GIARDIASIS AND CRYPOSPORIDIOSIS IN THE URBAN-RURAL SPECTRUM
By
Rebecca L. Ives

A THESIS
Submitted to
Michigan State University
in partial fulfillment of the requirements
for the degree of
MASTER OF SCIENCE
Fisheries and Wildlife
2011

ABSTRACT
GIARDIASIS AND CRYPOSPORIDIOSIS IN THE URBAN-RURAL SPECTRUM
By
Rebecca L. Ives

This study explored occurrence of the environmentally transmissible forms of
Cryptosporidium spp. and Giardia spp. and disease incidence patterns in rural and urban land
cover types. The risks of infection and illness caused by Cryptosporidium spp. and Giardia spp.
in Combined Sewer Overflow (CSO) discharge receiving waters were also examined.
Significant differences in patterns of giardiasis and cryptosporidiosis were found between urban
and rural zip codes. The correlation of population with both diseases suggests that population
levels are important factors of transmission in the study area. Although differences in the pattern
of disease occurrence between urban and rural areas were found, no evidence of a difference was
found in post-hoc testing of the occurrence of the environmentally transmissible forms of the
parasites between the surface waters of the Grand River and River Raisin watersheds.
Health risks due to Cryptosporidium spp. and Giardia spp. in the Grand River at the point
of CSO discharge and at recreational sites downstream were assessed. A hypothetical swimming
scenario at the discharge point represents moderate daily health risk to a swimmer while multiple
day recreation at the discharge point over the recreation season represents a high health risk.
Despite CSO discharges, recreational use of the Grand River is predicted to meet the
recommended freshwater recreational criteria of 8 illnesses per 1000 swimmers at least 92.4% of
the time for both cryptosporidiosis and giardiasis at the three recreational sites examined.

ACKNOWLEDGEMENTS

This study was funded in part by the National Oceanic and Atmospheric
Administration (NOAA) grant #89043. The author acknowledges S. Devin McLennan for
assistance in plug flow modeling and personnel at the City of Grand Rapids Wastewater
Plant for providing access to the MARB facility and sampling assistance.

iii

TABLE OF CONTENTS

List of Tables...................................................................................................................... vi
List of Figures .................................................................................................................... ix
Chapter One: Introduction.................................................................................................. 1
1.1 Significance of Cryptosporidium and Giardia to Public Health ............................. 1
1.2 Environmental Transmission of Cryptosporidium and Giardia .............................. 4
1.3 Sources of Environmental Contamination .............................................................. 5
1.4 Cryptosporidiosis and Giardiasis: Reporting of Human Disease Cases ................. 8
1.5 Epidemiological use of Geographic Information Systems (GIS) ......................... 13
1.6 Research Objectives .............................................................................................. 14
Chapter Two: Material and Methods ............................................................................... 16
2.1 Creation of ArcGIS Database................................................................................ 16
2.2 Statistical Analysis – Land Cover and the Incidence of Cryptosporidiosis and
Giardiasis .............................................................................................................. 23
2.2.1 Statistical Comparison of Urban and Rural Demographic Attributes .......... 23
2.2.2 Statistical Comparison of Disease Occurrence ............................................ 24
2.3 Sample Collection – Grand River Watershed ....................................................... 26
2.3.1 Recreational Areas ....................................................................................... 26
2.3.2 Waste Treatment Systems ............................................................................ 27
2.4 Sample Collection – River Raisin watershed ........................................................ 31
2.5 Sample Processing ................................................................................................ 33
2.6 Recovery Efficiency .............................................................................................. 35
2.7 Statistical Analysis: Grand River and River Raisin Watershed Comparison ....... 37
2.7.1 Meaningful Comparisons ............................................................................. 37
2.7.2 Test of Normality and Equal Variance Assumptions ................................... 37
2.7.3 Statistical Analysis of Grand River Watershed Recreational Sites .............. 38
2.7.4 Statistical Analysis of Grand River Watershed Recreational Sites and
MARB .......................................................................................................... 39
2.7.5 Statistical Analysis of River Raisin Watershed ........................................... 39
2.7.4 Statistical Analysis of River Raisin vs Grand River Watershed .................. 40
2.8 Risk Simulation and Statistical Analysis .............................................................. 41
Chapter Three: Land cover and the Incidence of Cryptosporidiosis and Giardiasis......... 46
3.1 Results ................................................................................................................... 46
3.2 Discussion ............................................................................................................. 55

iv

3.3 Conclusion ............................................................................................................ 59

Chapter 4. Grand River and River Raisin Watershed Comparison: Cryptosporidium
and Giardia Occurrence ............................................................................... 61
4.1 Parasite Occurrence Results .................................................................................. 62
4.2 Recovery Efficiency Results ................................................................................. 69
4.2 Statistical Analysis Results: Grand River Watershed Recreational Sites ............. 74
4.3 Statistical Analysis Results: Grand River Watershed Recreational Sites and
MARB ................................................................................................................... 74
4.4 Statistical Analysis Results: River Raisin Watershed ........................................... 75
4.5 Statistical Analysis Results: River Raisin vs Grand River Watershed ................. 76
4.6 Discussion ............................................................................................................. 78
4.6.1 River Raisin Watershed................................................................................ 78
4.6.2 Grand River Watershed ................................................................................ 78
4.6.3 River Raisin vs Grand River Watershed ...................................................... 80
4.7 Conclusion ............................................................................................................ 82
Chapter 5: Recreational Health Risks of Combined Sewer Overflows ........................... 83
5.1 Risk Assessment Results ....................................................................................... 83
5.2 Discussion ............................................................................................................. 95
5.3 Conclusion .......................................................................................................... 100
APPENDICES ................................................................................................................ 101
REFERENCES................................................................................................................ 146

v

List of Tables

Table 1.1 Reported Cases of Cryptosporidiosis and Giardiasis in the United
States and Michigan ............................................................................ 11
Table 1.2 CDC Criteria for Cryptosporidiosis Case Confirmation ....................... 12
Table 2.1 Land Cover Reclassification Scheme ................................................... 19
Table 2.2 ArcGIS Database Source Files .............................................................. 21
Table 3.1 Attributes of Study Area ....................................................................... 49
Table 3.2 Comparison of Disease Cases by Urban or Rural Designation. ........... 52
Table 3.3 Statistical Comparison of Disease Occurrence. .................................... 53
Table 3.4 Correlation Assessment......................................................................... 54
Table 4.1. Water Samples Positive for Cryptosporidium and Giardia ............... 64
Table 4.2 Geometric Mean of Parasite Occurrence by Sampling Location ........ 68
Table 4.3. Parasite Concentrations in CSO Samples .......................................... 69
Table 4.4. Cryptosporidium and Giardia Recovery in Matrix Spike
Samples ............................................................................................... 71
Table 4.5. Cryptosporidium and Giardia Recovery in Ongoing Precision
and Recovery (OPR) Samples............................................................. 72
Table 5.1. Risk Assessment Distribution Parameters. ......................................... 84
Table 5.2. Predicted Concentration Distributions of Cryptosporidium and
Giardia at the CSO Discharge Point ( CRIVER)..................................... 88
Table 5.3. Estimated Risk of Health Outcomes for Children Swimming at
Recreational Sites and at Point of CSO Discharge in the Grand
River Watershed .................................................................................. 91

vi

Table 5.4. Likelihood of Meeting Recreational Criteria of 8 Illnesses in
1,000 Swimmers (0.8%)...................................................................... 95
Table A1 Demographics: Population .................................................................. 103
Table A2 Demographics: Area (km2) ................................................................. 104
Table A3 Demographics: Population Density (People/km2) .............................. 105
Table A4 Demographics: Percentage of Individuals Below the Poverty
Level .................................................................................................. 106
Table A5 Demographics: Median Household Income (US Dollars) .................. 107
Table A6 Demographics: Median Age (Years) .................................................. 108
Table A7 Demographics: Mean Travel Time to Work (Minutes) ...................... 109
Table A8 Demographics: % Caucasian............................................................... 110
Table B1 Disease Comparison: Non-normalized ................................................ 112
Table B2 Disease Comparison: Population Normalized..................................... 113
Table B3 Disease Comparison: Area Normalized .............................................. 114
Table B4 Disease Comparison: Population Density Normalized ....................... 115
Table C1 Occurrence Data: River Raisin Categorized ....................................... 117
Table C2 Occurrence Data: River Raisin Pooled................................................ 118
Table C3 Occurrence Data: Grand River ............................................................ 119
Table C4 Occurrence Data: Grand River and MARB ........................................ 121
Table C5 Occurrence Data: River Raisin (Pooled) and Grand River ................. 124
Table D1 Detection Limits: River Raisin Categorized ....................................... 127
Table D2 Detection Limits: Grand River ............................................................ 128
Table D3 Detection Limits: Grand River and MARB ........................................ 129

vii

Table D4 Detection Limits: Grand River and River Raisin ................................ 130
Table E1 Matrix Spike: River Raisin Categorized Data ..................................... 132
Table E2 Matrix Spike: River Raisin Pooled Data ............................................. 133
Table E3 Matrix Spike: Grand River .................................................................. 134
Table E4 Matrix Spike: Grand River and MARB............................................... 135
Table E5 Matrix Spike: Grand River and River Raisin ...................................... 136
Table F1 OPR: River Raisin ............................................................................... 139
Table F2 OPR: Grand River ................................................................................ 140
Table F3 OPR: Grand River and MARB ............................................................ 141
Table F4 OPR: Grand River and River Raisin .................................................... 143

viii

List of Figures

Figure 2.1 Land Cover Reclassification Decision Tree ........................................... 18
Figure 2.2 ArcGIS Analysis Process Data Model .................................................... 22
Figure 2.3 Grand River Watershed Sampling Locations ......................................... 30
Figure 2.4 River Raisin Watershed Sampling Locations ......................................... 32
Figure 3.1 Land Cover Classifications of Study Area ............................................. 48
Figure 4.1 Geometric Mean of Parasite Occurrence Versus Site............................. 67
Figure 4.2 Watershed Land Cover in Ottawa and Lenawee County, MI ................. 73

ix

Chapter One: Introduction
1.1 Significance of Cryptosporidium and Giardia to Public Health
Cryptosporidium and Giardia are parasitic protozoa that infect the gastrointestinal
tract of animals and humans. These parasite infections may progress to clinical diseases,
called cryptosporidiosis and giardiasis, respectively. Cryptosporidium and Giardia are
generally considered zoonotic pathogens, meaning that the life cycle of the organism can
involve a human and a non-human animal host. This capability leads to reservoirs of
active infections outside the human population, with the potential for pathogen
transmission between human and animal populations. Both parasites are obligate
parasites; therefore they cannot reproduce outside of the host. However, both Giardia
and Cryptosporidium produce environmentally resistant stages, called cysts and oocysts,
respectively. These resistant stages are shed in the feces of infected individuals and
allow the organism to survive in the environment outside of the host’s body. This
environmental resistance increases the likelihood that a suitable host will encounter the
organism and possibly become infected. There are several routes of transmission of
Giardia and Cryptosporidium; the routes of interest in this study involve waterborne
transmission. Transmission may involve recreational contact with contaminated
waterbodies or transport through water that is then ingested in some manner; such as
contaminated drinking water and contaminated water used for irrigation of crops or
washing of foods that are eaten raw.
In the human population, Cryptosporidium infection occurs in both the
immunocompetent and immunocompromized, in both adults and children, and in both
the developing and industrialized world (18). The species of Cryptosporidium that cause

1

the majority of disease cases in humans are C. parvum and C. hominis. Less commonly,
disease cases in humans are also caused by C. meleagridis, C. felis, C. canis, C. suis, and
C. baileyi, as well as the cervine genotype of Cryptosporidium (18). Among the
immunocompetent, infection most commonly presents as gastrointestinal illness with
acute diarrhea after an incubation period of 3-14 days (18). Other clinical presentations
include persistent diarrhea as well as asymptomatic infection (18). The site of infection
is the intestinal tract among the immunocompetent, with diarrheal symptoms usually
lasting 6-14 days and the disease is generally self-limiting (18). Among the
immunocompromised population, diarrheal symptoms can be much more severe, with
daily stool outputs between 2 to 17L (18). Infection in the immunocommpromised
population may be persistent or chronic and may develop into infection at extraintestinal
sites, such as the biliary tract (18). Cryptosporidium infection is associated with
increased morbidity and mortality in the immunocompromised population (18).
The infective stage of Cryptosporidium, the oocyst, is shed with the feces from an
infected individual. The peak intensity of oocyst shedding coincides with the peak
intensity of symptoms, during this period an infected individual sheds an average
6

concentration of 10 oocyst per gram of feces (4,56). The duration of oocyst shedding
varies and may continue after diarrheal symptoms have cleared. In a study of 33
immunocompetent individuals (31 children), Baxby et al found that the majority (85%)
shed oocysts after diarrhea had ceased (4). Within this study, when measured from the
start of diarrheal symptoms, 100% of the infected individuals shed oocysts for 1 week,
82% for 2 weeks, 42% for 3 weeks, and 21% were still shedding oocysts after 4 weeks
(4,56)

2

Like crypotosporidiosis, giardiasis is a common protozoal infection worldwide,
infecting both children and adults in both the industrialized and developing world (23).
The prevalence of Giardia in stool specimens submitted for parasite analysis is between
2% -5% in industrialized countries and higher (20% - 30%) in developing nations (41).
Of the five recognized Giardia species, only Giardia lamblia (syn. G. duodenalis, G.
intestinalis) has been found to infect humans (1). Like Cryptosporidium, Giardia infect
the intestinal tract, primarily the duodenum and upper jejunum (41). Giardia infections
may be asymptomatic, acute, or chronic and infections are generally self-limiting in the
immunocompetent population (23). Asymptomatic infections, in which infected
individuals either have transient diarrhea that is unnoticed or no symptoms, may be the
most common (23). The incubation time between initial infection and shedding of
Giardia cysts is 12-19 days and symptoms generally occur between 6-15 days after
infection, but may appear between 1-75 days after infection (5,30,43,56). Clinical
symptoms of acute Giardiasis are predominantly diarrhea, which is characteristically
watery during the initial onset of symptoms (23). Some Giardia infections will become
chronic with intermittent diarrhea. Weight loss may also occur in chronic infections, up
to 10-20% of body weight (23). Complications of Giardiasis include nutrient
deficiencies, which are generally reversed after the infection is cleared (either
spontaneous or with treatment) (23). However, nutrient deficiencies in children with
chronic infection may negatively affect growth and development (23).

3

1.2 Environmental Transmission of Cryptosporidium and Giardia
The environmentally transmissible forms of Cryptosporidium and Giardia are the
oocyst and the cyst, respectively. Since both Cryptosporidium and Giardia are obligate
parasites, the original source of the oocysts and cysts in the environment is a fecal
contamination event. Individuals shed these oocysts and cysts during the patent period of
Cryptosporidium and Giardia infections and, as previously discussed, this shedding
period may be of long duration in some cases. High numbers of Cryptosporidium
oocysts and Giardia cysts may be shed by infected individuals. In infected humans, for
8

6

example, 10 Cryptosporidium oocysts per gram and/or 10 Giardia cysts per gram of
feces may be shed (4,46,56).
Cryptosporidium oocysts and Giardia cysts have several qualities that allow them to
be effectively transmitted through the environment. Oocysts and cysts are
environmentally resistant, protecting the organism from adverse conditions outside the
host. Both Cryptosporidium oocysts and Giardia cysts are resistant to chlorine, one of
the most commonly used disinfectants in water treatment, although Giardia is less
resistant than Cryptosporidium. Oocysts and cysts are immediately infectious, no
maturation is required before they can initiate infection in a new host. Additionally, low
doses of oocysts and cysts are capable of initiating infection. In humans, Giardia
lamblia infections may result from the ingestion of 10 cysts and doses of 10 to 25 cysts
in feeding studies caused infection in 36.4% of 22 volunteers (43). Feeding studies using
several C. parvum isolates have demonstrated that there is a range of infectivity among
isolates (38). However, the ID50 (the dose at which 50% of a population is infected) of
some isolates is as low as 9 ingested oocysts and infection may potentially be initiated
4

by a single oocyst (38). Also of importance, the waterborne transmission of
Cryptosporidium and Giardia from a single contamination event can potentially infect
large numbers of people (23).

1.3 Sources of Environmental Contamination
Various species of Cryptosporidium and Giardia infect humans, domestic livestock,
pets, and wildlife. This leads to multiple potential sources of contamination by these
parasites in the environment (18,49), although not all species may be medically
important in terms of human disease. Human sources of Cryptosporidium and Giardia
contamination may include improperly treated sewage, discharges of untreated sewage
via sanitary sewer overflows or combined sewer overflows (CSOs), land application of
biosolids and septage, and leaking sewer or septic systems. Animal sources may also
contribute to environmental contamination in a variety of ways. Runoff from domestic
livestock operations or fields fertilized using animal manure may carry Cryptosporidium
and Giardia into waterbodies. Defecation by pet animals in the environment and
defecation of wildlife also may provide sources of these parasites. However, although
genotyping studies of Cryptosporidium and Giardia have demonstrated that both hostadapted and zoonotic strains are present in wildlife (1,22), most strains are host-adapted
(22). Feng et al conclude that while wildlife can contribute to Cryptosporidium
contamination in the water, wildlife sources may not have major public health
significance (22). Studies on Cryptosporidium and Giardia concentrations have also
found higher concentrations in domestic animals than in wildlife (17,26). Fecal
contamination of waterways from human sources and domestic animals, therefore, may
be of greater concern than contamination by wildlife.

5

Cattle, sheep, pigs, and horses are susceptible to Cryptosporidium and Giardia
infections. Cattle are of particular interest due to high prevalence of Cryptosporidium
among some herds and high intensity shedding patterns. In studies examining cattle, the
percentage of animals infected with Cryptosporidium range from 0.17% to 69.2% (3,58).
In studies examining cattle, the percentage of animals infected with Giardia range from
11% to 73% (39,40,58). In a shedding study of calves, individual animals have been
documented as shedding up to 26 million Cryptosporidium oocysts per gram of feces
and 4.2 million Giardia cysts per gram of feces (58). No molecular typing of
Cryptosporidium or Giardia was performed in the shedding study, although
Cryptosporidium oocysts were morphologically similar to C. parvum, and therefore
assumed to be pathogenic to humans. In the study of shedding patterns in calves,
Cryptosporidium shedding was found to peak earlier and with higher maximum
individual concentrations per gram of feces than Giardia shedding. However, once the
calves reached four weeks of age, Giardia shedding intensity was four orders of
magnitude greater than the Cryptosporidium shedding intensity. The earliest date of
shedding was four days for both Giardia and Cryptosporidium. All sampled animals
had at least one Cryptosporidium shedding episode and all sampled animals had at least
one Giardia shedding episode (58).
The parasites produce infections that lead to disease states in livestock, however,
asymptomatic animals have also been found to shed Cryptosporidium and Giardia. In a
1997 study of asymptomatic sheep, pigs, cattle, and horses by Olsen et al, differences in
the population age and shedding were found between Cryptosporidium and Giardia (40).
In cattle, detection of Cryptosporidium tended to be more frequent in young animals (<6

6

months of age) than older animals, whereas in sheep, pigs and horses, detection tended
to be more frequent in older animals(>6 months of age) than younger animals. With
Giardia, detection in both cattle and sheep tended to be more frequent in young animals
than older animals. Detection of Giardia in pigs and horses tended to be more frequent
in older animals than younger animals. Olsen et al found that the difference in
prevalence between younger and older animals tended to be significant (P=0.05) except
in the case of sheep, in which no significant difference was found between older and
younger animals (40). Another study of asymptomatic and symptomatic cattle found no
statistical difference in prevalence of Cryptosporidium infection among diarrheic (63.3%
prevalence) and non-diarrheic (69.2% prevalence) animals (58).
The shedding of parasites by asymptomatic animals presents a challenge to reduction
of environmental loading of the parasites by management practices on the farm. If only
symptomatic animals shed the parasites, animals presenting with symptoms could
conceivably be isolated from the rest of the herd and manure from these isolated animals
more heavily managed to reduce environmental loading. In a seven-state study,
estimates of environmental loading of Cryptosporidium by feedlot steers ranged from
4

5

2.8 x 10 to 1.4 x 10 oocysts/steer per day for the arithmetic mean estimates and 9.1 x
3

4

10 to 3.7 x 10 oocysts/steer per day for the geometric mean estimates (3).
In a 2005 review of zoonotic transmission pathways by Hunter et al, studies
indicated that C. hominis is transmitted only between humans but domestic livestock,
especially cattle, are reservoirs of C. parvum (28). These studies also indicated that
zoonotic disease transmission results from direct contact with cattle and indirectly
through contamination of drinking water (28). In the case of giardiasis, however,

7

evidence from studies does not support zoonotic transmission as a major transmission
route (28).
In regards to the parasites Cryptosporidium and Giardia, one of the challenges to
management of animal manures for protection of human and animal health is the
stability of these parasites. In a study of C. parvum inoculated into various animal
manures/slurries, estimates of the amount of time for a 1 log decrease in infectivity (as
measured by DAPI and PI vital staining) ranged from three months to almost 1 year (29).
Even these long time periods may underestimate of the time required for inactivation, as
the vital staining methods are not as sensitive to low numbers of infective (oo)cysts as
animal infectivity or, in the case of Cryptosporidium, cell culture infectivity. The long
time periods required for inactivation through storage present logistical challenges to
manure management, as storage space may be limited. This may lead to spreading of
manures that still contain infectious (oo)cysts onto agricultural lands in order to clear
storage space for more recently deposited fecal material.

1.4 Cryptosporidiosis and Giardiasis: Reporting of Human Disease
Cases
Human cases of cryptosporidiosis and giardiasis are currently on the list of nationally
notifiable diseases compiled by the United States Center for Disease Control (CDC) and
public health officials (10). According to the CDC,
―A notifiable disease is one for which regular, frequent, and timely information on
individual cases is considered necessary for the prevention and control of the disease.‖
(16)

National reporting for cryptosporidiosis began in 1995 (2,972 cases, 27 states) (15).
National reporting for giardiasis officially began in 2002 (21,206 cases, 46 states) (11),
8

however, beginning in 1992 there are earlier cases reported from a smaller subset of
states which have been summarized by the CDC (12). Disease reporting at the state and
local levels is mandatory, however state reporting to the CDC is voluntary, therefore the
data used in summary reports generated by the CDC may be incomplete (10). Table 1.1
summarizes the reported cases of cryptosporidiosis and giardiasis from onset of national
reporting to 2008, the most recently published surveillance data. Cryptosporidiosis in
the United States has been trending upwards and recent data (2005-2008) show the
highest number of reported cases. Cryptosporidiosis in Michigan has also been
increasing. However, the cryptosporidiosis trend in Michigan has been fairly stable over
the period from 2001-2008. Giardiasis in the United States and Michigan peaked in
1996 and 1995, respectively. In the last five years of published data, the giardiasis
trends in both the United States and Michigan have been fairly stable.
The reported data for both cryptosporidiosis and giardiasis are likely to be
underestimates of the disease burden in the United States. For example, extrapolating
from the estimates of salmonellosis reporting (1% - 5%), in 2002 the true disease burden
in the Unites States could have been between 60,320 - 301,600 cases of
cryptosporidiosis and 424,120 -2,120,600 cases of giardiasis (reported cases: 3,016 and
21,206, respectively) (11).
Nationally reported surveillance data may contain both confirmed cases and probable
cases of giardiasis and cryptosporidiosis, depending on how a jurisdiction (i.e. State)
reports the data. The CDC has used four definitions for a laboratory confirmed case of
cryptosporidiosis from 1995 through 2011 as methods have developed. The criteria used
for these definitions are shown in Table 1.2. A case must meet one of the required

9

criteria (specific to definition publication interval) to be defined as a confirmed case of
cryptosporidiosis. The current (2011) CDC definition of a laboratory confirmed case of
cryptosporidiosis is ―the detection of Cryptosporidium organisms or DNA in stool,
intestinal fluid, tissue samples, biopsy specimens, or other biological sample‖ which
includes direct fluorescent antibody tests but excludes other immunodiagnostic tests (6).
The CDC has defined a laboratory confirmed case of giardiasis as ―the detection of
Giardia organisms, antigen, or DNA in stool, intestinal fluid, tissue samples, biopsy
specimens or other biological sample‖(7). A probable case is a case that is
epidemiologically linked with a confirmed case with symptoms that match with the
clinical description of disease (9). The epidemiological linkage of a probable case with
a confirmed case is necessary, due to the similarity of cryptosporidiosis and giardiasis
symptoms with many other gastrointestinal diseases.

10

Table 1.1 Reported Cases of Cryptosporidiosis and Giardiasis in the United States and Michigan
Disease

Year
1992
1993
1994
1995
1996
1997
1998
1999
2000
2001
2002
2003
2004
2005
2006
2007
2008

Cryptosporidosis
United States
Michigan
Number of
Reported
Reported
reporting
cases
cases
states
*
*
*
*
*
*
*
*
*
2,972
27
*
2,426
42
*
2,566
40
46
3,793
42
39
2,769
46
52
3,128
46
97
3,785
49
187
3,016
50
135
3,506
49
152
3,911
49
157
8,269
50
112
6,479
50
145
11,657
50
215
10,500
50
284

Reference

(15)
(15)
(14)
(15)
(11)
(11)
(11)
(11)
(9)
(9)
(9)
(8)
(8)
(8)

Giardiasis
United States
Michigan
Number of
Reported
Reported
reporting
cases
cases
states
12,793
22**
1,333
19,565
36**
1,199
26,346
42**
1,370
25,571
43**
1,435
27,778
45**
1,290
25,389
43**
1,212
24,204
42
1,172
23,245
43
1,166
21,772
44
1,135
19,659
42
1,003
21,206
46
923
19,718
45
781
20,655
45
718
19,789
45
783
18,958
45
715
19,421
45
620
18,913
45
611

* No Data Available
** Reports do not distinguish between zero cases and non-reporting of data

11

Reference
(12)
(12)
(12)
(12)
(12)
(12)
(11)
(11)
(11)
(11)
(11)
(9)
(9)
(9)
(8)
(8)
(8)

Table 1.2 CDC Criteria for Cryptosporidiosis Case Confirmation
Required Criteria
(X)
1995
Microscopic
demonstration of
X
Cryptosporidium
oocysts in stool
Microscopic
demonstration of
Cryptosporidium
X
in intestinal fluid
or small-bowel
biopsy specimens
Demonstration of
Cryptosporidium
antigen in stool by
X
immunodiagnostic
test
Demonstration of
Cryptosporidium
antigen in
intestinal fluid by
immunodiagnostic
test
Demonstration of
Cryptosporidium
nucleic acid by
PCR in stool,
intestinal fluid, or
tissue samples or
biopsy specimens
Demonstration of
Cryptosporidium
reproductive
stages in tissue
preparations

Definition Publication Year
1998
2009

2011

X

X

X

X

X

X

X

X

X

X

X

X

X

X

The Michigan Public Health Code (333.5111 of the Michigan Compiled Laws,
Public Health Code, Act No. 368 of the Public Acts of 1978) requires that certain
conditions and agents of infection, including Giardia lamblia and Cryptosporidium, be
reported to the local health authorities. All physicians and laboratories are required to

12

report cases of these infections or conditions. Other health professionals are authorized
to report, but are not required to do so. Reports of individual infection are required to
contain the following information: the patient’s full name, residential address, telephone
number, date of birth (or age), and gender. Reports must also include the specific
laboratory test used for diagnosis, date performed, test results, the name and address of
the reporting clinical laboratory, and the name, address, and telephone number of the
ordering person. This surveillance data can be used for identifying outbreaks, tracking
incidence of disease over time, and other epidemiological uses to protect the public
health.

1.5 Epidemiological Use of Geographic Information Systems (GIS)
Spatial statistical methods and GIS have been used to investigate Cryptosporidium
and Giardia epidemiology in a few studies. Odoi et al used GIS to explore the role of
livestock density / manure application in giardiasis rates in Canada but didn’t find strong
evidence that these were important epidemiological factors (37). Pollock et al examined
the association of cryptosporidiosis with a number of variables including human
population density, livestock density (cattle, sheep, farmed deer, goat, and pig), ratio of
farms to human inhabitants and two orthogonal axes (West-East and South-North) in
Scotland (42). Pollock et al found increased rates of C. parvum infection in areas with
lower human population densities, in areas with higher ratios of farms to humans, and
higher ratios of private water supplies to human populations, all of which can be
considered rural indicators (42). In a GIS case study in England and Wales, Lake et al
found higher rates of cryptosporidiosis in rural areas (defined by housing density), areas

13

with more agricultural manure application, and areas with poorly treated water supplies
(31). These studies demonstrate the usefulness of GIS in epidemiological investigations.

1.6 Research Objectives
Much of the research on Cryptosporidium and Giardia infection to date has been
focused on individuals or groups of individuals in order to determine the parasitic life
cycles, the symptoms of the diseases they produce, host specificity, resistance to
disinfectants, and medical interventions. Environmental research on Cryptosporidium
and Giardia tends to focus on occurrence in water and disinfection by water treatment.
Few studies are available regarding land cover and occurrence of Cryptosporidium spp.
and Giardia spp. or risk of infection/illness by Cryptosporidium spp. and Giardia spp in
recreational waters.
This work focused on parasite occurrence patterns in rural and urban land cover
types, as well as recreational health risks. The specific objectives were:
1. Establish a prototype GIS database for land cover attributes, demographic
attributes, incidence of cryptosporidiosis, and incidence of giardiasis in Hillsdale,
Lenawee, Ottawa, and Kent Counties.
2. Examine patterns between urban, rural/agricultural land cover and incidence of
giardiasis and cryptosporidiosis
3. Assess patterns between urban, rural/agricultural land cover and occurrence of
Cryptosporidium spp. and Giardia spp. in surface waters.
4. Assess the health risks due to Cryptosporidium spp. and Giardia spp. in
combined sewer overflow (CSO) receiving waters

14

Chapter two describes the creation of the prototype GIS database (research objective
one), environmental sampling, sample processing procedures, and data analysis methods.
In chapter three, research objective two is addressed, exploring the patterns between
disease and land cover type. In chapter three, the null hypothesis that disease levels are
the same in urban and rural areas was tested. Chapter four is a descriptive analysis of
the Cryptosporidium spp. and Giardia spp occurrence in two watershed segments with
differences in urban and rural/agricultural land cover. In chapter four, the null hypothesis
that parasite occurrence values was the same in the Grand River watershed and in the
River Raisin watershed was tested. Chapter five addresses objective four in a health risk
assessment of Cryptosporidium spp. and Giardia spp. from exposure via recreation in
combined sewer overflow (CSO) receiving waters.

15

Chapter Two: Material and Methods
2.1 Creation of ArcGIS Database
To investigate patterns between land cover type and the incidence of giardiasis and
cryptosporidiosis, a database was constructed in ArcGIS (ArcGIS 9.2, ESRI). The
database was populated with publicly available information from the Michigan
Geographic Data Library including census tracts from the counties of interest (Hillsdale,
Lenawee, Ottawa, and Kent) and the 2001 land cover raster image of Michigan’s Lower
Peninsula (30m x 30m cell size). The database also included information obtained from
the United States Census Bureau on the five digit ZIP code tabulation areas (ZCTAs) for
the year 2000 (51). Information on giardiasis and cryptosporidiosis cases was obtained
from the Michigan Department of Community Health. Case information over the time
span of January 2000 to December 2008 was summed by ZIP code and included in the
database. Data on additional attributes for each ZIP code in the study area were obtained
from the 2000 United States Census. These attributes were:
1.

Population

2.

Percentage of individuals below poverty level

3.

Median household income

4.

Median age

5.

Mean travel time to work

6.

Percentage of Caucasian/white individuals

A case information table with cases of cryptosporidiosis and cases of giardiasis
summed by ZIP code was created based on information from the Michigan Department

16

of Community Health (34). Using the ―Union‖ tool in ArcGIS, the census tract
information for Kent, Ottawa, Lenawee, and Hillsdale counties was combined into an
output file. Using the ―Extract by mask‖ tool with the census tract output file and the
raster file for the 2001 land cover of Michigan’s Lower Peninsula, an output file
containing the land cover of Kent, Ottawa, Lenawee, and Hillsdale counties was created.
Land cover was reclassified to allow hypothesis testing of the differences between
urban and rural land cover. The land cover of the counties was reclassified following the
reclassification scheme shown in Table 2.1 to produce a file of Kent, Ottawa, Lenawee,
and Hillsdale counties with land cover classified as ―rural‖, ―urban‖, and ―other‖. The
decision tree with the criteria used to reclassify all land cover categories other than the
―Parks / Golf courses‖ is shown in Figure 2.1. The ―Parks / Golf courses‖ is an
ambiguous category as the parks could range from city parks well connected to sewer
infrastructure and drinking water distribution networks located in high population
density areas to State parks with limited sewer and drinking water access. In this case,
―Parks / Golf courses‖ was reclassified as ―urban‖ since this category is associated with
human activity rather than agricultural livestock.

17

Figure 2.1 Land Cover Reclassification Decision Tree

YES

Urban

YES

Is land cover category designated as urban?

Urban

YES

Rural

YES

Urban

NO

Does land cover category involve
transportation infrastructure (roads,
airports, etc)?

NO

Can non-human animal waste (terrestrial,
excluding pet animals) be generated or
land-applied in this land cover category?

NO

Is this land cover category likely to support
high density human habitation?

NO

Other

18

Table 2.1 Land Cover Reclassification Scheme

2001 Land cover category

Reclassified category

Non-vegetated Farmland
Forage Crops/Non-tilled Herbaceous
Agriculture
Orchards/Vineyard/Nursery
Herbaceous Openland
Upland Shrub / Low density Trees
Oak Association
Aspen Association
Other Upland Deciduous
Mixed Upland Deciduous
Pines
Other Upland Conifers
Mixed Upland Conifers
Upland Mixed Forest
Lowland Deciduous Forest
Lowland Coniferous Forest
Lowland Mixed Forest
Lowland Shrub
Emergent Wetland
Mixed Non-Forest Wetland
Low Intensity Urban
High Intensity Urban
Airport
Road / Parking Lot
Parks / Golf Courses
Floating Aquatic
Sand / Soil
Water
Exposed Rock
Mud Flats
Other Bare / Sparsely vegetated

Rural
Rural
Rural
Rural
Rural
Rural
Rural
Rural
Rural
Rural
Rural
Rural
Rural
Rural
Rural
Rural
Rural
Rural
Rural
Urban
Urban
Urban
Urban
Urban
Other
Other
Other
Other
Other
Other

19

The reclassified raster file was then converted to a vector file using the spatial
analyst tools. The census tract output file was also used with the ―Clip‖ tool and the
Census 2000 5-Digit ZIP Code Tabulation Areas (ZCTA file) to produce an output file
with both the census tract information and the ZIP code for Kent, Ottawa, Lenawee, and
Hillsdale counties. Using the ―Join attributes from a table‖ tool, the case information
table was added to the output. The join was based on the ZCTA field. The spatial
analyst tool ―Tabulate area‖ was used with the vector land cover file and the output
containing the case information to produce a field with information on the percentage of
―rural‖, ―urban‖, and ―other‖ land cover for each ZIP code in the counties of interest.
Using the ―Join‖ tool, this field was added to the database.

The end product was a

database that has records of ZIP codes, percentage of ―rural‖ area, percentage of ―urban‖
area, percentage of ―other‖ area, and numbers of cases of giaridasis and numbers of
cases of cryptosporidiosis during the period from January 2000 to December 2008. ZIP
codes that were ≥50% rural by area were designated as rural ZIP codes. ZIP codes that
were ≥50% urban by area were designated as urban ZIP codes. Information on data files
used in the creation of the ArcGIS database is shown in Table 2.2. Figure 2.2 shows the
data model of the analysis process.
Using the database, ZIP codes that overlapped into counties other than Kent, Ottawa,
Lenawee, and Hillsdale counties were identified. Data from these ZIP codes were not
used in statistical analysis of disease occurrence and land cover type.

20

Table 2.2 ArcGIS Database Source Files

Data Type

Coordinate System
Geographic Projection

Case information
by ZIP code

N/A

N/A

2001 Land cover
Lower Peninsula
(Raster)

North
American
Datum of 1983

Michigan
GeoRef

River Raisin
Watershed
Grand River
Watershed
Census
2000 5-Digit ZIP
Code Tabulation
Areas (ZCTAs)
Census 2000
Tracts Hillsdale
county
Census 2000
Tracts Ottawa
county
Census 2000
Tracts Kent
county
Census 2000
Tracts Lenawee
county

North
American
Datum of 1983
North
American
Datum of 1983
North
American
Datum of 1983
North
American
Datum of 1983
North
American
Datum of 1983
North
American
Datum of 1983
North
American
Datum of 1983

Age of
Dataset
01/01/2000 12/31/2008
Constructed
from image
dates 1997 2001

Reference
34

36

Michigan
GeoRef

1998

36

Michigan
GeoRef

1998

36

Michigan
GeoRef

1/1/2000

51

Michigan
GeoRef

5/1/2009

36

Michigan
GeoRef

5/1/2009

36

Michigan
GeoRef

5/1/2009

36

Michigan
GeoRef

5/1/2009

36

21

Figure 2.2 ArcGIS Analysis Process Data Model
Census Tract
Kent
Union

Land cover

Output

Reclassify

Output

Census Tract
Ottawa
Union

Output

Census Tract
Lenawee
Union

Output

Census Tract
Hillsdale

Clip

Extract
by Mask

Output

ZCTA

Output

Join

Tabulate
Area

Output

Join

Output

Table: Case
Information

Output

Intermediate output file

Source data file

Final output file

ArcGIS tool

22

2.2 Statistical Analysis – Land Cover and the Incidence of
Cryptosporidiosis and Giardiasis
There were 102 ZIP codes identified in the Kent, Ottawa, Lenawee, and Hillsdale
counties and 54 of these ZIP codes were excluded from the statistical analysis. Two of
the excluded ZIP codes had neither >50% urban area or >50% rural area, five ZIP codes
were generic 3 digit ZIP codes which had no population associated with them (i.e. areas
located alongside interstate highways), two ZIP codes had no area associated with them
in the ArcGIS database, and 45 of the ZIP codes extended beyond the borders of
Hillsdale, Lenawee, Kent, and Ottawa county. Thus, 48 ZIP codes of the 102 ZIP codes
in Kent, Ottawa, Lenawee, and Hillsdale counties were analyzed.

2.2.1 Statistical Comparison of Urban and Rural Demographic Attributes
Information on study area attributes for each of the 48 ZIP codes was obtained from
either the constructed ArcGIS database or the 2000 United States Census (51) for
statistical comparison. These attributes were:
1.

Population

2.

Area

3.

Population density

4.

Percentage of individuals below poverty level

5.

Median household income

6.

Median age

7.

Mean travel time to work

8.

Percentage of Caucasian/white individuals

23

These attributes were compared between the urban and rural ZIP codes and examined for
statistically significant differences. Datasets for statistical analysis were not normally
distributed. The ―Median household income‖ and ―Percentage of individuals below
poverty level‖ attributes could be transformed to produce normally distributed datasets
using a log transformation. For these attributes, t-tests were used to examine the datasets
for statistically significant differences.
No transformation was found that produced normally distributed datasets for the other
demographic attributes. The non-parametric Mann-Whitney Rank Sum Test was used to
examine these datasets for statistical differences.

2.2.2 Statistical Comparison of Disease Occurrence
Statistical comparisons of classified ZIP codes:
Statistical analysis of urban cryptosporidiosis, rural cryptosporidiosis, urban
giardiasis, and rural giardiasis was performed using ZIP codes that had been classified as
either urban or rural. Four treatments of the four datasets were analyzed:
1.

Analysis of case data by ZIP code for each disease as extracted from
the ArcGIS database.

2.

Analysis of case data by ZIP code for each disease with number of
cases divided by the area of the ZIP code.

3.

Analysis of case data by ZIP code for each disease with number of
cases divided by the population of the ZIP code.

4.

Analysis of case data by ZIP code for each disease with number of
cases divided by the population density of the ZIP code.

24

Datasets for statistical analysis were not normally distributed and no transformation
was found that produced normally distributed datasets. Therefore, statistical analysis
using non-parametric tests were used. For each analysis, a Kruskal-Wallis One Way
Analysis of Variance (ANOVA) on Ranks was performed (SigmaPlot 11, Systat
Software, Inc), followed by a Dunn's test to evaluate multiple pairwise differences.

Correlation assessment:
ZIP codes (non-classified into urban or rural) were examined for correlation between
the cases of giardias or cases of cryptosporidiosis and nine other factors. These factors
were:
1.

Percentage of urban area

2.

Percentage of rural area

3.

Population

4.

Population density

5.

Percentage of Caucasian/white individuals

6.

Percentage of individuals below poverty level

7.

Median household income

8.

Median age

9.

Mean travel time to work

25

The percentage of urban area and percentage of rural area of ZIP codes in the study area
(non-classified into urban or rural) were examined for correlation between seven other
factors. These factors were:
1.

Population

2.

Population density

3.

Percentage of Caucasian/white individuals

4.

Percentage of individuals below poverty level

5.

Median household income

6.

Median age

7.

Mean travel time to work

The statistical test used for all correlation assessments was Spearman correlation on
ranks.

2.3 Sample Collection – Grand River Watershed
2.3.1 Recreational Areas
The Grand River (Michigan, USA) is 420 km long with a 14,431 km² drainage area
and an average discharge of 108 m³/s (50). Within the lower Grand River watershed,
two sites along the Grand River, Deer Creek Park and Riverside Park, were monitored
for Cryptosporidium and Giardia. North Beach Park, a Lake Michigan beach to the
north of the mouth of the Grand River was also monitored for Cryptosporidium and
Giardia. Deer Creek Park, Riverside Park, and North Beach Park are 29.2 km, 38.6 km,
64.9 km downstream from the City of Grand Rapids Wastewater Treatment Plant,
respectively (see Figure 2.3a). Surface water samples were collected from the banks of
26

the Grand River at Riverside Park and Deer Creek Park at wadeable depths
(approximately 30 cm). At North Beach Park, samples were collected from wadeable
depths (approximately 30 cm). One 20L grab sample was collected per site during each
sampling event. Between April 2005 and August 2006, 21 samples were collected from
Deer Creek Park and 19 samples each were collected from Riverside Park and North
Beach Park. Riverside Park, Deer Creek Park, and North Beach Park are all publicly
maintained recreational areas.

2.3.2 Waste Treatment Systems
Samples were collected from the City of Grand Rapids, Michigan sewage treatment
system. Samples were collected from multiple points in the Grand Rapids sewage
treatment system including the Grand Rapids Market Ave Retention Basin (MARB), and
combined sewer overflow (CSO) weirs during CSO events. Sampling locations are
shown in Figure 2.3a and 2.3b.
Figure 2.3a shows the recreational area sampling locations of North Beach Park,
Riverside Park, and Deer Creek Park within the Grand River watershed, Michigan USA.
Figure 2.3b shows the sampling locations (within City of Grand Rapids city limits) of
the Market Avenue Retention Basin (MARB), combined sewer weir at the Ionia and
Stevens intersection, and #2 Goodrich Grand River sampling sites within the Grand
River watershed, Michigan USA.
Composite samples were collected from the Grand Rapids Market Ave Retention
Basin (MARB, see Figure 2.3b) during eight rainfall events between March 2008 and
August 2008. The MARB retains a combination of stormwater and sewage during storm
events when the treatment capacity of the Grand Rapids sewage treatment plant is

27

exceeded. MARB helps reduce the occurrence of CSO discharges, however, CSO
discharges still occur in this system. MARB samples provide an estimate of
Cryptosporidium spp. and Giardia spp. loads that may be discharged from a combined
sewer. An autosampler collects the MARB influent after passage through a bar screen,
but prior to entry into the MARB catchment area. This sampling location allows flows
from individual rain events to be sampled without compositing with flows from previous
rain events held in the MARB catchment area. The autosampler drew a 1.6L sample
every 5 minutes until a total volume of 10L was collected.
Samples were collected from one CSO event in February 2006 and one event in July
2006 from a weir at the Ionia and Stevens intersection in the City of Grand Rapids (see
Figure 2.3b). During the February 2006 event, 1.443 million gallons were released from
the Ionia and Stevens outfall. The July 2006 event occurred during a storm that caused a
total overflow of 25.331 million gallons from the Grand Rapids combined sewer system,
with 1.443 million gallons released from the Ionia and Stevens CSO outfall. The Grand
Rapids CSO weir locations are accessed through manholes in city streets. Installation
and powering an autosampler is difficult at these locations, so samples were collected by
Grand Rapids Sewer treatment personnel.
Additionally, one sample was collected during the July 2006 CSO event at the #2
Goodrich site ( see Figure 2.3b) which is a Grand River sampling site 1.6 km
downstream of the discharge point for the Ionia and Stevens CSO. The #2 Goodrich site
was sampled to determine if Cryptosporidium and Giardia concentrations in the river
could remain, elevated downstream of a CSO discharge. During each event 20L grab
samples were collected.

28

CSO events are often of short duration, which makes it difficult for personnel to
reach the site in time to collect overflow. Entering and exiting manholes located in
street intersections during storm events present fall hazards and traffic hazards. To
improve safety of sampling and to collect samples that allow estimation of parasite loads
that enter surface waters, the majority of combined stormwater/sewage samples were
collected at the Grand Rapids Market Ave Retention Basin (MARB).

29

Figure 2.3 Grand River Watershed Sampling Locations

30

2.4 Sample Collection – River Raisin Watershed
The River Raisin (Michigan, USA) is 216 km long with a 2,776 km² drainage area
(20). Within the Lenawee County, Michigan portion of the River Raisin watershed the
main branch of the River Raisin, four tributary creeks, and one field drainage structure
were sampled between June 2004 and February 2005. The locations of the sampling
sites are shown in Figure 2.4.
Five sites in the main branch River Raisin were sampled; the main branch River
Raisin at Crockett Rd, the main branch River Raisin at Deerfield Rd, and the main
branch River Raisin at the drinking water plants of the City of Adrian, City of Deerfield,
and the City of Blissfield. Two sites each were sampled from the Stoney Creek and Bear
Creek tributaries and one site was sampled from the Black Creek and Wolf Creek
tributaries.
The River Raisin surface waters provide drinking water to the cities of Adrian,
Blissfield, and Deerfield. Rice Lake Drain is a drainage ditch adjacent to a Confined
Animal Feeding Operation (CAFO). Four of the monitored sites in the River Raisin
watershed are downstream of CSO outfalls. These sites include the main branch River
Raisin at Deerfield Rd, the main branch River Raisin at Crockett Rd, the intake to the
Deerfield water treatment plant, and the intake to the Blissfield water treatment plant. A
total of 39 samples from 12 sites in the River Raisin watershed were collected. Samples
were collected as 20L grab samples from each site.

31

Figure 2.4 River Raisin Watershed Sampling Locations

1. Wolf Creek at Forrister Rd – 3 samples
2. Rice Lake Drain – 5 samples
3. Stoney Creek at Senecca Rd – 2 samples
4. Bear Creek at Medina Rd – 1 sample
5. Stoney Creek at Gorman Rd – 2 samples
6. Bear Creek at Morse Rd – 1 sample
7. Main Branch River Raisin at Adrian Water Works (raw influent) – 5 samples
8. Main Branch River Raisin at Deerfield Rd – 2 samples
9. Main Branch River Raisin at Deerfield Water Works (raw influent) – 5 samples
10. Main Branch River Raisin at Crockett Rd – 2 samples
11. Main Branch River Raisin at Blissfield Water Works (raw influent) – 6 samples
12. Black Creek at Crockett Rd – 5 samples

32

2.5 Sample Processing
After collection, samples were placed on ice and transported to the laboratory for
analysis. Samples that met the holding period criteria of ≤4 days were analyzed using
U.S. Environmental Protection Agency (U.S.EPA) Method 1623, which is a standard
method designed to meet survey and monitoring requirements of the EPA (54). The
method has been validated for use in surface waters since February 1999, and has the
capability of identifying the genera Cryptosporidium and Giardia. The method cannot
determine species, host species of origin, or viability/infectivity of (oo)cysts.
In brief, parasite detection was performed by filtering water through Envirochek
HV filters (Pall Gelman Laboratories, Ann Arbor, MI, USA). After filtration, the filter
was eluted and the eluate concentrated by centrifugation according to U.S.EPA Method
1623. Since high turbidity samples may clog the filter after a low volume of sample has
passed through the filter, for samples with turbidity >35 NTU the filtration step was
omitted and the sample directly concentrated by centrifugation. After a concentrated
pellet was obtained, the pellet was resuspended and divided into subsamples based on
the original pellet size size so that each subsample had 0.5ml of pellet per 10mL.
Parasites were separated from the resuspended materials using the Dynal
Immunomagnetic Separation Technique (IMS) (Dynabeads® CG-combo Kit, Dynal
Biotech, Inc., Lake Success, NY, USA) in Leighton tubes. Modifications of the 1623
protocol included a second hydrochloric acid wash step and neutralization of the IMS
concentrate within a microcentrifuge tube rather than on a glass slide. Modifications
were primarily made to reduce the concentration of sample debris that was not removed

33

by IMS, as sample debris interferes with microscopic analysis. Quality control tests
were performed and acceptance criteria were met for all modifications as required in
U.S.EPA Method 1623 (54). The (oo)cyst suspension was placed on slides and allowed
to dry before samples were fixed with methanol and stained. Samples were stained with
DAPI to help visualize nucleic acid content then stained using an immunofluorescent
assay (IFA) method. This method uses monoclonal antibodies (EasyStain,
Biotechnology Frontiers, Australia) tagged with fluorescein isothiocyanate to
specifically stain the (oo)cyst walls. Microscopic examination of the slides after IFA
resulted in total counts of oocysts and cysts in the sample. Positive staining controls
consisted of slides with purified Cryptosporidium and Giardia (EasyStain,
Biotechnology Frontiers, Australia). Negative staining controls consisted of slides
prepared with phosphate buffered saline in place of the sample. These control slides
were fixed, stained, and read with each set of samples processed.
Matrix specific alterations, primarily to address differences in particulate matter, are
described below. When particulate matter resulted in >0.5mL of concentrated pellet
following centrifugation, the pellet was divided into subsamples as required by U.S.EPA
Method 1623. In this study, equivalent volumes less than 10L are due to amounts of
particulate matter that resulted in subsampling.
MARB samples
Samples were mixed on a stirring plate using a magnetic stir bar. Subsample
volumes of 75mL to 250mL were withdrawn and centrifuged to produce a concentrated
0.5ml pellet. The sample was remixed between withdrawal of additional subsample
volumes. The pellet was processed as described above. Equivalent volumes of 75ml to
34

250ml were examined on slides. A total of 27 subsamples from 8 storm events were
analyzed.

CSO samples
Samples were analyzed as described above. Equivalent volumes of 1L to 11.19L
were examined on slides. A total of 2 samples were analyzed from the Ionia and Stevens
CSO discharge.
Grand River watershed samples
Samples were analyzed as described above. A total of 22, 19, and 19 samples were
analyzed for Deer Creek Park, Riverside Park, and North Beach Park, respectively.
Equivalent volumes of 5.54L to 13.3L were examined on slides from Deer Creek Park,
Riverside Park, and North Beach Park. An equivalent volume of 1 liter from the #2
Goodrich sample was analyzed.
River Raisin watershed samples
Samples were analyzed as described above. Equivalent volumes of 0.66L to 11.05L
were examined on slides.

2.6 Recovery Efficiency
Ongoing Precision and Recovery analyses
Recovery efficiencies in laboratory reagent water were assessed by seeding with a
known concentration of Cryptosporidium and Giardia. River Raisin watershed samples
were seeded with Colorseed (Biotechnology Frontiers, Australia). MARB and Grand
River watershed samples were seeded with EasySeed (Biotechnology Frontiers,

35

Australia). These ongoing precision and recovery (OPR) samples were processed as
described above. When sample matrices required centrifugation the OPR samples were
also processed by centrifugation. After processing, counts of Cryptosporidium and
Giardia were compared to the number of seeded organisms and a method blank of
laboratory reagent water containing no seeded Cryptosporidium and Giardia to calculate
the method’s efficiency. Acceptance criteria for Cryptosporidium and Giardia recovery
were 13-111% and 15-118%, respectively. In model simulations, recovery efficiency
was not used to extrapolate values for model simulation. A total of 23 OPR samples
were performed during sampling of the Grand River watershed. A total of four OPR
samples were performed during sampling of the River Raisin Watershed. A total of five
OPR samples were performed during sampling of MARB.
Matrix Spikes
To determine recovery efficiencies in sample matrices, duplicate samples were
seeded with a known concentration of Cryptosporidium and Giardia. River Raisin
watershed samples were seeded with Colorseed (Biotechnology Frontiers, Australia).
MARB and Grand River watershed samples were seeded with EasySeed
(Biotechnology Frontiers, Australia). These matrix spike samples were concentrated
and processed as described above. After processing, counts of Cryptosporidium and
Giardia were compared to the number of seeded organisms and the number of naturally
occurring Cryptosporidium and Giardia in the associated field sample to calculate the
method’s efficiency in the environmental matrices. Recovery efficiencies were examined
in six MARB samples, two North Beach Park samples, two Riverside Park samples, two
Deer Creek Park samples, and 14 River Raisin watershed samples. At least one matrix
36

spike was performed with every lot of samples analyzed from the River Raisin
Watershed. Recovery efficiencies were not examined for CSO samples.

2.7 Statistical Analysis: Grand River and River Raisin Watershed
Comparison
2.7.1 Meaningful Comparisons
Statistical analyses were conducted to identify meaningful comparisons within and
between sites. In this study, meaningful comparisons are those which compare:
1. Giardia versus Cryptosporidium within the same site.
2. Cryptosporidium versus Cryptosporidium between multiple sites
3. Giardia versus Giardia between multiple sites.
In this study, statistical differences in other comparisons will not be discussed.
Detection limits were analyzed to check for bias that could be introduced due to
differences in analysis sensitivity. Recovery efficiencies were analyzed to check for
differences in method performance.

2.7.2 Test of Normality and Equal Variance Assumptions
Probability distributions were assessed for normality using the Shapiro-Wilk test
(α = 0.05) and normal distributions were assessed for equal variance using the Levine
Median test (α = 0.05) in SigmaPlot (SigmaPlot v11.0, Systat Software, Inc).
The probability distributions of Cryptosporidium and Giardia occurrence data and
detection limits all failed the normality test. Data was adjusted by adding one to each
data point and then transformation of the data to produce normal distributions was
attempted. Natural logarithm, log10, square root, square, and exponential

37

transformations failed the normality test. Using the reciprocal transformation, only the
Cryptosporidium dataset for the River Raisin category downstream of the CSO outfall
passed the normality test. As most occurrence and detection limit datasets violate the
assumption of normality, all further statistical analyses on these datasets were conducted
using non-parametric tests.
Cryptosporidium and Giardia recovery efficiency in sample matrices (matrix spike
analyses) follow normal distribution patterns. However, all matrix spike recovery
efficiency comparisons involving the River Raisin watershed samples failed the equal
variance test. Therefore, all statistical tests involving matrix spike recovery efficiency in
the River Raisin watershed were conducted using non-parametric statistical tests. All
other matrix spike efficiency comparisons were conducted with parametric statistical
tests. Cryptosporidium and Giardia recovery efficiency in laboratory reagent water
(OPR analyses) follow normal probability distributions with equal variance. All
statistical analyses on these datasets were conducted using parametric tests.

2.7.3 Statistical Analysis of Grand River Watershed Recreational Sites
Kruskal-Wallis one way ANOVA on ranks was used to compare parasite occurrence
datasets from North Beach Park, Riverside Park, and Deer Creek Park. A pairwise
multiple comparison using Dunn’s method was performed to examine the groups that
differed from the others. Kruskal-Wallis one way ANOVA on ranks was used to
compare detection limit datasets from North Beach Park, Riverside Park, and Deer
Creek Park. A t-test was used to compare recovery efficiency of Giardia versus
recovery efficiency of Cryptosporidium in matrix spike samples. A t-test was used to

38

compare recovery efficiency of Giardia versus recovery efficiency of Cryptosporidium
in OPR samples.

2.7.4 Statistical Analysis of Grand River Watershed Recreational Sites and
MARB
Kruskal-Wallis one way ANOVA on ranks was used to compare parasite occurrence
from MARB, North Beach Park, Riverside Park, and Deer Creek Park datasets. A
pairwise multiple comparison using Dunn’s method was performed to examine the
groups that differed from the others. Kruskal-Wallis one way ANOVA on ranks was
used to compare detection limit datasets from North Beach Park, Riverside Park, Deer
Creek Park, and MARB. A pairwise multiple comparison using Dunn’s method was
performed to examine the groups that differed from the others. Statistical differences
between detection limits at the #2 Goodrich site (single sample) and North Beach Park,
Deer Creek Park, and Riverside Park were not examined. One way ANOVA was used
to compare parasite recovery efficiency in matrix spike samples from MARB and the
combined North Beach Park, Riverside Park, and Deer Creek Park recovery efficiency
dataset.
One way ANOVA was used to compare parasite recovery efficiency in OPR samples
from MARB and the combined North Beach Park, Riverside Park, and Deer Creek Park
recovery efficiency dataset (p=0.005).

2.7.5 Statistical Analysis of River Raisin Watershed
Kruskal-Wallis one way ANOVA on ranks was used to compare parasite
occurrence datasets from sites upstream of CSO outfalls against sites downstream of
CSO outfalls. Parasite occurrence datasets from upstream and downstream sites were

39

pooled and Cryptosporidium occurrence was compared to Giardia occurrence using the
Mann-Whitney rank sum test. Detection limit datasets from upstream and downstream
sites were compared using the Mann Whitney rank sum test. Kruskal-Wallis one way
ANOVA on ranks was used to compare parasite recovery efficiency in matrix spike
samples between sites upstream of CSO outfalls and sites downstream of CSO outfalls.
A pairwise multiple comparison using Dunn’s method was performed to examine the
groups that differed from the others. Parasite recovery efficiency in matrix spike sample
datasets from upstream and downstream sites were pooled and compared using the
Mann-Whitney rank sum test. Recovery efficiencies of Giardia and Cryptosporidium in
OPR samples were compared using a t-test.

2.7.4 Statistical Analysis of River Raisin vs Grand River Watershed
Deer Creek Park and Riverside Park parasite occurrence datasets were pooled for
comparison with the River Raisin watershed sites. Upstream and downstream sites in
the River Raisin watershed were pooled for comparison with the Grand River watershed
sites.
Kruskal-Wallis one way ANOVA on ranks was used to compare parasite occurrence
in datasets from North Beach Park, from the combined Deer Creek Park- Riverside Park
dataset, and from the total River Raisin dataset. A pairwise multiple comparison using
Dunn’s method was performed to examine the groups that differed from the others.
Kruskal-Wallis one way ANOVA on ranks was used to compare detection limits in
datasets from North Beach Park, from the combined Deer Creek Park- Riverside Park
dataset, and from the total River Raisin dataset. A pairwise multiple comparison using
Dunn’s method was performed to examine the groups that differed from the others.

40

Kruskal-Wallis one way ANOVA on ranks was used to compare parasite recovery
efficiency in matrix spike samples from the total Grand River dataset (North Beach Park,
Deer Creek Park, and Riverside Park) and from the total River Raisin dataset A pairwise
multiple comparison using Dunn’s method was performed to examine the groups that
differed from the others.
One way ANOVA was used to compare parasite recovery efficiency in OPR samples
from the total Grand River dataset (North Beach Park, Deer Creek Park, and Riverside
Park) and from the total River Raisin dataset. A pairwise multiple comparison using the
Holm Sidak method was performed to examine the groups that differed from the others.

2.8 Risk Simulation and Statistical Analysis
A scenario for risk assessment involving children under the age of 16 swimming in
recreational areas was developed and analyzed using a Monte Carlo method. This risk
assessment focused on ingestion of water contaminated with Cryptosporidium and
Giardia and the probability of infection and illness. Probability distributions for
exposure factors, including the amount of time spent swimming and the volume ingested
during swimming, were developed using Crystal Ball™ (version 7.3.1, Oracle).
Probability distributions were also calculated for the fraction of ingested organisms that
survive ingestion to initiate infection (based on exponential dose response parameter).
Concentrations of parasites from samples collected from the three recreational areas
(Riverside Park, Deer Creek Park, North Beach Park) and the Market Avenue Retention
Basin were fit to distributions, which were ranked according to the KolmogorovSmirnov goodness-of-fit test. The highest ranking distribution was chosen as the best fit.
In distribution fitting, non-detects were replaced by the detection limit, producing the
41

most conservative (i.e. most protective) concentration estimates. All detected parasites
were assumed to be viable and capable of initiating infection in humans. Data on
duration and volume of CSO events reported by the Grand Rapids Wastewater
Treatment Plant (35) and the mean volumetric flow of the Grand River from January 1,
2002 to December 31, 2006 at the USGS gauging station located Grand Rapids (52)
were also fit to distributions using the Kolmogorov-Smirnov goodness-of-fit test. All
distributions were truncated at a minimum of zero. Table 5.1 contains the distributions
and distribution parameters.
Equation 1 is the daily ingested number of organisms ( dSITE ) at each recreational
site calculated as a function of the concentration of organisms at a site ( CSITE ),
ingestion volume during swimming ( I ), the amount of time spent swimming ( TSW IM )
and a constant to convert units of months to days.

dSITE was calculated for each of the

three recreational sites.

(1)

dSITE  0.033  CSITE  I  TSW IM
Equation 2 is the daily ingested number of organisms ( dRIVER) in the river at the

CSO discharge point calculated as a function of the concentration of organisms in the
river at the CSO discharge point ( CRIVER, see Equation 5), ingestion volume during
swimming ( I ), the amount of time spent swimming ( TSW IM ) and a constant to convert
units of months to days.

42

dRIVER 0.033  CRiver I  TSW IM

(2)

Equation 3 is the daily probability, or risk, of infection (

PI ) due to ingestion of a

r = fraction of ingested organisms that survive
to initiate infection (45). The value of r for each organism is shown in Table 5.1. In
volume of contaminated water where

equation 3, d =
site and d =

dSITE when calculating the probability of infection at each recreational

dRIVER when calculating the probability of infection at the CSO discharge

point.

(3)

PI  1  exp( r  d )
Equation 4 is the annual risk of infection due to ingestion ( PANNUAL).

a function of the daily risk where

PANNUAL is

N is the number of days of exposure during a

recreational season to the health hazard (25), in this case, parasites.

(4)

PANNUAL 1  (1  PI ) N
Equation 4 was used to calculate daily risk by setting

N = 1.

Seasonal risk of

infection was calculated by making the assumption that recreation occurs over a period
of 90 days in the summer ( N = 90).

43

Equation 5 is the concentration of parasites in the river at the CSO discharge point
( CRIVER).

(5)

C RIVER 

CCSO  VCSO
TCSO (VCSO  d 1 Q R )

Where CRIVER = Parasite concentration in the river

CCSO

VCSO
TCSO
QR

= Parasite concentration in the MARB influent (estimate of
CSO discharge concentration)
= Volume of CSO discharge
= Duration of the CSO discharge
= Volumetric flow rate of the river

The risk of daily illness and seasonal illness was calculated as a function of the daily
probability of infection multiplied by a morbidity ratio (see Table 5.1). The morbidity
ratio for Giardia was based on an indirect assessment of outbreak attack rates compared
with infectivity dose-response relationships (45). The morbidity ratio for
Cryptosporidium was based on dose-response studies in healthy adults (25).
Using Crystal Ball™ (version 7.3.1, Oracle), Monte Carlo analysis was conducted
by running 10,000 simulations based on values randomly generated from the parameter
distributions. Each of these randomly generated values was then used in the
calculations above to produce an estimated distribution of risk. Probability distributions
for daily risk of illness and seasonal risk of illness were compared with the EPA

44

recommended illness rate water quality criteria for recreational waters of 0.8% (8
illnesses in 1,000 swimmers) (21).
One way analysis of variance was performed on the simulation outputs using the
mean, standard deviation, and simulation run size of 10,000 (SigmaPlot v11.0, Systat
Software, Inc). The Holm-Sidak method for multiple pairwise comparison was used to
identify significant differences in risk of infection at Deer Creek Park, Riverside Park,
and at the point of CSO discharge to the river. The Holm-Sidak method for multiple
pairwise comparison was also used to identify significant differences in risk of illness at
Deer Creek Park, Riverside Park, and at the point of CSO discharge to the river.

45

Chapter Three: Land Cover and the Incidence of
Cryptosporidiosis and Giardiasis
In chapter three, the patterns between disease and land cover type were explored
using the prototype GIS database. The null hypothesis that disease levels are the same in
urban and rural areas was tested. Cryptosporidiosis and giardiasis were the diseases
chosen for this study.

3.1 Results
Figure 3.1 shows the land cover classifications of the study area. Table 3.1
summarizes attributes of the study area by rural and urban designation, including the
number of ZIP codes, area, population, population density. The majority of ZIP codes
in the study area were designated as rural ZIP codes. In the total study area, the rural
area and rural population were greater than the urban area and urban population.
However, the median population value was greater in the urban ZIP codes than in the
rural ZIP codes. The urban population density of the study area was greater than the
rural population density.
Table 3.1 also summarizes demographic information about the study area, including
the percentage of the population that is Caucasian/white, median age, mean travel time
to work, median age, median household income, and the percentage of individuals living
below the poverty level. When compared to urban ZIP codes, the rural ZIP codes had a
higher median percentage of Caucasian/white individuals in the population, the median
population age was older, the mean travel time to work was longer, and the median

46

household income was higher. Additionally, in the rural ZIP codes, a lower percentage
of individuals were living below the poverty level than in the urban ZIP codes.
Results of the statistical comparison of the study area attributes are shown in Table
3.1. Each attribute investigated showed a significant difference between the urban and
rural ZIP codes, with p values ranging from <0.001 to 0.031

47

Figure 3.1 Land Cover Classifications of Study Area

10 20

Kilometers

48

Table 3.1 Attributes of Study Area
t – Test

Attributes

Number of ZIP
codes analyzed
Number of ZIP
codes
Population*
2

Area (km )
Population
density
2
(People/km )
Percentage
individuals
below the
poverty level*
Median
household
income
(US dollars)*
Median age
(Years) *
Mean travel
time to work
(minutes) *
% Caucasian/
white*

ZIP code
designation

ZIP code
median
value

Total
study
area

Urban vs
Rural

MannWhitney
Rank Sum
Test
Urban vs
Rural

48
Urban
Rural
Urban
Rural

36,821.5

8
40
310,302

6,875

475,667

Urban
Rural

25.380

211.00

89.818

4529.0

Urban
Rural

1383

1469

61

105

Urban

Significant
difference
(P<0.001)

10.05

Rural

5.75

Urban
Rural

$40,127
$46,270

Urban
Rural

32.2

19.5

Urban
Rural

Significant
difference
(P=0.031)
Significant
difference
(P<0.001)
Significant
difference
(P=0.001)
Significant
difference
(P=0.001)

35.95

Urban
Rural

Significant
difference
(P<0.001)
Significant
difference
(P<0.001)
Significant
difference
(P<0.001)

79.4

23.9

97.0

*Source: 2000 US Census (51)

49

A comparison of the median number of disease cases per ZIP code and total disease
cases by urban or rural designation over the eight year study period is shown in Table
3.2. When comparing locations with the greatest amount of disease incidence, there was
a difference in disease pattern between the median case per ZIP code and the total
number of cases. The dataset of disease cases represents disease incidence, which is the
number of cases in a given amount of time. In this dataset, the total number of both
giardiasis and cryptosporidiosis cases were greatest in the rural areas. However, the ZIP
code median values of giardiasis and cryptosporidiosis case incidence were greatest in
the urban ZIP codes. Once the incidence datasets were normalized (by population, area,
or population density), the patterns of disease occurrence between the ―total number of
cases‖ and the ―median number of cases per ZIP code‖ categories were in agreement.
When the disease cases were divided by population values, the overall number of
cryptosporidiosis disease cases was greatest in the rural ZIP codes. Conversely, for
giardiasis, the greatest numbers of disease cases were in the urban ZIP codes. When the
disease cases were divided by area, the numbers of disease cases were greatest in the
urban ZIP codes for both cryptosporidiosis and giardiasis. When the disease cases were
divided by population density, the numbers of disease cases was greatest in the rural ZIP
codes for both cryptosporidiosis and giardiasis.
Results of the statistical comparison of disease between urban and rural designated
ZIP codes are shown in Table 3.3. All examined treatments resulted in significant
differences between the datasets examined. Results from the Dunn's test to evaluate
multiple pairwise differences are also shown in Table 3.3. The source of the statistical
difference differed between the four treatments of the dataset. The rural giardiasis vs

50

urban cryptosporidiosis and urban giardiasis vs rural cryptosporidiosis comparisons are
not shown in Table 3.3 as these comparisons are not particularly meaningful in
elucidating disease transmission.
Results of the correlation assessment are shown in Table 3.4. No significant
correlations with either the median household income or the percentage of individuals
below poverty level were found (p>0.15). Three strong correlations were found. Strong
positive correlations exist between giardiasis and population (r = 0.918, p<0.001) and
between the percentage of urban area and population density (r = 0.978, p<0.001). A
strong inverse relationship exists between the percentage of rural area and population
density (r = -0.977, p<0.001). Five moderately strong positive correlations and four
moderately strong inverse correlations were found in this assessment. The moderately
strong positive correlations consisted of giardiasis and population density (r =0.799,
p<0.001), giardiasis and percentage urban area (r=0.79, p<0.001), cryptosporidiosis and
population (r = 0.793, p<0.001), percentage urban area and population (r = 0.754,
p<0.001), and percentage rural area and percentage of Caucasian / white individuals (r =
0.763, p<0.001). The moderately strong inverse correlations consisted of giardiasis and
percentage rural area (r = -0.788, p<0.001), the percentage of urban area and percentage
of Caucasian / white individuals (r = -0.777, p<0.001), the percentage of rural area and
population (r = -0.752, p<0.001), and giardiasis and mean travel time to work
(r = -0.75, p<0.001).
The relationship with population produced the highest correlation coefficients for
cases of either giardiasis or cryptosporidiosis, although the correlation with giardiasis is

51

strong and the correlation with cryptosporidiosis is only moderate. For both giardiasis
and cryptosporidiosis, the next strongest relationship was with population density.

Table 3.2 Comparison of Disease Cases by Urban or Rural Designation.
Disease

ZIP code
designation

Cryptosporidiosis Urban

Cryptosporidiosis Rural

Giardiasis

Giardiasis

a

Urban

Rural

Treatment of
dataset

Disease Cases
Disease Cases
a
Population
Disease Cases
Area
Disease Cases
Population
density
Disease Cases
Disease Cases
a
Population
Disease Cases
Area
Disease Cases
Population
density
Disease Cases
Disease Cases
a
Population
Disease Cases
Area
Disease Cases
Population
density
Disease Cases
Disease Cases
a
Population
Disease Cases
Area
Disease Cases
Population
density

Calculated as the number of cases per capita

52

Median
number of
cases per ZIP
code
5.0

Total
number of
cases
39

0.000136

0.00013

0.184

0.185

0.00330

0.027

1.5

113

0.000184

0.00024

0.0145

0.025

0.0156

1.076

53.0

405

0.00117

0.00131

1.96

1.92

0.0340

0.276

5.0

467

0.000788

0.00098

0.0392

0.100

0.0729

4.448

Table 3.3 Statistical Comparison of Disease Occurrence.
Treatment
ANOVA
Multiple Pairwise Occurrence
a
of dataset
on Ranks
(Dunn’s Test)
Cryptosporidiosis
Giardiasis
Urban

Rural

Urban vs Rural

Urban vs Rural

Cryptosporidiosis vs
Giardiasis

Cryptosporidiosis
vs Giardiasis

None
(incidence)

Significant
difference
(P<0.001)

No significant difference

Significant
difference
(P<0.05)

No significant
difference

No significant
difference

Population
normalized

Significant
difference
(P<0.001)

No significant difference

No significant
difference

Significant difference
(P<0.05)

No significant
difference

Area
normalized
(disease
intensity)
Population
density
normalized

Significant
difference
(P<0.001)

Significant difference
(P<0.05)

Significant
difference
(P<0.05)

No significant
difference

No significant
difference

Significant
difference
(P=0.002)

No significant difference

No significant
difference

No significant
difference

Significant
difference (P<0.05)

a

Results of rural giardiasis vs urban cryptosporidiosis and urban giardiasis vs rural cryptosporidiosis comparisons not shown.

53

Table 3.4 Correlation Assessment.
Factor
Giardiasis
Cryptosporidiosis
%Urban
r
p value
% Rural
r
p value
Population
r
p value
Population
density
r
p value
% Caucasian
/ white
r
p value
Median age
r
p value
Mean Travel
Time to
Work
r
p value
Median
Household
Income
r
p value
%
Individuals
Below the
Poverty
Level
r
p value

%Urban

% Rural

0.79
<0.001

0.595
<0.001

---

---

-0.788
<0.001

-0.591
<0.001

---

---

0.918
<0.001

0.793
<0.001

0.754
<0.001

-0.752
<0.001

0.799
<0.001

0.609
<0.001

0.978
<0.001

-0.977
<0.001

-0.694
<0.001

-0.554
<0.001

-0.777
<0.001

0.763
<0.001

-0.638
<0.001

-0.49
<0.001

-0.554
<0.001

0.557
<0.001

-0.75
<0.001

-0.663
<0.001

-0.65
<0.001

0.648
<0.001

0.207
0.157

0.121
0.413

0.0939
0.524

-0.0843
0.567

0.115
0.434

-0.0525
0.722

0.177
0.227

-0.19
0.194

54

3.2 Discussion
In this study, after assigning ZIP codes a rural or urban designation, the study area
attributes were examined for differences that might produce significant differences
between disease in urban and rural areas. Significant differences were found in all
attributes examined. Three of these attributes (population, area, and population density)
were used to normalize the cases of disease. The attributes of population and area both
had higher values in rural areas than in urban areas. In contrast, the population density
was higher in urban areas than in rural areas. The three normalized datasets and the nonnormalized dataset were used to test the null hypothesis that disease levels in urban and
rural areas are the same.
Based on the statistically significant differences between disease occurrence in ZIP
codes designated as urban and those designated as rural (p<0.001), the null hypothesis
was rejected in favor of the research hypothesis that there were significant differences
between patterns of disease in urban and rural areas. Statistically significant differences
were present between disease occurrence in ZIP codes designated as urban and those
designated as rural (p<0.001). This statistical difference was maintained even when
disease occurrence was normalized by different attributes present in the study area (ie
population, area, or population density of the ZIP codes) although the cause(s) of this
difference changed as the datasets were normalized. For example, in the non-normalized
dataset, the cause of the significant difference between urban disease and rural disease
was due to the difference between urban giardiasis and rural giardiasis. However, when
the dataset was population normalized the cause of the significant difference between

55

urban disease and rural disease was the difference between urban cryptosporidiosis and
urban giardiasis.
As there was no obvious pattern in the pairwise comparisons in the normalized or
non-normalized datasets that suggested particular transmission routes, the disease cases
in all ZIP codes were examined for correlations with the study area attributes. For the
correlation assessment, instead of using the urban and rural designations for the ZIP
codes, the percentages of urban area and rural area for each ZIP code were used. Of the
resulting correlations, the direct correlation with population was strongest for both
giardiasis and cryptosporidiosis. The strong direct correlation of population with
giardiasis (r=0.918, p<0.001) and the moderate direct correlation of population with
cryptosporidiosis (r=0.793, p<0.001) suggest that population levels were important
factors of transmission of these parasitic diseases in the study area. Since the percentage
of urban area had a direct positive relationship with the population levels (r = 0.754,
moderate) and population density (r = 0.978, strong), positive relationships between the
percentage of urban area and disease cases were logical. The correlation between the
percentage of urban area and giardiasis is stronger than the correlation between the
percentage of urban area and cryptosporidiosis, however. The correlation of disease with
population, particularly for giardiasis, suggests that higher disease transmission is
mediated in some way by larger population levels. This may involve direct person to
person transmission, transmission through food or water contaminated with feces from
an infected human, or transmission through the environment (ie contaminated
recreational water).

56

Livestock density and manure land application were not included in the prototype
database, although these may be important factors in disease transmission. Information
on livestock density is only available at the county scale. Including this information into
the prototype database would require all comparisons to be done at the county scale,
potentially obscuring patterns that occur at more local levels. However, there are some
studies in the scientific literature that have examined associations between
cryptosporidosis / giardiasis and livestock density, manure application, and other
livestock related factors. These studies provide insight into which attributes would be
desirable to include in an expanded GIS database.
Previous studies of cryptosporidiosis and giardiasis in England, Scotland, and Wales
have demonstrated association with rural areas in some studies and associations with
urban areas in others. During the 2001 foot and mouth disease outbreak in livestock in
England and Wales, in which control measures were implemented including restriction
of access to farms, limiting movement of livestock for trade and between pasturage, and
culling of affected herds and flocks, there was a corresponding decrease in reported
cryptosporidiosis in humans ranging from 35-63% (47) throughout England and Wales
with declines of 81.8% in northwest England (23). During the interval of foot and
mouth disease control measures, the proportion of C. parvum cases decreased compared
to case incidence in 2000, suggesting a decrease in human infection due to a decrease in
exposure to pathogen reservoirs in livestock (47). No significant reduction in giardiasis
cases was observed during this interval, which may indicate a difference between
Giardia and Cryptosporidium transmission routes and reservoirs of infection (47).
Studies of area based cryptospordiosis rates in England and Wales using housing density

57

to define rural areas demonstrated higher illness rates in rural areas than urban areas,
higher rates in areas with more agricultural manure application, and higher rates in areas
with inadequate drinking water treatment (31, 32). A study of spatial epidemiology of
sporadic cases of human cryptosporidiosis in Scotland found increased rates of C.
parvum infection in areas with lower human density, a higher ratio of farms to humans,
and a higher ratio of private water supplies to the human population, indicating an
association of C. parvum infection with rural areas (42). Unlike C. parvum, C. hominis
was reported more often in the more heavily populated areas of south Scotland,
associating this genotype more strongly with urban areas (42). In a case control study of
cryptosporidiosis in the United Kingdom, the urban-rural gradient was not found to be a
significant variable in the full model of disease aetiology when both C. parvum and C.
hominis cases were included. However, when C. hominis cases were excluded,
cryptosporidiosis was negatively associated with urban areas and when C. parvum cases
were excluded, cryptosporidiosis was positively associated with urban areas, indicating
genotype specific transmission associated with the geographical classifications (32).
The differences in the association of cryptosporidiosis with urban or rural areas in the
scientific literature may therefore be partially due to the Cryptosporidium genotype
causing the infections.
In a spatial investigation of giardiasis in Canada that explored associations with
livestock density and land application of manure with disease patterns, low correlation
coeffients between giardiasis rates and cattle density (r=0.11) and between giardiasis
rates and land application (r=0.09) were observed when all geographic regions in the
study area were included (37). However, these correlation coefficients were higher in

58

certain regions of the study area when these areas were independently examined (37).
These results suggest that livestock density and land application of manure can
contribute to transmission of Giardia, but that other factors may be more important.
In the current study, the major caveat to the conclusion that significant differences in
the patterns of giardiasis and cryptosporidiosis exist between urban and rural ZIP codes
is the small number of ZIP codes that are designated as urban in the study area. Since
this method of examining health data appears promising, expanding the study area to
larger geographic regions of the United States (ie Midwest) in future work is
recommended. Expanding the geographic region would also allow for agricultural
census data to be incorporated into the study design by allowing analysis at the county
level, which is the minimum scale of the agricultural census data. Previous studies from
the scientific literature suggest that livestock density, animal transport frequency, the
number of farm workers/visitors, manure application rate, ratio of farms to humans, and
the Cryptosporidium genotype would be informative attributes in an expanded GIS
database. Utilizing GIS as a tool to integrate factors from the rural environment with
factors from the urbanized environment has the potential to be extremely useful to public
health agencies in targeting funds to reduce disease transmission in communities.

3.3 Conclusion
The prototype GIS database was a useful tool for testing hypotheses regarding
disease incidence and land cover type. The relatively small number of urban zip codes
versus the number of rural zip codes was the major caveat in hypothesis testing. During
the development of the prototype database, data gaps in agricultural information and data

59

scale issues were identified. Expansion of the study area, identified data gaps and scale
issues will be addressed in the next iteration of the GIS database.
Significant differences in the patterns of giardiasis and cryptosporidiosis exist
between urban and rural ZIP codes in the study area. The strong direct correlation of
population with giardiasis and the moderate direct correlation of population with
cryptosporidiosis suggest that population levels are important factors in determining
disease incidence in the study area. Giardiasis was moderately correlated with the
percentage of urban area and cryptosporidiosis was weakly correlated with the
percentage of urban area.

60

Chapter 4. Grand River and River Raisin Watershed
Comparison: Cryptosporidium and Giardia Occurrence
In this study, the association of land cover with the parasites Cryptosporidium and
Giardia was further examined by comparing parasite concentrations in surface water at
the watershed scale. Two watersheds with differences in the percentage of urban and
rural land cover were compared to see if occurrence patterns in these watersheds were
associated with land cover types. The metrics of this comparison differs from the
comparison in Chapter 3 in that the concentration of parasites in surface water rather
than disease is the factor of interest and the scale of the comparison changed from the
ZIP code scale to the watershed scale. Surface waters of the Grand River watershed and
River Raisin watershed were first individually examined for parasite occurrence and then
the parasite concentrations of each watershed compared against one another. As the
experimental design for these watersheds was not originally planned for this interwatershed comparison (post-hoc testing), detection limits and recovery efficiency were
considered alongside parasite occurrence. Differences in these factors could introduce
bias into comparisons of parasite occurrence.
The following datasets from the Grand River watershed in Ottawa County, MI were
assessed: North Beach Park, Riverside Park, Deer Creek Park, and combined datasets of
Riverside Park and Deer Creek Park. Occurrence data from the River Raisin watershed
in Lenawee County, MI (12 sampling sites) were divided into 3 categories: upstream of
Combined Sewer Overflow (CSO) outfalls, downstream of CSO outfalls, and total data.
Locations of sampling sites, sample collection and processing methods are described in
Chapter 2.

61

Figure 4.2 shows the location and land cover classifications of the watersheds in the
counties of interest. Land cover classification methods are described in Chapter 2. The
Grand River watershed in Ottawa County was 86.0% rural and 11.9% urban. The River
Raisin watershed in Lenawee County was 94.1% rural and 5.5% urban.

4.1 Parasite Occurrence Results
Cryptosporidium and Giardia were detected at all of the monitored sites during at
least one sampling event. Among the Grand River recreational sites, Cryptosporidium
and Giardia percentage occurrence was highest at Riverside Park, 73.7% and 89.5%,
respectively, and lowest at North Beach Park, 26.3% and 26.3%, respectively (Table 4.1).
At Deer Creek Park, Cryptosporidium was detected in 59.1% of the samples and Giardia
in 36.4%. The Market Avenue Retention Basin (MARB) samples were positive 62.5%
and 100% of the time for Cryptosporidium and Giardia, respectively. Cryptosporidium
occurred at a frequency equal or greater than occurrence of Giardia in North Beach Park
and Deer Creek Park samples. In Riverside Park and MARB samples, however, the
percentage of Giardia occurrence is higher than the percentage of Cryptosporidium
occurrence.
Among the River Raisin watershed sites, Cryptosporidium was detected at least 50%
of the time (Table 4.1). Giardia was detected in the River Raisin watershed less
frequently than Cryptosporidium. Cryptosporidium percentage occurrence was slightly
higher in the River Raisin watershed sites downstream from CSO outfalls (53.3%) than
in the River Raisin watershed sites upstream from CSO outfalls (50%). Conversely,
Giardia percentage occurrence was higher in the River Raisin watershed sites upstream

62

from CSO outfalls (29.2%) than in the River Raisin watershed sites downstream from
CSO outfalls (13.3%). When evaluating the combined Grand River recreational sites
(Deer Creek Park and Riverside Park) and all sites in the River Raisin watershed,
Cryptosporidium was detected more frequently than Giardia in the aggregated datasets.

63

Table 4.1. Water Samples Positive for Cryptosporidium and Giardia
Watershed

Number Total
of
number
sampling of sample
locations events

Site

North Beach
Park
Riverside
Park
Deer Creek
Grand River
Park
watershed
Grand River Sites
(Deer Creek &
Riverside Park)

Giardia
Cryptosporidium
Positive Events
Positive Events (%) (%)

1

19

5 (26.3)

5 (26.3)

1

19

14 (73.7)

17 (89.5)

1

22

13 (59.1)

8 (36.4)

2

41

27 (65.9)

25 (61.0)

1

8d

5 (62.5)

8 (100)

River Raisin River Raisin
watershed
b
River Raisin

8

24

12 (50.0)

7 (29.2)

4

15

8 (53.3)

2 (13.3)

c

13

39

20 (51.3)

9 (23.1)

MARB
a

River Raisin

Cryptosporidium and Giardia occurrence values are presented as positive events and as percentage.
a

River Raisin watershed sites upstream of CSO outfalls

b
c
d

River Raisin watershed sites downstream of CSO outfalls
All River Raisin watershed sites
Rain dependent sampling

64

Figure 4.1 and Table 4.2 show the geometric mean of parasite concentrations at the
monitored sites. While the parasites were detected frequently, the concentrations were
consistently low except at MARB. When MARB samples were excluded, the geometric
mean Cryptosporidium concentrations found in the River Raisin sites were higher than
those in the Grand River watershed. There was no consistent pattern to Giardia
occurrence between the River Raisin and Grand River sites, however. The parasite
concentrations from two CSO discharge events (sampled directly from the combined
sewer weir at the Ionia and Stevens intersection and the #2 Goodrich site) were highly
variable (Table 4.3). Cryptosporidium concentrations from these events were below
detectable levels and Giardia ranged from a low concentration similar to the Grand
River and River Raisin samples on the 2/16/2006 event, to a concentration that fell
between that of the Grand and River Raisin concentration and the MARB concentration
during the 7/18/2006 event.
Of the routinely monitored sites, MARB samples had the highest geometric mean
-1

-1

concentrations for both Cryptosporidium and Giardia, 6 oocysts L and 1350 cysts L ,
respectively. The next highest geometric mean parasite concentration detected, the
Cryptosporidium concentration found in the River Raisin sites upstream from CSO
-1

outfalls (0.615 oocysts L ), was approximately 10 fold lower than the MARB
Cryptosporidium concentration. After MARB, in decreasing order of Giardia
-1

-1

concentration were Riverside Park (0.29 cysts L ), River Raisin sites (0.2 cysts L , all
-1

site combinations), and Deer Creek Park and North Beach Park (0.1 cysts L , each).
Within the Grand River watershed, the highest geometric mean for Cryptosporidium

65

-1

concentrations was found in Deer Creek Park (0.27 oocysts L ), while the highest
geometric mean for Giardia was detected in water samples from Riverside Park (0.29
-1

cysts L ; range 0.1 to 0.452). For Cryptosporidium, the combined Deer Creek Park and
-1

Riverside Park geometric mean (0.25 cysts L ) was less than the overall geometric
-1

mean from all River Raisin watershed sites (615 cysts L ). However, the geometric
mean for Giardia from the combined Deer Creek Park and Riverside Park samples
-1

(0.20 cysts L ) was similar to the overall geometric mean from all the River Raisin
-1

watershed sites (0.2 cysts L ) . For all river sites, the geometric mean value was below
-1

1 (oo)cyst L , although the Cryptosporidium detection in the River Raisin watershed
-1

was variable, producing an upper 95% confidence limit of 17.2 oocysts L and 10.5
-1

oocysts L for River Raisin sites upstream of CSO outfalls and all the River Raisin
watershed sites, respectively. This variability is largely due to the Rice Lake Drain
sampling location in which Cryptosporidium concentrations ranged from 0.3 to 599
-1

oocysts L .

66

Figure 4.1 Geometric Mean of Parasite Occurrence Versus Site

Parasite occurrence: Geometric mean
10000

100
Cryptosporidium

10

Giardia

1

River Raisinc
n=39

River Raisinb
n=15

River Raisina
n=24

MARB
n=8

Grand River Sites
n=41

Deer Creek Park
n=22

0.01

Riverside Park
n=19

0.1
North Beach Park
n=19

(Oo)cysts L

-1

1000

Figure 4.2 Geometric mean of parasite occurrence versus site. Error bars indicate the 95% confidence interval.
Detection limits were used to calculate concentrations in samples where parasites were not detected.
a
River Raisin watershed sites upstream of CSO outfalls
b
c

River Raisin watershed sites downstream of CSO outfalls
All River Raisin watershed sites

67

Table 4.2 Geometric Mean of Parasite Occurrence by Sampling Location

Watershed

Grand
River
watershed

River
Raisin
watershed

Sampling
Location

Number of
Sampling
Events

North Beach
Park
Riverside Park
Deer Creek
Park
Grand River
Sites
MARB
River Raisin
River Raisin
River Raisin

Geometric Mean  95% Confidence
Interval
Cryptosporidium
Giardia
-1
-1
(oocysts L )
(cysts L )

19

0.1  0.008

0.1  0.01

19

0.2  0.03

0.29  0.11

22

0.27  0.17

0.1  0.02

41

0.25  0.091

0.20 0.055

8

a

24

b

15

c

39

63
0.615  17.2
0.3  0.04
0.440  10.5

1350  274
0.2  0.04
0.2  0.02
0.2  0.03

Table 4.2 Geometric mean of parasite occurrence by sampling location. Detection limits were used
to calculate concentrations in samples where parasites were not detected.
a
b
c

River Raisin watershed sites upstream of CSO outfalls
River Raisin watershed sites downstream of CSO outfalls
All River Raisin watershed sites

68

Table 4.3. Parasite Concentrations in CSO Samples

Site

Ionia and
Stevens
Ionia and
Stevens
#2 Goodrich

Date

Cryptosporidium
-1
(oocyst L )

Giardia
-1
(cyst L )

2/16/2006

<0.09

0.09

7/18/2006

<1

184

7/18/2006

<1.0

102

4.2 Recovery Efficiency Results
For North Beach Park matrix spike samples, the mean percentage and standard
deviation of Cryptosporidium and Giardia recovery was 72.2 ± 2.1% and 67.3± 28.8%,
respectively (Table 3.1). For Deer Creek Park matrix spike samples, the mean
percentage and standard deviation of Cryptosporidium and Giardia recovery was 66.2
±16.9% and 42.5± 25.2%, respectively. For the single Riverside Park matrix spike
sample, Cryptosporidium and Giardia recovery was 73.0% and 43.4%, respectively.
All the Grand River watershed matrix spike and OPR sample recovery percentages were
acceptable according to the criteria (OPR criteria) used to assess Cryptosporidium and
Giaridia recovery in laboratory reagent water (Table 4.4 and Table 4.5).
Recoveries dropped for MARB matrix spike samples, the percentage of
Cryptosporidium recovery ranged from 11% to 73% with a mean recovery of 33.7% and
a standard deviation of 22.6 (Table 4.4). The percentage of Giardia recovery for MARB
samples, ranged from 9.5% to 79.5% with a mean recovery of 42.1% and a standard
deviation of 24.9. Two of the seeded MARB samples had recovery efficiencies below

69

the criteria used to assess recovery in laboratory reagent water (OPR criteria). These
low recoveries are most likely due to matrix interference as the recoveries from OPR
samples were acceptable (Table 4.5).
For all River Raisin watershed matrix spike samples, the mean percentage and
standard deviation of Cryptosporidium and Giardia recovery was 49.6± 24.2% and 9.7±
10.6%, respectively (Table 4.4). The percentage of Cryptosporidium recovery ranged
from 16.5% to 90.4%. All Cryptosporidium recovery percentages in matrix spike
samples were acceptable according to the criteria used to assess Cryptosporidium
recovery in laboratory reagent water (OPR criteria). The percentage of Giardia recovery
ranged from 0% to 34%. Only two of the River Raisin watershed matrix spike samples
met the criteria for acceptable Giardia recovery in laboratory reagent water (OPR
criteria). The recoveries from all Giardia OPR samples were acceptable (Table 4.5),
indicating probable matrix interference with Giardia recovery in the matrix spike
samples.

70

Table 4.4. Cryptosporidium and Giardia Recovery in Matrix Spike
Samples
Location

Cryptosporidium
Recovery %
Mean 
Individual
standard
recovery %
deviation

River Raisin
Watershed
All Sites

Upstream
of CSO

Downstream
of CSO

Grand River
Watershed
Recreational Sites
North Beach Park
Deer Creek Park
Riverside Park

MARB

a

Giardia Recovery %
Individual
recovery %

49.6  24.2
39.1
36.5
26.15
73.45
62.1
16.5
90.4
80.3
20.0
54.4
60.1
24.2
37.8
73.3

49.4  27.7

50.0  19.2

9.7  10.6
3.8
12.4
4.8
13.5
0
2.75
5.3
0
0
34
8.17
30.3
9.1
11.8

69.9  9.20
73.7
70.7
78.1
54.2
73.0
11.0
a
14.0
53.0
73.0
32.0
35.0
18.0

72.2 ± 2.1
66.2 ±16.9
NA

33.7 ± 22.6

Matrix spike duplicate

71

Mean 
standard
deviation

4.7  5.1

18.7  12.4

52.6  23.3
46.9
87.6
24.7
60.3
43.4
9.50
a
46.0
40.5
12.0
59.0
48.0
80.0

67.3± 28.8
42.5± 25.2
NA

42.1± 24.9

Table 4.5. Cryptosporidium and Giardia Recovery in Ongoing
Precision and Recovery (OPR) Samples
Location

River Raisin
Watershed
All Sites

Grand River
Watershed
Recreational Sites

MARB

Cryptosporidium Recovery %
Mean 
Individual
standard
recovery %
deviation
49
30.6
41.3  7.8
44
41.6
56
51
80.8
40.4
40.4
48.5
56.6
37.4
49.5
64.6
63.6
37.4
55.2  11.7
63.6
68.7
69.7
60.6
60.6
50.5
70.7
45.5
47.5
46.1
60.6
44.9
36.7
37.0
47.3  13.6
48.0
70.0

Giardia Recovery %
Mean 
Individual
standard
recovery %
deviation
55
35.4
42.1  10.5
46.05
31.8
44
22.8
34
29
32
33.7
46.9
42.9
44.9
41.8
23.5
34.7
43.9  15.7
16.3
51.0
31.6
72.4
65.3
41.8
73.5
51
62.2
56.1
59.2
55.6
97.0
55.5
67.0  17.3
62.0
65.0

Note: Acceptance criteria for Cryptosporidium and Giardia recovery were 13-111%
and 15-118%, respectively.

72

Figure 4.2 Watershed Land Cover in Ottawa and Lenawee County, MI

73

4.2 Statistical Analysis Results: Grand River Watershed Recreational
Sites
Comparison of parasite occurrence datasets from North Beach Park, Riverside Park,
and Deer Creek Park identified significant differences in these datasets.
Cryptosporidium occurrence at North Beach Park was significantly less than
Cryptosporidium occurrence at either Deer Creek Park or Riverside Park (p<0.05).
North Beach Park Giardia occurrence was significantly less than Giardia occurrence at
Riverside Park (p<0.05). No other meaningful comparison of parasite occurrence was
found to be statistically significant.
Comparison of parasite detection limit datasets from North Beach Park, Riverside
Park, and Deer Creek Park found no significant difference between these sites (p=0.700).
No significant difference between Giardia and Cryptosporidium recovery efficiency
in matrix spike samples was found (p=0.160), however the power of this test was low
(power = 0.173). Therefore, there could be significant differences between these
recovery efficiencies that this test cannot detect.
When recovery efficiency in OPR samples was examined, significantly less Giardia
was recovered compared to Cryptosporidium (p=0.008)

4.3 Statistical Analysis Results: Grand River Watershed Recreational
Sites and MARB
There was a statistically significant difference between parasite occurrence from
MARB, North Beach Park, Riverside Park, and Deer Creek Park datasets (p<0.001).
Cryptosporidium occurrence in MARB was significantly greater than Cryptosporidium
occurrence in both Deer Creek Park (p<0.05) and North Beach Park (p<0.05) datasets.
74

MARB Giardia occurrence was significantly greater than Giardia occurrence in both
Deer Creek Park (p<0.05) and North Beach Park (p<0.05) datasets. No other
meaningful comparison of parasite occurrence was found to be statistically significant.
Detection limit datasets from MARB, North Beach Park, Riverside Park, and Deer
Creek Park were significantly different (p<0.001). The detection limits from MARB
were significantly higher than those of the North Beach Park, Riverside Park, and Deer
Creek Park detection limits.
No significant differences between recovery efficiencies in matrix spike samples
from MARB and the combined North Beach Park, Riverside Park, and Deer Creek Park
were found (p=0.055), however the power of this test was low (power = 0.429).
Therefore, there could be significant differences between these recovery efficiencies that
this test cannot detect.
There were significant differences between parasite recovery efficiency in OPR
samples from MARB and the combined North Beach Park, Riverside Park, and Deer
Creek Park recovery efficiency dataset (p=0.005). Significantly fewer Giardia than
Cryptosporidium were recovered in the combined North Beach Park, Riverside Park, and
Deer Creek Park dataset. Significantly less Giardia were recovered in MARB dataset
than in the combined North Beach Park, Riverside Park, and Deer Creek Park dataset.

4.4 Statistical Analysis Results: River Raisin Watershed
No significant difference (p=0.282) in parasite occurrence was found between sites
upstream of CSO outfalls and sites downstream of CSO outfalls. No significant
differences were found when detection limits between upstream and downstream sites
were compared (p=0.444). No significant difference between Cryptosporidium and

75

Giardia concentration was found (p=0.082) when parasite occurrence datasets from
upstream and downstream sites were pooled.
There were significant differences in parasite recovery efficiency in matrix spike
samples between sites upstream of CSO outfalls and sites downstream of CSO outfalls
(p<0.001). The recovery of Cryptosporidium was significantly greater than recovery of
Giardia in sites upstream from CSO outfalls (p<0.05). No other meaningful comparison
of parasite recovery in matrix spike samples was found to be statistically significant.
When parasite recovery efficiency in matrix spike sample datasets from upstream and
downstream sites were pooled and compared Cryptosporidium recovery was
significantly greater than Giardia recovery (p<0.001).
No significant differences were found when recovery efficiencies of Giardia and
Cryptosporidium in OPR samples were compared (p=0.911), however the power of this
test was low (power = 0.05). Therefore, there could be significant differences between
these recovery efficiencies that this test cannot detect.

4.5 Statistical Analysis Results: River Raisin vs Grand River
Watershed
As no significant differences were found between parasite occurrence between Deer
Creek Park and Riverside Park, these datasets were pooled for comparison with the
River Raisin watershed sites. As no significant differences were found between
upstream and downstream sites in the River Raisin watershed, these datasets were
pooled for comparison with the Grand River watershed sites.
There were significant differences between parasite occurrence in datasets from
North Beach Park, from the combined Deer Creek Park-Riverside Park dataset, and from

76

the total River Raisin dataset (p<0.001). North Beach Park Cryptosporidium occurrence
was significantly less than Cryptosporidium occurrence in the combined Deer Creek
Park-Riverside Park dataset, and significantly less than Cryptosporidium occurrence in
the total River Raisin dataset. North Beach Park Giardia occurrence was also
significantly less than Giardia occurrence in the Deer Creek Park-Riverside Park dataset,
and significantly less than Giardia occurrence in the total River Raisin dataset.
There were significant differences between detection limits in datasets from North
Beach Park, from the combined Deer Creek Park and Riverside Park dataset, and from
the total River Raisin dataset (p<0.001). The detection limits in the total River Raisin
dataset were significantly higher than detection limits in the North Beach Park dataset
and significantly higher than detection limits in the combined Deer Creek ParkRiverside Park dataset.
There were significant differences between parasite recovery efficiency in matrix
spike samples from the total Grand River dataset (North Beach Park, Deer Creek Park,
and Riverside Park) and from the total River Raisin dataset (p<0.001). Cryptosporidium
recovery in matrix spike samples was significantly higher than Giardia recovery in the
total River Raisin dataset. Giardia recovery in the total Grand River dataset was
significantly higher than Giardia recovery in the total River Raisin dataset.
There were significant differences between parasite recovery efficiency in OPR
samples from the total Grand River dataset (North Beach Park, Deer Creek Park, and
Riverside Park) and from the total River Raisin dataset (p<0.001). Cryptosporidium
recovery in OPR samples was significantly higher than Giardia recovery in the total
Grand River dataset.

77

4.6 Discussion
4.6.1 River Raisin Watershed
During the time period that sampling was conducted in the River Raisin watershed,
four CSO discharges occurred (35). However, the lack of significant differences
between sites upstream and downstream of CSO outfalls suggests that CSO discharges
were not significantly impacting parasite occurrence in the watershed during the
sampling period.
Matrix spike recovery efficiency was significantly greater for Cryptosporidium than
Giardia in upstream sites, which may explain why geometric mean occurrence of
Cryptosporidium in sites upstream of CSO outfalls was higher (although not
significantly) than in sites downstream of COS outfalls.
Pooled upstream and downstream sites did not have significantly different
concentrations of Cryptosporidium and Giardia, however the pooled matrix spike
recovery efficiencies were significantly different, indicating a potential bias against
Giardia. As no significant differences in detection limits or in recovery efficiency of
OPR samples was found, these factors are not introducing bias into intra-watershed
comparisons.

4.6.2 Grand River Watershed
The Grand River watershed recreational sampling sites included both river and lake
sampling locations. Comparisons of parasite occurrence between the river and lake
locations could be affected by dilution effects, as the Cryptosporidium occurrence at
North Beach Park was significantly lower than Cryptosporidium occurrence at either
Deer Creek Park or Riverside Park. However, Giardia occurrence was not significantly

78

different between North Beach Park and Deer Creek Park, while it was significantly
higher at Riverside Park than at North Beach Park. This pattern suggests that there may
be an additional source of Giardia contamination between Deer Creek Park and
Riverside Park. As no significant differences in detection limits or in recovery
efficiency of matrix spike samples was found, these factors are not considered to be
introducing bias into intra-watershed comparisons. However, significantly less Giardia
than Cryptosporidium was recovered in OPR samples, indicating a method bias against
detection of Giardia.
The Grand River watershed study area also contains the Market Avenue Retention
Basin (MARB) and other CSO outfalls that may impact the recreational water quality of
the Grand River recreation areas during CSO discharges. Cryptosporidium occurrence
in MARB was not significantly different from Cryptosporidium occurrence in Riverside
Park and Giardia occurrence in MARB was not significantly different from Giardia
occurrence in Riverside Park. This information, combined with the indication of a
contamination source located between Deer Creek Park and Riverside Park suggests that
this suspected source is a frequently discharging CSO outfall, a sewage treatment plant
discharge with low physical removal of Cryptosporidium and Giardia, or an illicit
discharge of untreated wastewater.
The detection limits in MARB samples were significantly higher than those of the
recreational Grand River sites. Since Giardia was detected in 100% of MARB samples,
this detection limit difference is not a problem when comparing Giardia occurrence
values. However, the limit of detection may introduce bias when comparing
Cryptosporidium values as the geometric mean is calculated using the detection limit

79

when no parasite is detected. This difference could produce artificially higher
Cryptosporidium values in MARB samples.
Additionally, there were significant differences between parasite recovery efficiency
in OPR samples from MARB and the combined North Beach Park, Riverside Park, and
Deer Creek Park recovery efficiency dataset. This could introduce bias against Giardia
occurrence values in the Grand River recreational sites as they are compared to MARB
occurrence values. As no significant differences in recovery efficiency of matrix spike
samples was found, this factor is not considered to be introducing bias into intrawatershed comparisons.

4.6.3 River Raisin vs Grand River Watershed
When the inter-watershed comparison of the River Raisin and Grand River
watershed was conducted, parasite occurrence from North Beach Park, the combined
Deer Creek Park- Riverside Park dataset, and the total River Raisin dataset were
compared. Both Cryptosporidium and Giardia occurrence in North Beach Park were
significantly different from Cryptosporidium and Giardia occurrence in the River Raisin
sites. Both Cryptosporidium and Giardia occurrence in North Beach Park were
significantly different from Cryptosporidium and Giardia occurrence in the combined
Deer Creek Park- Riverside Park dataset. However, no significant differences between
the combined Deer Creek Park- Riverside Park dataset, and the total River Raisin dataset
were found. As previously discussed, the difference between parasite occurrence at
North Beach Park and the combined Deer Creek Park- Riverside Park dataset may be
due to dilution effects. To address the question of whether parasite occurrence patterns
in the watersheds were associated with differences in land cover type, the most

80

informative comparison is therefore that between the combined Deer Creek ParkRiverside Park dataset and the total River Raisin dataset. No significant difference in
parasite occurrence patterns between the combined Deer Creek Park- Riverside Park
dataset and the total River Raisin dataset was detected; therefore we fail to reject the null
hypothesis that the parasite occurrence in both watersheds is the same.
Potential biases exist that may affect this conclusion due to differences in detection
limits and matrix spike recoveries. The detection limits in the total River Raisin dataset
are significantly higher than detection limits in the combined Deer Creek ParkRiverside Park dataset which could artificially increase the parasite occurrence values in
the River Raisin dataset. In matrix spike samples, Giardia recovery in the total Grand
River dataset was significantly higher than Giardia recovery in the total River Raisin
dataset, which could produce artificially low Giardia occurrence values in the River
Raisin dataset. Significant differences were also found between OPR recovery
efficiencies, however these differences were between Cryptosporidium and Giardia
recovery in an intra-watershed comparison of the Grand River watershed and do not
affect the inter-watershed comparison of the Grand River and River Raisin watershed.
Although there were differences in the percentage of urban and rural land cover
between the Grand River watershed in Ottawa County and the River Raisin watershed in
Lenawee County, the parasite occurrence in both watersheds was statistically similar.
Parasite occurrence was not shown to vary with differences in land cover, unlike
parasitic disease incidence (Chapter 3). The difference in patterns between parasite
occurrence and parasitic disease could be a result of scale. Parasitic disease patterns
were examined by ZIP code which is a finer scale than the watershed and may resolve

81

patterns better. At coarser scales, to detect statistically significant differences in parasite
occurrence, the differences in the land cover percentages may need to be more extreme
than those found between the Grand River watershed in Ottawa County and the River
Raisin watershed in Lenawee County.
The third caveat in the inter-watershed comparison involves potential biases due to
differences in weather patterns between the watersheds. The original experimental
design for data collection in the River Raisin watershed was based on sampling after
rainfall to examine parasite concentrations after storm events. The River Raisin dataset
is therefore less representative of the yearly parasite concentration in surface water than
the Grand River dataset, which was collected at regular intervals over a longer time
period. The differences in seasonal weather patterns and differences between yearly
weather patterns may introduce biases into the comparison of parasite concentrations
between the River Raisin and Grand River watersheds, although this is beyond the scope
of the study.

4.7 Conclusion
Although differences in the pattern of disease occurrence between urban and rural
areas were found in Chapter 3, no evidence was found in the post-hoc testing of parasite
occurrence between the Grand River and River Raisin watersheds that suggests a
difference in parasite occurrence between urban and rural areas. Caveats to this
conclusion include differences in detection limits and matrix spike recoveries. The data
used to drawn this conclusion may also be affected by weather patterns during sampling,
as the Grand River watershed and River Raisin watershed were sampled in different
years and over different seasons, although this was beyond the scope of the study.

82

Chapter 5: Recreational Health Risks of Combined Sewer
Overflows
Combined sewers are sewers that are connected to storm drains. During dry weather,
they carry sewage to the sewage treatment plant. During rain events, or during seasonal
thaws of ice and snow, water from storm events enters these sewers as well. Combined
sewers are designed with overflow points (or outfalls) where combined sewage can be
discharged, untreated, into the environment. CSO discharges occur when either the
capacity of the sewage treatment plant to treat incoming sewage or when the capacity of
the sewage pipe network to transport sewage is exceeded. Discharges of untreated
sewage into recreational waters can lead to pathogen exposure, infection, and illness in
the human population. In this study, health risks caused by exposure to Cryptosporidium
and Giardia in children under the age of 16 swimming in recreational areas that receive
inputs from CSOs were investigated.

5.1 Risk Assessment Results
Table 5.1 contains the distributions and distribution parameters used in the Monte
Carlo analysis of the health risks. The predicted concentration of Cryptosporidium and
Giardia at the CSO discharge point ( CRIVER) is shown in Table 5.2.

83

Table 5.1. Risk assessment distribution parameters.
Exposure Factor

Distribution

Distribution
Parameter

Unit

Source

Cryptosporidium
concentration
a
at recreational area
North Beach Park

Weibull

Location = 0.07519
Scale = 0.04541
Shape = 0.94238
organism ·
Location = 0.07874
Scale = 0.29437
Shape = 0.52793

Deer Creek Park

Weibull

Riverside Park

Beta

Minimum = 0.09858
Maximum = 0.84573
Alpha = 0.37251
Beta = 1.1563

Maximum
Extreme

Likeliest = 0.10853
Scale = 0.03819

-1

This

L

study

Giardia concentration
a
at recreational area
North Beach Park

Deer Creek Park

Riverside Park

Location = 0.07653
Shape = 1.5981
Pareto

Lognormal

Mean = 0.41996
Standard
Deviation = 0.44272

84

organism ·
-1

L

This
study

Table 5.1 (cont’d).

Exposure Factor
Cryptosporidium
concentration
in MARB
a
influent

Giardia
concentration
in MARB
a
influent

Distribution Distribution Parameter

Triangular

Beta

Duration of CSO
discharge
(374 events)

Weibull

Volumetric flow
rate of Grand
River

Gamma

Unit

Source

Minimum = 0.77
Likeliest = 13.33
Maximum = 30.56

organism
-1
·L

This
study

Minimum=478.15
Maximum=5,220.23
Alpha=0.50131
Beta=1.57452

organism
-1
·L

This
study

minute

This
study

L
-1
·minute

This
study

Location = -1.8981 x10
Scale = 91.461
Shape = 0.52722
Location = 363.57
Scale = 1126.5
Shape = 1.2229

85

-72

Table 5.1 (cont’d).
Exposure Factor
Ingestion volume
during swimming

Distribution

Unit

Source

Minimum = 0
Likeliest = 0.000833
Maximum = 0.00166

L·
-1
minute

(53)

Minimum = 0
Likeliest = 137.6
Maximum = 181

minute
-1
·month

(53)

Constant

0.00419

Unitless

(25)

Constant

0.01982

Unitless

(45)

Unitless

(25)

Unitless

(45)

Triangular

(I )
Amount of time
spent swimming
( TSW IM )
r
Fraction of ingested
Cryptosporidium
that survive to
initiate infection
r
Fraction of ingested
Giardia that survive
to initiate infection

Triangular

Morbidity Ratio :
Triangular
Cryptosporidium
Morbidity Ratio:
Normal
Giardia

Distribution Parameter

Minimum = 0.19
Likeliest = 0.39
Maximum = 0.62
Mean = 0.505
Standard
Deviation = 0.1902

86

Table 5.1 (cont’d).

Exposure

Distribution

Distribution Parameter

Beta

Minimum = -61738914.76
Maximum = 62078291.44
Alpha = 100
Beta = 100

Beta

Unit

Minimum = -894803.55
Maximum = 902736.90
Alpha = 100
Beta = 100

Source

Factor
Volume of
CSO
discharge at
b
outfall
Grand River

Grand River
via
Coldbrook
Drain

(35),
L

Mean = 5051.00
Scale = 97680.66
Grand River
via
Plaster Creek

Logistic

This
study

Mean = 5061.39
Scale = 696103.24

Grand River
Logistic
via
Market Avenue
Retention
Basin
a

Based on empirical values with non-detects represented using the detection limit.
Use produces the most conservative (ie most protective) estimate of risk
b

Based on time interval from January 1, 2002 to December 31, 2006

87

Table 5.2. Predicted Concentration Distributions of Cryptosporidium
and Giardia at the CSO Discharge Point ( CRIVER).

Summary
statistics of
distribution

Cryptosporidium
-1
(oocyst L )

Mean
Median
Standard
Deviation
Variance
Minimum
Maximum
Lower (5%)
confidence
limit
Upper (95%)
confidence
limit

Giardia
-1
(cyst L )

13.96
13.66

1,502.22
1,048.94

6.03
36.34
1.28
30.18

1,127.78
1,271,881.27
95.07
5,097.08

4.33

459.35

24.93

3,915.89

The estimated health risks to children under the age of sixteen from exposure to the
parasites Cryptosporidium and Giardia while swimming are shown in Table 5.3. The
mean daily risks of Cryptosporidium infection at North Beach Park, Deer Creek Park,
-6

-6

Riverside Park, and at the CSO point of discharge were 1.50 x10 , 7.41 x10 ,
-6

-4

3.61 x10 , and 1.78 x10 , respectively. The mean daily risks of Giardia infections at
North Beach Park, Deer Creek Park, Riverside Park, and at the CSO point of discharge
-6

-6

-6

-2

were 1.58 x10 , 2.48 x10 , 5.18 x10 , and 1.80 x10 , respectively. The mean
seasonal risks of Cryptosporidium infection at North Beach Park, Deer Creek Park,
-4

-4

Riverside Park, and at the CSO point of discharge were 1.35 x10 , 6.66 x10 ,
-4

-2

3.24 x10 , and 1.58 x10 , respectively. The mean seasonal risks of Giardia infection at

88

North Beach Park, Deer Creek Park, Riverside Park, and at the CSO point of discharge
-4

-4

-4

-1

were 1.42 x10 , 2.23 x10 , 4.66 x10 , and 6.31 x10 , respectively.
The mean daily risks of cryptosporidiosis (illness) at North Beach Park, Deer Creek
Park, Riverside Park, and at the CSO point of discharge were about 40% of the
-7

-6

-6

-5

infectivity and were estimated at 6.02 x10 , 2.98 x10 , 1.43 x10 , and 7.03 x10 ,
respectively. The mean daily risks of giardiasis at North Beach Park, Deer Creek Park,
-7

-6

Riverside Park, and at the CSO point of discharge were 7.90 x10 , 1.29 x10 ,
-6

-3

2.62 x10 , and 9.39 x10 , respectively. Again seasonal risks went up by about 90
times. The mean seasonal risks of cryptosporidiosis at North Beach Park, Deer Creek
-5

-4

Park, Riverside Park, and at the CSO point of discharge were 5.42 x10 , 2.68 x10 ,
-4

-3

1.29 x10 , and 6.24 x10 , respectively. The mean seasonal risks of giardiasis at North
Beach Park, Deer Creek Park, Riverside Park, and at the CSO point of discharge were
-5

-4

-4

-1

7.11 x10 , 1.16 x10 , 2.36 x10 , and 3.25 x10 , respectively. Additional summary
statistics including the mean, median, standard deviation, variance, minimum, maximum,
and upper and lower confidence limits of the risk distributions for daily and seasonal risk
of infection and daily and seasonal risk of illness are shown in Table 5.3.
Giardia risks were significantly higher than Cryptosporidium risks at the CSO
discharge site. (P<0.001, α = 0.05). At the recreational sites the risks were not
statistically different between Giardia and Cryptosporidium (P≥0.694, α = 0.05). The
daily risk of infection, daily risk of illness, seasonal risk of infection, and seasonal risk
of illness associated with Giardia at the CSO discharge point were all significantly

89

higher than the respective risk associated with Giardia at the recreational sites (P<0.001,
α = 0.05). The seasonal risk of infection and seasonal risk of illness associated with
Cryptosporidium at the CSO discharge point were both significantly higher than the
respective risk associated with Cryptosporidium at the recreational sites (P<0.001, α =
0.05).
Using the Monte Carlo analysis, the probability of meeting recreational water
quality criteria for illness rates was computed and the results are shown in Table 5.4.
These are based on the probability distributions produced from 10,000 trials. Table 5.4
shows the likelihood of meeting the EPA recommended freshwater recreational criteria
of 8 illnesses in 1000 swimmers (0.8%).
Using the estimate of daily risk, the water quality of all recreational sites met the
freshwater recreational criteria in 99.9% of cases for both cryptosporidiosis and
giardiasis. The water quality at the point of the CSO discharge was also predicted to
meet the recreational criteria for cryptosporidiosis in 99.9% of cases. However, the
water quality at the point of CSO discharge was predicted to meet the freshwater
recreational criteria for giardiasis in only 62.7% of cases.
Using the estimate of seasonal risk, the water quality of North Beach Park and
Riverside Park met the recommended freshwater recreational criteria 99.9% of the time
for both cryptosporidiosis and giardiasis. The water quality of Deer Creek Park met the
freshwater recreational criteria 92.4% of the time for cryptosporidiosis and 99.9% of the
time for giardiasis. The water quality at the point of CSO discharge met the recreational
criteria in 73.0% of cases for cryptosporidiosis and in 0.46% of cases for giardiasis.

90

Table 5.3. Estimated Risk of Health Outcomes for Children
Swimming at Recreational Sites and at Point of CSO Discharge in the
Grand River Watershed
North Beach Park Cryptosporidium
Summary
statistics of risk
distribution

Daily
Risk of
Infection

Mean

1.50 x10

Median
Standard
Deviation

1.29 x10

Variance

1.06 x10
1.12 x10

Seasonal Risk
of Infection
-6
-6
-6

-12

Minimum

5.97 x10

Maximum
Lower (5%)
confidence limit
Upper (95%)
confidence limit

8.49 x10
2.78 x10
3.52 x10

-8
-6
-7
-6

1.35 x10
1.16 x10
9.51 x10
9.04 x10
5.38 x10
7.64 x10
2.50 x10
3.17 x10

-4
-4
-5
-9
-6
-4
-5
-4

Daily
Risk of Illness
6.02 x10
4.85 x10
4.67 x10
2.18 x10

-7
-7
-7

-13

2.23 x10
3.80 x10
9.92 x10
1.50 x10

-8
-6
-8
-6

Seasonal Risk
of Illness
5.42 x10
4.36 x10
4.20 x10
1.76 x10
2.01 x10
3.42 x10
8.93 x10
1.35 x10

-5
-5
-5
-9
-6
-4
-6
-4

North Beach Park Giardia
Summary
statistics of risk
distribution
Mean
Median
Standard
Deviation
Variance

Daily
Risk of
Infection
-6
1.58 x10
1.33 x10
1.14 x10
1.29 x10

-6
-6

-12

Minimum

2.26 x10

Maximum
Lower (5%)
confidence limit
Upper (95%)
confidence limit

7.79 x10
2.60 x10
3.68 x10

-9
-6
-7
-6

Seasonal Risk
Daily
Seasonal Risk
of Infection
Risk of Illness
of Illness
-4
-7
-5
1.42 x10
7.90 x10
7.11 x10
1.19 x10
1.02 x10
1.05 x10
2.03 x10
7.00 x10
2.34 x10
3.31 x10

91

-4
-4
-8
-7
-4
-5
-4

6.27 x10
6.43 x10
4.14 x10

-7
-7

-13

1.78 x10
4.05 x10
1.01 x10
2.03 x10

-9
-6
-7
-6

5.64 x10
5.79 x10
3.35 x10
1.60 x10
3.65 x10
9.11 x10
1.83 x10

-5
-5
-9
-7
-4
-6
-4

Table 5.3. (cont’d)
Deer Creek Park Cryptosporidium
Summary
Daily
statistics of risk
Risk of
Seasonal Risk
Daily
distribution
Infection
of Infection
Risk of Illness
-6
-4
-6
7.41 x10
6.66 x10
2.98 x10
Mean
Median
Standard
Deviation
Variance
Minimum

2.31 x10
1.58 x10
2.51 x10

-6
-5

-10

7.85 x10

-8
-4

2.18 x10
Maximum
Lower (5%)
-7
3.74 x10
confidence limit
Upper (95%)
-5
3.23 x10
confidence limit
Deer Creek Park Giardia
Summary
statistics of risk
distribution
Mean
Median
Standard
Deviation
Variance

Daily
Risk of
Infection
-6
2.48 x10
1.48 x10
4.70 x10
2.21 x10

-6
-6

-11

Minimum

2.20 x10

Maximum
Lower (5%)
confidence limit
Upper (95%)
confidence limit

9.99 x10
2.84 x10
7.65 x10

-8
-5
-7
-6

2.08 x10
1.42 x10
2.01 x10
7.06 x10
1.94 x10
3.37 x10
2.90 x10

-4
-3
-6
-6
-2
-5
-3

9.13 x10
6.40 x10
4.10 x10

-7
-6

-11

3.76 x10
7.56 x10
1.42 x10
1.31 x10

-8
-5
-7
-5

Seasonal Risk
Daily
of Infection
Risk of Illness
-4
-6
2.23 x10
1.29 x10
1.33 x10
4.22 x10
1.78 x10
1.98 x10
8.95 x10
2.55 x10
6.89 x10

92

-4
-4
-7
-6
-3
-5
-4

6.78 x10
2.98 x10
8.90 x10

-7
-6

-12

4.96 x10
6.47 x10
1.11 x10
4.08 x10

-9
-5
-7
-6

Seasonal
Risk
of Illness
-4
2.68 x10
8.21 x10
5.73 x10
3.29 x10
3.39 x10
6.75 x10
1.28 x10
1.18 x10

-5
-4
-7
-6
-3
-5
-3

Seasonal
Risk
of Illness
-4
1.16 x10
6.10 x10
2.68 x10
7.17 x10
4.46 x10
5.80 x10
9.98 x10
3.67 x10

-5
-4
-8
-7
-3
-6
-4

Table 5.3 (cont’d).
Riverside Park Cryptosporidium
Summary
statistics of risk
distribution
Mean
Median
Standard
Deviation
Variance

Daily
Risk of
Infection
-6
3.61 x10
2.43 x10
3.51 x10
1.23 x10

-6
-6

-11

Minimum

8.45 x10

Maximum
Lower (5%)
confidence limit
Upper (95%)
confidence limit

2.08 x10
4.38 x10
1.06 x10

-8
-5
-7
-5

Seasonal Risk
Daily
of Infection
Risk of Illness
-4
-6
3.24 x10
1.43 x10
2.19 x10
3.16 x10
9.99 x10
7.60 x10
1.87 x10
3.94 x10
9.51 x10

-4
-4
-8
-6
-3
-5
-4

-7

9.37 x10

-6

1.45 x10
2.11 x10

-12
-8

3.28 x10

-6

9.47 x10

-7

1.57 x10

-6

4.31 x10

Seasonal
Risk
of Illness
-4
1.29 x10
8.44 x10
1.31 x10
1.71 x10
2.95 x10
8.51 x10
1.42 x10
3.88 x10

-5
-4
-8
-6
-4
-5
-4

Riverside Park Giardia
Summary
statistics of risk
distribution

Daily
Risk of
Infection

Mean

5.18 x10

Median
Standard
Deviation

3.27 x10

Variance

5.84 x10
3.41 x10

Seasonal Risk
of Infection
-6
-6
-6

-11

Minimum

3.31 x10

Maximum
Lower (5%)
confidence limit
Upper (95%)
confidence limit

5.09 x10
4.05 x10
1.62 x10

-8
-5
-7
-5

4.66 x10
2.94 x10
5.25 x10
2.76 x10
2.97 x10
4.57 x10
3.64 x10
1.46 x10

93

-4
-4
-4
-7
-6
-3
-5
-3

Daily
Risk of Illness
-6

2.62 x10

-6

1.54 x10

-6

3.29 x10
1.09 x10

-11
-8

1.78 x10

-5

3.03 x10

-7

1.64 x10

-6

8.85 x10

Seasonal
Risk
of Illness
2.36 x10
1.39 x10
2.96 x10
8.77 x10
1.60 x10
2.72 x10
1.47 x10
7.96 x10

-4
-4
-4
-8
-6
-3
-5
-4

Table 5.3 (cont’d).
CSO Point of Discharge: Cryptosporidium
Summary
statistics of risk
distribution
Mean

Daily
Risk of
Infection
-4
1.78 x10

Median
Standard
Deviation

1.43 x10

Variance

1.80 x10

Minimum

3.97 x10

1.34 x10

-4
-4
-8
-7
-4

7.24 x10
Maximum
Lower (5%)
-5
2.34 x10
confidence limit
Upper (95%)
-4
4.45 x10
confidence limit
CSO Point of Discharge: Giardia
Summary
statistics of risk
distribution
Mean

Daily
Risk of
Infection
-2
1.80 x10

Median
Standard
Deviation

1.18 x10

Variance

3.33 x10

Minimum

2.84 x10

1.82 x10

-2
-2
-4
-5

Seasonal Risk
Daily
of Infection
Risk of Illness
-2
-5
1.58 x10
7.03 x10
1.27 x10
1.18 x10
1.40 x10
3.57 x10
6.31 x10
2.10 x10
3.93 x10

-2
-2
-4
-5
-2
-3
-2

5.56 x10
5.61 x10
3.14 x10
1.62 x10
3.70 x10
9.74 x10
1.85 x10

-5
-5
-9
-7
-4
-6
-4

Seasonal Risk
Daily
of Infection
Risk of Illness
-1
-3
6.31 x10
9.39 x10
6.57 x10
2.70 x10
7.28 x10
2.56 x10

-1

-1
-1
-2
-3
0

-3

5.60 x10

-2

1.09 x10
1.19 x10
2.07 x10

-4
-5
-2

Seasonal
Risk
of Illness
-3
6.24 x10
4.97 x10
4.93 x10
2.44 x10
1.46 x10
3.23 x10
8.76 x10
1.63 x10

-3
-3
-5
-5
-2
-4
-2

Seasonal
Risk
of Illness
-1
3.25 x10
3.05 x10
1.95 x10
3.80 x10
1.86 x10

-1
-1
-2
-3
-1

1.26 x10
1.00 x10
8.82 x10
9.90 x10
Maximum
Lower (5%)
-3
-1
-4
-2
1.91 x10
1.58 x10
7.43 x10
5.65 x10
confidence limit
Upper (95%)
-2
-1
-2
-1
5.48 x10
9.94 x10
3.16 x10
6.87 x10
confidence limit
Note: Non-detects in the dataset assigned the value of the detection limit for
simulation purposes

94

Table 5.4. Likelihood of Meeting Recreational Criteria of 8 illnesses
in 1,000 swimmers (0.8%).
Site

Daily Risk

Seasonal Risk

North Beach Park

>99.9

>99.9

North Beach Park Giardia

>99.9

>99.9

Riverside Park

>99.9

>99.9

Riverside Park Giardia

>99.9

>99.9

Deer Creek Park

>99.9

92.4

Deer Creek Park Giardia

>99.9

>99.9

CSO discharge point

>99.9

73.0

62.7

0.460

Cryptosporidium

Cryptosporidium

Cryptosporidium

Cryptosporidium

CSO discharge point
Giardia

Note: Non-detects in the dataset assigned the value of the detection limit for
simulation purposes

5.2 Discussion

95

It has long been known that untreated sewage can contribute numerous enteric
pathogens at fairly significant concentrations to receiving waters and that this poses a
risk to recreational users if they are exposed to these discharges. EPA has estimated that
between 3,448 to 5,576 cases of illness annually at state recognized beaches in the
United States and U.S. territories are attributable to recreational exposure to water
contaminated by combined sewer overflows and sanitary sewer overflows (55).
Epidemiological studies performed at Great Lake beaches have shown increased illness
rates in children associated with sewage impacted beaches (56). However, health
surveillance as a passive system is unable to detect all the potential illnesses, particularly
as cases may intermittently occur over long time periods. Therefore, an alternative
approach has been to use water quality monitoring and quantitative risk assessment
methods to examine potential public health impacts and risk reduction strategies.
Great expenditures are being made to control combined sewer overflows to the
Grand River in Michigan. The Grand River discharges to Lake Michigan and has the
potential to affect a number of recreational beaches, including North Beach. This study
was undertaken to address parasite contamination of the Grand River associated with
CSOs and to estimate the potential risk to swimmers. Previous studies on wastewater
have shown varying concentrations of Cryptosporidium oocysts (averaging about 1
-1

-1

oocyst L ) and higher concentrations of Giardia ranging from 10 to 100 cysts L

(33,

44). Overall, in this investigation low concentrations of Cryptosporidium were found in
samples of combined stormwater and sewage from the Market Avenue Retention Basin.
This may be due to low levels of infection in the human population in the sanitary sewer
service area or dilution effects of stormwater. Concentrations of Giardia were much

96

higher, suggesting that Giardia infection is more prevalent in the area than
Cryptosporidium. Detection of higher Giardia concentrations than Cryptosporidium
concentrations agrees with the findings of other sewage and CSO studies (2, 55).
The MARB system was designed to capture the majority of combined stormwater
and sewage for eventual return to the sewage treatment plant. This study demonstrates
that the river is currently being protected from loading of billions of (oo)cysts during
most storm events for which the majority of the CSO flows are captured by the MARB.
However, using highly conservative estimates of risk, the remaining flows of combined
sewage and storm water can still produce risk levels of cryptosporidiosis and giardiasis
that exceed the EPA recommended freshwater recreational water quality criteria for
illness at the point of discharge. Due to the higher concentrations of Giardia found in
the combined sewage/stormwater, larger fraction that survives to initiate infection, and
larger morbidity ratio, exceeding the recommended criteria at the point of CSO
discharge is more likely for Giardia than for Cryptosporidium. Multiple day exposure to
the water at the point of discharge is required to produce a cryptosporidiosis risk level
that exceeds the water quality criteria. The daily risk of giardiasis at the CSO discharge
point is predicted to exceed the recommended freshwater recreational water quality
criteria for illness approximately 37% of the time. These risk estimates are highly
conservative due to the assumption that all detected (oo)cysts are viable and capable of
initiating infection in humans.
Although no public recreational facilities are located at the point of CSO discharge to
the Grand River, a hypothetical swimming scenario at this point represents moderate
daily health risk to a swimmer. Using conservative risk estimates developed in this

97

study, multiple day recreation at the discharge point over the recreation season
represents a high health risk to a swimmer. At the point of CSO discharge, the river is of
sufficient size to allow boating, water skiing, and the use of personal watercraft.
Therefore, recreational exposure is possible even in the absence of public recreational
facilities at the point of discharge. The elevated Giardia concentrations at the #2
Goodrich sampling location suggests that parasite concentrations released in CSO
discharge may increase health risks at locations greater that 1.6 km from the CSO
discharge point.
Parasite concentrations from the Market Avenue Retention Basin may overestimate
the concentration of parasites entering the river. Comparisons between the predicted
concentrations at the CSO discharge points ( CRIVER, Equation 5) and the
concentrations detected at during CSO discharge show the predicted concentrations are
higher than the detected parasite concentrations. This discrepancy may be due to
different ratios of sewage and stormwater that may be discharged at an outfall location
compared to that entering the retention basin. These ratios are likely to be affected by
the amount of precipitation entering the storm drain in the area of the outfall and the
number of sewer connections served by that line. The influent into the retention basin
may represent a higher number of sanitary service connections than the CSO outfalls.
Donovan et al found Giardia concentrations at 1,860 cysts per liter when sampling from
a New Jersey CSO discharge and 798 cysts per liter 10 feet downstream of the CSO
discharge (19). These Giardia values fall within the 95% confidence interval predicted
-1

at the CSO discharge point in the current study (459.35 to 3,915.89 cyst L , Table 5.2).
No Cryptosporidium was detected by Donovan et al at either location (19). A study of

98

CSO discharge from two CSO events at three CSO outfall sites in Atlanta, GA and
Louisville, KY found that geometric means ranged from 59 to 100 Cryptosporidium
oocysts per 100L and 200 to 30,000 Giardia cysts per 100L (2). These ranges are below
the 95% confidence interval predicted in the current study for both Cryptosporidium
-1

-1

(4.33 to 24.93 oocyst L , Table 5.2) and Giardia (459.35 to 3,915.89 cyst L , Table
5.2). A study of parasite concentration in CSO discharge by Gibson et al (24) reported
geometric mean and median values of Giardia that fell below the 95% confidence
interval of predicted concentration at the CSO discharge point in the present study.
Gibson et al (24) also reported arithmetic mean, geometric mean, and median values of
Cryptosporidium that exceed 95% confidence interval of predicted concentration at the
CSO discharge point. States et al (48) reported geometric mean values for Giardia that
fell below the 95% confidence interval and geometric mean values for Cryptosporidium
that fell within the 95% confidence interval predicted at the CSO discharge point. When
comparing parasite concentrations in CSO discharges from different geographic areas it
is possible that value discrepancies may also result from different levels of infection in
the populations.
In this study, the low health risks associated with Cryptosporidium and Giardia at
the recreational areas suggest that the impacts of CSO discharges at these sites are being
reduced by the retention basin, dilution, and other environmental transport processes.
Cryptosporidium concentrations in the retention basin samples were about 20 times
higher than Cryptosporidium concentrations at Deer Creek Park and Riverside Park
downstream from CSO discharge points, but Giardia cyst concentrations were
approximately 2000 times greater in MARB samples than in the river. The concentration

99

of (oo)cysts associated with routine sewage effluent discharges, non-point sources and
CSOs may be reduced in the water column via dilution and sedimentation. The Grand
River’s currents, mixing zones, and flow rates may have an impact on the concentrations
of parasites found downstream of the CSO discharge point and may influence the
likelihood that parasites will be carried downstream in the water column versus
sedimentation. The factors of sedimentation versus dilution by tributary streamflow are
difficult to separate due to the presence of only one stream flow gage in this area of the
Lower Grand River Watershed. Sedimentation may remove parasites from the water
column and either temporarily or permanently sequester them in the sediments.
Tributary streams add additional volumes to the river, potentially diluting contaminant
concentrations further. Both of these processes may reduce the exposure of an
individual to these pathogens, with an accompanying decrease in risk of infection or
illness.

5.3 Conclusion
Despite the occasional untreated sewage discharges, recreational use of the Grand
River met the EPA recommended freshwater recreational criteria of 8 illnesses per 1000
swimmers at least 92.4% of the time for both cryptosporidiosis and giardiasis at the three
recreational sites examined. Reporting of CSO discharges can provide further protection
of public health, especially when reporting is done in real time. This study demonstrates
the value of monitoring and a quantitative microbial risk approach for examining
specific pathogens and recreational waters.

100

APPENDICES

101

APPENDIX A: Statistical analysis of study area demographics

102

Table A1 Demographics: Population
Demographic data: Mann-Whitney Rank Sum Test

Rural
Urban

Group
A
B

N
40
8

Missing
0
0

Median
6875.000
36821.500

25%
2165.500
32915.500

75%
20032.500
40132.000

Mann-Whitney U Statistic= 16.000
T = 340.000
P <0.001

n(small)= 8

n(large)= 40

The difference in the median values between the two groups is greater than would be expected by chance; there is a
statistically significant difference (P <0.001)

103

2

Table A2 Demographics: Area (km )
Demographic data: Mann-Whitney Rank Sum Test

Rural
Urban

Group
A
B

N
40
8

Missing
0
0

Median
89.818
25.380

25%
60.660
18.941

75%
156.283
31.648

Mann-Whitney U Statistic= 25.000
T = 61.000
P <0.001

n(small)= 8

n(large)= 40

The difference in the median values between the two groups is greater than would be expected by chance; there is a
statistically significant difference (P <0.001)

104

2

Table A3 Demographics: Population Density (people/km )
Demographic data: Mann-Whitney Rank Sum Test

Rural
Urban

Group
A
B

N
40
8

Missing
0
0

Median
61.187
1382.606

25%
26.107
1268.094

75%
196.894
1806.801

Mann-Whitney U Statistic= 0.000
T = 356.000
P <0.001

n(small)= 8

n(large)= 40

The difference in the median values between the two groups is greater than would be expected by chance; there is a
statistically significant difference (P <0.001)

105

Table A4 Demographics: Percentage of Individuals Below the Poverty Level
Demographic data: t – Test

Rural
Urban

Group
A
B

N
40
8

Missing
0
0

Mean
0.755
1.050

Std Dev
0.213
0.195

SEM
0.0337
0.0689

Difference
-0.295
t = -3.620 with 46 degrees of freedom.
P <0.001
95 percent confidence interval for difference of means: -0.459 to -0.131
The difference in the mean values of the two groups is greater than would be expected by chance; there is a statistically
significant difference between the input groups (P <0.001).
Power of performed test with alpha = 0.050: 0.943

106

Table A5 Demographics: Median household income (US dollars)
Demographic data: t - Test

Rural
Urban

Group
A
B

N
40
8

Missing
0
0

Mean
4.677
4.607

Std Dev
0.0823
0.0718

SEM
0.0130
0.0254

Difference
0.0697
t = 2.228 with 46 degrees of freedom.
P = 0.031
95 percent confidence interval for difference of means: 0.00675 to 0.133
The difference in the mean values of the two groups is greater than would be expected by chance; there is a
statistically significant difference between the input groups (P = 0.031).
Power of performed test with alpha = 0.050: 0.486

107

Table A6 Demographics: Median Age (Years)
Demographic data: Mann-Whitney Rank Sum Test

Rural
Urban

Group
A
B

N
40
8

Missing
0
0

Median
35.950
32.200

25%
34.400
30.250

75%
37.450
33.300

Mann-Whitney U Statistic= 41.500
T = 77.500
P =0.001

n(small)= 8

n(large)= 40

The difference in the median values between the two groups is greater than would be expected by chance; there is a
statistically significant difference (P =0.001)

108

Table A7 Demographics: Mean Travel Time to Work (Minutes)
Demographic data: Mann-Whitney Rank Sum Test

Rural
Urban

Group
A
B

N
40
8

Missing
0
0

Median
23.900
19.450

25%
20.500
19.300

75%
29.000
19.900

Mann-Whitney U Statistic= 42.000
T = 78.000
P =0.001

n(small)= 8

n(large)= 40

The difference in the median values between the two groups is greater than would be expected by chance; there is a
statistically significant difference (P =0.001)

109

Table A8 Demographics: % Caucasian
Demographic data: Mann-Whitney Rank Sum Test

Rural
Urban

Group
A
B

N
40
8

Missing
0
0

Median
96.950
79.400

25%
94.600
68.900

75%
97.650
84.200

Mann-Whitney U Statistic= 10.000
T = 46.000
P <0.001

n(small)= 8

n(large)= 40

The difference in the median values between the two groups is greater than would be expected by chance; there is a
statistically significant difference (P <0.001)

110

APPENDIX B: Statistical comparison of disease between urban and rural areas

111

Table B1 Disease Comparison: Non-normalized

Disease Comparison Data: Population Normalized: Kruskal-Wallis One Way Analysis of Variance on Ranks
Group N
Missing
Median
25%
75%
Urban giardiasis
A
8
0
53.000
34.000
65.500
Urban cryptosporidiosis
B
8
0
5.000
4.000
5.500
Rural giardiasis
C
40
0
5.000
0.500
19.000
Rural cryptosporidiosis
D
40
0
1.500
0.000
4.000
H = 29.106 with 3 degrees of freedom.
P < 0.001
The differences in the median values among the treatment groups are greater than would be expected by chance; there
is a statistically significant difference (P < 0.001)
Disease Comparison Data: Population Normalized: All Pairwise Multiple Comparison Procedures
(Dunn's Method)
Comparison:

Difference of
Q
P<0.05
Ranks
A vs D
55.575
5.151
Yes
A vs C
38.750
3.592
Yes
A vs B
35.375
2.540
No
B vs D
20.200
1.872
No
B vs C
3.375
0.313
Do Not Test
C vs D
16.825
2.701
Do Not Test
Note: The multiple comparisons on ranks do not include an adjustment for ties.

112

Table B2 Disease Comparison: Population Normalized
Disease Comparison Data: Population Normalized: Kruskal-Wallis One Way Analysis of Variance on Ranks
Group
N
Missing Median
25%
75%
Urban giardiasis
A
8
0
0.00117
0.000950
0.00172
Urban cryptosporidiosis
B
8
0
0.000136
0.000113
0.000154
Rural giardiasis
C
40
0
0.000788
0.000164
0.00118
Rural cryptosporidiosis
D
40
0
0.000184
0.000
0.000347

H = 27.502 with 3 degrees of freedom
P < 0.001
The differences in the median values among the treatment groups are greater than would be expected by chance; there
is a statistically significant difference (P < 0.001)
Disease Comparison Data: Population Normalized: All Pairwise Multiple Comparison Procedures
(Dunn's Method)
Comparison:

Difference of
Q
P<0.05
Ranks
A vs B
49.875
3.581
Yes
A vs D
46.112
4.274
Yes
A vs C
24.013
2.226
No
C vs B
25.862
2.397
No
C vs D
22.100
3.548
Do Not Test
D vs B
3.763
0.349
Do Not Test
Note: The multiple comparisons on ranks do not include an adjustment for ties.

113

Table B3 Disease Comparison: Area Normalized
Disease Comparison Data: Area Normalized: Kruskal-Wallis One Way Analysis of Variance on Ranks
Group
N
Missing
Median
25%
Urban giardiasis
A
8
0
1.961
1.406
Urban cryptosporidiosis
B
8
0
0.184
0.167
Rural giardiasis
C
40
0
0.0392
0.00440
Rural cryptosporidiosis
D
40
0
0.0145
0.000

75%
2.522
0.250
0.172
0.0360

H = 38.089 with 3 degrees of freedom.
P < 0.001
The differences in the median values among the treatment groups are greater than would be expected by chance; there
is a statistically significant difference (P < 0.001)
Disease Comparison Data: Area Normalized: All Pairwise Multiple Comparison Procedures (Dunn's Method)
Comparison:
Difference of
Q
P<0.05
Ranks
A vs D
58.612
5.433
Yes
A vs C
43.263
4.010
Yes
A vs B
18.625
1.337
No
B vs D
39.987
3.706
Yes
B vs C
24.638
2.284
No
C vs D
15.350
2.464
No
Note: The multiple comparisons on ranks do not include an adjustment for ties.

114

Table B4 Disease Comparison: Population Density Normalized
Disease Comparison Data: Population Density Normalized: Kruskal-Wallis One Way Analysis of Variance on
Ranks
Group
N
Missing
Median
25%
75%
Urban giardiasis
A
8
0
0.0340
0.0230
0.0435
Urban cryptosporidiosis
B
8
0
0.00330
0.00252
0.00422
Rural giardiasis
C
40
0
0.0729
0.0165
0.146
Rural cryptosporidiosis
D
40
0
0.0156
0.000
0.0438
H = 15.228 with 3 degrees of freedom.
P = 0.002
The differences in the median values among the treatment groups are greater than would be expected by chance; there
is a statistically significant difference (P = 0.002)
Disease Comparison Data: Population Density Normalized: All Pairwise Multiple Comparison Procedures
(Dunn's Method)
Comparison:

Difference of
Q
P<0.05
Ranks
C vs B
33.438
3.099
Yes
C vs D
19.425
3.118
Yes
C vs A
9.688
0.898
No
A vs B
23.750
1.705
No
A vs D
9.737
0.903
Do Not Test
D vs B
14.013
1.299
Do Not Test
C vs B
33.438
3.099
Yes
Note: The multiple comparisons on ranks do not include an adjustment for ties.

115

APPENDIX C: Statistical comparison of occurrence data

116

Table C1 Occurrence Data: River Raisin Categorized
Occurrence data: Kruskal-Wallis One Way Analysis of Variance on Ranks
Group
N
Missing
Cryptosporidium occurrence:
A
15
0
River Raisin sites downstream of CSO outfalls
Giardia occurrence:
B
15
0
River Raisin sites downstream of CSO outfalls
Cryptosporidium occurrence:
C
24
0
River Raisin sites upstream of CSO outfalls
Giardia occurrence:
D
24
0
River Raisin sites upstream of CSO outfalls

Median
0.194

25%
0.149

75%
0.489

0.157

0.126

0.193

0.262

0.131

2.629

0.191

0.131

0.316

H = 3.815 with 3 degrees of freedom.
P< 0.282
The differences in the median values among the treatment groups are not great enough to exclude the possibility that
the difference is due to random sampling variability; there is not a statistically significant difference (P = 0.282)

117

Table C2 Occurrence Data: River Raisin Pooled
Occurrence data: Mann-Whitney Rank Sum Test

Cryptosporidium occurrence:
all River Raisin samples
Giardia occurrence:
all River Raisin samples

Group
A

N
39

Missing Median
0
0.240

25%
0.132

75%
0.635

B

39

0

0.128

0.276

n(small)=
39

n(large)= 39

0.161

Mann-Whitney U Statistic= 586.000
T = 1715.000
P = 0.082
The difference in the median values between the two groups is not great enough to exclude the possibility that the
difference is due to random sampling variability; there is not a statistically significant difference (P = 0.082)

118

Table C3 Occurrence Data: Grand River
Occurrence Data: Kruskal-Wallis One Way Analysis of Variance on Ranks
Group
N
Missing
Median
Cryptosporidium occurrence:
A
19
0
0.1000
North Beach Park
Giardia recovery :
B
19
0
0.109
North Beach Park
Cryptosporidium occurrence:
C
19
0
0.181
Riverside Park
Giardia recovery :
D
19
0
0.280
Riverside Park
Cryptosporidium occurrence:
E
22
0
0.176
Deer Creek Park
Giardia recovery :
F
22
0
0.112
Deer Creek Park

25%
0.0905

75%
0.133

0.0936

0.136

0.137

0.379

0.152

0.434

0.1000

0.510

0.1000

0.200

H = 25.884 with 5 degrees of freedom
P < 0.001
The differences in the median values among the treatment groups are greater than would be expected by chance;
there is a statistically significant difference (P = <0.001)

119

C3 (cont’d).
Occurrence Data: All Pairwise Multiple Comparison Procedures (Dunn's Method)
Comparison:
Difference of
Q
P<0.05
Ranks
D vs A
44.789
3.969
Yes
D vs B
38.316
3.395
Yes
D vs F
30.713
2.819
No
D vs E
12.258
1.125
Do Not Test
D vs C
8.579
0.760
Do Not Test
C vs A
36.211
3.209
Yes
C vs B
29.737
2.635
No
C vs F
22.134
2.032
Do Not Test
C vs E
3.679
0.338
Do Not Test
E vs A
32.531
2.986
Yes
E vs B
26.057
2.392
Do Not Test
E vs F
18.455
1.760
Do Not Test
F vs A
14.077
1.292
No
F vs B
7.603
0.698
Do Not Test
B vs A
6.474
0.574
Do Not Test
Note: The multiple comparisons on ranks do not include an adjustment for ties.

120

Table C4 Occurrence Data: Grand River and MARB
Occurrence Data: Kruskal-Wallis One Way Analysis of Variance on Ranks
Group
N
Missing
Cryptosporidium occurrence:
A
8
0
Market Avenue Retention Basin (MARB)
Giardia recovery :
B
8
0
Market Avenue Retention Basin (MARB)
Cryptosporidium occurrence:
C
19
0
North Beach Park
Giardia recovery :
D
19
0
North Beach Park
Cryptosporidium occurrence:
E
19
0
Riverside Park
Giardia recovery :
F
19
0
Riverside Park
Cryptosporidium occurrence:
G
22
0
Deer Creek Park
Giardia recovery :
H
22
0
Deer Creek Park

Median
5.000

25%
3.333

75%
18.333

1541.667

732.400

2040.000

0.1000

0.0905

0.133

0.109

0.0936

0.136

0.181

0.137

0.379

0.280

0.152

0.434

0.176

0.1000

0.510

0.112

0.1000

0.200

H = 61.849 with 7 degrees of freedom.
P < 0.001
The differences in the median values among the treatment groups are greater than would be expected by chance;
there is a statistically significant difference (P = <0.001)

121

Table C4 (cont’d).
Occurrence Data: All Pairwise Multiple Comparison Procedures (Dunn's Method)
Comparison:
Difference of Ranks
Q
P<0.05
B vs C
94.395
5.684
Yes
B vs D
87.921
5.294
Yes
B vs H
80.318
4.937
Yes
B vs G
61.545
3.783
Yes
B vs E
58.184
3.504
Yes
B vs F
49.526
2.982
No
B vs A
9.063
0.460
Do Not Test
A vs C
85.332
5.138
Yes
A vs D
78.859
4.748
Yes
A vs H
71.256
4.380
Yes
A vs G
52.483
3.226
Yes
A vs E
49.122
2.958
No
A vs F
40.464
2.437
Do Not Test
Note: The multiple comparisons on ranks do not include an adjustment for ties.

122

Table C4 (cont’d).
Occurrence Data: All Pairwise Multiple Comparison Procedures (Dunn's Method)
Comparison:
Difference of Ranks
Q
P<0.05
F vs C
44.868
3.510
Yes
F vs D
38.395
3.003
No
F vs H
30.792
2.495
Do Not Test
F vs G
12.019
0.974
Do Not Test
F vs E
8.658
0.677
Do Not Test
E vs C
36.211
2.832
No
E vs D
29.737
2.326
Do Not Test
E vs H
22.134
1.794
Do Not Test
E vs G
3.361
0.272
Do Not Test
G vs C
32.849
2.662
Do Not Test
G vs D
26.376
2.137
Do Not Test
G vs H
18.773
1.580
Do Not Test
H vs C
14.077
1.141
Do Not Test
H vs D
7.603
0.616
Do Not Test
D vs C
6.474
0.506
Do Not Test
Note: The multiple comparisons on ranks do not include an adjustment for ties.

123

Table C5 Occurrence Data: River Raisin (Pooled) and Grand River
Occurrence Data: Kruskal-Wallis One Way Analysis of Variance on Ranks
Group
N
Missing
Cryptosporidium occurrence:
A
19
0
North Beach Park
Giardia recovery :
B
19
0
North Beach Park
Cryptosporidium occurrence:
C
41
0
Deer Creek Park and Riverside Park pooled samples
Giardia recovery :
D
41
0
Deer Creek Park and Riverside Park pooled samples
Cryptosporidium occurrence:
E
39
0
all River Raisin samples
Giardia occurrence:
F
39
0
all River Raisin samples

Median
0.1000

25%
0.0905

75%
0.133

0.109

0.0936

0.136

0.181

0.1000

0.500

0.143

0.1000

0.315

0.240

0.132

0.635

0.161

0.128

0.276

H = 30.060 with 5 degrees of freedom.
P < 0.001
The differences in the median values among the treatment groups are greater than would be expected by chance; there is a
statistically significant difference (P = <0.001)

124

Table C5 (cont’d).

Occurrence Data: All Pairwise Multiple Comparison Procedures (Dunn's Method)
Comparison:
Difference of
Q
P<0.05
Ranks
E vs A
72.244
4.506
Yes
E vs B
62.455
3.896
Yes
E vs D
25.779
2.011
No
E vs F
19.397
1.495
Do Not Test
E vs C
15.133
1.181
Do Not Test
C vs A
57.112
3.591
Yes
C vs B
47.322
2.976
Yes
C vs D
10.646
0.841
Do Not Test
C vs F
4.265
0.333
Do Not Test
F vs A
52.847
3.296
Yes
F vs B
43.057
2.686
No
F vs D
6.381
0.498
Do Not Test
D vs A
46.465
2.922
No
D vs B
36.676
2.306
Do Not Test
B vs A
9.789
0.527
Do Not Test
Note: The multiple comparisons on ranks do not include an adjustment for ties.

125

APPENDIX D: Statistical analysis of detection limits

126

Table D1 Detection Limits: River Raisin Categorized
Detection limits: Mann-Whitney Rank Sum Test
Group
N
River Raisin
A
15
sites
downstream of
CSO outfalls
River Raisin
B
24
sites upstream
of CSO outfalls

Missing
0

Median
0.157

25%
0.126

75%
0.193

0

0.168

0.119

0.316

Mann-Whitney U Statistic= 153.000
T = 273.000
n(small)= 15
n(large)= 24
P = 0.444
The difference in the median values between the two groups is not great enough to exclude the possibility that the difference is
due to random sampling variability; there is not a statistically significant difference (P = 0.444)

127

Table D2 Detection Limits: Grand River
Detection limits: Kruskal-Wallis One Way Analysis of Variance on Ranks
Group
N
Missing
Median
North Beach
A
19
0
0.1000
Park
Riverside Park
B
19
0
0.1000
Deer Creek
C
22
0
0.1000
Park

25%
0.0915

75%
0.129

0.0948
0.0944

0.133
0.129

H = 0.715 with 2 degrees of freedom
P = 0.700
The differences in the median values among the treatment groups are not great enough to exclude the possibility that the
difference is due to random sampling variability; there is not a statistically significant difference (P = 0.700)

128

Table D3 Detection Limits: Grand River and MARB
Detection limits: Kruskal-Wallis One Way Analysis of Variance on Ranks
Group
N
Missing
Median
North Beach
A
19
0
0.1000
Park
Riverside Park
B
19
0
0.1000
Deer Creek Park C
22
0
0.1000
MARB
D
8
0
5.000

25%
0.0915

75%
0.129

0.0948
0.0944
3.333

0.133
0.129
18.333

H = 21.495 with 3 degrees of freedom.
P <0.001
The differences in the median values among the treatment groups are greater than would be expected by chance; there is a
statistically significant difference (P = <0.001)
Detection limits: All Pairwise Multiple Comparison Procedures (Dunn's Method)
Comparison:
Difference of
Q
P<0.05
Ranks
D vs A
36.500
4.380
Yes
D vs C
33.795
4.140
Yes
D vs B
31.737
3.808
Yes
B vs A
4.763
0.742
No
B vs C
2.059
0.332
Do Not Test
C vs A
2.705
0.437
Do Not Test
Note: The multiple comparisons on ranks do not include an adjustment for ties.

129

Table D4 Detection Limits: Grand River and River Raisin
Detection limits: Kruskal-Wallis One Way Analysis of Variance on Ranks
Group
N
Missing
Median
Deer Creek Park A
41
0
0.1000
and Riverside
Park
River Raisin
B
43
0
0.161
total sites
North Beach
C
19
0
0.1000
Park

25%
0.0947

75%
0.130

0.128

0.241

0.0915

0.129

H = 34.233 with 2 degrees of freedom.
P <0.001
The differences in the median values among the treatment groups are greater than would be expected by chance; there is a
statistically significant difference
(P <0.001)
Detection limits: All Pairwise Multiple Comparison Procedures (Dunn's Method)
Comparison:
Difference of
Q
P<0.05
Ranks
B vs C
37.729
4.584
Yes
B vs A
33.402
5.122
Yes
A vs C
4.327
0.522
No
Note: The multiple comparisons on ranks do not include an adjustment for ties.

130

APPENDIX E: Statistical analysis of matrix spike recovery efficiency

131

Table E1 Matrix Spike: River Raisin Categorized Data
Matrix Spike Sample Recovery Efficiency: Kruskal-Wallis One Way Analysis of Variance on Ranks
Group
N
Missing
Median
25%
Cryptosporidium recovery:
A
5
0
54.400
34.400
River Raisin sites downstream of CSO outfalls
Giardia recovery:
B
5
0
11.800
8.867
River Raisin sites downstream of CSO outfalls
Cryptosporidium recovery:
C
9
0
39.100
24.612
River Raisin sites upstream of CSO outfalls
Giardia recovery:
D
9
0
3.800
0.000
River Raisin sites upstream of CSO outfalls

75%
63.4
31.225
75.163
7.075

H =19.276 with 3 degrees of freedom
P <0.001
The differences in the median values among the treatment groups are greater than would be expected by chance; there is a
statistically significant difference (P <0.001)
Matrix Spike Sample Recovery Efficiency: All Pairwise Multiple Comparison Procedures (Dunn's Method)
Comparison:
Difference of
Q
P<0.05
Ranks
A vs D
15.333
3.342
Yes
A vs B
8.600
1.653
No
A vs C
0.111
0.0242
Do Not Test
C vs D
15.222
3.926
Yes
C vs B
8.489
1.850
Do Not Test
B vs D
6.733
1.468
No
Note: The multiple comparisons on ranks do not include an adjustment for ties.

132

Table E2 Matrix Spike: River Raisin Pooled Data
Matrix Spike Sample Recovery Efficiency: Mann-Whitney Rank Sum Test
Group
Cryptosporidium A
recovery :
all River Raisin
associated
matrix spike
samples
Giardia
B
recovery :
all River Raisin
associated
matrix spike
samples

N
14

Missing
0

Median
46.750

25%
26.150

75%
73.300

14

0

6.735

2.750

12.400

Mann-Whitney U Statistic= 8.000
T = 293.000
P <0.001

n(small)= 14

n(large)= 14

The difference in the median values between the two groups is greater than would be expected by chance; there is a
statistically significant difference (P = <0.001)

133

Table E3 Matrix Spike: Grand River
Matrix Spike Sample Recovery Efficiency: t - Test
Group
Cryptosporidium A
recovery :
all recreational
Grand River site
associated
matrix spike
samples
Giardia
B
recovery :
all recreational
Grand River site
associated
matrix spike
samples

N
5

Missing
0

Mean
69.940

Std Dev
9.198

SEM
4.113

5

0

52.580

23.344

10.440

Difference
17.360
t = 1.547 with 8 degrees of freedom.
P = 0.160
95 percent confidence interval for difference of means: -8.516 to 43.236
The difference in the mean values of the two groups is not great enough to reject the possibility that the difference is due to
random sampling variability. There is not a statistically significant difference between the input groups (P = 0.160).
Power of performed test with alpha = 0.050: 0.173
The power of the performed test (0.173) is below the desired power of 0.800.
Less than desired power indicates you are less likely to detect a difference when one actually exists. Negative results should be
interpreted cautiously.

134

Table E4 Matrix Spike: Grand River and MARB
Matrix Spike Sample Recovery Efficiency: One Way Analysis of Variance
Group
N
Name
Cryptosporidium recovery :
A
7
all MARB associated matrix spike samples
Giardia recovery :
B
7
all MARB associated matrix spike samples
Cryptosporidium recovery :
C
5
all recreational Grand River site associated
matrix spike samples
Giardia recovery :
D
5
all recreational Grand River site associated
matrix spike samples

Missing

Mean

Std Dev

SEM

0

33.714

22.625

8.552

0

42.071

24.852

9.393

0

69.940

9.198

4.113

0

52.580

23.344

10.440

Source of Variation
DF
SS
MS
F
Between Groups
3
4183.007
1394.336
3.000
Residual
20
9295.383
464.769
Total
23
13478.390
P = 0.055
The differences in the mean values among the treatment groups are not great enough to exclude the possibility that the
difference is due to random sampling variability; there is not a statistically significant difference (P = 0.055).
Power of performed test with alpha = 0.050: 0.429
The power of the performed test (0.429) is below the desired power of 0.800.
Less than desired power indicates you are less likely to detect a difference when one actually exists. Negative results should be
interpreted cautiously.

135

Table E5 Matrix Spike: Grand River and River Raisin
Matrix Spike Sample Recovery Efficiency: Kruskal-Wallis One Way Analysis of Variance on Ranks
Group
N
Missing
Median
25%
Cryptosporidium recovery :
A
5
0
73.000
66.575
all recreational Grand River site associated matrix
spike samples
Giardia recovery :
B
5
0
46.900
38.725
all recreational Grand River site associated matrix
spike samples
Cryptosporidium recovery :
C
14
0
46.750
26.150
all River Raisin associated matrix spike samples
Giardia recovery :
D
14
0
6.735
2.750
all River Raisin associated matrix spike samples

75%
74.800

67.125

73.300
12.400

H = 24.144 with 3 degrees of freedom.
P<0.001
The differences in the median values among the treatment groups are greater than would be expected by chance; there is a
statistically significant difference (P = <0.001)

136

Table E5 (cont’d).
Matrix Spike Sample Recovery Efficiency: All Pairwise Multiple Comparison Procedures (Dunn's Method)
Comparison:
Difference of
Q
P<0.05
Ranks
A vs D
22.786
3.936
Yes
A vs C
6.500
1.123
No
A vs B
5.400
0.768
Do Not Test
B vs D
17.386
3.003
Yes
B vs C
1.100
0.190
Do Not Test
C vs D
16.286
3.877
Yes
Note: The multiple comparisons on ranks do not include an adjustment for ties.
A result of "Do Not Test" occurs for a comparison when no significant difference is found between two means that enclose that
comparison. Note that not testing the enclosed means is a procedural rule, and a result of Do Not Test should be treated as if
there is no significant difference between the means, even though one may appear to exist.

137

APPENDIX F: Statistical analysis of ongoing precision and recovery (OPR) efficiency

138

Table F1 OPR: River Raisin
OPR: t - Test

Cryptosporidium
recovery :
all River Raisin
site associated
OPR samples
Giardia recovery :
all River Raisin
site associated
OPR samples

Group
A

N
4

Missing
0

Mean
41.300

Std Dev
7.771

SEM
3.885

B

4

0

42.063

10.535

5.268

Difference
-0.763
t = -0.116 with 6 degrees of freedom.
P = 0.911
95 percent confidence interval for difference of means: -16.779 to 15.254
The difference in the mean values of the two groups is not great enough to reject the possibility that the difference is
due to random sampling variability. There is not a statistically significant difference between the input groups (P =
0.911).

Power of performed test with alpha = 0.050: 0.050
The power of the performed test (0.050) is below the desired power of 0.800.
Less than desired power indicates you are less likely to detect a difference when one actually exists. Negative results
should be interpreted cautiously.

139

Table F2 OPR: Grand River
OPR: t - Test
Group
Cryptosporidium A
recovery :
all recreational
Grand River site
associated OPR
samples
Giardia
B
recovery :
all recreational
Grand River site
associated OPR
samples

N
23

Missing
0

Mean
55.230

Std Dev
11.741

SEM
2.448

23

0

43.939

15.681

3.270

Difference
11.291
t = 2.764 with 44 degrees of freedom.
P = 0.008
95 percent confidence interval for difference of means: 3.059 to 19.523
The difference in the mean values of the two groups is greater than would be expected by chance; there is a statistically
significant difference between the input groups (P = 0.008).
Power of performed test with alpha = 0.050: 0.720

140

Table F3 OPR: Grand River and MARB
OPR: One Way Analysis of Variance
Group
Name
A

Cryptosporidium recovery :
all MARB associated OPR samples
Giardia recovery :
B
all MARB associated OPR samples
Cryptosporidium recovery :
C
all recreational Grand River site associated OPR
samples
Giardia recovery :
D
all recreational Grand River site associated OPR
samples

N

Missing

Mean

Std Dev

SEM

5

0

47.320

13.601

6.083

5

0

67.020

17.257

7.718

23

0

55.230

11.741

2.448

23

0

43.939

15.681

3.270

Source of Variation
DF
SS
MS
F
Between Groups
3
2909.014
969.671
4.861
Residual
52
10373.439
199.489
Total
55
13282.454
P = 0.005
The differences in the mean values among the treatment groups are greater than would be expected by chance; there is a
statistically significant difference (P = 0.005).
Power of performed test with alpha = 0.050: 0.799

141

Table F3 (cont’d).

OPR: All Pairwise Multiple Comparison Procedures (Tukey Test):
Comparison
Diff of Means
p
q
P
P<0.050
B vs. D
23.081
4
4.684
0.009
Yes
B vs. A
19.700
4
3.119
0.135
No
B vs. C
11.790
4
2.392
0.338
Do Not Test
C vs. D
11.291
4
3.834
0.044
Yes
C vs. A
7.910
4
1.605
0.670
Do Not Test
A vs. D
3.381
4
0.686
0.962
No
Note: A result of "Do Not Test" occurs for a comparison when no significant difference is found between two means that
enclose that comparison. Note that not testing the enclosed means is a procedural rule, and a result of Do Not Test should be
treated as if there is no significant difference between the means, even though one may appear to exist.

142

Table F4 OPR: Grand River and River Raisin
OPR: One Way Analysis of Variance
Group
Name
A

Cryptosporidium recovery :
all recreational Grand River site associated OPR
samples
Giardia recovery :
B
all recreational Grand River site associated OPR
samples
Cryptosporidium recovery :
C
all River Raisin site associated OPR samples
Giardia recovery :
D
all River Raisin site associated OPR samples

N

Missing

Mean

Std Dev

SEM

23

0

55.230

11.741

2.448

23

0

43.939

15.681

3.270

4

0

41.300

7.771

3.885

4

0

42.063

10.535

5.268

Source of Variation
DF
SS
MS
F
Between Groups
3
1893.032
631.011
3.523
Residual
50
8956.390
179.128
Total
53
10849.422
P = 0.021
The differences in the mean values among the treatment groups are greater than would be expected by chance; there is a
statistically significant difference (P = 0.021).
Power of performed test with alpha = 0.050: 0.585

143

Table F4 (cont’d).
OPR: All Pairwise Multiple Comparison Procedures (Holm-Sidak method)
Comparison
Diff of Means
t
Unadjusted P
Critical Level
A vs. B
11.291
2.861
0.006
0.009
A vs. C
13.930
1.921
0.060
0.010
A vs. D
13.168
1.816
0.075
0.013
B vs. C
2.639
0.364
0.717
0.017
B vs. D
1.877
0.259
0.797
0.025
D vs. C
0.763
0.0806
0.936
0.050
Note: The multiple comparisons on ranks do not include an adjustment for ties.

144

P<0.050
Yes
No
No
No
No
No

REFERENCES

145

REFERENCES
1. Applebee, A.J. Thompson, R.C., Olson, M.E., 2005. Giardia and
Cryptosporidium in mammalian wildlife – current status and future needs. Trends
in Parasitology 21: 370-376.
2. Arnone, R.D., Walling. J.P., 2006. Evaluating Cryptosporidium and Giardia
concentrations in combined sewer overflow. Journal of Water and Health
4(2):157-65.
3. Atwill, E.R., Pereira, M.D., Alonso, L.H., Elmi, C., Epperson, W.B., Smith, R.,
Riggs, W., Carpenter, L.V., Dargatz, D.A., Hoar, B., 2006. Environmental Load
of Cryptosporidium parvum Oocysts from Cattle Manure in Feedlots from the
Central and Western United States. Journal of Environmental Quality 35:200-206.
4. Baxby, D., Hart, C.A., Blundell, N., 1985. Shedding of oocysts by
immunocompetent individuals with cryptosporidiosis. Journal of Hygiene
95:703–709.
5. Brodsky, R.E., Spencer, H.C., Schultz M.G., 1974. Giardiasis in American
travelers to the Soviet Union. Journal of Infectious Diseases 130:319–323.
6. Centers for Disease Control and Prevention, 2011. Cryptosporidiosis
(Cryptosporidium spp.) case definition. Atlanta, GA: US Department of Health
and Human Services, CDC. Available at
http://www.cdc.gov/osels/ph_surveillance/nndss/casedef/cryptosporidiosis_curren
t.htm. Accessed May 5, 2011.
7. Centers for Disease Control and Prevention, 2011. Giardiasis (Giardia lamblia)
case definition. Atlanta, GA: US Department of Health and Human Services,
CDC. Available at
http://www.cdc.gov/osels/ph_surveillance/nndss/print/giardiasis_current.htm.
Accessed May 5, 2011.
8. Centers for Disease Control and Prevention, 2010. Cryptosporidiosis Surveillance
– United States 2006-2008 and Giardiasis Surveillance – United States 2006-2008.
In: Surveillance Summaries, June 11, 2010. MMWR 59 (No. SS-6).
9. Centers for Disease Control and Prevention, 2007. Cryptosporidiosis
Surveillance – United States 2003-2005 and Giardiasis Surveillance – United
States 2003-2005. In: Surveillance Summaries, September 7, 2007. MMWR 56
(No. SS-7).

146

10. Centers for Disease Control and Prevention, 2006. Summary of notifiable diseases
- United States, 2006. MMWR 55(No. 53).
11. Centers for Disease Control and Prevention, 2005. Cryptosporidiosis
Surveillance – United States 1999-2002 and Giardiasis Surveillance – United
States 1998-2002. In: Surveillance Summaries, January 28, 2005. MMWR 54
(No. SS-1).
12. Centers for Disease Control and Prevention, 2000. Giardiasis Surveillance --United States, 1992—1997. In: Surveillance Summaries August 11 2000.
MMWR 49 (No SS-07).
13. Centers for Disease Control and Prevention, 1998. Summary of notifiable diseases,
United States, 1998. MMWR 47(53): 5.
14. Centers for Disease Control and Prevention, 1998. Summary of notifiable diseases,
United States, 1997. MMWR 46(54): 5.
15. Centers for Disease Control and Prevention, 1996. Summary of notifiable diseases,
United States, 1996. MMWR 45(53):vii.
16. Centers for Disease Control and Prevention, 1993. Summary of notifiable
diseases - United States, 1993. Published October 21, 1994, for MMWR 42(53).
17. Cox, P., Griffith, M., Angles, M., Deere, D., Ferguson, C., 2005. Concentrations
of Pathogens and Indicators in Animal Feces in the Sydney Watershed. Applied
and Environmental Microbiology 71: 5929-5934
18. Cryptosporidium and Cryptosporidiosis, 2
2008 CRC Press. Boca Raton, Fl.

nd

edition. Fayer, R and Xiao, L., ed.

19. Donovan, E., Unice, K., Roberts, J.D., Harris ,M., Finley, B., 2008. Risk of
gastrointestinal disease associated with exposure to pathogens in the water of the
lower Passaic River. Applied and Environmental Microbiology 74: 994-1003.
20. Erickson, D.L., 1995. Rural land use and land cover change: Implications for
local planning in the River Raisin watershed. Land Use Policy. 12(3): 223-236
21. Federal Register, 2004. Water quality standards for coastal and Great Lakes
recreation waters; final rule. Federal Register 69(220): 67217-67243.
22. Feng, Y., Alderisio, K.A., Yang, W., Blancero, L.A., Kuhne, W.G., Nadareski,
C.A., Reid, M., Xiao, L., 2007. Cryptosporidium Genotypes in Wildlife from a
New York Watershed. Applied and Environmental Microbiology 73: 6475-6483.

147

23. Giardia: From Molecules to Disease. Thompson, R.C.A, Reynoldson, J.A.,
Lymbery, A.J., ed. 1994. CAB International, United Kingdom.
24. Gibson, C.J., Stadterman, K.L., States, S., Sykora, J., 1998. Combined sewer
overflows: a source of Cryptosporidium and Giardia? Water Science and
Technology 38(12): 67-72.
25. Haas, C.N., Crockett, C.S., Rose, J.B., Gerba, C.P., Fazil, A.M., 1996. Assessing
the risk posed by oocysts in drinking water. American Water Works Association
Journal 88(9): 131-136.
26. Heitman, T.L., Frederick, L.M., Viste, J.R., Guselle, N.J., Morgan, U.M.,
Thompson, R.C.A., Olson, M.E., 2002. Prevalence of Giardia and
Cryptosporidium and characterization of Cryptosporidium spp. isolated from
wildlife, human, and agricultural sources in the North Saskatchewan River Basin
in Alberta, Canada. Canadian Journal of Microbiology 48:530-541.
27. Hunter, P.R. Thompson, R.C., 2005. The zoonotic transmission of Giardia and
Cryptosporidium. International Journal for Parasitology. 35(11-12):1181-90.
28. Hunter, P.R., Chalmers, R.M., Syed, Q., Hughes, L.S., Woodhouse, S., Swift, L.,
2003. Foot and Mouth Disease and Cryptosporidiosis: Possible interaction
between two emerging infectious diseases. Emerging Infectious Diseases
9(1):109-112.
29. Hutchison, M.L., Walters, L. D., Moore, A., Avery, S.M., 2005. Declines of
zoonotic agents in liquid livestock wastes stored in batches on-farm. Journal of
Applied Microbiology 99:58-65.
30. Jokipii, A.M.M., Hemila, M., Jokipii, L., 1985. Prospective study of acquisition
of Cryptosporidium, Giardia lamblia, and gastrointestinal illness. Lancet,
2(8453):487–489.
31. Lake, I.R., Nichols, G., Harrison, F.C.D., Bentham, G., Kovats, R.S., Grundy, C.,
Hunter, P.R., 2009. Using infectious intestinal disease surveillance data to explore
illness aetiology; a cryptosporidiosis case study. Health and Place 15(1):333-339.
32. Lake, I.R., Harrison, F.C.D., Chalmers, R.M., Bentham, G., Nichols, G., Hunter,
P.R., Kovats, R.S., Grundy, C., 2007. Case-control study of environmental and
social factors influencing cryptosporidiosis. European journal of epidemiology
22(11): 805-811.
33. Levine, A.D., Harwood, V.J., Farrah, S.R., Scott, T.M., Rose, J.B., 2008.
Pathogen and indicator organism reduction through secondary effluent filtration:
implications for reclaimed water production. Water Environment Research 80(7):
596-608.

148

34. Michigan Department of Community Health. Personal communication.
35. Michigan Geographic Data Library. http://www.mcgi.state.mi.us/mgdl/
Accessed January 2010
36. Michigan Department of Environmental Quality (MDEQ). Online Services:
Combined Sewer Overflow and Sanitary Sewer Overflow System.
http://www.deq.state.mi.us/csosso Accessed October 2008.
37. Odoi, A., Martin, S.W., Michel, P., Middleton, D., Holt, J., Wilson, J., 2004.
Investigation of clusters of giardiasis using GIS and a spatial scan statistic.
International Journal of Health Geographics 3(11).
38. Okhuysen, P.C., Chappell, C.L., Crabb, J.H., Sterling, C.R., DuPont, H.L., 1999.
Virulence of three distinct Cryptosporidium parvum isolates for healthy adults.
The Journal of Infectious Diseases 180(4): 1275-1281.
39. Olson, M.E., Guselle, N.J., O'Handley, R.M., Swift, M.L., McAllister, T.A.,
Jelinski, M.D., Morck.D.W., 1997. Cryptosporidium in dairy calves in British
Columbia. The Canadian Veterinary Journal 38(11):703-706.
40. Olson, M.E., Thorlakson, C.L., Deselliers, L., Morck.D.W., McAllister, T.A.,
1997. Giardia and Cryptosporidium in Canadian farm animals. Veterinary
Parasitology 68: 375-381.
41. Ortega, Y.R., Adam, R.D., 1997. Giardia: Overview and Update. Clinical
Infectious Diseases 25:545-549.
42. Pollock, K.G.J., Ternent, H.E., Mellor, D.J., Chalmers, R.M., Smith, H.V.,
Ramsay, C.N., Innocent, G.T., 2010. Spatial and Temporal Epidemiology of
Sporadic Human Cryptosporidiosis in Scotland. Zoonoses and Public Health 57:
487-492.
43. Rendtorff, R.C., 1954. The experimental transmission of human intestinal
protozoan parasites: Giardia intestinalis cysts given in capsules. American
Journal of Hygiene 59: 209–220.
44. Rose, J.B., 2007. Water reclamation, reuse and public health. Water Science &
Technology 55(1-2): 275–282.
45. Rose, J.B., Haas, C.N., Regli, S., 1991. Risk assessment and control of
waterborne giardiasis. American Journal of Public Health. 81(6): 709-13
46. Roxstrom-Lindquist, K.,Palm, Reiner, D., Ringqvist, D.E., Svard, .G., 2006.
Giardia immunity – an update. Trends in Parasitology 22(1): 26-31.

149

47. Smerdon, W.J., Nichols, T., Chalmers, R.M., Heine, H., Reacher, M.H., 2003.
Foot and mouth disease in livestock and reduced cryptosporidiosis in humans,
England and Wales. Emerging infectious diseases. 9(1): 22-28.
48. States, S., Stadterman, K.L., Ammon, L., Vogel, P., Baldizar, J., Wright, D.,
Conley, L., Sykora, J., 1997. Protozoa in river water: sources, occurrence, and
treatment. Journal of the American Water Works Association 89(9):74-87.
49. Thompson, R.C.A., Hopkins, R.M., Homan, W.L., 2000. Nomenclature and
genetic groupings of Giardia infecting mammals. Parasitology Today 16: 210213
50. U.S. Army Corps of Engineers (USACE), 1972. Comprehensive Water Resource
Study of the Grand River Basin Michigan.
51. U.S. Census Bureau. http://www.census.gov/ Access date: January 2010.
52. U.S. Department of the Interior, U.S. Geological Survey. USGS Surface-Water
Data for Michigan. http://waterdata.usgs.gov/mi/nwis/sw? Site 04119000
Access date: November 2008
53. U.S. Environmental Protection Agency (U.S.EPA), 2008. Child-specific exposure
factors handbook. National Center for Environmental Assessment, Washington,
DC; EPA/600/RP-060/096F.
54. U.S. Environmental Protection Agency (U.S.EPA ), 2005. Method 1623:
Cryptosporidium and Giardia in water by filtration/IMS/FA. United States
Environmental Protection Agency. Office of Water. Washington, D.C.
EPA 815-R-05-002
55. U.S. Environmental Protection Agency (U.S.EPA), 2004. Report to Congress.
Impacts and control of CSOs and SSOs. EPA 833-R-04-001. U.S.
Environmental Protection Agency, Washington, D.C.
56. Wade, T.J., Calderon, R.L., Brenner, K.P., Sams, E., Beach, M., Haugland, R.,
Wymer, L., Dufour, A.P., 2008. High sensitivity of children to swimmingassociated gastrointestinal illness. Epidemiology 19(3): 375–383.
57. World Health Organization. Guidelines for Drinking-Water Quality, 2nd Edition.
1996. Addendum: Microbiological agents in drinking water. Protozoan Parasites
(Cryptosporidium, Giardia, Cyclospora). World Health Organization, Geneva.
58. Xiao, L., Herd, R.P., 1994. Infection patterns of Cryptosporidium and Giardia in
Calves. Parasitology Today 55:254-262.

150

59. Zisan, E., Alabay, B.M., Fidanci, H., Düzgün, A., Cerci, H., 1998. Prevalence of
Cryptosporidium spp. infection and its relation to other enteric pathogens
(Escherichia coli K99 and rotavirus) in cattle in Ankara, Turkey. Turkish Journal
of Veterinary and Animal Sciences 22: 453-457.

151