Abstract
Background
Keywords
Findings
Authors
What is already known on this subject?
What does this study add?
Background
Record linkage is commonly used in health research to fill data gaps. This study summarizes the linkage of the 2006 Census of Population (excluding Quebec) to hospital data from the Discharge Abstract Database (DAD).
Data and methods
Hierarchical deterministic exact matching was employed to link 2006 Census and DAD (2006/2007, 2007/2008 and 2008/2009) data, based on linkage keys derived from three variables common to both files—date of birth, postal code and sex. The full census file (short-form; 23.4 million) was used for record linkage; the 20% file (long-form; 4.65 million) representing the study cohort was used for validation. Linked files were compared across jurisdictions, years and other selected covariates in terms of eligibility for linkage, keys linked, and linkage and coverage rates.
Results
Overall, 80% of linkage keys identified in the DAD were linked to the 2006 Census. The percentage of long-form census respondents linked to at least one hospital record ranged between 5% and 8% across jurisdictions; linkage rates were higher among known high users of hospital services: older age groups, lower-income individuals, and Aboriginal people. In general, the linked census file represents the majority of hospital events that occurred during the study period. Coverage rates (weighted/unweighted) varied by geography and age group, with lower weighted rates for the territories and some younger age groups.
Interpretation
With hierarchical deterministic exact matching, census data can be linked to multiple years of DAD data. Incorporation of updated postal codes from tax files reduced linkage rate attrition over time. Lower coverage rates for the territories and younger age groups suggest that these populations may be underrepresented in the linked files.
Keywords
Coverage, data collection, data linkage, databases, medical records, hospital records
Findings
Record linkage, the process of matching records across or within databases, is commonly used in health research to fill data gaps and create a dataset with broad applications. Most health-related linkages in Canada have relied on health insurance numbers (HINs) from provincial health registries, which are unique to individuals. However, HINs are not available in most databases (for instance, mortality, census, tax), and access to provincial registries is limited. [Full Text]
Authors
Michelle Rotermann (Michelle.Rotermann@statcan.gc.ca) and Claudia Sanmartin are with the Health Analysis Division, Richard Trudeau (Richard.Trudeau@statcan.gc.ca) is with Special Surveys Division, and Hélène St-Jean is with the Household Survey Methods Division at Statistics Canada, Ottawa, Ontario.
What is already known on this subject?
- Record linkage is a cost-effective way to add value to administrative and survey datasets.
- Combining complementary data sources permits analyses that would otherwise not be possible.
- The use of provincial health insurance registries to facilitate record linkages for population health research is well established.
- Registry and non-registry approaches to linking health administration and 2006 Census information have yielded research-quality data for the provinces of Ontario and Manitoba for 2006/2007.
What does this study add?
- This study presents the results of a hierarchical deterministic exact matching approach to link 2006 Census data to 2006/2007-to-2008/09 hospitalization records in the Discharge Abstract Database (DAD) for all Canadian jurisdictions (excluding Quebec,) based on date of birth, postal code and sex.
- Use of tax data helped to ensure that the postal codes of census respondents were up-to-date, thereby minimizing missed links.
- The linked cohort represents approximately 80% of hospitalizations that occurred during the 2006/2007-to-2008/2009 period.
- The nationally representative linked census-DAD file fills a data gap and offers new research opportunities.
- Date modified: