Discussion

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This page has been archived on the Web.

The results of this study demonstrate that the CCR can be linked to hospital data, and thereby, can be used to report on selected types of cancer treatment. Linkage rates to the DAD and NACRS (Ontario) using HINs were 94% or more for breast and colorectal tumours in all provinces and years; rates for prostate cancer varied from between 75% to 79% in Manitoba to more than 90% in Ontario. Other linked data sources created using HINs that reported linkage rates at 75% have been judged to be of sufficient quality for use in research.^{6,45,46,47,48,49,50,51}

Given that the majority of cancer tumours in this analysis were eligible for linkage, linkage rates were unbiased. The resulting linked data are comprehensive and accurate representations of total single malignant tumour cohorts for breast, colorectal and prostate cancers for 2005 through 2008. The patterns of treatment rates by age and stage generally reflected expected outcomes. For all three cancers, the likelihood of linkage to a treatment tended to be lower among the oldest patients and for stage IV cancers—results consistent with expectation. For example, the lowest rates of radical prostatectomy and lower rates of colon surgery have previously been reported for patients aged 75 or older.^50,51

However, the feasibility of reporting treatment rates by type of procedure varied, based on the availability of data for record linkage. The CCR linked data could feasibly be used to report on treatments such as surgery occurring as day or acute care hospital procedures, but the data were less complete for treatments, such as radiation and pharmacotherapy, that may occur outside of acute care and same-day surgery hospital settings. The exception was Ontario where information from oncology centres and clinics was available via the NACRS data. This limitation to treatment linkage reflects the nature of treatment capture available in the DAD/NACRS, rather than weakness in the linkage methodology.

Comparisons with published population treatment rates suggest that, for at least two  cancers (breast and colorectal), the CCR linked data yielded expected rates for surgical treatment. The results of this study indicated that for breast cancer, surgical treatments such as breast-conserving surgery and mastectomy were the most prevalent types of treatment. This finding was confirmed by other published sources, which highlight these surgeries as integral to breast cancer treatment.^{38,52,53,54,55,56,57} Linked results resemble surgery rates previously reported for Ontario, Manitoba and Prince Edward Island^54,55 when linked  mastectomy and breast-conserving surgery/mastectomy rates are combined. Some differences were noted; for example, the surgery rates in Manitoba were generally below Cancer Care Manitoba’s breast cancer surgery indicator for 2010,⁵⁶ but similar to the mastectomy rate reported by the Canadian Partnership Against Cancer.⁵⁷ Surgery rates for colorectal cancer in Ontario were similar to published rates.⁵⁰

Conversely, surgical rates for prostate cancer derived from the linked data differed from other published reports. For example, surgical rates were lower than radical prostatectomy rates previously reported for Ontario⁵¹ and Manitoba.⁵⁶ However, surgical rates in this study were similar to prostatectomy rates reported for Ontario’s cancer system performance quality indicator⁵² and for the U.S. population.⁶²

Rates for treatments that commonly occur in other settings, such as radiation and pharmacotherapy, were consistently low in all provinces except Ontario where the NACRS data were used to capture treatments in oncology clinics. This is confirmed when rates for these treatments are compared with published provincial cancer control and system performance indicators.^52,56,65 The inability to accurately report rates for these types of treatments is less an issue of data linkage and more related to the comprehensiveness of available administrative data in capturing the full range of cancer treatments that occur in a variety of  settings. With the continued expansion of the NACRS data in Canada, comprehensiveness in reporting radiation and pharmacotherapy therapy may improve. Nonetheless, some types of treatment will not be captured through linkage to hospital records—for example, “watchful waiting” or “active surveillance” for low-risk prostate cancer cases,⁶¹ which involve regular blood testing for prostate-specific antigen and digital rectal exams that occur elsewhere.

Other factors warrant consideration in assessments of the feasibility of using linked data to report cancer treatment. The utility of linked information in reporting treatment depends on the comprehensiveness and accuracy of the data used in the linkage process, namely, the DAD and the NACRS. Turner et al.¹² concluded that information about breast surgery was more complete in the Manitoba Registry than in the DAD, which suggests that the linked CCR-to-DAD/NACRS breast treatment rates may be underestimates. By contrast, other researchers concluded that although physician billing data were more accurate than the DAD data, definitive breast surgery information in the DAD accurately reflected original patient charts.¹³

Even if information is comprehensively reported to the DAD/NACRS, other limitations in the hospital administrative data may hinder the accurate reporting of treatment rates using the linked CCR data. These limitations arise from existing standards of coding practice in the DAD/NACRS. For example, treatments involving lymph node interventions generally had a low frequency of occurrence. In the DAD, axillary lymph node procedures for breast cancer may not be recorded distinctly to radical mastectomy procedures, with the result that rates for procedures involving axillary lymph nodes may be under-reported.¹²

Similarly, for prostate, lymph node removal is not always captured when radical prostatectomy is recorded, so it is possible that only the radical prostatectomy was coded even though regional lymph nodes were removed.³⁴ Again, low rates for those diagnostic or surgical treatments reflect the nature of the information available in the DAD/NACRS, rather than weakness in the linkage methodology. Therefore, obtaining comprehensive or absolute counts of lymph node interventions from hospital data is not feasible, and rates for treatment groupings that included lymph node procedures may be underestimated. Furthermore, given existing coding standards, available diagnostic and procedure codes cannot be used to distinguish between breast-conserving surgery and open excisional biopsy in the DAD/NACRS; therefore, treatment rates for one or the other may have been underestimated.

Another potential limitation of the hospital administrative data sources is the varying completeness in volume of reported ambulatory visits in Ontario’s NACRS across years. Volume varied after changes to provincial mandatory reporting requirements in 2007.^26,27 Despite the decrease in the volume of submitted records of ambulatory visits, centres reporting chemotherapy and radiation were still required to submit abstracts to the NACRS. Consequently, the availability of records for linkage to day surgery, radiation and chemotherapy would likely not have been affected by decreased volume of reporting.⁶⁶

An important limitation of these results pertains to the scope of data sources used in the linkage. Nova Scotia NACRS records were not included, although starting in 2003, some hospital facilities started to report day surgery visits to the NACRS rather than to the DAD.^22,23,67 As a result, treatment reporting for Nova Scotia may be incomplete in the DAD. Also, for Manitoba, breast treatments may have been underestimated based on linkage to the DAD alone, because out-of-hospital lymph node diagnostic procedures and breast-conserving surgery are available elsewhere.⁶⁸

The study did not include in situ and multiple tumours, so the generalizability of the results of this study does not extend to such tumours. Future studies should attempt to include them to determine the feasibility of using linked data to report on treatment rates for these tumours.

An advantage of this linkage is that some inherent limitations of the DAD/NACRS were partly overcome through linkage to the CCR. The accuracy of reporting treatment for patients with dual cancer diagnoses was improved. Because coincident cancers can have overlapping treatment guidelines—for example, prostate and bladder cancers—incorrect attribution of treatment to a given tumour could occur using unlinked DAD/NACRS information. However, linked CCR-DAD/NACRS data yielded information about dual cancer diagnoses and the timing of tumour diagnosis, so the likelihood of incorrect attribution of treatment was reduced.

Another advantage of this methodology is the ability to locate information about treatment outside a patient’s province of residence before and after diagnosis, because the CCR was linked to the national DAD using HINs. Linkage depended on the patient’s using the same HIN and on out-of-province facilities’ recording this HIN correctly. The power of this methodology was demonstrated by prostate radiation treatment of residents of Prince Edward Island—almost 100% of their brachytherapy information was submitted by hospitals in New Brunswick, yet these linked successfully to prostate tumours in patients from Prince Edward Island.