Ischemic heart disease (IHD) is one of the leading causes of death in high-income countries worldwide 1 . IHD is the most costly cardiovascular disease (CVD) in terms of physician services, hospitalizations, and lost productivity due to premature death. It accounted for over CDN $8 billion in direct and indirect health care costs in Canada in 2000 2 . IHD related hospitalizations and deaths in Canada have declined markedly since the 1970s 2 , possibly due to a combination of prevention, improved detection and treatment of the early stages of disease, improved management of IHD, and more timely and effective treatment of acute events 3 4 5 6 7 . While rates have declined, the number of people needing treatment has remained high as the percentage of elderly in the population grows 2 .

Currently no formal mechanism to track prevalence and incidence of IHD exists in Canada. Data sources such as surveys underestimate the prevalence of chronic health conditions since the information is self-reported and institutionalized elderly are commonly not surveyed 8 9 10 11 . Also, survey questions often lack the precision to adequately define IHD. Routinely collected administrative hospital data are commonly used for the surveillance of many health conditions in Canada 12 13 14 15 . While useful for assessing system performance and evaluating the quality of health care at the hospital level 16 17 , hospital discharge data alone may be inadequate for estimating population burden and trends of IHD and other chronic disease conditions that are often initially diagnosed and managed in the outpatient setting. Therefore, combining two administrative health data files: 1) hospital discharge abstract database, and 2) physician billing database may provide a more comprehensive description of diagnosed IHD and may be more suitable for monitoring the disease burden in the population.

The validity of IHD diagnostic codes in administrative data has already been evaluated using medical record reabstraction 18 19 20 21 22 23 24 25 or survey data as the gold standard 8 26 . A more recent validation study in Ontario used a sample of family physician Electronic Medical Records (EMR) as the reference standard 27 . Based on this study, we chose to assess three case definitions (one hospital discharge abstract and two or three physician claims within a one-year period). We selected a one-year period for the second or third physician billing code because allowing for a two- or three-year period had little impact on the sensitivity and specificity of the case definition 27 and the shorter duration allowed for more timely reporting of surveillance findings. In addition, since previous research on heart disease surveillance in Canada has relied on hospital discharge data alone 14 , we decided to include this definition for comparison purposes.

Our objective is to describe the process and methods used to establish national population-based surveillance of diagnosed IHD using administrative health databases from seven provinces covering approximately 95% of Canada’s total population. Specifically, estimates of the incidence and prevalence of IHD in the Canadian population using different combinations of hospitalization and/or physician billing claim databases and their consistency over time and jurisdiction are presented.

This population-based study demonstrated that the prevalence and incidence of IHD were underestimated using hospital data alone compared with a combination of hospital and physician records. Many IHD cases complaining of chest pain may not initially go to the hospital and instead consult a family physician for their symptoms 31 32 33 . Thus, hospital data are likely to capture severe cases, or individuals who are in the later stages of the disease. Physician claims not only capture services provided by family physicians but also services provided by specialists, regardless of the service location, including inpatient consultations. Therefore, these two databases are an invaluable source of information for chronic disease surveillance.

Accurate estimates of IHD occurrence depends on data quality. Following the clinical path of the family physician to specialist and then back to the family physician, we defined IHD in two ways (i.e. at least one hospitalization or two physician claims within one year and at least one hospitalization or three physician claims within one year) and found that 1H3P generated slightly lower prevalence and incidence estimates than 1H2P. Tu et al. assessed the validity of these two methods using physician chart review information in Ontario 27 . They found a sensitivity of 77% for 1H2P vs. 72% for 1H3P and a positive predictive value (PPV) of 75% for 1H2P and 77% for 1H3P. Given a 5% higher sensitivity with 1H2P compared to 1H3P and minimal impact on PPV, we propose adopting the 1H2P algorithm for the national surveillance of IHD in the CCDSS.

Compared to self-reported data from the 2007-2008 Canadian Community Health Survey (CCHS), crude IHD prevalence rates estimated with the 1H2P case definition ranged by province from 6.0% (95% CI: 6.0-6.0) to 9.2% (95% CI: 9.1-9.3) in 2006/07, while CCHS estimates ranged by province from 3.7% (95% CI: 3.2-4.2) to 7.1% (95% CI: 6.4-7.8) in 2007-2008. This discrepancy is likely because individuals who live in collective dwellings such as long-term care settings and those who are hospitalized are not captured by the CCHS, but are by administrative provincial/territorial databases. Moreover, survey data are sensitive to recall bias; some people may not recall having been told by a physician that they have IHD.

This study was strengthened by the use of a uniform definition to ascertain diagnosed IHD across seven provinces, as well as the availability of over ten years of data that allowed a sufficient clearance time to identify incident cases. Furthermore, we used a definition that included outpatients for which the validity has been demonstrated 27 .

A few limitations are worth noting. First, although the inclusion of physician claims may capture earlier stages of disease, our case definition does not capture undiagnosed cases or individuals who die of IHD prior to receiving a diagnosis, due to the absence of linkage of hospitalization data to vital statistics. In fact, sensitivity analyses conducted in Nova Scotia and Québec demonstrated that capturing fatal AMI events (ICD-9 410 or ICD-10 I21-I22) using vital statistics would increase incidence rates of AMI by 16% in Nova Scotia and 18% in Québec. Despite this likely underestimation of the existence of IHD in the population, the methods established here and the consistency of results across jurisdictions are valuable for monitoring trends over time. Second, incidence and prevalence indicators may be sensitive to changes in rules for reimbursement of health insurance schemes over time and across provinces or territories. It will be important to monitor these changes over time and to conduct periodic revalidation to ensure the integrity of ongoing national surveillance. Third, while our case definition was validated for the estimation of prevalence, we had only 12 years of historical information for the Canadian population. Thus, some cases identified as 'incident’ may have been previously diagnosed, leading to an overestimation of incidence in some categories.

The findings of this study indicate that chronic disease surveillance can benefit from using multiple data sources, including physician claims and hospital administrative data. The CCDSS has since been expanded to include IHD as part of its ongoing national surveillance, and now has the ability to provide Canadian-wide estimates of IHD prevalence, incidence and all-cause mortality.

The authors declare that they have no competing interests.

All authors contributed to the conception and design of the study and the interpretation of its findings. CR, CB (PHAC) and SD drafted the manuscript. CR, CB (PHAC) and CW performed the statistical analyses, coordinated the data acquisition from the provinces, and integrated the data at PHAC. CB (PHAC) and SD participated in the coordination of the study. KT, CB (INSPQ), CP, MS, LWS, KR, JC, and RP carried out the data derivation from provincial administrative data sources and reviewed the manuscript iterations. DR, HJ, YG, LML, and HQ contributed to the review of the manuscript iterations. All authors read and approved the final manuscript.

PHAC funded the data extraction and linkage for the seven provinces participating in this study.

Data derivation in Ontario was supported by the ICES, which is funded by an annual grant from the Ontario Ministry of Health and Long-Term Care (MOHLTC). The opinions, results and conclusions reported in this paper are those of the authors and are independent from the funding sources. No endorsement by ICES or the Ontario MOHLTC is intended or should be inferred.

The results and conclusions are those of the authors, and no official endorsement by Manitoba Health is intended or should be inferred.

Dr. Karen Tu is supported by a Canadian Institutes of Health Research Fellowship Award in Primary Care Research.

Dr. Hude Quan is supported by Alberta Innovates - Health Solutions.