                A PILOT EPIDEMIOLOGICAL STUDY OF HEALTH RISKS
                   ASSOCIATED WITH SWIMMING AT KUHIO BEACH



                            David M. Morens
                            Kimberly K. Roll
                            Roger S. Fujioka







                    Project Completion Report KSDS-5







                               March 1994






                             PREPARED FOR
                           State of Hawaii
                        Department of Health
                   Contract No.:  ASO Log No. 92-613
             Project Period: 1 April 1992-31 December 1993
                 Principal Investigator: David M. Morens


                    WATER RESOURCES RESEARCH CENTER
                     University of Hawaii at Manoa
                       Honolulu, Hawaii 96822


I.	MOTIVATION FOR STUDY

A.  Need for an Epidemiological Study.  There is widespread
belief in Hawai`i that the State's waters are polluted with
sewage.  In recent lawsuits against the City and County of
Honolulu, which operates O`ahu's treatment plants, Hawai`i
residents have claimed illnesses resulting from swimming in
O`ahu's recreational waters.  These claims are difficult to
evaluate without information on public health risks. But at
present no studies of health risk are underway, nor to our
knowledge are any being planned.  Moreover, there may be
potential sources of water contamination or even health risks
other than sewage discharge, including soil and storm drain
runoff, animals, and contamination of beach waters by high bather
densities.

	Other data suggest that indicator organisms are poor markers
of sewage contamination and of health risk, inasmuch as they may
be found in environmental sources that have not been contaminated
with sewage.  Furthermore, the same organisms alleged to indicate
sewage contamination are found in storm drains that are not
contaminated with sewage.  Reasonable questions arising from this
situation include what originating sources indicator organisms
actually reflect, whether they indicate a health risk independent
of their source, and whether, in Hawai`i, they are of any value
in monitoring health risks.  No matter how many studies of "water
quality" are completed, public health decision-making will remain
clouded by controversy until the actual human risk posed by
Hawai`i's marine waters has been scientifically measured.  While
standard measurements of the impact of sewage discharge can have
important ecologic and aesthetic implications, and while
measurements of indicator organisms may at best allow crude (but
conceivably erroneous) inferences about health risks, only
studies of exposed humans can potentially resolve the ultimate
controversies surrounding Hawai`i's waters.  Epidemiological
study of health risks posed by ostensibly contaminated waters is
thus, in our opinion, the single most important type of study
that needs to be undertaken.

	Evaluation of marine health risks associated with the
outfall of a storm drain, into which sewage is not released,
constitutes an important opportunity to examine some of these
questions by controlling for sewage contamination.   The ultimate
aim of this type of research, for which the present study is a
pilot evaluation, is thus to identify and measure those risks,
and to provide the public and public officials with accurate data
that can serve as a basis for informed discussion and public
health decision making.


B.  Assessment of Previous Water Quality Studies.  The waters of
Mamala Bay, including in some cases waters near the sites of this
study, have for many years been monitored by City and County of
Honolulu, the State of Hawai`i, and various research
investigators.  Extensive data collected over prolonged periods
document water quality, including study of bacterial indicator
organisms.  However, as noted, microbial organisms assayed as
water quality indicators may have little or no relationship to
health risks.  One problem is that many of the indicator
organisms are not human pathogens.  The notion, implicitly
endorsed by the Environmental Protection Agency (EPA), that
counts of either non-pathogenic or potentially-pathogenic
indicator organisms in seawater directly correlate with potential
health risk may not be valid.  The studies that seemed to
generate this notion are flawed methodologically and have
unfortunately been subjected to considerable over-interpretation
(vide infra).

	Of significance, the environmental conditions in Hawaii
differ from those of the continental USA.  In Hawai`i it has been
shown that the same organisms regarded as indicating sewage
contamination of seawater are found in high concentration in soil
and streams.  In fact, freshwater streams and canals, as well as
ocean water fed by such sources, may typically contain higher
levels of indicator organisms than seawater allegedly
contaminated by sewage outfalls.  For example, a 1990 report of
co-investigator R. Fujioka (Hawai`i Department of Health Contract
No. 88-465 [WQ/P-1]) showed that while Hawai`i's recreational
beach waters were largely free of microbial indicator organisms,
fresh water streams and brackish waters such as canals and
lagoons contain them in high numbers:  nearly half of all
freshwater sources sampled by Fujioka's team had > 2,000
coliforms and/or > 1,000 E. coli per 100 ml GMT.

	Although Hawai`i's recreational standard has changed from
fecal coliform to enterococci, this new indicator may be no
better: the same data showed that the great majority of fresh
water sources in Hawaii (79%) had greater than 35 enterococci per
100 ml GMT, more than 5 times the safety ceiling set by the State
(7 per 100 ml).  Most had extremely high levels, as many as 6,000
to 7,000 per 100 ml GMT.  Preliminary data suggest that storm
drain run-off is also heavily laden with indicator organisms.
Thus Hawai`i lacks not only data on human health risks from
marine recreational waters, but a reliable way to distinguish
between contamination by sewage and other sources of microbial
indicator organisms.   This distinction is critical because
without knowledge of the source of indicator organisms there can
be no rationale for developing public health interventions.

	Subsumed under the larger question of the safety of
Hawai`i's waters are other questions of importance.  Is sewage-
contamination a principal human risk, or is it even a risk at
all?  Do other sources of microbial contamination (e.g., soil,
storm drains) pose a health risk to beachgoers?  To what extent
do indicator organisms reflect contamination from these various
sources, and are any of them correlated with human risk?  All of
these questions remain unanswered in Hawai`i.  Risk data from
other locales are of little help in resolving these issues in
Hawai`i.  Over the years many studies in the United States and in
other nations have looked at health risks to swimmers in marine
waters, but, as discussed below, the results have been
conflicting and difficult to generalize.


C.  Assessment of Previous Epidemiological Studies.  As early as
1953, published epidemiologic studies identified and predicted
some of the difficult methodologic problems that complicate
interpretation of studies four decades later.  In that year,
Stevenson reported a prospective study of fresh water swimmers
(1) that led to two apparently contradictory conclusions: 1) that
swimmers had higher incidence rates of illness than non-swimmers,
and 2) that the increased risk was not associated with fecal
borne infection, but with skin and respiratory infections.  To
most epidemiologists today this is not surprising -- persons who
choose to participate in outdoor activities such as swimming are
different in many ways from those who do not; these differences
may well account for illness risks irrespective of the presence
or absence of infectious organisms in the water.

	In most of the subsequent studies (2-22), including some
that did and some that did not appear to show a risk,
investigators typically failed to control or adjust for biases
associated with inherently different risks of persons who chose
exposure or non-exposure.  For example, the most influential
studies to date (those which formed the basis of EPA guidelines
for marine indicator organisms) have been repeatedly
misinterpreted.  In the multi-site EPA study of Cabelli et al
(6), swimmers had generally higher rates of gastro-intestinal
symptoms (including "highly credible" symptoms) than non-
swimmers, and in some cases the differences were statistically
significant.  However, the differences most often touted as
measuring risk were based on the wrong comparisons -- swimmers v.
non-swimmers at "contaminated" beaches, instead of swimmers at
contaminated v. uncontaminated beaches and non-swimmers at
contaminated v. uncontaminated beaches or, even more
appropriately, swimmers v. swimmers and non-swimmers v. non-
swimmers at the same beaches during times of both high and low
contamination.  Although often ignored, the EPA data show clearly
that swimmers had higher illness rates even at uncontaminated
beaches, suggesting that "swimming proclivity", rather than
swimming in contaminated water could have explained illness risk.
Moreover, when using the published figures to make the more
correct comparisons in the same EPA data set -- simultaneously
comparing swimmers with swimmers and non-swimmers with non-
swimmers at the same beaches at times of high and low
contamination, the risk differences appear to either become less
pronounced or to disappear entirely.  For example, although the
EPA data demonstrate that swimmers had a statistically
significant fourfold higher risk than non-swimmers of highly
credible gastrointestinal symptoms after swimming in the Lake
Ponchartrain levee at a mean enterococcus density of 495 per 100
ml (p < 0.01), swimmers also had a significantly increased risk
when swimming in the same levee at 44 per 100 ml enterococcus
mean density, and there was no difference in risk for either
swimmers v. swimmers or non-swimmers v. non-swimmers during times
of low and high density.  The same data set shows similar
problems with E. coli density correlations.  Thus, while these
data might fairly be used to suggest that swimmers have higher
rates of illness than non-swimmers, they do not demonstrate that
the risk is associated with water contamination.

	Additional problems with existing epidemiologic data on
health risks of swimming in sewage contaminated water are more
difficult to exclude by attention to study design and
interpretation.  Two of the most serious potential problems are
those related to bather-bather transmission and to community
transmission of infectious organisms that also appear in sewage
outflow.  Most studies have not been able to control for the
first problem: potentially deleterious effects of nearby bathers
on seawater.  Persons who excrete infectious enteric or skin
organisms into seawater at a crowded beach may conceivably expose
surrounding bathers to markedly higher titers of microorganisms
than could ever be achieved by outfall drift.  For example, rough
calculations based on available data on sewage outfall (in this
case, Honouliuli), drift, and organism decay suggest that an
individual who excreted as little as 0.1 gram of fecal material
containing 105 organisms per ml into 1 cubic meter of water would
potentially expose an adjacent swimmer to from 1-10 billion-fold
more infectious organisms than would a community epidemic of 100
persons excreting 100 grams of the same infectious fecal material
into the sewage system daily.  Bather density may confound not
only counts of organisms in the water but also background health
risks: it is not hard to imagine that a crowded beach might lead
not only to more organisms in the water, but also a higher risk
of illness in general, as severe crowding has been associated
with all types of infectious agents transmitted by the (so-
called) "fecal-oral" and respiratory routes.

	The second potential problem is that if infectious enteric
organisms are being transmitted in the community, it is likely
that they will show up in the community's sewage.  Studies that
attempt to examine risks of swimmers from sewage contaminated
waters should ideally distinguish between illness acquired in the
water and illness in the community, since the two may be highly
correlated.  In most of the published studies, persons who swim
in a community's waters also live in the community and are
exposed to the community's transmissible diseases, greatly
complicating inferences about source of illness for study
subjects who become ill.

	There are many other pitfalls in epidemiological studies as
well. At the outset, Hawai`i has few reliable data upon which to
base decisions about either health risks or the potential
efficacy of risk reduction measures for users of recreational
waters.  Obtaining such information would require well designed
and well conducted studies that address Hawai`i's unique
environmental situations.  Thus a pilot study of beachgoers at a
storm drain outfall is a small but necessary preliminary step in
understanding health risks in Hawai`i.



II.  GOALS, OBJECTIVES AND LIMITATIONS OF STUDY

A.  Expectations of Study.  This is a pilot study, necessarily
limited in scope by funding and time constraints.  It was
accepted and clearly stated from the outset that it was unlikely
that conclusive data on health risks would be generated.
Therefore, the primary objective was to conduct a feasibility
study, to specifically determine and measure critical  parameters
that would allow more refined estimates of the scope of work, and
the amount of funds, required to obtain health risk data
considered definitive at any pre-selected level of certainty.  As
components of the feasibility study, we also sought to obtain the
following four types of data:

1.	Unit cost data on the expense involved in obtaining and
analyzing health risk information, in a form and manner that
could be readily extrapolated to estimate the costs of
future health risk studies in Hawaii marine waters

2.	Background rates of indicator illnesses (principally
gastrointestinal illnesses, ear infections, and skin
conditions) in study subjects from and not from Hawai`i

3.	Evaluation of certain novel methodologic "corrections"
incorporated into the study design (see below), intended to
overcome problems encountered with other studies of marine
water health risks, including those of EPA

4.	Development and assessment of an approach to marine water-
associated health risk assessment that could be applied to
future studies in Hawai`i


B.  Additional Objectives.

1.	To determine whether storm drain discharge poses a
measurable health risk to users of recreational water

2.	To determine whether any detected health risks are
associated with indicators of water quality

3.	To generate information useful to public officials and
health planners concerned with health and sanitation



III.  METHODOLOGY

A.  Study Site and Experimental Design.  The study site, Kuhio
Beach, is a popular Waikiki beach into which the Kapahulu storm
drain feeds (Figure 1).  The water sampling sites include the
enclosed portions of Kuhio Beach Site 1 and Site 2.  Site 5 is
directly adjacent to the outfall, near Kuhio Beach 1.  Queen's
Surf Beach (Site 7) was used as an "exposure control" beach.  The
study involves correlation of illnesses in swimming and non-
swimming beach users with regularly-obtained water quality
sampling from selected sites in the vicinity (Figure 1).  A
nearby "control" beach (Queen's Surf Beach) was also monitored,
as were waters near the outfall (Site 5).

	In keeping with the design of the EPA epidemiological study,
water samples from the interview beaches were taken twice a day
(morning and afternoon) to determine the microbiological quality
of the water during the testing period, as described in Standard
Methods for the Examination of Water and Wastewater (23).
Concentrations of enterococci were monitored because the EPA
study claimed a direct correlation between the concentrations of
enterococci in recreational waters and incidence of swimming
associated diarrheal diseases (6), and because enterococci are
the principal indicators used in Hawai`i to estimate potential
health risks.

	Since this epidemiological study was conducted in Hawai`i,
we decided that data applicable to Hawai`i should be
incorporated.  Thus, the waters were sampled for two other
indicator bacteria shown to be useful in Hawai`i.  The first
alternative indicator was Clostridium perfringens, a more
reliable indicator of sewage pollution of streams in Hawai`i than
fecal coliforms, E. coli, or enterococci (24).  C. perfringens
was assayed using the methods as descibed by Bisson and Cabelli
(25).  The second alternative indicator was aerobic bacillus
bacterium as a marker for soil contribution in water samples.

	Twice-daily samples were tested for pH, reactive phosphorus,
salinity, and concentrations of selected indicator organisms:
fecal coliforms, E. coli, enterococci, bacillus spores and C.
perfringens.  These test methods are detailed in companion reoprt
by Roll et al.  Up to three interviewers encountered and
administered face-to-face structured interviews to beach users,
inquiring about health, beach use and other experiences in the
prior three days.  Each subject was then re-contacted by
telephone three days later, to provide information on incident
illnesses and subsequent exposures (see questionnaire, Appendix
A).  The study was thus prospective.  It also attempted to
control for exposures that occurred before and after the index
exposure.

B.  Case Definitions.  Gastrointestinal illnesses were of primary
interest, although illnesses of the eye, ear, nose, skin and
respiratory tract were also surveyed to obtain a better overview
of the epidemiology of illnesses related to the general beach
environment (see Appendix A).

	Gastrointestinal symptoms included vomiting, diarrhea,
stomach ache nausea, gas, cramps, and anorexia.  Highly credible
gastrointestinal symptoms, as defined by EPA (6), include any one
of the following:

	(1)	Vomiting
	(2)	Diarrhea with fever
	(3)	Diarrhea with disabling condition (remained home,
         remained in bed or sought medical advice), or
	(4)	Nausea or stomach ache accompanied by fever.

	Otic symptoms include earache or ear infection.  Eye
symptoms include sore eye, discharge, itching, watering or
redness.  Skin symptoms include rash, exclusive of sunburn.
Respiratory symptoms include sore throat, cough, and runny nose.
These symptoms were defined in the marine and fresh recreational
water quality studies conducted by the EPA (6).  The same symptom
definitions have been used in other water quality and swimming-
related illness studies (17).

	Categories of exposure status by study site included non-
swimmers, swimmers who immersed the head but do not swallow
water, and swimmers who swallowed water.  As noted, each exposure
category can be stratified on "swimming proclivity", and on three
days prior and three days subsequent beach exposures.

	In addition to the interview data and water quality sampling
data (see below), the interviewers estimated hourly bathing
density as follows:  the number of persons in the water in a
designated area of measured size, bounded by easily identified
landmarks, was directly counted.  Interview and microbial
sampling emphasized the morning hours, when bather contamination
of water should be minimal, and also peak hours, when bather
contamination should be maximal.

	To control for the potential effect of community
transmission of organisms also found in seawater, we chose as
subjects a mix of residents and non-residents (tourists).  The
latter subjects presumably are exposed to relatively fewer
community risks since they do not live, work, or go to school in
the community.  Because many tourists speak Japanese, all of our
interviewers were bilingual, speaking fluent Japanese and
English.  Due to budget constraints, we were unable to monitor
community transmission of enteric organisms.

	To attempt to minimize the effects of adjacent bather
variables we conducted bather density assessments four times each
sampling day.  We also focused on early morning interviews and
samples, since the beaches are uncrowded for many hours
overnight.  Unfortunately, because of budget limitations, it was
not possible to regularly assay for human skin organisms such as
staphylococci.  The study design was devised to include the
following improvements:

     1. control of bather density variables

     2. improved correlation between epidemiological and
        microbiological sampling

     3. inclusion of newer, more sensitive nonpathogen
        indicator organisms

     4. improved reference group use in analyses


C.  Development of Survey Questions.  During the summer of 1992 a
preliminary questionnaire was developed and refined.  This
questionnaire was put together from pre-existing questionnaires
and data sets available to Naowarat Charoenca, formerly of the
WRRC.  This was expanded and developed by the investigators to
incorporate questions elicited by other comparable studies of
marine risks in the U.S. and elsewhere.  An attempt was made to
include questions that would elicit information directly
comparable with data of the EPA studies, to allow for comparison.
In September 1992 an initial study questionnaire was field-tested
at Kuhio Beach.  After administration to approximately 100
subjects it was slightly revised, field-tested a second time, re-
revised, and then made available for use.  The second set of
questionnaire revisions was largely for ease of coding and data
entry.

	The questionnaire (Appendix A) was designed to elicit
information about demographic characteristics, including place of
residence, about past and present history of various symptoms of
illness, and about exposures to other recreational waters in the
past three days.  It also elicited information on time, place,
and phone numbers for the follow-up phone interview, conducted
three calendar days after the initial face-to-face interview.
The follow-up interview elicited information about incident
illnesses and symptoms, as well as recreational water exposures
that may have occurred after the first interview.  The three day
time intervals before and after the index exposure were designed
to cover the incubation periods of the most common
gastrointestinal and dermal conditions without being so long as
to introduce recall biases.

	The symptoms covered in the interviews included all
gastrointestinal symptoms from the EPA studies (including the
"highly credible" symptoms -- vide infra), as well as some not
found in these and other studies.  Some symptoms not associated
with swimming in the EPA and other studies were deleted to
shorten the interview process.  In distinction to the EPA and
some other studies, we did not exclude persons who used
recreational waters before or after the initial interview, but we
did obtain information on their other exposures so that they
could be stratified in analysis.  Among the reasons for their
inclusion is the fact that many Hawai'i beachgoers, including
tourists, frequent beaches on one or more occasion:  their
exclusion would render the study more time-consuming and more
expensive, because the majority of encountered subjects would
later have to be dropped.  It would also potentially introduce
biases associated with sampling a highly unrepresentative group -
- significant findings would therefore be less generalizable.
More importantly, however, previous published studies (vide
supra) suggest that persons who are regular swimmers may be at
higher risk of illnesses independent of water contamination.  The
questionnaire elicited information on general frequency of
swimming and beachgoing to control for this phenomenon.  In
addition, the study proportionally sampled local residents and
tourists, an important mechanism for eliminating the possible
confounding effects of community transmission of the same or
similar illnesses (for reasons cited above, tourists should be at
relatively lower risk of community- transmitted enteric
infections).


D.  Conducting the Survey.  The survey was conducted by from one
to three bilingual (English/Japanese) University of Hawai`i
students working concurrently.  Each student was trained in
questionnaire administration and study methodology.  Overall
supervision was provided by Dr. Morens.  Ms. Roll managed the
students on a day-to-day basis, and co-ordinated water sampling
with questionnaire administration.  A graduate student team
leader, Ms. Yurie Sakakibara, worked with the other interviewers
to coordinate interview schedules, arranged back-up coverage when
an interviewer was unavailable to work at a scheduled time,
compiled and reported weekly statistics on interview completion,
etc.  A fourth graduate student was responsible for data entry,
editing and (supervised) univariate analysis.  Interviews were
conducted relatively evenly throughout a 12 month year (September
1992 to September 1993).  A special attempt was made to interview
beachgoers within 24 hours after rainfall.  The water sampling
was performed twice daily, in the morning and afternoon, on each
day of interviewing.  Each of the interviewers regularly rotated
beach sites on the same and subsequent days.

	Beach users were sequentially encountered as they left the
beach, or while they were at the beach.  All children aged 5-19,
along with all accompanying adults, and every other third adult
or adult couple were encountered (in an attempt to over-sample
children, who are at higher risk for many enteric infectious
diseases, and who may also be more likely to swallow water while
swimming).  At the encounter, persons who agreed to participate
(verbal informed consent -- see consent statement, Appendix B)
were administered the questionnaire (Appendix A) in standardized
face-to-face interview.  The interview generally took about ten
minutes.  Participants were given Hawai`i postcards as
compensation for their time.  Persons who were ill with a
"credible" gastrointestinal illness (see below), or who intended
to leave O`ahu within the next three days, were excluded from the
study.






IV.  RESULTS & DISCUSSION

A.  Cost Data.  Based on the methods and scope of work in the
Kuhio Beach study, including training, field-testing, and start-
up time, we estimate that five full-time interviewer equivalents
(10 graduate students each working 20 hours per week for 12
months) would generate approximately 21,600 completed
questionnaires at a cost of $80,000.   This compares favorably
with the EPA studies, which elicited information on 25,242
subjects over a five year period, but presumably incurred a much
greater cost in 1992 dollars.  Cost data for microbial studies
are much more highly dependent on which tests are chosen, but a
reasonable sampling scheme of three days per week, with two
samplings per day, for five indicator organisms, would incur an
expense of approximately $50,000.  Professional time,
miscellaneous expenses, and overhead expenses are not included in
these figures.

	Using data generated from this study (vide infra), a cost
algorithm for similar health risks studies is discussed below.
It is apparent that with even a low background incidence rate of
symptoms of interest, well-planned future studies could achieve
ample statistical power to detect moderate differences (e.g.,
two-fold increases) in incidence rates between exposed and
unexposed groups.


B.  Descriptive Data on Study Participants.  We administered
questionnaires to 3721 persons and completed first and follow-up
interviews on 2556 persons (68.8% completion rate) using 1.2
person-year of interview time (an estimated 2154 completed
interviews per interviewer-year).  The completed
questionnaires/phone interviews provided data from 2,556
subjects, representing 7,668 person-days, or 21.0 person-years,
of follow-up.  There were 1,681 visits to Kuhio Beach site 1
(65.8%), 590 visits to Kuhio Beach site 2 (23.1%), 113 visits to
Queen's Surf Beach (4.4%), and 172 visits to Site 5 (6.7%).  The
decision to "under-sample" the latter two sites was based on the
need to target limited resources to the sites of greatest
interest (vide infra).  The participants included 51.4% males and
48.6% females.  Participant ages ranged from 2 to 85 years, with
the majority in the 20's and 30's.  Of the 2,556 respondents,
1,334 (52.2%) were residents of Japan, 621 (24.3%) were residents
of United States other than Hawai`i, 273 (10.7%) were residents
of O`ahu, 20 (0.8%) were residents of other Hawai`ian islands,
and the remaining 308 participants (12.1%) were from other
countries, prominently including Canada (176 participants; 6.9%).
The majority of the participants (1,462; 57.2%) identified
themselves as being of Asian ethnicity; 878 (34.4%) identified
themselves as Caucasian, 74 (2.9%) as Pacific Islander, 21 (0.8%)
as Hispanic, and eight (0.3%) as black.  Ninety of the remaining
110 persons (81.8%) identified themselves as being of mixed
ethnicity.


C.  Behavioral Risk Data.  Of the 2,556 subjects followed for
three days, only three developed otic conditions, and only three
developed dermal conditions exclusive of sunburn, which had been
chosen as a "control" diagnosis (Table 1).  All three who
"developed" otic conditions had had otic complaints in the three
days prior to beach use, as had two of the three persons who
developed dermal conditions, leaving respective incidence rates
of 0.00 and 0.05 cases per person-year of follow-up.  Because the
numbers are too small for meaningful analysis, these conditions
are not considered further.  Usable data on incident illnesses
fell into these separate categories:  gastrointestinal illnesses,
ophthalmic conditions, and constitutional/respiratory conditions.
Each of these is considered separately, below.


	1.   Gastrointestinal Complaints.  Fifty-one of 2,556
subjects (2.0%) experienced one or more gastrointestinal symptom
in the three days following interview.  However, a larger number
(65 persons) had experienced such symptoms in the three days
before interview, suggesting that the frequency of
gastrointestinal symptoms following beach use was not increased
(Table 1).  Regarding "highly credible gastrointestinal illness"
(HCGI), as defined in various EPA reports (4), we identified only
one incident case, corresponding to 0.05 cases per person-year.
This single individual had diarrhea and vomiting with fever, but
had not swum or swallowed ocean water.

	Adjusting for prevalent (before interview) gastrointestinal
symptomatology in the 51 persons with one or more gastro-
intestinal symptom, 34 persons experienced incident vomiting
and/or diarrhea in the three days after interview, representing
1.62 incident cases per person-year.  The frequency of incident
illness did not differ by beach site, gender, age, place of
residence, or ethnic identity (data not shown).

	For further analysis we stratified study participants on
whether they had, or had not, visited the same or other beaches
in any of the three days prior to encounter, and examined the
frequency of gastrointestinal illness ("incident gastrointestinal
illness") in non-swimmers, swimmers who did not swallow water,
and swimmers who did swallow water.

	As the data did not differ between Kuhio Beach sites 1 and
2, nor between strata for one-time and multiple beach visitors,
the two sites are combined in collapsed-strata analysis and
designated as "Kuhio Beach".  The frequency of "incident
gastrointestinal illness" in Kuhio Beach swimmers, as defined
above, was 28 cases per 1,677 beach visits, or 2.03 cases per
person-year.  This was more than twice as high as in non-swimmers
(4 cases per 595 visits, 0.82 cases per person-year), but the
difference was not statistically significant (p=0.21, chi
square).  Moreover, there was no association between swallowing
water and developing "gastrointestinal illness":  the risk of
illness was actually lower in persons who swallowed water, 1.06
v. 2.42 cases per person-year, relative risk 2.28, though not
significantly so (p=0.72, Fisher Exact Test).  There was no
qualitative or trend difference in frequency of illness by number
of times (from one to four) that water was swallowed.
   2.  Ophthalmic Illness.  Forty-one persons reported one or
more ophthalmic complaint (1.95 cases per person-year).  There
were no significant differences in frequency of ophthalmic
complaints by beach visited, gender, age, ethnic background, or
residence.  Only 30 of the 41 cases were incident, the remainder
having been prevalent in the three days before visiting the
beach.  This corresponds to an incidence rate of 1.43 cases per
person-year.  There was no relationship between either swimming
or swallowing water and frequency of ophthalmic complaints.
   3.  Fever and Constitutional Symptoms.  Only nine of 2,556
persons (0.43 cases per person-year) reported fever in the three
days following interview.  As noted, one of these individuals
also reported vomiting and diarrhea.  All of the rest had had
fever in the three days prior to interview, suggesting that their
subsequent fevers may not have represented incident conditions.
In any case, there was no significant relationship between
swimming or swallowing water and frequency of fever in the three
days after visiting the beach.  When examining other
"constitutional" symptoms such as headache and bodyache, there
were no significant differences.  Swimmers who swallowed water
had a marginally increased frequency of experiencing headache in
the three days after visiting the beach (p = 0.08, chi square).


D.  Microbial Risk Data.  In the EPA study (26), data were
reported on microbial sampling for indicator organisms at various
sites.  Despite the absence of a detectable risk associated with
swimming or swallowing water, we sought to correlate indicator
counts with human illness on the theory that an actual risk
association might exist but be "buried" in the behavioral data.
(This might occur in any of several ways.  For example, swimming
or water swallowing might constitute a risk only if some
threshold level of organisms was exceeded, or only on specific
occasions such as following rainfall, or only during times of
high bather density.  We therefore used logistic regression to
associate the ordinal exposure variables (counts of specific
indicator organisms, numbers of bathers) with the categorical
outcome variables of presence or absence of the human illnesses
of interest.  Water quality data listing concentration of all
indicator bacteria at each site are summarized and discussed in
companion report by Roll and Fujioka (27).

	Analysis of the data show a notable lack of association of
any indicator organism with risk of illness (Tables 2 and 3).
Neither enterococci, fecal coliforms, C. perfringens, nor any of
the other organisms we studied appeared to be correlated with
illness risk.  This lack of association held when studying
swimmers only (Table 3), or swimmers who swallowed water (not
shown).  Although rainfall was associated with indicator
organisms, it was not associated with human risk for any of the
parameters studied.  These results support the subsequent EPA
study (Calderon et al., 21) which showed that when the source of
the indicator between the recreational water is non-point source
rather than sewage, there is no correlation with bacterial counts
in water and increased incidence of human enteric diseases.

	It was interesting to note some apparent correlation between
bather density and staphylococcus counts in limited pilot
sampling at the beginning of the study.  However, as noted above,
we were unable to continue staphylococcus sampling throughout the
study period.  In any case, logistic regression analysis revealed
no association between bather density and risk of
gastrointestinal illness, constitutional symptoms, or eye
disorders (as defined above; Tables 2 and 3, summarized in Table
4).






V. SUMMARY, CONCLUSIONS & RECOMMENDATIONS

A.  Feasibility and Cost Considerations for Related Health
Studies. The pilot epidemiological study provided important
feasibility information on assessment of human health risks in
Hawai`i waters.  The information in this study was collected in
what we believe to be a highly cost-effective manner, although we
have no comparison figures to support this belief.  Nevertheless,
it is apparent that even with economical methodology, detection
of relatively rare health outcomes associated with common
exposures becomes prohibitively expensive as the desired level of
certainty regarding study validity is increased.  This is
demonstrated in Table 5, which provides crude cost estimates, in
1992 dollars, for health risk studies of the type conducted here.
The Table is constructed to provide cost estimates for studies
conducted under pre-specified scientific expectations, which must
be chosen, from among many options, beforehand.

	In reading the Table, it is helpful to think of the first
three columns as representing preliminary decisions about the
features of the study, and the last four columns as estimates of
the scope of work required as a consequence of these choices.
The scope of work is thus dictated by the features that have been
chosen: it is represented in the Table by the numbers of subjects
required to meet the study requirements and, consequently, the
study cost.
The first column records the magnitude of the risk difference
between exposed and non-exposed persons (e.g., swimmers v. non-
swimmers) this study will try to detect, and is selected based on
public health perceptions of the importance of a high v. low
"attributable risk per cent".  For example, if public health
officials feel that only a doubling of gastroenteritis risk for
swimmers would be of sufficient concern to warrant public health
action, then the 100% category might be chosen.  If, on the other
hand, it was felt that even a 25% increase in risk was sufficient
to justify action, then the 25% category might be selected.

	The second column represents the statistical "power" of the
study to correctly detect a health risk when a health risk
exists.  This column could be interpreted as answering the
question "If swimming at the beach really does double a person's
chance of getting gastroenteritis, how likely is the study to
detect this increased risk, with acceptable certainty, in a study
of X people?"  This is of course the opposite of asking how
likely it is that a true risk difference will be missed merely
because too few people have been studied.  It can be seen that
raising the level of certainty from only 80% to 99% more than
doubles the study size and cost.  "Power" thus represents a
potential trade-off, and its selection may depend on the public
health consequences of failing to take preventive action because
a true health problem is erroneously believed not to exist.

	The third column represents the "probability value" or "p
value".  This figure reflects the likelihood that a detected risk
difference is real, and not just a chance statistical artifact.
For example, one might ask the question "If our study finds that
swimmers are twice as likely as non-swimmers to get
gastroenteritis, how likely is it that that difference is not
real, but a fluke arising from statistical chance?"  Obviously,
the lower the "p value" the better, but as was true for raising
statistical power, lowering the p value increases study costs
considerably.  The p value is selected based on public health
assessment of the consequences of falsely concluding that a risk
exists when, in reality, it does not.  When public health action
is required, and especially when large expenditures would be
involved, the lowest p value possible is desirable.

	From the first three columns of Table 5, it can be seen that
a number of public health considerations need to be factored into
any determination of whether it is desirable to conduct a health
risk study and, if it is, what study size to recommend.  Ideally,
any study undertaken would be highly able (power, column 2) to
detect even a small risk increment (column 1) with reasonable
assurance (column 3) that the detected risk was real.
Unfortunately, satisfying all three of these criteria, for water
risk assessment or for any study of uncommon conditions, may be
extremely expensive.

	Certain other considerations should also be pointed out in
cost assessment calculations.  First, the figures used in Table
5, provided by this pilot study, are situation-specific.  The
unit cost of completing an encounter and follow-up interview,
including data entry and analysis, was estimated at $3.70 in
direct costs.  This cost might vary considerably in other
studies, depending on such factors as availability of subjects,
length of the questionnaire, required level of training and
experience for interviewers, etc.  It is our opinion that it
would be difficult to achieve or reduce this figure in other
studies because, in an effort to overcome the expected problem of
insufficient funds to achieve optimal sample size, we used an
abbreviated questionnaire, concentrated on an extremely crowded
beach where potential study subjects were nearly always
available, even after rains, and hired and trained comparatively
inexperienced student interviewers.  Furthermore, in part for
methodologic reasons (see below), we used a three day follow-up
time, shortening administration of the second questionnaire, and
reducing the number of subjects lost to follow-up.

	An even more important consideration is that the sample size
calculations used to provide estimates in column 4, Table 5, were
taken from background risk figures generated by this study.  To
be as accurate as possible in estimating what is ultimately
unknowable, sample size calculations require a best guess of the
baseline risk of the disease in question.  When multiple outcomes
need to be detected, the sample size must normally be further
increased.  We selected an intermediate background incidence
estimate (i.e., neither conservative nor generous) of about 15
incident cases per 2,556 persons, the approximate risk for
serious but not "highly credible" gastrointestinal symptoms in
non-swimmers (0.0087), and also for swimmers who swallowed water
(0.0087), before correction for prevalent illnesses.  Such
figures may or may not be representative of risks associated with
other situations.  For most of the incident outcomes we sought to
detect, relatively few cases were ultimately identified.  For
example, we found only about 30 persons with even loosely-defined
gastrointestinal illnesses.  Sample size estimates vary widely as
the background occurrence of the disease varies.  Thus the sample
size required to detect risk associated with a "healthier" site
would be comparatively larger than the sample size needed to
detect risk associated with a "less healthy" site.  A reasonable
estimate of the "background risk" of a particular study site is
perhaps the single most important factor in estimating study
size, scope and cost.  It should also be pointed out that the
figures in Table 5 do not include indirect or overhead costs,
include only nominal professional fees, and are expressed in 1992
dollars.


B.  Risk to Swimmers at Kuhio Beach.   We found no evidence of
human health risk associated with recreational use of Kuhio Beach
waters known to receive discharge from the Kapahulu Storm Drain.
The results clearly indicate that human health risk was not high.
However, it should be obvious from the above discussion, that it
cannot be concluded with certainty that no health risk exists.
There are several reasons for caution in interpretation of these
data.

	First and perhaps the most important reason is that the
study may not have been large enough to detect a health risk that
was real, but small.  While it may be reassuring that this study
appears to rule out the possibility of a major health risk, it is
still possible that swimmers in Kuhio Beach waters are at low but
increased risk that would only be detectable with larger studies.
As can be seen from Table 5, even if our study size had been more
than tripled (with an approximate tripling of cost), there would
still have been a 20% chance of missing a doubling of
gastroenteritis risk for swimmers (albeit from extremely low risk
to very low risk), under the assumption of the generic background
rate we originally selected (15 cases per 2,556 beachgoers).  As
it was, we observed a slightly higher background rate than the
one selected, and can therefore be about 90% certain that risk of
gastroenteritis for beachgoers was not increased two-fold or
more, about 95% certain that the risk for swimmers was not
increased by two-fold or more, and about 95% certain that the
risk for water-swallowers was not increased by  two-fold or more.
However, since the background rate of illness can never actually
be known until the study is finished, to ensure that studies will
generate usable results it is normally necessary to select a
conservative estimate for the purpose of study size calculations,
thereby increasing the expected cost.  In some situations,
however, a sequential occurrence algorithm can be set up in order
to terminate the study if the desired level of certainty is
reached earlier than expected based on conservative planning
assumptions.

	Secondly, because of limited resources we decided to focus
more on Kuhio Beach and less on the comparison beach (Queen's
Surf).  The latter beach was generally much less crowded, making
it more difficult and more costly to enroll subjects at that
site.  Furthermore, although we would have liked to enroll more
subjects swimming in the Site 5 area, few persons entered or swam
in that area, limiting availability of subjects.  As a
consequence, the majority of subjects were from Kuhio Beach Site
1 and Site 2, reducing the  possibility of detecting risks in
outfall v. non-outfall water in favor of the more important, but
incomplete, comparison of risks at times of high and low organism
counts at the same sites.  We consider this an unfortunate but
necessary trade-off.

	Thirdly, although we did consider "bather density" we were
unable to monitor skin organisms such as staphylococci.  Since
humans harbor, exude, and excrete high concentrations of
potentially pathogenic skin and enteric organisms into the water,
it would seem plausible that any risk associated with swimming in
crowded waters would be more likely to come from adjacent
swimmers than from distant environmental sources.  However, this
possibility could neither be studied nor controlled, except in a
crude way with density estimates.

	Fourthly, it is possible that true microbial risk was
periodic and therefore submerged in the low background risk.
This might theoretically occur in any of several ways.  Risk
associated with only a high threshold of indicator organisms
would escape detection, as could risk not associated with
indicator organism counts.  It is also conceivable that some
particular combination of physical parameters we did not
specifically look for (e.g., rainfall, pH, and tide) was
associated with a true but submerged risk.  These kinds of
situations would be difficult or impossible to detect in any
scientific study.

	Finally, we were unable to control for community
transmission of organisms.  However, we doubt this harmed our
study for two reasons.  Since the Kapahulu Storm Drain presumably
does not contain sewage, there would be at best limited
opportunity for such human pathogens to enter it.  Also, we
encountered primarily tourists, who presumably had less contact
with prevalent endemic organisms than persons who live and work
in the community.  We found no differences whatsoever, in any
parameter, between tourists and local residents.


C.  Evaluation of Study.  Among the unique aspects of this study,
we followed persons for only three days, as compared to eight to
10 days in the EPA studies.  It might be argued that our study
could have missed detecting a true risk by selecting an
artificial cut-off time to detect incident illness that was less
than the incubation period for many of them.  Aside from trade-
off considerations in conducting a study with limited resources,
we believe the three day cut-off period is justified and possibly
optimal.  It is widely recognized that while many different types
of organisms may cause waterborne illness, the principal cause is
a group of viruses known as "Norwalk agents" (apparently enteric
caliciviruses).  These viruses are associated with an incubation
period of 12-36 hours.  Thus virtually all incident Norwalk
disease, as well as some, but not all, other diseases associated
with swimming, would be picked up in our three day interval.
Furthermore, the three day period greatly limits the possibility
and magnitude of recall biases, which plague virtually all health
studies that rely on subjects' memories.  Recall biases appear to
rise substantially as the duration of the recall period
increases, and are particularly problematic for persons who have
developed illnesses.  Such persons may differentially recall
exposures ("rumination bias").  Finally, the three day period
allowed us to enroll many tourists; most Japanese visitors to
Hawai`i, and tourists from many other places as well, stay on
O`ahu for only a few days.  Specifying long follow-up periods
would have forced us to impose highly restrictive selection
criteria on beachgoers, limiting the generalizability of
findings, eliminating our ability to control for the potentially
confounding effect of community transmission, and greatly
increasing study costs.

	It is interesting to note that the background rate of
gastrointestinal illnesses we observed (between 10 and 20 per
1,000 persons per three day follow-up) is around 10%-50% that of
the total (i.e., non-"highly credible") in the EPA studies, which
were associated with follow-up periods of from two to 10 days.
When we applied the EPA definition for "highly credible"
symptoms, the background rate fell to about one twentieth that
found in the EPA studies.  The reason for this is unknown:  it
could be due to follow-up differences or to lower community
illness rates.  That the EPA background illness rates tended to
be higher in long v. short follow-up could reflect incident
illnesses with long incubation periods, but may more likely
represent an unvarying background risk observed over a longer
interval.  This raises the question of whether high level endemic
transmission in the EPA study communities could have confounded
results by leading to simultaneous community acquisition of
infection and appearance of sewage-borne community organisms in
the water, thereby falsely implicating water as the source of
illness.

	We generated sufficient outcome data to evaluate risks for
three separate categories of illness in beachgoers:
gastrointestinal illnesses, ophthalmic illnesses, and
constitutional/respiratory illnesses.  Of these, the category of
greatest interest is the gastrointestinal illnesses, which have
been epidemiologically associated with epidemics caused by eating
contaminated food and drinking contaminated beverages.
Furthermore, the association of many different behaviors with
gastrointestinal illnesses has become fixed in the public's mind.
As noted, the background rate of gastrointestinal illnesses we
found in three day follow-up was lower than that found in the EPA
studies.  Moreover, based on a looser definition of outcome than
that used in some of the EPA computations (vomiting and/or
diarrhea of no other known cause), we did not detect increased
risk in swimmers, nor in swimmers who swallowed water.  Using the
stricter of the EPA definitions, we identified only one person
with "highly credible" gastro- intestinal symptoms.  It thus
appears that the background rate of illness in our subjects in
1992-1993 is lower than for EPA study subjects in the 1970s.
While this is good news, it also means that detecting the
presence or absence of water-associated health risks will likely
be much more difficult and more expensive in Hawai`i; measuring
those risks, if they exist, may be even more problematic.

	The value of seeking to detect "highly credible"
gastrointestinal symptoms in Hawai`i is questionable.  Although
scientists agree that highly specific case definitions are
preferable to less specific definitions, which may misclassify
persons with other diseases as having diseases of interest, in
this situation they are probably of greater value in studies with
long versus short follow-up, because after 10 days recall of
minor complaints like stomach ache may be prone to significant
error.  On the other hand, tourists with novel vacation
itineraries to prompt recall, and probably also with increased
salience of somatic occurrences, would probably be more likely to
accurately recall, and less likely to erroneously remember,
symptoms occurring within the past three days.

	It is curious to note that the incidence rate of
gastrointestinal illness in swimmers was higher (though not
significantly so) than in non-swimmers, while swimmers who
swallowed water were at lower risk (again, insignificantly) than
swimmers who did not.  While it must be assumed that these
findings represent statistical chance, in planning similar health
studies it should also be considered that persons who swim might
be at inherently different risk of illness than those who do not.
To cite a purely speculative but illustrative example, swimmers
might be more likely to be "athletic risk takers" and non-
swimmers "cautious couch potatoes".  Each of these lifestyle
characterizations might, independent of any particular instance
of swimming or water-swallowing, predict a different background
risk of gastrointestinal symptoms.  For example, the former group
of "athletic risk takers" might tend to overeat, drink more
alcoholic beverages, consume too many chili peppers, be less
cautious in hand washing, diaper changing, etc., and thereby be
at elevated risk of gastrointestinal illness before even
approaching the water to swim.  As discussed above, we believe
this phenomenon probably operated in the EPA and some other
studies, and may well have led to misinterpretation of the
findings of those studies.

	Skin and ear infections are well known risks of swimming,
but for the most part are not thought to be associated with water
contamination.  Even persons who swim in highly chlorinated pool
water are at risk of acquiring ear and skin infections due to
skin maceration, which encourages growth of dermal bacteria and
fungi that may be part of the individual's normal flora.  Viral
conjunctivitis has occasionally been associated with swimming
pool outbreaks in which chlorination has lapsed, but is not
commonly thought to be waterborne and, like skin and eye
infection, would be suspected of resulting from autoinfection
rather than exogenous marine organisms.  Constitutional symptoms
are by nature non-specific and therefore make poor outcome
indicators for studies of this type.  The constellation of fever,
headache, bodyache, and respiratory symptoms is characteristic of
influenza and many influenza-like illnesses, but few are thought
capable of water-borne transmission.  Thus, it is not surprising
that the risk of these conditions was not greatly different in
swimmers.  It is of interest that the three-days frequencies of
the individual and composite symptoms in these categories were
about half those detected in EPA studies with eight to 10 day
follow-up.  We found no evidence to suggest that either these
symptoms or related illnesses would be of value for outcome
monitoring.

	In conclusion, this epidemiological pilot study generated
realistic cost estimates for possible future studies of this
type, and also indicated that no major health risk to swimmers
appeared to exist.  While this study could not, because of its
size and scope, completely rule out all health risks, it was
suggested that such risks, if they exist, must be very small.  We
hope that these data, interpreted in conjunction with other
public health data on the severity, impact, preventability, and
priority of illnesses associated with potentially water-borne
microorganisms, will be useful to health officials and policy
makers.

Acknowledgements:  We thank Andrew Grandinetti for assistance in
statistical analysis; Etsuko Chida and Achara Thawatwiboopol also
assisted in data entry and analysis.  Interviews were conducted
by Etsuko Chida, Atsushi Nishihata, Yurie Sakakibara, and Satoru
Yamamoto.




REFERENCES

1.  Stevenson AH.  Studies of bathing water quality and health.
Am J Publ Health 1953;43:529-38.

2.  Moore B, Blowers AR, Crone PB, et al.  Sewage contamination
of coastal bathing waters in England and Wales.  A bacterial and
epidemiological study.  J Hyg 1959;57:435-71.

3.  Cabelli VJ, Levin, MA, Dufour AP, McCabe LJ.  The development
of criteria for recreational waters.  In:  Discharge of Sewage
From Sea Outfalls, London:  Pergamon Press; 1975;63-73.

4.  Cabelli VJ, Dufour AP, Levin MA, McCabe LJ, Haberman PW.
Relationship of microbial indicators to health effects at marine
bathing beaches.  Am J Public Health 1979;69:690-6.

5.  D'Alessio DJ, Minor TE, Allen CI, Tsiatis AA, Nelson DB.  A
study of the proportions of swimmers among well controls and
children with enterovirus-like illness shedding or not shedding
an enterovirus.  Am J Epidemiol 1981;113:533-41.

6.  Cabelli VJ, Dufour AP, McCabe LJ, Levin MA.  Swimming-
associated gastroenteritis and water quality.  Am J Epidemiol
1982;115:606-16.

7.  Mujeriego R, Bravo JM, Feliu MT.  Recreation in coastal
waters: Public health implications.  VIes Journes tud
Pollutions 1982:585-94.

8.  El Sharkawi F, Hassan MNER.  The relation between the state
of pollution in Alexandria swimming beaches and the occurrence of
typhoid among bathers.  Bull High Inst Public Health Alexandria
1982;12:337-51. [Or 1983?]

9.  Foulon G, Maurin J, Quoi NN, Martin-Bouyer G.  tude de la
morbidit humaine en relation avec la pollution bacteriologique
des eaux de baignade en mer.  Rv Fran Sci L'Eau 1983;2:127-43.

10. Seyfried PL, Tobin RS, Brown NE, Ness PF.  A prospective
study of swimming-related illness.  I. Swimming-associated health
risk.  Am J Publ Health 1985;75:1068-70.

11. Fattal B, Peleg-Olevsky E, Yoshpe-Purer Y, Shuval HI.  The
association between morbidity among bathers and microbial quality
of seawater.  Water Sci Tech 1986;18:59-69.

12. Fattal B, Peleg-Olevsky E, Agursky T, Shuval HI.  The
association between seawater pollution as measured by bacterial
indicators and morbidity among bathers at Mediterranean bathing
beaches of Israel.  Chemosphere 1987;16:565-70.

13. Cheung WHS, Kleevens JWL, Chang KCK, Hung RPS.  Health
effects of beach water pollution in Hong Kong.  Proc Inst Water
Env Management 1988:376-83.

14. Ferley JP, Zmirou D, Balducci F, et al.  Epidemiological
significance of microbial pollution criteria for river
recreational waters.  Int J Epidemiol 1989;18:198-205.

15. Lightfoot NE.  A prospective study of swimming related
illness at six freshwater beaches in Southern Ontario.  Ph.D.
Dissertation, University of Ontario, 1989.

16. Alexander LM.  Health risks associated with exposure to
seawater contaminated with sewage: The Blackpool beach survey
1990.  Lancaster University Environmental Epidemiology Research
Unit, 1991.

17. Cheung WHS, Chang KCK, Hung RPS, Kleevens JWL.  Health
effects of beach water pollution in Hong Kong.  Epidemiol Infect
1990;105:139-62.

18. Zmirou D, Ferley JP, Balducci F, et al.  valuation des
indicateurs microbiens du risque sanitaire li aux baignades en
rivire.  Rev pidm Sant Publique 1990;38:101-10.

19.  Balaraja R, Soni, Raleigh V, Yuen P, Wheeler D, Machin D,
Cartwright R.  Health risks associated with bathing in sea water.
BMJ 1991;303:1445-6.

20.  Brown JM, Campbell EA, Rickards AD, Wheeler D.  Sewage
pollution of bathing water.  Lancet 1991;2:1208-9.

21. Calderon RL, Mood EW, Dufour AP.  Health effects of swimmers
and nonpoint sources of contaminated water.  Int J Env Health Res
1991;1:21-31.

22.  Brown JM, Jones F, Kay D, Stanwell-Smith R, Wyer M, Morano
R.  Water and non-water-related risk factors for gastroenteritis
among bathers exposed to sewage-contaminated marine waters.  Int
J Epidemiol 1993;22:689-708.

23. American Public Health Association, American Water Works
Association, Water Pollution Control Federation.  Standard
Methods for the Examination of Water and Wastewater, 17th
Edition, APHA:1989.

24. Fujioka RS, Shizumura LK.  Clostridium perfringens, a
reliable indicator of stream water.  J Water Pollution Cont Fed
1985;57:986-92.

25. Bisson JW, Cabelli VJ.  Membrane filter enumeration for
Clostridium perfringens.  Appl Env Microbiol 1979;37:55-66.

26. Cabelli, V. J.  1983.  Health effects criteria for marine
recreational waters, EPA-600/1-80-031, U.S. Environmental
Protection Agency, Washington, D.C. 98 p.

27. Roll K.K., Fujioka R.S.  Microbiological characterization of
the water and sediment in Kapahulu Storm Drain and Kuhio Beach.
Project Completion Report KSDS-4. March 1994.




Table 1.  Signs and symptoms in 2,556 beachgoers in the three
days before and after interview, four beach sites combined,
Kapahulu Storm Drain risk study, Honolulu, Hawai`i, 1992-1993.


	                  Three Days       Three Days
                       Before           After

Gastrointestinal
Nausea                   13               4
Vomiting                  6               2
Diarrhea                 27              39
Stomach ache             21               3
Cramps                    5               1
Gas                      26               6
Anorexia                  9               4

Ophthalmic
Red eyes                 30              38
Itchy eyes                1               3
Watery eyes               6              11
Eye discharge             6               3
Eye pain                  6               7
Photophobia               0               1

Optic
Ear ache                  9               2
Ear infection             3               1

Dermal
Rash                     10               3
Sunburn                  63              60

Other
Fever                    15               9
Headache                 88              38
Body aches               41              19
Rhinorrhea               48              28





Table 2.  Association between selected indicator organisms and
occurrence or gastrointestinal (GI), constitutional, and eye
disorders, as defined in text, Kuhio Beaches 1 and 2, all
beachgoers, logistic regression analysis, Honolulu, Hawai`i,
1992-1993

GI disorder

Indices	            *beta     Standard    Odds    Confidence Interval
                      (b)     Error       Ratio    Lower    Upper

Enterococci          0.0059    0.0091     1.006    0.988    1.024
fecal coliform       0.0044    0.0047     1.004    0.995    1.014
C. perfringens       0.0032    0.012      1.003    0.980    1.027
E. coli              0.0055    0.007      1.006    0.992    1.020
bather density      -0.0029    0.0065     0.997    0.984    1.010


Constitutional disorders

Indices	            *beta     Standard    Odds    Confidence Interval
                      (b)     Error       Ratio    Lower    Upper

Enteroterococci     0.0005    0.0016      1.000    0.997    1.004
fecal coliform     -0.0003    0.0006      1.000    0.999    1.001
C. perfringens      0.0072    0.0310      1.007    0.948    1.070
E. coli             0.0053    0.0021      1.005    1.001    1.009
Bather density     -0.0035    0.0069      0.997    0.983    1.010

Eye Disorders

Indices	            *beta     Standard    Odds     Confidence Interval
                      (b)     Error       Ratio    Lower    Upper

enterococci         0.0015    0.0044      1.002    0.993    1.010
fecal coliform      0.0004    0.0013      1.000    0.998    1.003
C. perfringens      0.0014    0.0091      1.001    0.984    1.019
E. coli             0.0008    0.0026      1.001    0.996    1.006
bather density      0.0085    0.0083      1.009    0.992    1.025

* 1 - b = power of a test


Table 3.  Association between selected indicator organisms and
occurrence of gastrointestinal (GI), constitutional, and eye
disorders, as defined in text, Kuhio Beaches 1 and 2, swimmers
who immersed heads in water, logistic regression analysis,
Honolulu, Hawai`i, 1992-1993


Indices	            *beta     Standard    Odds     Confidence Interval
                      (b)     Error       Ratio    Lower    Upper

enterococci         0.0168    0.0177      1.017    0.982    1.053
fecal coliform      0.0059    0.0061      1.006    0.994    1.018
C. perfringens      0.1023    0.1664      1.108    0.799    1.535
E. coli             0.0079    0.0091      1.008    0.990    1.026
bather density     -0.0029    0.0069      0.997    0.984    1.011


Constitutional disorders

Indices	            *beta     Standard    Odds     Confidence Interval
                      (b)     Error       Ratio    Lower    Upper

enterococci         0.0007    0.003       1.001    0.995    1.007
fecal coliform      0.0001    0.0014      1.000    0.997    1.003
C. perfringens      0.0008    0.0079      1.001    0.985    1.016
E. coli             0.0002    0.0025      1.000    0.995    1.005
bather density     -0.0038    0.0072      0.996    0.982    1.010


Eye Disorders

Indices	            *beta     Standard    Odds     Confidence Interval
                      (b)     Error       Ratio    Lower    Upper

enterococci         0.0001    0.0014      1.000    0.997    1.003
fecal coliform     -0.0006    0.0007      0.999    0.998    1.001
C. perfringens      0.0097    0.0545      1.010    0.907    1.124
E. coli             0.0001    0.0022      1.000    0.996    1.004
bather density      0.0118    0.0090      1.012    0.994    1.030

* 1 - b = power of a test


Table 4.  Logistic regression of bather density on illness
outcome for three types of illnesses, Kuhio Beaches 1 and 2,
Honolulu, Hawai`i, 1992-1993.  The illnesses are defined above in
the report.

Illness               beta (b)    Standard      Odds Ratio   95% Confidence
                                  Error                      Interval
Gastrointestinal      -0.0029     0.0065        0.9971       0.9845-1.0099
Constitutional        -0.0035     0.0069        0.9965       0.9831-1.0101
Ophthalmic             0.0085     0.0083        1.0085       0.9923-1.0251


