (n ¼ 203)
Therefore, as a single test for unimproved sources, Petrifilm Manufactured HACH test 79 9 12
yields a much less accurate prediction than predicting that and purchased (n ¼ 203)
all unimproved sources are contaminated. For improved
sources, the addition of Petrifilm significantly reduced the
error by 39%, with an error of 39%. and unprotected open dug wells, rainwater, shallow and
deep bore wells and chlorinated and un-chlorinated house-

H2S test results hold taps) and to test 40 samples in Cambridge,

Massachusetts: 38 from the Charles River, and 2 de-ionized

The frequency distribution table for the H2S bacteria test water samples.

results compared to Quanti-Tray test results are presented When comparing the H2S laboratory-made reagents,

in Table 13. The corresponding TR, FP and FN values are surprisingly, the 20-mL test gave slightly more TR (84%)

presented in Table 14 for samples collected in Capiz and than the 100-mL (80%) and 10-mL tests (80%). As would

Cambridge. be expected, the percentage of FP results was highest for

The H2S medium was used to test 163 water samples in the 100-mL test (16%) and lowest for the 10-mL test (9%);

Capiz Province from different sources (springs, protected whereas the percentage of FNs was highest for the 10-mL
test (11%) and lowest for the 100-mL test (4%).
In general, it was noted that as the sample volume of the
Table 13 | Frequency distribution tables for H2S bacteria tests
H2S test increased, FP increased (in fact, becoming overly
sensitive, with 16% FP results) and FN decreased (becoming
10-mL H2S test Quanti-Tray
sensitive).
Presence Absence
The 20-mL HACH PathoScreen test had results that
10-mL H2S test Presence 112 19
were very similar to the 10-mL H2S test, although it still
Absence 22 50
proved to be the least accurate of all the H2S tests: it had
20-mL H2S test Quanti-Tray
the lowest percentage of true results (79%) and although
Presence Absence
its percentage of FPs was low (9%), it had the highest per-
20-mL H2S test Presence 122 20
centage of FNs (12%).
Absence 12 49
The high percentage of FP results (16%) for the 100-mL
HACH Pathoscreen Quanti-Tray
test is probably due to the H2S bacteria test detecting H2S
Presence Absence
that may not come from faecal bacteria. In groundwater in
HACH Pathoscreen Presence 110 19
particular, there is the strong possibility of H2S being pre-
Absence 24 50
sent due to natural hydrogeological sources and to
100-mL H2S test Quanti-Tray
anthropogenic impacts other than faecal contamination,
Presence Absence
both of which would lead to FP results (Sobsey & Pfaender
100-mL H2S test Presence 126 32
). This phenomenon is important in this study since
Absence 8 36
most drinking water samples (136 samples) tested using
 78 P. Chuang et al. | Comparison of four field-based microbiological tests Journal of Water, Sanitation and Hygiene for Development | 01.1 | 2011

the H2S test, were groundwater collected from wells and Table 15 | Error and proportional reduction in error for H2S bacteria tests, Capiz samples
only
spring sources. Furthermore, it has been shown that the
H2S test detects bacteria other than coliforms that are Unimproved sources Improved sources

associated with faecal contamination, such as Clostridium Test Error (%) λ n1 Error (%) λ n2

perfringens, which is one of the most resistant indicators of 10-mL H2S test 9.1 0.0 33 24.6 51.5 130
faecal contamination. Therefore, it is possible that the H2S 20-mL H2S test 9.1 0.0 33 20.0 60.6 130
test can yield a positive result even if no coliforms are pre- 100-mL H2S test 9.1 0.0 33 28.7 43.9 129
sent (Sobsey & Pfaender ). HACH test 21.2 133 33 23.1 54.6 130
Of great concern with microbiological tests in general is 1
Sample size for unimproved sources tested for given H2S test.
the potential for FNs, in other words not detecting faecal 2
Sample size for improved sources tested for given H2S test.

contamination when it is present. The percentage of FNs

was relatively low for the 10-mL sample (11%), but was chlorinated and un-chlorinated household taps) and to test
reduced by about half in the 100-mL sample (4%). The 40 samples in Cambridge, Massachusetts: 38 from the
higher percentage of FPs versus FNs in the 20- and 100- Charles River, and 2 de-ionized water samples.
mL tests is favourable because it errs on the side of caution. The Easygel test had a relatively high percentage of TRs
For the 10-mL H2S test and HACH tests, the percentage of (81%), few FPs and a high proportion of FNs. This means
FNs was higher than the percentage of FPs. that the Easygel test is a particularly good indicator of the
Table 15 presents the error and proportional reduction presence of E. coli contamination, but not of the absence
in error values for the H2S bacteria test for Capiz samples of contamination.
only, since the water source used in Cambridge was not a Table 18 presents the error and proportional reduction
drinking water source and therefore could not be deemed in error for the Easygel test for Capiz samples only, since
‘unimproved’ or ‘improved’. the water source used in Cambridge was not a drinking
It is interesting to note that errors for the H2S test were water source and could therefore not be deemed an ‘unim-
greater for improved sources than for unimproved sources. proved’ or ‘improved’ water source.
For unimproved sources, the addition of the 10-, 20- and For unimproved sources, the addition of Easygel did not
100-mL H2S test did not change the error (λ ¼ 0%); how- reduce our error, but in fact increased it (λ ¼ –100%). There-
ever, the addition of the HACH test increased the error by fore, as a single test for unimproved sources, the Easygel test
133%. Therefore, as a single test for unimproved sources, yields a less accurate prediction than predicting that all
the laboratory-made H2S test is no better than simply pre- unimproved sources are contaminated. For improved
dicting that all unimproved sources are contaminated. For sources, the addition of Easygel reduced our error by
improved sources, the addition of all H2S tests (laboratory 51.5%, with an error of 24.6%. Fisher’s exact test on the
made and HACH test) reduced the error on average by
53%, with an average error of 24%.
Table 16 | Frequency distribution table for Easygel test

Easygel test results Quanti-Tray

Presence Absence

The frequency distribution table for the Easygel test results Easygel Presence 49 1
compared to Quanti-Tray test results is presented in Absence 14 17
Table 16. The corresponding TR, FP and FN values are pre-
sented in Table 17 for Capiz and Cambridge samples Table 17 | TR, FP and FN results for Easygel test

combined.
True results False positives (%) False negatives (%)
The Easygel test was used to test 41 water samples in
Capiz Province from different sources (springs, protected Easygel 81 1 17
(n ¼ 83)
and unprotected open dug wells, deep bore wells and
 79 P. Chuang et al. | Comparison of four field-based microbiological tests Journal of Water, Sanitation and Hygiene for Development | 01.1 | 2011

contingency table for unimproved and improved sources in Test combinations

Capiz showed that these results are not statistically signifi-
cant, due to their small sample size. The accuracy of different testing combinations of presence/
The 3  3 frequency distribution table for the Easygel absence test (Colilert and H2S bacteria tests) and enumera-
test compared to Quanti-Tray test results is presented in tive test (Petrifilm and Easygel) was analysed. These
Table 19. This table presents the Easygel and Quanti-Tray combinations were compared statistically to Quanti-Tray
test results in three categories: the WHO Risk Levels (Con- using the 3  3 frequency distribution table and again look-
formity/Low, Intermediate and High/Very High) for Capiz ing at the error and proportional reduction in error. The
and Cambridge samples. test combinations are presented in Table 20.
The majority of samples (51) were identically classified Test results and corresponding WHO Risk Levels were
by the Easygel test and Quanti-Tray. The true results percen- set up for the different test combinations and are presented
tage (i.e. results that lie in the same WHO Risk Level for the here for EC-Kit (Colilert þ Petrifilm) (Table 21), H2S test þ
Easygel test and Quanti-Tray) for this 3  3 frequency distri- Petrifilm (Table 22), Colilert þ Easygel (Table 23) and
bution table is 64%. However, what is most important here is H2S test þ Easygel (Table 24). The corresponding WHO
that the WHO Risk Level for a given sample, obtained by the Risk Levels for Easygel were for a sample volume of 5 mL.
Easygel test, corresponds to the same or a lower-risk WHO It should be noted that the associated WHO Risk Levels
Risk Level (shaded region in Table 19). In this light, the true should not be taken as ‘absolutes’ but, rather, as an initial
results percentage (i.e. results that lie in the same or higher benchmark with which to compare test combinations to
WHO Risk Level for the Easygel test than Quanti-Tray) is Quanti-Tray results.
75%. Again, such misclassifications err on the side of caution The 3  3 frequency distribution table for the EC-Kit test
as it can result in the rejection of water that may be safe to results is presented in Table 25. The 3  3 contingency tables
drink as opposed to acceptance of water that is unsafe. for the testing kit combinations compared Quanti-Tray to
improved sources and unimproved sources separately. The cor-
responding error and proportional reduction in error (λ) are

Table 18 | Error and proportional reduction in error for Easygel test, Capiz samples only
presented in Table 26. This table presents values for Capiz
samples only, since the water source used in Cambridge was
Unimproved sources Improved sources not a drinking water source and could therefore not be
Test
Error λ n1 Error λ n2
deemed an ‘unimproved’ or ‘improved’ water source.
Easygel 28.6% –100% 14 24.6% 51.5% 28
1
Sample size for unimproved sources.
2
Sample size for improved sources.

Table 20 | Testing combinations of new, field-based tests

Test combinations
Table 19 | 3  3 frequency distribution table for Eaysgel test, Capiz and Cambridge
samples
Colilert þ Petrifilm (EC-Kit)

Quanti-Tray
10-mL H2S test þ Petrifilm
Low/ High/very 20-mL H2S test þ Petrifilm
conformity Intermediate high
100-mL H2S test þ Petrifilm
Easygel Low/ 22 9 0 20-mL HACH test þ Petrifilm
conformity1
Colilert þ Easygel
Intermediate1 3 5 11
High/very 0 7 24 10-mL H2S test þ Easygel
high1 20-mL H2S test þ Easygel
1
The WHO risk levels were determined based on the sample volume used in the Easygel 100-mL H2S test þ Easygel
test (5 mL) compared to the actual risk levels based on a 100 mL sample. 20-mL HACH test þ Easygel
Low/Conformity: 0 CFU/5 mL, Intermediate: 1 to 4 CFU/5 mL, High/Very High: .5 CFU/5 mL.
 80 P. Chuang et al. | Comparison of four field-based microbiological tests Journal of Water, Sanitation and Hygiene for Development | 01.1 | 2011

Table 21 | WHO risk levels and corresponding EC-Kit test results. Adapted from WHO Table 25 | 3  3 frequency distribution table for EC-Kit tests, Capiz samples (n ¼ 521
(1997), replacing ‘thermotolerant bacteria’ with ‘E. coli’ samples)

E. coli in sample Colilert Petrifilm Quanti-Tray

WHO risk level (CFU/100 mL) E. coli result E. coli result
Low/ High/very
conformity Intermediate high
Conformity ,1 Clear 0
1
Low 1–10 Clear 0 EC- Low/conformity 230 13 4
Kit Intermediate1 76 34 15
Intermediate 10–100 Blue fluorescence 0
High/very high1 13 30 106
High 100–1000 Blue fluorescence 1–10
Very high .1000 Blue fluorescence .10
explained above, these calculations are not as useful for
water quality testing because we are not so much concerned
Table 22 | WHO risk levels and corresponding H2S test þ Petrifilm test results
with exactly matching Quanti-Tray so much as providing
information on water sources unsuitable for consumption.
WHO risk level H2S test result Petrifilm result (CFU/mL)
The second method (which includes overestimates of risk
Conformity Yellow 0 level, Table 6) is a preferable statistical analysis approach.
Low Yellow 0 This is because an overestimate of the risk level still yields
Intermediate Black 0 useful information for someone interested in knowing
High Black 1–10 their water quality. In other words, it errs on the
Very high Black .10

Table 26 | Error, proportional reduction in error (λ) and sample number, n for unimproved
and improved samples collected in Capiz and compared to Quanti-Tray

Table 23 | WHO risk levels and corresponding Colilert and Easygel test results
Unimproved sources Improved sources

WHO risk level Colilert result Easygel result (CFU/5 mL) Error (%) λ n1 Error (%) λ n2

Conformity Clear 0 EC-Kit (Colilert þ 25 –63 521 29 54 521

Low Clear 0 Petrifilm (EXACT
MATCH)
Intermediate Blue fluorescence 0–4
EC-Kit (Colilert þ 6 63 521 6 60 521
High Blue fluorescence 5–50 Petrifilm)
Very high Blue fluorescence .50 10-mL H2S test þ 9.1 82 33 3.5 93 114
Petrifilm
20-mL H2S test þ 12 –33 33 2.4 95 126
Table 24 | WHO risk levels and corresponding H2S test and Easygel test results Petrifilm
100-mL H2S test þ 6.1 33 33 1.6 97 125
WHO risk level H2S test result Easygel (CFU/5 mL) Petrifilm

Conformity Yellow 0 20-mL HACH test þ 15 –67 33 1.6 97 125

Petrifilm
Low Yellow 0
Colilert þ Easygel 0.0 100 13 0.0 100 28
Intermediate Black 0–4
10-mL H2S test þ 0.0 100 4 0.0 100 18
High Black 5–50
Easygel
Very high Black .50
20-mL H2S test þ 0.0 100 4 0.0 100 19
Easygel
100-mL H2S test þ 0.0 100 3 0.0 100 19
As previously mentioned, the calculations for error for
Easygel
EC-Kit were done in two ways. The ‘exact match’ method
20-mL HACH test þ 0.0 100 3 0.0 100 22
is useful information for the verification of EC-Kit, and the Easygel
results indicate that the EC-Kit has a 25–29% error for 1
Sample size for unimproved sources.
exactly matching the Quanti-Tray results. However, as 2
Sample size for improved sources.
 81 P. Chuang et al. | Comparison of four field-based microbiological tests Journal of Water, Sanitation and Hygiene for Development | 01.1 | 2011

conservative side of over-prediction rather than under-pre- EC-Kit

diction. Given these facts, the addition of both tests in the
form of an EC-Kit would improve predictions by 60–63% Currently, EC-Kits are being assembled and disseminated by
for both improved and unimproved water sources. Susan Murcott, Senior Lecturer in the Civil and Environ-
In general, for unimproved and improved sources, the mental Department at MIT, as part of a research and
combination of tests yielded better prediction of faecal con- mapping project. These kits are sold at cost.
tamination than single tests, with the exception of 20-mL At this time, four models (Model A through D) are avail-
H2S test þ Petrifilm (λ ¼ –33% for unimproved sources) and able. Every model contains sterile sample bags, individually
20-mL HACH test þ Petrifilm (λ ¼ –67% for unimproved wrapped sterile pipettes, a portable UV lamp with 4 AA bat-
sources). In other words, the 20-mL H2S test alone or the teries, an insulated cooler bag, waist-belt incubator, ice pack
assumption of contamination based on source type without and laminated instructions. The price of each kit model are
any testing are better predictors than the 20-mL H2S test þ given in Table 27 including the price per test (without the
Petrifilm or 20-mL HACH test þ Petrifilm combinations. rest of the kit contents) of $1.50/Colilert test and $1.25/
It is interesting to note that all combinations that Petrifilm test. Also, these costs do not include the cost of
included Easygel reduced the error by 100%, such that domestic US postage, which can range from $5 to $20
error ¼ 0%. This proportional reduction in error is much depending on the kit size and speed of delivery; or the
larger than the proportional reduction in error obtained cost of international postage.
for Easygel alone (–100% for unimproved and 51.5% for
improved). This large difference in λ can be attributed to
the properties of the P/A tests that were combined with H2S test
Easygel. As a matter of fact, the Easygel test yielded few
FP results (1%) and many FN results (17%). On the other The H2S test, or the H2S paper strip test, requires the use of
hand, the H2S tests that were combined with Easygel had readily available laboratory reagents, distilled or de-ionized
many FP results (9 to 16%) and few FN results (4 to 11%). water, paper towels or toilet paper and vials or sterile
This could mean that the two tests effectively complement sampling bags. The cost of the paper strip (non-toxic absorb-
one another, such that the λ value of the combined tests is ing paper) is included in Table 28, but the cost of the vials or
significantly greater than the λ value of a single test. sampling bags is not included.
Finally, it must be mentioned that the sample size The cost for the H2S paper strip test (laboratory-made M2
for these Easygel combinations was particularly small, reagent) was calculated based on the price of reagents
especially for unimproved sources (3 to 4 samples). required to make 2.5 L of H2S reagent solution (5000 tests
for the 10-mL H2S test, 2500 for the 20-mL H2S test and
Cost analysis 1000 for the 100-mL H2S test). The price and units for all
reagents were taken from Sigma Aldrich (www.sigmaaldrich.
The cost summaries presented here are only for the new, com), except for sodium thiosulfate, for which the price and
field-based microbiological tests: EC-Kit (Colilert and units were obtained from VWR (www.vwr.com). Prices
Petrifilm), H2S bacteria test and Easygel. The cost of the Stan- were obtained for orders based in the United States and in
dard Methods tests (Quanti-Tray and membrane filtration) the Philippines: $344 and $830, respectively. It is important
were not included in this paper because, throughout this pro- to note that the price of reagents in the Philippines is almost
ject, these tests were used for verification purposes and as the 2.5 times higher than the price of the same reagents in the
Standard Methods against which to test the field-based United State (The prices shown in Table 28 assume purchase
methods. More specifically, not only are these tests expensive in the US).
(Quanti-Tray tests ranged from $6 in USA to $21 in Philip- The HACH P/A media is available directly from the
pines per sample) but they also require the use of other HACH website. Typically, one pouch (or ‘powder pillow’)
expensive equipment (sealer and incubator). is used as the test reagent for a 20-mL sample volume. A
 82 P. Chuang et al. | Comparison of four field-based microbiological tests Journal of Water, Sanitation and Hygiene for Development | 01.1 | 2011

Table 27 | Contents and cost of EC-Kit Model A, B, C and D

Cost ($)/test Cost ($)/test

Kit contents Total cost ($) Number of tests Complete kit Test reagents only

Model A (C-10) • 10 Colilert tests 32.00 10 3.20 1.50

Model B (CP-25) • 25 Colilert tests 104.00 25 4.16 2.75
• 25 3M Petrifilm (1 pack)
• Incubator belt
• 2 ice packs
• 10 cardboard squares
• 20 rubber bands
Model C (CP-50) • 50 Colilert tests 187.00 50 3.74 2.75
• 50 3M Petrifilm (1 pack)
• Incubator belt
• 2 ice packs
• 20 cardboard squares
• 40 rubber bands
Model D (CP-100) • 100 Colilert tests 349.00 100 3.49 2.75
• 100 3M Petrifilm (1 pack)
• Incubator belt
• 2 ice packs
• 30 cardboard squares
• 60 rubber bands

pack of 50 powder pillows is $29.39, or approximately 59¢ tests (less than 60¢ each), excluding the initial cost of glass
per test. vials, 100-mL sterile sampling bags and paper strips.
Table 28 presents the average cost per test for the three The Easygel tests, however, the H2S, the Easygel and Model
different laboratory-made H2S tests and HACH tests, from a A of the EC-Kit are for individual tests.
2.5 L reagent solution for just the test reagents themselves, The H2S (20 ml) þ the Easygel is a two  test system, as
not including the sample vial or sampling bag. it the EC-Kit, Models B, C and D. The EC-Kit Models B, C
or D is $1.00 more expensive per test set that than the
Easygel H2S (20 ml) þ the Easygel test set.

The Easygel test requires a specially pre-treated Petri dish and

the Easygel media. These are sold as a test kit (one kit is com- Table 28 | Average cost per test for different H2S test sample volumes, from a 2.5 L
reagent solution
prised of one medium bottle and one treated Petri dish) from
Micrology Laboratories (www.micrologylabs.com) and are Laboratory-made H2S
test sample volume
available in sets of ten tests for $21.25/set if one to nine sets HACH test
10 mL 20 mL 100 mL 20-mL
are purchased and for $16.25/set if more than ten sets are pur-
Reagent volume/test (mL) 0.5 1.0 2.5 n/a
chased. This means that individual tests range from $1.63 to
No. of samples tested 5000 2500 1000 n/a
$2.13.
United States – Average 0.07 0.14 0.35 0.59
cost/test 1 ($)
Cost comparison Philippines – Average 0.17 0.33 0.83 n/a
cost/test 1 ($)

Table 29 compares the cost of each microbiological test, assum- n/a: not applicable.
1 The average cost per test was calculated based on the cost of laboratory reagents for
ing purchase in the US. The H2S tests (10-, 20-, 100-mL and 2.5 L of solution, adding the cost of the paper towels. The cost of vials/bottles and
HACH) were by far the least expensive of the microbiological sampling bags was not included in the average cost/test.
 83 P. Chuang et al. | Comparison of four field-based microbiological tests Journal of Water, Sanitation and Hygiene for Development | 01.1 | 2011

Table 29 | Cost/test of H2S test, Easygel and EC-Kit in the USA

H2S test Easygel

H2S test (20 ml) þ H2S test (100 ml) þ EC-Kit: Colilert þ
Test 10 mL1 20 mL1 100 mL1 HACH 1–9 sets 10þ sets Easygel Petrifilm Petrifilm

Cost/test ($) 0.07 0.14 0.35 0.59 2.13 1.63 1.77 1.60 2.75
1
The cost data presented in this table for the H2S test reflects cost incurred in the United States in order to provide an adequate comparison with the Easygel and EC-Kit

Other considerations reasonably good results (with 88% true results and 4% and
8% of false positives and false negatives, respectively).
Finally, the figures cited here represent an approximate cost However, once assumptions are made regarding the
of each test. It is important to consider that prices differ water source type (e.g. all doubtful sources are unimproved),
greatly from country to country, and that costs of reagents then from the calculations of error and proportional
and laboratory supplies are usually more expensive in devel- reduction in error for unimproved and improved water
oping countries. Also, one may need to consider freight/ sources, it is possible to improve predictions with the use
transportation costs associated with shipping the reagents of a single test.
to remote locations worldwide. The statistical analyses also showed that the EC-Kit
performs very well in predicting water contamination (6%
error). The 100-mL H2S test in conjunction with Petrifilm
RECOMMENDATIONS also yielded highly promising results (6.1% error for unim-
proved sources and 1.6% error for improved sources).
Table 30 provides a summary of the TR, FP and FN values However, the most accurate testing combination was the
for the individual, field-based tests. H2S tests (10-mL, 20-mL, 100-mL and 20-mL HACH)
Given the statistical analyses from each of the individual combined with Easygel. Although these combinations all
microbiological tests, one individual test cannot be rec- yielded the same error and proportional reduction in error
ommended as an accurate and/or suitable microbiological (0% and 100%, respectively) the 20-mL H2S test þ Easygel
test for drinking water. In fact, although Petrifilm had 88% combination was chosen as the best option based on the accu-
true results, it must be noted that Petrifilm has a high detec- racy of the individual 20-mL H2S test. However, it must be
tion limit (i.e. it can only detect microbiological noted that the both the 20-mL H2S tests and Easygel were per-
contamination in samples that have a concentration of 100 formed on a very small sample size (23 samples only: 4
CFU/100 mL, and are above the WHO Intermediate Risk unimproved sources and 19 improved sources), therefore it
Level); therefore it is unsuitable for water sources with low is recommended that the 20-mL H2S tests and Easygel combi-
E. coli concentrations. The 20-mL H2S test also provided nation be verified in future studies, on a much larger scale. It is
also recommended that due to the limitations present in the
Table 30 | Summary of TR, FP and FN values for individual tests detection limits for Quanti-Tray and Petrifilm, future projects
verifying the EC-Kit should use Quanti-Tray/2000.
True False False
results (%) positives (%) negatives (%) n

Colilert 75 6 19 521
CONCLUSIONS
Petrifilm 88 4 8 521
10-mL H2S test 80 9 11 203
The primary objective of this study was to verify and
20-mL H2S test 84 10 6 203
assess the suitability of four, low-cost, microbiological
100-mL H2S test 80 16 4 203
field-based tests for drinking water quality testing. More
20-mL HACH test 79 9 12 203
specifically, the study examined the 10-mL pre-dispensed
Easygel 81 1 17 83
Colilert test, the Petrifilm test, the laboratory-made P/A
 84 P. Chuang et al. | Comparison of four field-based microbiological tests Journal of Water, Sanitation and Hygiene for Development | 01.1 | 2011

H2S test (for 10-, 20- and 100-mL sample volume), the test ($0.35 for the 100-mL H2S test reagent only and $1.25
HACH PathoScreen P/A H2 S test (20-mL sample for Petrifilm, if purchased in the United States), and the
volume) and the Easygel test. The different tests 20-mL H2S test and Easygel cost $1.77/test ($0.14 for the
were verified and compared for accuracy. Accuracy 20-mL H2S test reagent only and $1.63 for Easygel, if pur-
was measured by comparing the field test results to chased in the United States). These prices are less
the results obtained using a Standard Methods test expensive than the EC-Kit two-test combination, which
(Quanti-Tray). costs $2.75/test ($1.75 for Colilert and $1.25 for Petrifilm
The drinking water samples used in this study were Colilert if only the tests themselves are considered and if
collected in different municipalities throughout Capiz purchased in the United States).
Province, Philippines, in January 2010, and from the Although the 20-mL H2S test and Easygel combination
Charles River in Cambridge, Massachusetts, in April 2010. was the most promising testing combination, it cannot yet
In total, 521 samples were tested using the Colilert and be recommended for extensive use in microbiological testing
Petrifilm tests; 203 samples were tested using the of drinking water in developing countries since the sample
H2S tests; and 83 samples were tested using the Easygel test. size for this particular combination was very small (23
The tests were evaluated as single tests, and as a combi- samples). Therefore, it is recommended that the 20-mL
nation of tests (i.e. Colilert and Petrifilm, H2S test and H2S tests and Easygel combination be verified in future
Petrifilm, H2S test and Easygel) to determine the best testing studies, on a much larger scale. It is also recommended
combination. that due to the limitations present in the detection limits
Based on the statistical analyses conducted, the results for Quanti-Tray and Petrifilm, future projects verifying the
from this study do not support the use of one single micro- EC-Kit should use Quanti-Tray/2000 as the standard for
biological test as a suitable method for determining comparison.
drinking water quality accurately. That said, the single
20-mL laboratory-made H2S test yielded the best results
with 88% true results, 4% false positives and 8% false
negatives. REFERENCES
Test combinations proved to be promising alternatives
to single tests for accurate results. In fact, combining tests CIA  CIA The World Factbook. Retrieved February 20, 2010,
from Central Intelligence Agency: https://www.cia.gov/
such as the Colilert and Petrifilm (EC-Kit) yielded very library/publications/the-world-factbook/geos/rp.html.
good results, with an overall error of 6% (compared to Gerba, C. P.  Indicator Microorganisms. In: Environmental
25% and 12% for the single tests, Colilert and Petrifilm, Microbiology (ed. R. Maier, I. Pepper & C. Gerba). Academic
Press, San Diego, pp. 491–503.
respectively). And although statistical analyses comparing
IDEXX  IDEXX February 2010 Newsletter. Retrieved
the EC-Kit with Quanti-Tray revealed that the results from February 20, 2010, from IDEXX: www.idexx.com/view/
both tests did not exactly match, the EC-Kit’s low overall xhtml/en_us/water/newsletter/201002.jsf.
error of 6% means that it is indeed very useful in assessing Manja, K., Maurya, M. & Rao, K.  A simple field test for the
detection of faecal pollution in drinking water. Bull. WHO
at-risk water sources.
60, 797–801.
The 100-mL H2S test and Petrifilm test combination also Micrology Laboratories  Coliscan Easygel. Retrieved February
yielded highly promising results (6.1% error for unimproved 21, 2010, from Micrology Laboratories: http://micrologylabs.
sources and 1.6% error for improved sources), but the most mennonite.net/Home/Our_Methods/Coliscan_Media/
Coliscan_Easygel.
accurate testing combination was determined to be the 20- Micrology Laboratories  Coliscan Easygel Guide, Goshen, IN.
mL H2S test and Easygel combination, since it yielded an NSO  Press Release – Census – 2002. Retrieved February 27,
error of 0% and since the 20-mL H2S test was the most accu- 2010, from National Statistics Office of the Philippines:
http://www.census.gov.ph/data/pressrelease/2002/
rate single test evaluated here.
pr0276tx.html.
These two testing combinations were also inexpensive Pillai, J., Mathew, K., Gibbs, R. & Ho, G.  H2S paper strip
(Table 29). The 100-mL H2S test and Petrifilm cost $1.60/ method – A bacteriological test for faecal coliforms in
 85 P. Chuang et al. | Comparison of four field-based microbiological tests Journal of Water, Sanitation and Hygiene for Development | 01.1 | 2011

drinking water at various temperatures. Water Sci. Tech. H2S strip test. J. Water Supply Res. Technol. – Aqua 43 (6),
40 (2), 85–90. 311–314.
Sobsey, M. & Pfaender, F.  Evaluation of the H2S method for WHO  Guidelines for Drinking-Water Quality, vol. 3,
detection of fecal contamination of drinking water. WHO, Surveillance and control of community supplies, Geneva,
Geneva. Switzerland.
United Nations  UN Millennium Development Goal 7, Target WHO/UNICEF  Progress on Sanitation and Drinking-Water
7C. Retrieved October 11, 2010. http://www.un.org/millennium 2010 Update. WHO, Geneva.
goals/environ.shtml. WHO/UNICEF Joint Monitoring Programme  Water for Life:
Venkobachar, C., Kumar, D., Talreja, K., Kumar, A. & Iyengar, L.  Making it Happen. WHO, Geneva. Retrieved March 21, 2011.
Assessment of bacteriological water quality using a modified http://www.unicef.org/media/files/JMP_05_text.pdf.

First received 17 October 2010; accepted in revised form 4 February 2011