Data were obtained from the Ontario Registered Persons Database (RPDB), the Ontario Health Insurance Plan (OHIP) database and the Canadian Institute for Health Information (CIHI) database. The RPDB contains demographic information on all Ontario residents ever covered under OHIP. The OHIP database records all physician claims in Ontario. The CIHI database contains diagnostic and procedural information on all patients discharged from hospitals and same-day surgery units. These databases have high rates of reliability for demographic, diagnostic and procedural information.27
Inclusion and exclusion criteria for the study cohort are summarized in . People aged 50 to 70 years living in Ontario on Jan. 1, 1997, who had a valid OHIP number were identified in the RPDB. Those who in the previous 5 years had had CRC, inflammatory bowel disease or colorectal investigation were excluded. The remainder approximated a cohort at average risk of CRC. People with missing geographic or suppressed income data were excluded.
A longitudinal record was created, and each person was followed until death or Dec. 31, 2001. People were classified as living in urban or rural areas according to 1996 Statistics Canada Census definitions28
applied to the postal code of the primary address. Health status on Jan. 1, 1997, was assessed with the Deyo score, as follows. Weights were assigned for specific comorbid conditions (e.g., myocardial infarction, diabetes mellitus without complications, renal disease), and a score was calculated for each person in the cohort with the use of codes of the International Classification of Diseases, 9th revision, recorded in the hospital discharge abstracts between Jan. 1, 1992, and Dec. 31, 1996, according to the method described by Deyo, Cherkin and Ciol.29
With this approach, a score of 0 indicates no comorbidity, and higher scores denote higher levels of comorbidity.
People who received any colorectal investigation (FOBT, barium-enema radiography, sigmoidoscopy, colonoscopy) were identified in both the OHIP and CIHI databases; Appendix 1
lists the procedural and diagnostic codes. Using administrative data, one cannot distinguish between screening and diagnostic investigations; however, identifying these procedures in an average-risk population approximates identifying screening procedures.
Because personal income is not included in Ontario administrative databases, we used 1996 Statistics Canada Census data to calculate the mean household income of the enumeration area in which the person lived. An enumeration area is the geographic area canvassed by 1 census representative; in 1996 the number of dwellings in each enumeration area varied from 125 to 440. The enumeration areas were classified into quintiles, each containing approximately 20% of the Ontario population, that were based on the mean household income: $34 370 for the first quintile, $47 792 for the second, $56 244 for the third, $65 949 for the fourth and $95 066 for the fifth. Census data have previously been used as a surrogate for personal income.21,22,23,26,30,31
We used analysis of variance (for continuous variables) and χ2 tests (for categorical variables) to determine if any statistically significant difference existed across income quintiles and between the group that received colorectal investigations and the group that did not. We used multivariate logistic regression to evaluate the independent association of income and receipt of any investigation, adjusting for age, sex, location of primary address (rural or urban) and the Deyo score; we calculated 95% confidence intervals (CIs) for each value. We performed separate analyses for people who received colonoscopy and those who did not. The level of statistical significance was set at p < 0.05.
We obtained approval for the study from the Sunnybrook and Women's College Health Sciences Centre Research Ethics Board.