Methods

View the most recent version.

Archived Content

Information identified as archived is provided for reference, research or recordkeeping purposes. It is not subject to the Government of Canada Web Standards and has not been altered or updated since it was archived. Please "contact us" to request a format other than those available.

Archived

This page has been archived on the Web.

Record linkage

Different record linkage approaches were used to accommodate differences in the availability of linkage keys across jurisdictions. Health insurance number, a variable used to link the CCR and DAD/NACRS, was available on the CCR for Nova Scotia and Prince Edward Island, but not for Ontario and Manitoba. Consequently, an indirect linkage method was used for Ontario and Manitoba, and a direct method was used for Nova Scotia and Prince Edward Island.

Indirect linkage method
The indirect method used to link the CCR to the DAD and the NACRS for Ontario and Manitoba involved a two-step process. First, the 2008 CCR was probabilistically linked to each respective provincial health insurance plan client registry to obtain a HIN. The Fellegi-Sunter probabilistic method was applied, based on name, date of birth, sex and postal code as linkage variables because they were common to both databases. The probabilistic linkage methods have been reported in detail elsewhere.²⁸ The result of the probabilistic linkage was a set of keys containing both the CCR unique patient identifier and the associated HIN. These keys were then used to deterministically link CCR records with national DAD records from fiscal 2004/2005 to 2010/2011 and Ontario NACRS records from fiscal 2004/2005 to 2010/2011.

Direct linkage method
A direct approach was used to link the CCR and DAD data in Nova Scotia and Prince Edward Island, based on the HIN available in both datasets. The HIN was used in combination with the tumour-reporting province as proxy for province/territory that issued the HIN, since HINs are not unique across provinces. These two pieces of information were used to deterministically link CCR records submitted by Nova Scotia and Prince Edward Island to the national DAD. Additional information about the direct linkage is available elsewhere.²⁰

Validation

To validate both the indirect and direct linkage methods, the eligibility of cancer tumours for linkage and linkage rates are reported in Table 2. For Ontario and Manitoba, the rate for eligibility for linkage is defined as the percentage of tumours that linked to the provincial registries of all tumours. For Nova Scotia and Prince Edward Island, eligibility is defined as the percentage of tumours with a valid HIN of all tumours. This analysis also provides information about potential biases among non-eligible records. Linkage rates between the CCR and the DAD/NACRS for each cancer type are reported by province, age and hospital service type in Table 3. Rates are defined as the percentage of eligible tumours that link to at least one hospitalization (any cause).

Cohort selection

Cancers included for feasibility assessment were female breast, colorectal and prostate. These cancers are the leading causes of new cancers in Canada^29,30 and therefore, are expected to provide large sample sizes for analysis. Furthermore, these cancers have different treatment pathways, which allow assessment of the feasibility of using linked data to capture a range of treatment types.

International Classification of Diseases for Oncology, Third Edition (ICD-O-3)³¹ codes were used to define all tumour cohorts; these were grouped using Surveillance, Epidemiology, and End Results (SEER) Program grouping definitions.³² Total counts by province and year (2005 to 2008) are reported in Table 1.

Breast cancer
Breast tumour ICD-O-3 codes were site/topography C50.0-C50.9, and where behaviour is malignant. The following morphologies were excluded: mesothelioma (9050-9055), Kaposi sarcoma (9140), and Hodgkin and non-Hodgkin lymphomas (9590-9992).

Colorectal cancer
ICD-O-3 codes to define colorectal cancer were: colon site/topography C18.0, C18.2-C18.9; C26.0; and rectum/Rectosigmoid Junction site/topography C19.9, C20.9; both with malignant behaviour. Histology for both sites excluded: mesothelioma (9050-9055), Kaposi sarcoma (9140), and Hodgkin and non-Hodgkin lymphomas (9590-9992).

Prostate cancer
ICD-O-3 code used to define prostate cancer was site/topography C61.9; and where behaviour is malignant. Histology included code ranges 8000-9049, 9056-9139, 9141-9589.

Treatment codes and groups

A comprehensive list of potential treatments was developed for each cancer type based on published sources, including cancer treatment guidelines^{33,34,35,36,37,38,39,40,41,42} and Facility Oncology Registry Data Standards (FORDS).⁴³ The lists were reviewed by members of the national advisory committee, technical experts at the provincial cancer agencies, and clinical experts when required. After consultations, a final set of treatments were selected and grouped to reflect the following treatment categories: lymph node and diagnostic interventions, surgical interventions, radiation, and pharmacotherapy. Appendix A presents lists of interventions included in each treatment category.

The Canadian Classification of Health Interventions versions 2003, 2006 and 2009^33,34,35 were used to define treatments in the DAD/NACRS for each treatment category (Appendix A).  For each tumour, all linked hospitalizations were used to determine whether at least one instance of a given treatment occurred during the follow-up period. All available intervention fields in the linked DAD (20) and NACRS (10) records were used. This was done independently for each treatment type because multiple treatments in a single hospital admission are captured as separate treatment events.

It is possible that cancer patients with co-morbid conditions requiring the same interventions as certain selected cancer treatments might inflate cancer treatment rates since it was not possible to definitively determine which health condition was being treated. Treatment guidelines for different health conditions may overlap―for example, excision of testes in cases of prostate or testicular cancers. Therefore, linked CCR-DAD information was used to identify cases with relevant dual cancer diagnoses, specifically, prostate and testicular cancers, or colitis and colorectal cancer. For dual diagnoses cases involving prostate and testicular cancers, if the testicular cancer CCR-based diagnostic date fell within one year before the prostate linked excision of testes, that surgery was not used in the compilation of the overall selected prostate surgery treatment rates. This was to reduce the risk of including surgeries to treat the testicular cancer, rather than prostate, in treatment rates calculations.

Follow-up periods

The follow-up period for inclusion of hospital records and treatment events depended on cancer type. The CCR-based tumour diagnostic date was the start date for calculating the follow-up period. For colorectal and female breast tumours, hospital records were included if the admission date was within 31 days before, or one year (365 days) after the linked tumour diagnosis date. For prostate tumours, the inclusion criteria were an admission date within 31 days before and two years (730 days) after the diagnosis date. The admission date of a tumour-linked hospital record was used to determine if the admission occurred during the follow-up period.

Service type

The DAD and the NACRS represent different frames of hospital services that are expected to have an impact on reporting treatment rates. The DAD includes all discharges from acute care facilities and same-day surgery visits, except in Ontario, and for several hospitals in Nova Scotia since 2003, where day-surgery visits are recorded in the NACRS.⁴⁴ For Ontario, the NACRS data include visits to a broader range of services including: emergency room visits, day/night surgery, oncology clinics, Cancer Care Ontario for oncology care, and other types of ambulatory care (for example, renal dialysis clinics).^{23,24,25,26,27} The Ontario NACRS data provided an opportunity to link tumours to treatment data that were comparatively unavailable for the other three provinces; as a result, overall treatment rates are expected to be higher in Ontario. To adjust for differences in data availability, linkage and treatment rates are reported overall and by service type. For all provinces except Ontario, this represents the combined day surgery and acute care visits. For Ontario, this also includes all other service types of hospitalizations captured in the NACRS. For treatment rates, results for Ontario are reported in two ways: overall treatments rates, which include treatments identified in both the DAD and the NACRS, and also, rates using acute/day surgery only to allow comparisons with the other provinces. Day surgery records for Ontario included both DAD and NACRS records.

To classify hospital records by service type, for the DAD, the analytical institution type code field was used to indicate acute care (1) or day surgery (A).²¹ For Ontario, the NACRS Management Information System (MIS) Functional Centre Account Codes were used to classify each record as day surgery, oncology centres/clinics,^{23,24,25,26,27} or “other” ambulatory care. The last consisted of records that were neither acute care nor day surgery records.

Treatment rates

Treatment rates for each cancer are reported by province, year and type of treatment. The numerator is the number of tumours having at least one occurrence of a selected treatment during the follow-up period. The denominator is the number of tumours eligible for linkage to the DAD and/or the NACRS.