Two approaches to linking census and hospital data.

Area of Science:

• Health Informatics
• Population Health Research
• Data Linkage Methodologies

Background:

• Comparing data linkage strategies for Canadian health research.
• Utilizing 2006 Census of Population data for linkage.
• Linking census data to the Discharge Abstract Database (DAD).

Purpose of the Study:

• To compare registry and non-registry linkage approaches.
• To assess the effectiveness of linking census and hospital data.
• To determine the feasibility of creating research-quality datasets without provincial registries.

Main Methods:

• Probabilistic and deterministic linkage using health insurance numbers (HINs) for the registry approach.
• Hierarchical deterministic exact matching on common variables for the non-registry approach.
• Comparison based on linkage rates, coverage, sensitivity, specificity, and HIN consistency.

Main Results:

• Similar linkage and coverage rates between registry (7% Manitoba, 5% Ontario) and non-registry (6% Manitoba, 5% Ontario) approaches.
• Registry approach linked 84% of hospital admissions; non-registry linked 82%.
• Both methods produced comparable results for linking census and DAD data.

Conclusions:

• A non-registry linkage approach is a viable alternative for creating research-quality datasets.
• This method is particularly useful when access to provincial health insurance registries is limited.
• Enables robust population health research using linked administrative data.