THE EFFECT OF THE COLON CANCER CHECK PROGRAM ON COLORECTAL CANCER SCREENING IN ONTARIO

by

Gladys Honein

A thesis submitted in conformity with the requirements for the degree of Doctor of Philosophy in Health Services Research Graduate Institute of Health Policy Management and Evaluation University of Toronto

© Copyright by Gladys Honein, 2012

Thesis Abstract Thesis Title: The Effect of the Colon Cancer Check Program on Colorectal Cancer Screening in Ontario Student: Gladys Honein Degree: Doctor of Philosophy in Health Services Research Graduate Department: Institute of Health Policy Management and Evaluation University: University of Toronto Year of convocation: 2012 Background: This thesis is composed of three studies testing the effect of the Colon Cancer Check (CCC) program, the organized screening program for colorectal cancer in Ontario, on screening participation. In the first paper, we described the trends of participation to Fecal Occult Blood Test (FOBT) and endoscopy, and the trend of ‘up-to-date’ consistent with guidelines, overall and stratified by demographic characteristics between 2005 and 2011. In the second paper, we tested the effect of physician’s recommendation on FOBT participation and disparities in participation. In the third paper, we measured the effect of the CCC program on FOBT participation using an interrupted time series. Methods: We identified six annual cohorts of individuals eligible for CRC screening in Ontario between 2005 and 2011 by linking the Registered Persons Database to Ontario Health Insurance Plan and 2006 Census from Statistics Canada. We used descriptive statistics to describe the trends of participation. The effect of physician’s recommendation on screening participation was tested using multiple logistic regression analysis. The effect of the CCC program on FOBT participation was tested using segmented regression analysis. Results: An increasing trend in FOBT participation and ‘up-to-date’ status was observed across all demographic characteristics. The disparity gaps persisted over time by gender, income, recent registrant and age. The rural/urban gap was removed. Physician’s recommendation tripled the likelihood of FOBT participation (prevalence rate ratio=3.23, CI= 3.22-3.24) and mitigated ii

disparities. The CCC led to a temporary increase in level (8.2‰ person-month) in FOBT participation followed by a decline in trend and then a plateau. The increase in level was significant across all population sub-groups. Conclusions: We found that CRC screening has increased in Ontario across all subgroups of the population but remained suboptimal. Disparities in screening participation were identified. Proposed strategies to improve performance include interventions to increase the rate of physician’s recommendation at the practice level, tailored interventions to motivate under-users and public media campaigns.

iii

Acknowledgments

There are a number of people I would like to thank whose contributions made this thesis possible: First, I would like to thank Arlene Bierman, my supervisor and mentor during the PhD. Her insightful comments and advice kept me on track and encouraged me to go forward. To my Thesis Committee (Rahim Moineddin, David Urbach, and Arlene Bierman) for their input and guidance at every stage of this thesis. To my advisory committee who kept me on my toes with their pragmatic and insightful comments: Nancy Baxter, Linda Rabeneck, Lawrence Paszat, and Jill Tinmouth. To the Institute for Clinical Evaluative Sciences (ICES) for providing access to data and facilities to conduct the analysis. To ICES personnel who were supportive at every step of the analysis: Refik Saskin, Pam Slaughter, Lucy Gerry and Park Jin. To the Canadian Institutes of Health Research, Institute for Gender and Health for supporting me over the last three years of my thesis. To the Institute of Health Policy Management and Evaluation (Rhonda Cockerill, Whitney Berta, Christina Lopez, Mariana Vardaei) for their administrative support and guidance for all thesis related matters. To my family members for their understanding and encouragement all along. And most important to my wonderful loving husband, Mounir AbouHaidar who provided advice, support and unequivocal support at every stage of this arduous journey.

iv

Table of Contents

Table
of
Contents
 Acknowledgments ................................................................................................................. iv
 Table
of
Contents.................................................................................................................... v
 List
of
Tables .......................................................................................................................... ix
 List
of
Figures .......................................................................................................................... x
 List
of
Appendices .................................................................................................................. xi
 CHAPTER
1:
BACKGROUND .....................................................................................................1
 Colorectal
cancer
epidemiology...........................................................................................................................................1
 Risk
factors
for
colorectal
cancer........................................................................................................................................2
 Screening
for
colorectal
cancer............................................................................................................................................3
 Screening
testing
in
Canada...................................................................................................................................................4
 Colon
Cancer
Check
program:
an
organized
screening
program..........................................................................5
 The
Colon
Cancer
Check
program
in
Ontario.................................................................................................................6
 Thesis
overview ..........................................................................................................................................................................8
 Specific
objectives......................................................................................................................................................................9
 Significance
of
this
thesis........................................................................................................................................................9
 Conceptual
framework ......................................................................................................................................................... 10
 A
literature
map....................................................................................................................................................................... 13
 Intrapersonal
level.................................................................................................................................................................. 13
 Interpersonal
level .................................................................................................................................................................. 15
 Social
level .................................................................................................................................................................................. 17
 Political
level.............................................................................................................................................................................. 17
 Ethics
statement ...................................................................................................................................................................... 18
 Figure
for
background
chapter ......................................................................................................................................... 19


CHAPTER
2:
COHORT
IDENTIFICATION
AND
METHODS .......................................................... 20
 Study
design .............................................................................................................................................................................. 20
 Data
sources .............................................................................................................................................................................. 21
 v

Registered
Persons
Database
(RPDB) ............................................................................................................................ 21
 Ontario
Health
Insurance
Plan
(OHIP) .......................................................................................................................... 21
 The
Canadian
Institute
of
Health
Information­
Discharge
Abstracts
Database
(CIHI­DAD)................. 21
 The
Ontario
Cancer
Registry
(OCR)................................................................................................................................. 22
 2006
Census
Data .................................................................................................................................................................... 22
 Method ......................................................................................................................................................................................... 23
 Data
linkage............................................................................................................................................................................... 23
 Study
cohorts............................................................................................................................................................................. 24
 Definitions
of
variables ......................................................................................................................................................... 26
 Analyses ....................................................................................................................................................................................... 29
 Tracing
eligibility
and
calculating
person­month
in
each
quarter ................................................................... 31


CHAPTER
3:
TRENDS
AND
DISPARITIES
IN
COLORECTAL
CANCER
SCREENING
TESTS
 PARTICIPATION
IN
ONTARIO,
2005‐2011............................................................................... 33
 ABSTRACT.................................................................................................................................................................................. 33
 BACKGROUND .......................................................................................................................................................................... 35
 METHOD...................................................................................................................................................................................... 36
 Data
Sources .............................................................................................................................................................................. 36
 Cohort
Identification.............................................................................................................................................................. 36
 Measures..................................................................................................................................................................................... 38
 Demographic
variables......................................................................................................................................................... 39
 Statistical
Analyses ................................................................................................................................................................. 39
 RESULTS...................................................................................................................................................................................... 40
 Trends
in
FOBT
participation ............................................................................................................................................ 40
 Trends
in
endoscopy
participation .................................................................................................................................. 41
 Trends
in
‘up­to­date’
status............................................................................................................................................... 41
 DISCUSSION............................................................................................................................................................................... 42
 LIMITATIONS............................................................................................................................................................................ 45
 CONCLUSION............................................................................................................................................................................. 46
 Figures
for
chapter
3 ............................................................................................................................................................. 48
 Tables
for
chapter
3 ............................................................................................................................................................... 50


CHAPTER
4:
THE
INFLUENCE
OF
PHYSICIAN
RECOMMENDATION
ON
PARTICIPATION
IN
FECAL
 OCCULT
BLOOD
SCREENING
TEST
IN
ONTARIO
USING
POPULATION
BASED
DATA ................. 61
 ABSTRACT.................................................................................................................................................................................. 61
 vi

INTRODUCTION....................................................................................................................................................................... 62
 METHOD...................................................................................................................................................................................... 63
 Data
Sources .............................................................................................................................................................................. 63
 Cohort
Identification.............................................................................................................................................................. 64
 Statistical
Analyses ................................................................................................................................................................. 66
 RESULTS...................................................................................................................................................................................... 66
 Distribution
of
contact
with
physician
and
physician’s
recommendation
by
demographic
 characteristics........................................................................................................................................................................... 67
 Unadjusted
and
multivariate
regression
adjusted
prevalence
rate
ratio
of
FOBT
participation ........ 67
 DISCUSSION............................................................................................................................................................................... 68
 LIMITATIONS............................................................................................................................................................................ 71
 CONCLUSION............................................................................................................................................................................. 72
 Figures
for
chapter
4 ............................................................................................................................................................. 73
 Tables
for
chapter
4 ............................................................................................................................................................... 75


CHAPTER
5:
THE
EFFECT
OF
THE
COLON
CANCER
CHECK
PROGRAM
ON
FECAL
OCCULT
BLOOD
 TEST
PARTICIPATION
IN
ONTARIO:
AN
INTERRUPTED
TIME
SERIES
USING
SEGMENTED
 REGRESSION
ANALYSIS ......................................................................................................... 82
 ABSTRACT.................................................................................................................................................................................. 82
 BACKGROUND .......................................................................................................................................................................... 84
 METHOD...................................................................................................................................................................................... 85
 Data
Sources .............................................................................................................................................................................. 85
 Measures...................................................................................................................................................................................... 87
 Study
design ............................................................................................................................................................................... 88
 RESULTS...................................................................................................................................................................................... 90
 DISCUSSION............................................................................................................................................................................... 91
 STRENGTHS
&
LIMITATIONS ............................................................................................................................................ 94
 CONCLUSION............................................................................................................................................................................. 96
 Figures
for
chapter
5 ............................................................................................................................................................. 97
 Tables
for
chapter
5 ............................................................................................................................................................... 98


CHAPTER
6:

DISCUSSION
AND
CONCLUSION....................................................................... 102
 Thesis
Summary ................................................................................................................................................................... 102
 Implications
and
Recommendations ........................................................................................................................... 103
 Practice ......................................................................................................................................................................................103
 vii

Health
policy............................................................................................................................................................................104
 Research ....................................................................................................................................................................................105
 Thesis
Limitations................................................................................................................................................................ 106
 Future
studies........................................................................................................................................................................ 108
 Conclusion ............................................................................................................................................................................... 110


Appendices ......................................................................................................................... 111
 Appendix
1:
Characteristics
of
colorectal
cancer
screening
tests
used
in
Ontario.................................. 111
 Appendix
2:
Colon
Cancer
Check
physician
incentives........................................................................................ 112
 Appendix
3:
Data
linkage
flow
chart............................................................................................................................ 113
 Appendix
4:
Definition
of
demographic
variables ................................................................................................. 116
 Appendix
6:
Definition
of
explanatory
and
outcome
variables........................................................................ 118
 Appendix
7:
Segmented
regression
statistical
analysis....................................................................................... 120
 Appendix
8:
FOBT
participation
by
quarter
per
1000
person‐months
by
demographic
 characteristics,
Ontario,
2005‐2010 ............................................................................................................................ 122


References .......................................................................................................................... 124


viii

List of Tables Table 3.1: Population due for colorectal cancer screening by demographic characteristics, Ontario, 2005-2011 Table 3.2: Age standardized percent of Fecal Occult Blood Test participation by demographic characteristics, Ontario, 2005-2011 Table 3.3: Percent of Fecal Occult Blood Test participation by age group, Ontario, 2005-2011 Table 3.4: Age standardized percent of endoscopy participation by demographic characteristics, Ontario, 2005-2011 Table 3.5: Percent of endoscopy participation by age group, Ontario, 2005-2011 Table 3.6: Age standardized percent of ‘up-to-date’ status by demographic characteristics, Ontario, 2005-2011 Table 3.7: Percent of ‘up-to-date’ status by age group, Ontario, 2005-2011 Table 4.1: Population eligible for colorectal cancer screening by demographic characteristics, Ontario, 2008-2010 Table 4.2: Contact with physician by demographic characteristics and recommendation, Ontario, 2008- 2010 Table 4.3: Physician recommendation by demographic characteristics, Ontario, 2008-2010 Table 4.4: Unadjusted Prevalence Rate Ratio (PRR) of FOBT participation by demographic characteristics, Ontario, 2008-2010 Table 4.6: Multiple regression adjusted prevalence rate ratio of FOBT participation by demographic characteristics, Ontario, 2008-2010 Table 5.1: Population eligible for colorectal cancer screening by demographic characteristics, Ontario, 2005-2011 Table 5.2: Parameter estimates from the segmented regression analysis for the effect of the Colon Cancer Check on FOBT participation per 1000 person-months with and without adjustment for autocorrelation Table 5.3: Parameter estimates from the segmented regression analysis for the effect of the Colon Cancer Check program on FOBT participation per 1000 person-months by demographic characteristics

ix

List of Figures Figure 1.1: Thesis conceptual framework Figure 3.1. Age standardized percent of FOBT participation by demographic characteristics, Ontario, 2005-2011 Figure 3.2. Percent of FOBT participation by age, Ontario, 2005-2011 Figure
4.1.
Percent
of
physician
recommendation
by
demographic
characteristics,
Ontario,
 2008‐2010 Figure 4.2: Percent of FOBT participation by demographic characteristics, Ontario, 2008-2010 Figure 5.1: Quarterly rates of Fecal Occult Blood Test (FOBT) participation per 1000 personmonths, Ontario, 2005-2010 Figures in appendix 8: Figures 5.2-5.6: Quarterly rates of FOBT participation per 1000 personmonths by demographic characteristics, Ontario, 2005-2010

x

List of Appendices Appendix 1: Characteristics of colorectal cancer screening tests used in Ontario Appendix 2: Colon Cancer Check physician incentives Appendix 3: Data linkage flow chart Appendix 4: Definition of demographic variables Appendix 5: Person-month calculation flowchart Appendix 6: Definitions of outcome measures Appendix 7: Segmented regression statistical analyses Appendix 8: Quarterly rates of FOBT participation per 1000 person-months by demographic characteristics, Ontario, 2005-2010

xi

1

CHAPTER 1: BACKGROUND Colorectal cancer epidemiology Colorectal cancer (CRC) is the most common form of gastrointestinal cancer affecting the colon and the rectum. CRC almost always starts as a benign adenoma polyp on the inner wall of the colon or rectum. The cell linings of the colon or rectum become abnormal and divide rapidly forming a polyp. Certain kind of polyps called adenomas progress slowly towards a carcinoma (Bond, 2003; Kelloff, et al., 2004). Adenomas are very common, they occur in one third to one half of all individuals, but only 10% progress towards a carcinoma (Society, 2011). Carcinoma takes on average 10-15 years to progress from a polyp to a carcinoma (Bond, 2003). At first, cancer cells are contained on the surface of the polyp but with time they spread to the wall of the colon or rectum and then to blood and lymph nodes (Canada, 2011). If these polyps are detected at an early stage we can, not only detect cancer at an early stage, but also eliminate the disease at a benign stage. Colorectal cancer is a global burden. Worldwide, CRC is the third most common cancer in men (10% of total cancers) and the second in women (9.4%) (GLOBOCAN, 2008; I. A. f. R. o. C. IARC, 2011). Sixty percent of cases occur in developed countries. The higher incidence in developed countries may be due to disparate set of risk factors and diagnostic practices (Jemal, et al., 2011). Worldwide, CRC accounts for 8% of all cancer deaths, or an estimated 608,000 men and women die from the disease every year, making it the fourth most common cause of cancer deaths (GLOBOCAN, 2008). In 2011, 22,200 Canadians were estimated to be diagnosed with colorectal cancer and 8,900 to die from the disease, making it the second leading cause of cancer deaths in Canada (Canadian Cancer Soceity CCS, 2011). The east-west gradient in incidence and mortality rates is striking. The highest rates are in Atlantic Provinces and lowest in British Columbia and Alberta. The differences are attributed to variations in risk factors including genetic factors (Green, et al., 2007), lifestyle factors or screening practices (Canadian Cancer Statistics CCS, 2011). The incidence and mortality rates are higher in males than in females. One in 14 men is expected to develop colorectal cancer in a lifetime and one in 27 to die from the disease. One in 15 women

2

is expected to develop colorectal cancer in a lifetime and 31 to die from it (Canadian Cancer Statistics CCS, 2011). In Ontario, where this thesis is conducted, colorectal cancer is the second leading cause of cancer deaths (Care). In 2010, an estimated total of 8,200 of new CRC cases were diagnosed and there were 3,400 deaths from the disease. Colorectal cancer is the second most frequently diagnosed cancer for men and the third most frequent for women (13% and 12% respectively) (Canadian Cancer Statistics CCS, 2011; CSQI, 2011). The incidence has been decreasing overtime. In 2010, the age standardized incidence rates for men and women were estimated at 59.4 and 38.3 per 100,000 respectively, down from 62.8 and 42.6 in 1986. Mortality rates have decreased for both males (35 in 1986 to 25 per 100,000 in 2010) and females (22 in 1986 to 15 in 2010). The 5-year survival from CRC improved, from 57% in 2003 to 62% in 2007. The decrease in CRC mortality reflects improvement in prognosis due to screening, treatment or both (CSQI, 2011). The incidence of colorectal cancer is disparately distributed in Ontario. CRC incidence is 1617% higher in rural areas. The difference in incidence is attributed to differences in risk factors and screening for CRC. Individuals living in rural areas are more likely to smoke and less likely to consume the daily recommended servings of fruits and vegetables. Neighborhoods with higher percentage of immigrants have lower incidence rates, 22% compared to 24% in neighborhood with lower percentage of immigrants. The difference may reflect a lower risk of colon and rectum cancer in the countries of birth (CSQI, 2011) Risk factors for colorectal cancer The identifiable risk factors for colorectal cancer include hereditary factors, personal or family history of CRC, inflammatory bowel disease, racial and ethnic background, age, behavioral and social factors. These factors form the basis for stratifying individuals into high and average risk groups. Individuals at high risk of developing colorectal cancer are those who have hereditary, personal or family history of CRC, and those with history of inflammatory bowel disease. The presence of these risk factors necessitates screening on a more frequent basis and at an earlier age. About 5% of CRC cases are associated with inherited genetic factors. The most common genetic diseases are the Familial Adenomatous Polyposis (AFP) and hereditary non-polyposis colorectal

3

cancer (HNPCC). About 20-30% of CRC cases are associated with previous medical history and family history of the disease (Grady, 2003). First degree relatives of individuals with colon cancer have a two- fold probability of developing the disease compared to the general population (Grady, 2003). The risk increases when the disease occurs at a young age (CCAC, 2011). Ulcerative colitis and Crohn’s disease are the most common inflammatory bowel diseases increasing the lifetime risk for developing colorectal by 2% (Eaden, Abrams, & Mayberry, 2001), this risk increases with an early onset of the disease (Ekbom, Helmick, Zack, & Adami, 1990). Individuals at average risk of developing the disease are those who are asymptomatic, aged 50 and above with no biologic or hereditary history of the disease. The onset of CRC increases with age. The incidence rate rises from 36.7 per 100,000 for (45-49) age group to 64.13 for (50-54) age group (P. H. A. o. Canada) and increases gradually with each decade (CCAC, 2011). Dietary and lifestyle factors are associated with increased risk of developing the disease including high fat diet, physical inactivity, heavy alcohol consumption, and smoking (Care) (Huxley, et al., 2009). Occupational exposures to environmental carcinogens can also increase the risk for colon cancer (Bonner, et al., 2007). Obesity and diabetes increase the risk of developing CRC (Harriss, et al., 2009) (Moghaddam, Woodward, & Huxley, 2007) (Seow, Yuan, Koh, Lee, & Yu, 2006). Prevention of CRC for average risk individuals includes lifestyle behavior modifications and screening. Screening for colorectal cancer The purpose of screening for CRC is not only to early detect established cancerous cells but also to detect precancerous cells. Because it takes 10-15 years for the disease to progress from a benign polyp to a carcinoma (Bond, 2003), CRC screening tests are ideal not only to improve the prognosis but also to abort cancer while in its precancerous stage (D. Leddin, et al., 2004; Vinden & McAlister, 2005). Multiple screening options are available for colorectal cancer. The risk category determines the ideal type of screening. For average risk individuals, the most recent Canadian Association of Gastroenterology position statement (D. J. Leddin, et al., 2010) recommends two tests to be used in a population-based program: a stool test (high sensitivity- Guaiac Fecal Occult Blood Test (gFOBT) or preferably Fecal Immunochemical Test (FIT)) at least every two years and a

4

flexible simgoidoscopy every 10 years (previously was 5 years) (D. Leddin, et al., 2004). A colonoscopy can be used in an opportunistic screening and for follow-up of positive stool tests (D. J. Leddin, et al., 2010). Other screening tests for CRC include: air-contrast barium enema, Computerized Tomography Colonography (CTC), and altered DNA in stool (sDNA). The different screening options for CRC vary in their ability to detect cancerous or precancerous cells and in their ability to interrupt the adenoma-cancer sequence by removing polyps. A gFOBT, for example, is meant to detect small amount of blood in the stool resulting from bleeding vessels at the surface of polyps or adenomas, regardless of the site of bleeding. It is estimated that gFOBT can detect from 9-64% of polyps (CCAC, 2011). Colonoscopy, on the other hand, allows the doctor to see the inside of the rectum or colon and permits to remove the polyp during the procedure. Colonoscopy can detect 74% to 95% of cancer cells depending on the site and size, of adenoma cells (Baxter & Rabeneck, 2010; Rex, et al., 2009). Screening tests for CRC are also effective in reducing mortality from the disease. A regular screening using FOBT followed-up by colonoscopy for positive cases reduces mortality from CRC by 15% (Hardcastle, et al., 1996; Hewitson, Glasziou, Irwig, Towler, & Watson, 2007). (Appendix 1: summary characteristics for the most commonly used colorectal cancer screening tests in Ontario). Screening testing in Canada A few population-based studies done in provinces and sub-provincial regions in Canada revealed that since the introduction of the national screening recommendations a decade ago colorectal cancer screening participation was on the rise but remained suboptimal (McGregor, Hilsden, Li, Bryant, & Murray, 2007; Rabeneck & Paszat, 2004; Sewitch, Fournier, Ciampi, & Dyachenko, 2007; Wilkins, 2009; Zarychanski, Chen, Bernstein, & Hebert, 2007). In 2003, selfreported adherence to guidelines (i.e. FOBT every two years, colonoscopy/simgoidoscopy in past 10 years) in four provinces in Canada (Ontario, Newfoundland, Saskatchewan and British Columbia) was 30% (Sewitch, et al., 2007). In 2008, 40% of Canadians aged 50-74 selfreported having had an FOBT in the past two years or endoscopy (both colonoscopy and flexible simgoidoscopy) in the past five years (Wilkins, 2009). Geographic differences in CRC testing were systematically reported. In general the likelihood of CRC adherence to guidelines is less in Atlantic Provinces, Quebec and territories than the rest of

5

provinces. Moreover, geographic differences in the type of test were also reported. Atlantic provinces and Quebec were less likely to have an FOBT and more likely to have endoscopies than western provinces (Sewitch, et al., 2007; Wilkins, 2009). FOBT participation ranged from 10% in Quebec to 42% in Manitoba, while endoscopy participation varied from 11% in Yukon, 23% in Saskatchewan, to 30% in Ontario (Wilkins, 2009). Comparison over time revealed that the increase in trend was significant for all tests and in all provinces (Wilkins, 2009) The likelihood of participation in screening tests varied by socio-demographic characteristics. Individuals 65 or older, living in large metropolitan areas, Canadian born individuals, acculturated immigrants (20 years or more) and higher income were associated with increased participation(Wilkins, 2009). Gender association with CRC testing was inconsistent. The patterns of CRC screening in Ontario mirrors the overall pattern in Canada. Prior to 2004, a population-based study revealed that 20% of screen eligible individuals received CRC testing timyears period was 20% and in 2007-2008, rates rose to 30%. Colonoscopy rates increase was more modest but significant over time. A recent study showed that between 1996 and 2005, there was a four-fold increase in colonoscopy. The rates increased from 1.55 to 4.7% (Jacob, et al., 2011). The increase in FOBT participation was significant across all geographic regions of Ontario. In 2001/2002 rates for colonoscopy were highest in the north and central east as well as the Toronto regions in Ontario and lowest in the east (Schultz, Vinden, & Rabeneck, 2007). In 2007 and 2008, FOBT participation varied from 22% in Muskoka to 36% in Champlain (CCC, 2010). The likelihood of participation in screening tests varied by socio-demographic characteristics. Females were more likely to have an FOBT than males, and older adults (70-74) more than younger adults (50-55) (Ontario., 2010). Higher income were more likely to have an FOBT than lower income individuals (Krzyzanowska MK, 2009; Ramji, Cotterchio, Manno, Rabeneck, & Gallinger, 2005). Colon Cancer Check program: an organized screening program The high incidence rates of CRC and the suboptimal rates of screening in Ontario led to the establishment of the Colon Cancer Check program, the first organized province-wide screening program for colorectal cancer in Canada.

6

First, a brief background on opportunistic screening and organized screening. Opportunistic screening is a medical practice model targeting individuals (Rabeneck, 2007). Testing is embedded in routine primary care and occurs when the physician captures the opportunity to recommend and deliver the test and occasionally on individuals to request the test (Senore, Malila, Minozzi, & Armaroli, 2010). Opportunistic screening is the dominant approach in the United States. In contrast, an organized screening program is a public health model targeting the population at large (Rabeneck, 2007). Certain features characterize organized screening programs including a targeted population for screening, a specific test for screening and recommended interval for repeat testing, a management team for the implementation of the program, a health care team for the delivery of services, a standardized quality assurance program for the laboratory tests, and continuous performance measurement and monitoring(IARC, 2005a). The benefits of organized screening programs include reaching out to a large population, reduction of over use, higher quality of services, and a better follow-up process (Levin, et al., 2011). Organized screening programs for colorectal cancer are presently implemented through an integrated health care system in twelve Member States of European Union (I. A. f. R. o. C. IARC, 2011), in the US Veteran’s Administration and in Kaiser Permanente Northern California (Levin, et al., 2011), in Israel, Japan and the Republic of Korea (ICSN, 2011). In Canada, the first organized screening program was launched in Ontario in 2008. In 2010, Manitoba and Nova Scotia provinces phased in their programs to cover more than fifty percent of their communities. Other western provinces (British Columbia, Alberta and Saskatchewan) have expanded their program to cover between 10 and 50% of their communities. Eastern provinces (New Foundland, New Brunswick, Prince Edward Island and Quebec) were still at an evaluation and planning stage. No screening programs were yet organized for the Northern Territories (CCAC, 2011). The Colon Cancer Check program in Ontario Multiple efforts paved the way to the establishment of the organized screening program for colorectal cancer in Ontario. In 1999, Cancer Care Ontario set up an expert panel to develop recommendations for a population based colorectal cancer screening. The panel recommended an FOBT based program for CRC screening for average risk individuals between the ages of 50 and 75 (Screening, April 1999). From 2003 to 2005, the Ontario Ministry of Health and Long Term Care (MOHTLC) in collaboration with Cancer Care Ontario funded a pilot study to

7

identify effective methods for promoting and increasing screening in the community. In 2006, the pilot study report recommended the establishment of a primary care provider-led, organized screening program for CRC (Ontario, 2006). In January 2007, the government of Ontario officially announced the launch of the Colon Cancer Check program. From January 2007 to April 2008, the preparations for the public launch of the program were underway. On April 1st 2008, the CCC program was launched to the public. An intense media campaign accompanied the public launch. The dual goals of the program are to reduce the mortality from colorectal cancer and increase the capacity of primary care providers to participate in an organized screening program(CCC, 2010). The CCC is not a mail outreach program. The CCC is a provider-led program. Primary care physicians play a central role in the CCC program. They are responsible for counseling and dispensing FOBT kits to all eligible patients in their practices and for patients without primary care physicians (unattached patients), FOBT kits are made available through pharmacies or through Tele-Health Ontario and once they complete the test, they are attached to a primary care physician through the CCC. The CCC program incorporates most elements of an organized screening program features (Rabeneck, 2007). The CCC identifies average risk individuals aged 50-74 as the target population eligible for screening in Ontario, adopts the biennial Fecal Occult Blood Test as a primary test for screening, and colonoscopy to follow-up for positive cases. A capable management team leads the day-to-day activities of the program. Using an evaluation framework and indicators, the team is responsible for monitoring the uptake, and for measuring and reporting the performance. In addition, a province-wide primary care strategy is set to engage primary care physicians to participate in the program (Levitt & Lupea, 2009). The target of the Colon Cancer Check program is to increase FOBT participation from 17% in 2005 to 55% in 2013 (Care). Different strategies were used to achieve this target. An intense but temporary mass media campaign marked the launch of the program including television advertising, radio messages, newspapers clips and pamphlets in 22 languages across the province. The media campaign lasted for six months. After that, several continuous communitybased awareness programs took place in various regions of Ontario. Since it is a provider-led program, strategies to increase awareness and harness the support of primary care providers for the program were also implemented. An educational campaign targeting primary care providers was conducted between 2008 and 2010. The objectives of the campaign were to diffuse any

8

confusion among physicians about the evidence underlying anchoring the CCC program to FOBT and colonoscopy, to clarify the process of ordering FOBT kits and reporting results to the patient, to empower providers with counseling tools to assist them in education and provision of service (OCFP, 2009). In fall 2008, a province wide primary care strategy was launched to engage primary care physicians to participate in the program (Levitt & Lupea, 2009). Further, knowing that adoption of guidelines requires more steps than simple dissemination (Grimshaw, et al., 2004; Rabeneck, 2007; Vernon, 1997), the CCC program uses financial incentives to encourage physicians to expand their delivery of screening services (Chassin, 2006). The financial strategy of the CCC program includes rewarding physicians for providing the following services: 1- recommending and dispensing the FOBT kit to eligible individuals while visiting the practice; 2- calling in or sending a letter to individuals who did not visit the office asking them to book an appointment to discuss colorectal cancer screening; 3- for completion of the test and the follow-up thereafter; 4- and for accepting new unattached patients (Care) (Appendix 2: Colon Cancer Check physician incentives). To increase reliability and validity, only a few community laboratories that signed an agreement with the Ministry of Health and Long Term Care are allowed to perform the CCC FOBT test and claim the specimen collection fee set by the program (Care). Staff training and quality control are frequently implemented to reduce variability in examining and reporting the results (L. Rabeneck, 2007). Thesis overview The goal of this thesis is to evaluate the effect of the Colon Cancer Check program on colorectal cancer screening in Ontario. Three studies were conducted: 1. First study: Trends and disparities in colorectal cancer screening participation in Ontario, 2005-2011. 2. Second study: The influence of physician recommendation on participation in Fecal Occult Blood Screening test in Ontario using population-based data.

9

3. Third study: The effect of the Colon Cancer Check program on Fecal Occult Blood test participation in Ontario: An interrupted time series using segmented regression. Specific objectives 1. To describe the trends of FOBT screening rates, endoscopy screening rates and ‘up-to-date’ with CRC guidelines rates, overall and stratified by demographic characteristics, between April 1st 2005 and March 31st 2011. 2. To describe the pattern of physician’s recommendation in Ontario and test the effect of physician’s recommendation on FOBT participation and disparities in participation, between April 1st 2008 and March 31st 2010. 3. To estimate the effect of the Colon Cancer Check program on the trend of FOBT participation, overall and stratified by demographic characteristics between April 1st 2005 and September 30th 2010. Significance of this thesis This thesis addresses three policy issues. The first issue concerns the trend of disparities before and after the organized screening programs. When screening was opportunistic before the CCC, differences in participation by gender, age, income, and geographic region were observed (CCC, 2010). Two and a half years after, did the organized screening program reduce those disparities? In the first study, we highlight whether the disparity after the CCC narrowed, remained as is, or increased by gender, age, income, urban rural and we add recent registrant variable as a proxy measure for recent immigrant. The second policy issue concerns the effectiveness of a provider-led organized screening program in increasing participation and mitigating disparities. When the United Kingdom Bowel Cancer Screening Program (BSCP) started in 2006, they used mail outreach to invite individuals for screening. Thirty months after its implementation, von Wagner et al. found that the response rate for returning kits ranged from 32% in the most deprived quintile to 49% in the least deprived indicating that using the mailing outreach motivated higher income individuals to get screened more than those with lower incomes (von Wagner, et al., 2009). The Colon Cancer Check program is a provider-led program. Primary care physicians play a central role in

10

motivating their patients to be screened. In the second study, we test whether physician’s recommendation in a provider-led approach increases the likelihood of FOBT participation and compliance among all population subgroups. The third issue concerns the short-term effect of an organized screening program on participation and disparities in participation. Studies that have directly evaluated changes following the introduction of a public health intervention showed a rapid increase in participation at first. This increase stabilized forming a plateau for some time (D. Baker & Middleton, 2003; Victora, Vaughan, Barros, Silva, & Tomasi, 2000). These studies suggest that the intervention reached out first to the wealthiest (Victora, et al., 2000) and early adopters (Anderson, Mullins, Siahpush, Spittal, & Wakefield, 2009). Once the wealthiest and early adopters reach a level of improvement beyond which the intervention is likely to make progress, then the pattern stabilized forming a plateau. In the third study, we describe the secular trend of FOBT participation before the CCC, and then we estimate the effect of the CCC on the level of FOBT participation immediately after its implementation and the trend thereafter using an interrupted time series design. We compare the pattern the CCC had with that described in the literature and interpret the findings. Conceptual framework In formulation of a theoretical perspective for evaluating the effect of the Colon Cancer Check program on colorectal cancer screening in Ontario, a socio-ecological conceptual framework provides a useful prototype. Since the purpose of this thesis is to evaluate the effect of a new health policy on a health care system, then a system model would be an appropriate model. The Socio-Ecological Model (SEM) is a system model. Unlike individually or socially oriented models, SEM considers both the effect of individual and social environment on outcome. The central hypothesis of SEM posits that the behavior of an individual living in a system is determined by a dynamic and ‘reciprocal causation’ between the individual and the environment. A change in the social environment produces change in individuals and the support of individuals is essential for implementing social changes (McLeroy, Bibeau, Steckler, & Glanz, 1988). Alternatively, SEM differs from other traditional system models by viewing patterned behavior of individuals or aggregates as outcomes of interest. Researchers using SEM usually explore the

11

effect of the different levels of influence in the social environment on individual’s behavior. SEM depicts the different levels of social environment influence as overlaps of circles. The individual is at the center of the model and the other circles surround the individual indicating that the behavior of an individual is not acting alone in this universal system but rather all other levels of influence converge to shape an individual behavior. In an SEM, human behavior is determined by: •

Intrapersonal factors – Individual’s knowledge, attitudes, skills or intention to comply

with certain behavioral norms, which are usually the target for change. Demographic characteristics are often used as effect modifiers for these characteristics. •

Interpersonal factors – Formal and informal social network and social support including

family, friends, neighbors, coworkers and health care providers, which combined shape human’s behavior. •

Institutional factors – Social institutions with organizational characteristics to support

behavioral change. Institutions, such as school, work, church, professional or neighborhood groups, may have positive or negative effects on the health of their members. Since they are important sources and transmitters of social norms and values, organizations can provide the opportunity to build social support for a desirable behavior change. Organizational changes are needed to support long-term behavioral changes among individuals. •

Community Factors – Relationships between institutions, organizations and groups to

which an individual can belong. These "mediating structures," such as family, church, informal social networks, and neighborhoods, may provide social identity and resources. •

Public Policy - Regulatory policies, procedures and laws have been passed (national,

state or local) to help protect the health of communities. As a part of the policy development process, increasing the public's awareness of health and policy issues must be included. Socio-Ecological Models have been applied to a wide range of health issues including improving dietary habits (Robinson, 2008), tobacco control (X. Zhang, Cowling, & Tang, 2010), increasing physical activities among adolescents (Elder, et al., 2007), chronic disease management (Cassel, 2010) and increase compliance to screening services such as cervical cancer screening (Daley, et al., 2011), gastric cancer (Bae, et al., 2008). The Centre for Disease

12

Control adopted the Socio-Ecological Model to represent the colorectal cancer screening program approach to CRC prevention in the US(CDC, 2011). Explaining the social ecological model, McLeroy (McLeroy, et al., 1988) indicated that when a new policy is introduced into the system, it uses different ‘levels of interventions’ to cause a change in ‘outcomes among the target population’. And it is crucial to distinguish between level of intervention and target of the intervention. When the intervention takes the form of mass media, then the level of intervention is the community and the target is the individuals’ characteristics including their knowledge attitudes, and skills in an attempt to increase their compliance to the new behavior. Another intervention within the same policy can take the form of educational programs targeting practitioners, then the level of intervention is institutional and the target for the intervention are the members of that institution. But, it should be clear that by changing the latter (members of the institution), we are also making changes to the former (individuals). Hence, the converging circles depicted in the model, that all levels of influence act together to influence the individual who is at the center of the model. In the application of the SEM to this thesis, the five levels of influence identified by McLeroy (McLeroy, et al., 1988) will be defined in the following manner: 1. At the intrapersonal level, we have the individuals who belong to all walks of life. Participation to colorectal cancer screening using FOBT or endoscopy is the outcome or the desired individual behavior. 2. At the interpersonal level, we have the physician. The interaction between physician and individual (physician-patient dyad) is the expected behavior that motivates individuals to be screened. 3. At the organizational level, we have the educational program for physicians. The educational programs are meant to increase awareness and harness the support of physicians to discuss screening with their patients. The educational programs are expected to increase physician recommendation at the practice level. 4. At the community level, we have the mass media. Mass media is an intervention to increase awareness among the general public and also among health care providers about colorectal cancer disease and how screening save lives.

13

5. At the policy level, we have the Colon Cancer Check program. This health policy influences availability, accessibility and support for screening in all Ontario. The CCC is the mastermind intervention, which is expected to stir a wave of change in the system. With these specific variables, the formula for behavior, which was formulated by McLeroy would be adapted to read: the potential for screening participation to occur in Ontario is a function of individuals’ receptiveness to the mass media campaign. In addition, the effect of the educational program on physicians must be considered in relation to increase the person-toperson interaction with individuals as a motivator for screening. Finally, certain characteristics such as age, gender, income, recent immigration and living in rural neighborhoods may act as effect modifier of the program on the overall participation to screening. The following statement represents the underlying logic for designing and conducting this thesis. If individuals receive a) the broadcasted message over the radio and the TV meet and found it that it meets their perceived needs and susceptibility, b) their primary care physicians are endorsing this message again during their interpersonal encounter, c) and no other personal or structural barriers is keeping them from taking the test, then the level of screening participation in Ontario will attain a level never reached before. Figure 1.1: Thesis conceptual framework

A literature map Intrapersonal level Although all average risk individuals could benefit from screening tests for colorectal cancer, not everyone uses them. The limited use of CRC tests by certain sub-population groups is very common. Studies showing females are more likely to have an FOBT than males (Ioannou, Chapko, & Dominitz, 2003; Ko, Kreuter, & Baldwin, 2005; Krzyzanowska MK, 2009; McQueen, Vernon, Meissner, Klabunde, & Rakowski, 2006) suggested that because females have more frequent contacts with the health care system than men for reasons related to their own health and the health of their families, they have more opportunities to learn about screening than males (Ramji, et al., 2005). Other studies suggested that because females are exposed and participate

14

more to other preventive services than males, they have positive attitudes towards screening, thus increasing their likelihood to take a CRC test (Shapiro, Seeff, & Nadel, 2001; Slattery, Kinney, & Levin, 2004). In some studies, we found that males are more likely to have an endoscopy than females (colonoscopy, flexible simgoidoscopy) (Christman, et al., 2004; Janz, Wren, Schottenfeld, & Guire, 2003; Ko, et al., 2005; McQueen, et al., 2006) and the authors suggested that the attitude towards the test differs between females and males. Some females may be more concerned about the gender of the physician delivering the endoscopy and more embarrassed from bodily exposures than males hence the differences in participation by type of tests(Farraye, et al., 2004; Menees, Inadomi, Korsnes, & Elta, 2005; Stockwell, et al., 2003). Individuals aged 50-55 have the lowest FOBT participation rates and the lowest ‘up-to-date’ status as compared to other age groups (CCC, 2010; Canadian Cancer Statistics CCS, 2011) (CCC, 2010; Christman, et al., 2004; Fisher, et al., 2004; Honda, 2004; Meissner, Breen, Klabunde, & Vernon, 2006; Ramji, et al., 2005; Tessaro, Mangone, Parkar, & Pawar, 2006). The difference is attributed to the perceived risk for the disease, which increases with age. Screening services are secondary preventive measures aiming to reduce the burden of disease at a time when no sign that the disease is actually occurring(Rogers, 2002). Unlike acute care where the demand for the service follows the onset of the disease and is for immediate relief of symptoms, the demand for preventive care precedes the onset of the disease and is for selfprotection(Kenkel, 1994). Therefore, the perceived risk for the disease is low for a young healthy adult and screening is not an identified priority. As they get older, their health starts to show signs of illnesses, then their perceived risk for the disease increases and eventually their attitudes towards screening change and they become more likely to be screened. The levels of education and income influence participation in screening. Higher education is associated with higher uptake of FOBT screening (Meissner, et al., 2006; Pollack, Blackman, Wilson, Seeff, & Nadel, 2006; Thompson, Coronado, Neuhouser, & Chen, 2005; Ye, Xu, & Aladesanmi, 2009) .The level of education reflects a higher score of knowledge, which affects the cognitive functioning and make the individual more receptive to health messages or more able to communicate with and access appropriate health services (Galobardes, Shaw, Lawlor, Lynch, & Davey Smith, 2006). Level of education also facilitates better access to employment and income. Income by itself does not generate health; but what money can get affects health. Better income is associated with better access to quality material resources such as food and

15

shelter and allows access to improved health services (Galobardes, et al., 2006). The combined effect of health literacy, level of communication, affordability and accessibility provides a favorable environment for the client to participate in screening. Ethnicity also influences participation in screening. Undoubtedly, every individual in any society acknowledges the importance of prevention in order to remain healthy. But preventive practices are influenced by cultural and social factors. Screening is a western practice medicine. Older adults in some cultures tend to use alternative medicine as their primary source of preventive care and rely on western medicine when the former is ineffective (Garces, Scarinci, & Harrison, 2006). But, if they receive recommendation from their physicians, they may be more likely to comply to their physicians’ advices despite their beliefs and practices (Shokar, Carlson, & Weller, 2008). Country of birth also determines participation in screening. In Canada and the US immigrants, especially those who lived in the country for less than ten years, are less likely to have a CRC test than individuals born in Canada or US (Shih, Elting, & Levin, 2008; Wilkins, 2009). Several barriers contribute to this disparity. Recent immigrants tend to be more of low-income (Shokar, et al., 2008), low health literacy (Kreps & Sparks, 2008) and have less ability to speak the official language (Thomson & Hoffman-Goetz, 2010), leading to their marginalization and lack of acculturation. Acculturation is the process whereby one whose learning was in one culture then adopts values, attitudes and behavior of another culture (Suarez, 1994). Recent immigrants tend to form social networks with individuals from the same background, preserving their own cultural and social practices; and depending on the country of birth, they may not have heard about CRC screening or cancer fatalism in some cultures may keep them for being screened (Powe & Weinrich, 1999). But these factors tend to disappear with acculturation. For example, Hispanics in United States have lower CRC screening than whites (Beydoun & Beydoun, 2008), their odds for screening increases, however, with time since they immigrated, with speaking the language and their frequent utilization of health services (Afable-Munsuz, Liang, Ponce, & Walsh, 2009; Honda, 2004; Shah, Zhu, & Potter, 2006) Interpersonal level Neighborhood characteristics influence individuals’ attitudes towards colorectal cancer screening participation (89-95). Gresenz et al. (91) used the well-established sociological

16

principle of “homophily” by Lazarsfeld & Merton in 1954 (Lazarsfeld, 1954) to suggest that similar people flock together, which results in social networks that are homogenous with regards to socio-demographic and behavioral characteristics. Individuals who share the same ethnic origin tend to live in geographical proximity and are more likely to be of the same education, income, and language. This homogeneity in social network affects the information exchange, social values, beliefs and attitudes. Individuals living in proximity get together more often to exchange information about practically everything including sources of culturally competent care and transmit norms about the appropriate use of care. Interpersonal communication with the physician influences individuals’ attitudes towards colorectal cancer screening. A conversation with the physician increases perceived susceptibility to CRC, provides cues to action and ultimately leads to completion of screening(Fenton, Jerant, von Friederichs-Fitzwater, Tancredi, & Franks, 2011). Several barriers affect the interpersonal communication during an office visit. Some are behavioral others are organizational. Physicians may have ineffective patient-centered communication, which affects patient compliance to physicians’ recommendations(Stewart, et al., 2000). Following an interpersonal communication with the provider, patients perceive whether their physicians made an effort to listen and understand their concerns, responded to their needs and that their communication led to a common understanding. Following a patient-centered communication, patients are more satisfied with the services, have better compliance to care leading to improved efficiency and outcomes(Stewart, et al., 2000). Bias and stereotyping affect the amount of information delivered during a communication. Stereotypes are interpreted as “an adaptive cognitive strategy of making the world manageable by using categorizing and generalizing techniques” (van Ryn, Burgess, Malat, & Griffin, 2006). Using this strategy, individuals develop beliefs and expectations about these groups and mentally label them with specific characteristics. Because it is unintentional, most individuals deny engaging in stereotyping. But, stereotyping is common to all individuals and physicians are not immune to it. For example, physicians’ perceptions that certain ethnic minorities may have different illness beliefs, different social and cultural backgrounds, and different interest in engaging in screening, find it a challenge discussing screening with them (Fiscella & Epstein, 2008), which lead them to deliver less information to these ethnic groups (van Ryn & Fu, 2003).

17

The organizational characteristics of the practice affecting physician-provider interaction leading to physician’s recommendation of screening services include high volume practice (101103), and lack of allied health professionals to assist in the delivery of services (42, 98, 105). In team-based practices, the characteristics of team members, the skill mix, the ability of staff to participate in decision-making influence the prospect of discussing screening services. Other structural characteristics include office infrastructure such as electronic medical records and audio-visual resources, which increase the potentials for motivating patients to receive screening (Hogg, Rowan, Russell, Geneau, & Muldoon, 2008) Social level Mass media is one of the social factors influencing screening uptake (58, 107, 108). Rogers, E.M. suggested that mass media are effective in diffusing knowledge about an innovation and most individuals adopt that innovation not because of the scientific expert opinion but based on the subjective evaluations of near peers(Rogers, 2002). Similarly, Whynes et al. (Whynes, Philips, & Avis, 2007) found that screening is primarily driven by a search for reassurance, a sense of duty, and herd signaling (Catalano, Winett, Wallack, & Satariano, 2003; Jacobsen & Jacobsen, 2011). So, how does mass media influence behavior? The influence of mass media on health behavior comes in two forms. The first form is to increase health information on the topic of interest; second, to frame the issue as a public health problem and attract the attention of the audience on the tools to solve that problem(Randolph & Viswanath, 2004). In the case of mass media for colorectal cancer screening, a mass media campaign draws the attention on the high incidence of the disease and how screening could save lives. The effect of mass media on screening participation comes in stages. Diffusing that knowledge to a large audience encourages certain individuals to adopt screening faster than others. First, to early adopters constituting 13% of the population, followed by early majority who look for early adopters for advice and information about screening (34% of the population), then the late majority and the laggards constituting the rest of the population(Rogers, 2002). Political level Organized screening programs, such as the Colon Cancer Check program, are an example for a health policy influencing screening. Organized screening programs reflect a political will to support screening in the population. They are meant to enhance the quality of screening and to

18

provide population-wide coverage (Miles, Cockburn, Smith, & Wardle, 2004). They mitigate various barriers for screening. They remove the financial cost of purchasing the screening test. They ensure the test is available to all individuals targeted by the program. They set standards for the quality of screening tests. They reduce overuse and encourage under-users. Organized screening programs were proven to be effective in increasing cervical cancer participation in United Kingdom (Miles, et al., 2004), breast cancer and cervical cancer screening participation in Ontario (CSQI, 2011) Ethics statement This thesis only occurred after approval by the Institute for Clinical Evaluative Sciences at Sunnybrook Health Sciences Centre and University of Toronto institutional review boards. All data were uniquely labeled using encrypted health card numbers. No unique identifiers such as patient name, OHIP number, postal code or address were recorded. 


19

Figure for background chapter Figure 1.1: Thesis conceptual framework

20

CHAPTER 2: COHORT IDENTIFICATION AND METHODS Study design The studies conducted in this thesis were retrospective cohort studies using administrative data collected regularly at the Institute for Clinical Evaluative Sciences (ICES). Retrospective cohort studies are the most reliable design for observational studies exploring the effect of an intervention in a natural experiment. Because the data is collected in temporal sequence, it is easy to measure the prevalence of the outcome over time. Since retrospective cohorts use already collected data originally intended for other purposes, the advantages of using secondary data include quicker and cheaper access to data (Mann, 2003). Moreover, in retrospective cohort studies, it is also easier to estimate the relative risk of the outcome associated with other variables of interest especially when the outcome is quite prevalent among the population. The main disadvantage of retrospective cohort studies is that they may be missing important variables for the analysis. The main advantage of using population-based administrative data, on the other hand, is the external validity of the results. Findings usually apply to all defined population. Moreover, the quality and validity of ICES data are well documented. Raw data received at ICES undergo rigorous quality standards before it becomes available for research use. Data is collected regularly over time with no interruption, which enabled us to do a time series analysis. Finally, I was personally privileged to have all three of my thesis committee members ICES scientists, which enabled me to access the data. The main disadvantage of using ICES administrative data is that not all variables, especially demographic variables, are included in the original data collection. Perhaps, an alternative for administrative data would have been to use survey data such as the Canadian Community Health Survey (CCHS), which up to 2007 used to collect data every two years and after 2007 the data collection became yearly. CCHS uses a sampling frame that represents Canadians aged 12 and over. CCHS is a reliable and regular source of data used by many health services researchers in Canada. The advantage of the CCHS is that it includes a wide range of individual level data including data on ethnicity, immigration, and language. The disadvantages of CCHS include the recall bias associated with any survey data; external validity is less precise than population-based because it is based on a sample of the population. Finally,

21

for this particular study we could not use CCHS because the latest files available at ICES were the 2007-2008, the period just before the CCC, which did not serve the purpose of this thesis. Data sources We linked five data sources in order to identify individuals eligible for CRC screening in Ontario, their characteristics, and their screening status. The following data sources were used: Registered Persons Database (RPDB) The Registered Persons Database (RPDB) provides basic demographic information on those who have ever received an Ontario health card number. The Ministry of Health provides the information to ICES, where it is updated every month and enhanced by additional data. ICES analysts offset the completeness and inaccuracies of this database, which are largely due to the untimely updates sent to ICES on deceased, moved in or moved out of the province individuals, by using other databases to update the information. For example, if individuals had a contact with the health care system, they are most likely still alive and in the province and their addresses are considered up-to-date. In this study, we used two files from the RPDB database: 1- the contact YYYY file, which contains information on eligibility and contact with health care system; and 2- the PSTLYEAR file, which provides information on the best known postal code of the individual. Ontario Health Insurance Plan (OHIP) Ontario Health Insurance Plan database contains claims for services provided by eligible physicians, groups, and laboratories. OHIP database is a well valid and accurate database. Typically, about 97% of the claims having service dates in a given month arrive within 3 months, 99.7% arrive after 6 months and close to 100% arrive within 1 year. We used OHIP database to retrieve information on physician billings for CRC tests, counseling and management of CRC tests, for recent immigrant status and for physician providing the majority of care. The Canadian Institute of Health Information- Discharge Abstracts Database (CIHI-DAD) The CIHI-DAD is a database of information abstracted from hospital records. It includes patient

22

demographic data, acute and chronic hospital care, diagnostics tests, and other administrative information. The main data elements of the CIHI-DAD are: encrypted patient identifier, patient demographics (age, sex, residential postal code), diagnoses, procedures, and administrative information (length of stay, institution number). For the purpose of this study, we used the CIHI-DAD to identify individuals who were diagnosed with any inflammatory or bowel diseases prior to the beginning of each annual cohort included in the study. The Ontario Cancer Registry (OCR) The OCR database includes all Ontario residents who had cancer or who died from cancer since 1974. The sources of this database are from hospital discharge summaries, pathology reports, and records from regional cancer centers or Princess Margaret Hospital and from death certificates. We used OCR to identify and exclude individuals who were diagnosed with CRC prior to the beginning of each annual cohort covered in this study. 2006 Census Data The 2006 census contains a short questionnaire (eight questions), completed by 80% of households in Canada and a long questionnaire with 53 additional questions completed by 20% of the population. As a result, the 2006 census files at ICES contain aggregated data for Ontario and Canada that describe the general demographic information on 100% of the population including age, sex and marital status. The remaining information including income, education and ethnicity is taken from 20% of the population. We used Census data to identify the neighborhood characteristics of individuals living in Ontario. The postal code of an individual on the census was linked to the Dissemination Area (DA) using the Postal Code Conversion File 6+ (described below). In this study, we used the dissemination area (DA) as the standard geographic area for the analysis. DA’s have specific criteria that maximize their usefulness for data analysis(Statistics Canada). The dissemination area (DA) is the smallest standard geographic area for which one surveyor disseminates all census data thus increasing the reliability during the data collection. DAs cover all the territory of Canada unlike Census Tract (CT) that covers the metropolitan areas only. In most cases, DAs are uniform in terms of population size. They have between 400 and 700 individuals. They are stable over time thus can be used to compare between censuses. DA’s are compact in shape and composed of one or more adjacent dissemination blocks. They

23

can be added together or 'aggregated' to create any of the other standard geographic areas such as Census subdivisions (CSDs) and census tracts (CTs) (Statistics Canada) The Postal Code Conversion File is the digital file that provides correspondence between the Canada Post Corporation (CPC) six character postal code and Statistics Canada’s standard geographical areas for which census data and other statistics are produced. The link of the postal codes to the standard geographic areas using PCCF permits the integration of data from various sources. The PCCF is updated on a regular basis and is released every six months. The PCCF flags automatically the neighborhood income quintile of the individual and the urban/rural status (a full description is provided in the section on measuring neighborhood characteristics). Method Data linkage From the RPDB, cohorts of individuals eligible for health coverage in the Province of Ontario aged 50-74 were identified for each of the following years: •

April 1st, 2005- March 31st, 2006

•

April 1st 2006- March 31st 2007

•

April 1st 2007- March 31st 2008

•

April 1st 2008- March 31st 2009

•

April 1st 2009- March 31st 2010

•

April 1st 2010 – March 31st 2011.

For each cohort year, the following exclusions were applied: 1- Previous history of colorectal cancer since 1974 using the following codes from the OCR database •

ICD-9 codes: 153.0-153.4; 153.6-154.1

•

ICD-10 codes: C18, C19, C20, C21, C180, C182-C184, C186-C189

2- History of Crohn’s disease and Ulcerative colitis using CIHI database: •

ICD-9 codes: 556, 556.0 to 556.9 and 555, 555.0 to 555.9

•

ICD-10 codes: K500, K501, K508 to K515

24

3- Individuals excluded by their physician using OHIP billing code ‘Q142’. The exclusionary code for colorectal screening Q142 is used for the following: (i) Enrolled Patients with known cancer being followed by a physician; (ii) Enrolled Patients with known inflammatory bowel disease; (iii) Enrolled Patients who have had colonoscopies within five (5) years; (iv) Enrolled Patients with a history of malignant bowel disease; and (v) Enrolled Patients with any disease requiring regular colonoscopies for surveillance purposes. Using the encrypted numeric identifier (IKN), we linked each cohort on the RPDB list to OCR to identify and exclude those who had previous colorectal cancer, and to CIHI-DAD for Crohn’s disease or ulcerative colitis. Further, eligible individuals for screening were linked to OHIP to identify those who received screening tests in each year from 2005-2011 and those who received tests in previous five years from the beginning of the cohort. Using the postal code from RPDB, we linked the cohort to the PCCF file in order to assign the DA, neighborhood income quintile and the urban/rural status for each individual. (Appendix 3: Data linkage flow chart) Study cohorts First study For the first study, we were interested in describing the trend of FOBT and endoscopy, colonoscopy and flexible simgoidoscopy, participation among individuals ‘due’ for screening from April 1st 2005 till March 31st 2011. We used the physician billing codes to identify and exclude individuals who received CRC tests prior to the beginning of each annual cohort: we excluded those who received FOBT in previous 12 months, or a flexible simgoidoscopy and colonoscopy in previous 48 months using the following codes: •

FOBT: L181 (FOBT prior to CCC) & L179 (after the CCC) G004 (regular medical

check-up), Q152, Q118-Q123

25

•

Flexible simgoidoscopy: Z555 (without E740 or E741 or E747 or E705) or Z580

(flexible simgoidoscopy using 60 cm scope) •

Z555 with E740 (flexible simgoidoscopy or colonoscopy to splenic flexure)

•

Z555 with E740 and/or E 741 and/or E740 and/or E 741 and/or E 747 (incomplete

colonoscopy ) •

Z555 with E740 and E 741 and E 747 and E 705 (colonoscopy into terminal ileum)

In this thesis, we combined colonoscopy and flexible endoscopy into one test and named it endoscopy. This choice was based on the complexity of billing for colonoscopy seen above. This complexity stems from the fact that each combination of codes is used incrementally depending on the section of the colon reached. This resulted in an overlap of fee codes used for an incomplete colonoscopy and flexible simgoidoscopy (Z555 and E740). Therefore, it is almost impossible to discern whether the physician billed for a flexible simgoidoscopy or for an incomplete colonoscopy (Vinden C, 2004). The exclusions were based on the following criteria: 1- Canadian guidelines recommend a biennial FOBT for average risk individuals. In this thesis, individuals who received FOBT before 12 months from the beginning of each fiscal year were kept in the cohort because according to guidelines, they will be due for another test anytime during that fiscal year. 2- The 5-year cut-off point rather than the 10-year for individuals who received endoscopy was used for three reasons: first, each of colonoscopy and flexible simgoidoscopy has different recommendation for repeat test. Flexible simgoidoscopy is recommended every five years and colonoscopy every ten years for average risk individuals. A 5-year cut off point is a conservative measure for both tests. Second, other Canadian studies used the 5year cut-off point in their definitions, so we used it for comparability(31, 36). Third, given the small proportion of individuals who received endoscopies in the look back period between 1995-2001 (1.5-3%), excluding them would not affect our estimates (Jacob, et al., 2011) 3- Individuals who received endoscopy or barium enema before 48 months from the

26

beginning of each fiscal year were kept because they will be due for another test anytime during that year according to guidelines. Second study For the second study, we were interested in testing the influence of contact with physician and physician’s recommendation on FOBT participation. We used two annual cohorts following the CCC to be able to capture the physician’s incentives, which became effective after April 1st 2008. Physician’s recommendation codes were: •

Q150A for distribution of the kit and counseling during an office visit

•

Q005A for sending a letter or calling individuals inviting them to take an appointment to

discuss colorectal cancer screening. In each year, each individual who had at least one contact with a general practitioner/family physician in Ontario for a core primary service was assigned a physician on OHIP database. We used this variable to identify an individual who had at least one contact in one year. Third study For the third study, we used all six cohorts identified above. We traced individuals who were due for screening in each quarter of the year. We describe the algorithm for tracing eligibility in each quarter below. Definitions of variables Demographics Individual characteristics The RPDB provides individual information on the following variables: age, gender and eligibility for Ontario Health Insurance plan. We used these variables as follow: Sex: a dichotomous variable we referred to it as: Female (F) & Male (M) Age: was originally a continuous variable, we transformed into 5 years interval: 50-55; 56-60; 61-65; 66-70; and 71-74.

27

Recent registrant: In this study, we used recent registration on OHIP as a proxy measure for recent immigration. It is estimated that about 80% of recent registrants are recent immigrants (95). Recent registrant on Ontario Health Insurance Plan was calculated based on the date the individual started to be eligible for OHIP services. An individual whose eligibility for OHIP started within five years from the beginning of the cohort, i.e. for 2005 cohort, if eligibility started after the second quarter of the 2000 fiscal year, the individual was considered as recent registrant; otherwise, if eligibility started before the second quarter of the 2000 fiscal year then the individual was considered non-recent registrant. Neighborhood characteristics Neighborhood Income quintile and quintiles of population by neighborhood income Since the administrative data used in this study does not provide data on the income of individuals, the aggregate level was used to measure income commonly referred to as the Neighborhood Income Quintile. The method developed by Borugian et al. and used by the PCCF assigned a quintile for a neighborhood income for each dissemination area as follow(117): •Census data collects information from 20% of the population on the average household income and the number of individuals in a household in addition to their postal code. •Using this information, it calculates the average income per single person equivalent (SPE) in a dissemination area based on the following formula: Income per Single Person Equivalent in a DA= average household income x the number of households/ total number of Single Person Equivalent. How to derive the SPE? Each household in each DA is single, or two person, or three person and so on. A single person costs more to live than two or more people living together. To adjust for the household size, each person in the household is given a weight (multiplier) and the total number of single person equivalent is the sum of all the person weights in each DA. Statistics Canada have a more or less standard scale for person weight: “the oldest person in the family receives a factor of 1.0; the second oldest person in the family receives a factor of 0.4; all other family members aged 16 and over each receive a factor of 0.4; all other family members under age 16 receive a factor of 0.3” (S. Canada)

28

• This measure is captured in the PCCF and identified as the (IPPE) or Income Per Person Equivalent in a DA. • IPPE for all DA’s are ranked and divided into five quintiles of equal population size. • Because the cost of living within a large census metropolitan area are comparable within each census area but not between census areas, in order to create a standard to identify 1

those living below a “low income cut off (LICO) within each area, a quintile of the neighborhood income is calculated for each aggregation and then pooled across areas. For example, in Ontario, there are 15 CMA. For each CMA, a quintile of neighborhood income is derived and then all quintiles are pooled together. This measure is captured in the PCCF and identified as the QAIPPE (Quintile of neighborhood income per person equivalent within a CMA, CA or Residual Area). The QAIPPE was used as a measure of the neighborhood income quintile of the individual in this study. Urban/ Rural Index Several definitions for “rural” are available for national and provincial analysis using the databases at Statistics Canada(du Plessis, 2001). Each definition emphasizes different criteria such as population size, density and context. In this study, the rural definition used by the PCCF (du Plessis, 2001) which takes into consideration the size of the community is adopted. A rural area is defined in terms of a community size < 10,000. The PCCF has the capability of adding a flag for the corresponding urban/rural status.

1

LICO reflects a consistent and well-defined methodology that identifies those who are substantially worse off than average (Stats Canada). The Canada average family expenditure on food, shelter and clothing is calculated. This is expressed as a percentage of pre-tax income. When families spend 20 percentage points of their income more than the Canadian average on food, shelter and clothing, then these families live below the LICO. The FAMEX data is used to measure the amount spent on these necessities. The income levels, differentiated by size of area of residence, family size become the base year low-income cut-offs. The LICO are updated annually based on the consumer price index.

29

Analyses For the first study, we used descriptive statistics to measure the rates of CRC testing and logistic regression to test whether the change in trend over time was significant. For the second study, we used unadjusted and multivariate log binomial regression to examine the relation between contact with physician, physician’s recommendation and demographic factors on FOBT participation. Detailed descriptions for the first and second study are provided in chapter 3 and 4. Herein, we provide a detailed background for the interrupted time series and segmented regression used in the third study. Time Series Analysis Evaluation is an important tool used by policy makers to reveal whether an intervention achieved its intended objectives and identifies elements for improvement. Different types of evaluation can be done at different phases of the program and each is associated with its own method and statistical techniques (121, 122). Choosing the right method and statistical technique is an important step in the evaluation process(Cook TD, 1979) Health care evaluation research distinguishes between experimental and natural or quasiexperimental interventions in choosing the appropriate method and statistical analysis for evaluation(Cook TD, 1979). In an experimental setting, the randomization between study and control allows researchers to use randomized controlled trial (RCT) to control for bias. In natural experiments, where randomization to study and control is not possible, Interrupted Time Series (ITS) offer the strongest quasi-experimental design to control for potential biases (Cook TD, 1979; Ray, 1997). Interrupted time series are designed to compare the change in outcome before and after the introduction of a program (124, 125) using data collected over time for the outcome. Segmented regression analysis allows estimating in statistical terms the effect of an intervention on the outcome, immediately and over time (126, 127). Interrupted time series have been used to evaluate the effectiveness of population-based programs (P. R. Baker, Francis, Soares, Weightman, & Foster, 2011; Gillings, Makuc, & Siegel, 1981; Glenton, Scheel, Lewin, & Swingler, 2011; R. K. Heyding, A. M. Cheung, E. J. M. Mocarski, R. Moineddin, & S. W. Hwang, 2005; Morrato, et al., 2009; Novoa, Perez, Santamarina-Rubio, & Borrell, 2011; Oyo-Ita, Nwachukwu, Oringanje, & Meremikwu, 2011), the effect of health policies and guidelines on disease outcomes (Ansari, et al., 2003; Kastner, et

30

al., 2011; Mol, et al., 2005; Pande, Ross-Degnan, Zaslavsky, & Salomon, 2011; Peto, et al., 2008; Rodriguez-Bano, 2010; Valiyeva, Herrmann, Rochon, Gill, & Anderson, 2008; Walthour, Seymour, Tackett, & Perri, 2010), the impact of natural disasters on the health care system (Rodgers, St John, & Coleman, 2005; Wicke & Silver, 2009; Yang, Xirasagar, Chung, Huang, & Lin, 2005) and the effect of disease outbreaks on health system (Chang, et al., 2004; Huang, Lee, & Hsiao, 2009; Hwang, Cheung, Moineddin, & Bell, 2007; Moineddin, Nie, Domb, Leong, & Upshur, 2008). Heyding et al. (R. K. Heyding, A. M. Cheung, E. J. Mocarski, R. Moineddin, & S. W. Hwang, 2005) used interrupted times series to test the effect of community-based program on mammography utilization among disadvantaged women in Toronto and found that mammography participation increased from 4.7% pre-intervention to 29.2% post intervention (p