High completeness of the brest stroke registry evidenced by analysis of sources and capture-recapture method

Abstract

Background: Population-based stroke registries are necessary to evaluate the precise burden of stroke. The methodology used in the Brest Stroke Registry and an estimation of its completeness are described.

Methods: 'Hot pursuit' as well as 'cold pursuit' were used, and five sources of identification were included: emergency wards, brain imaging, practitioners, death certificates and hospital-based electronic research. Ascertainment for each case was certified by a neurologist. Inclusion criteria were: (1) age >15 years; (2) a stroke defined by WHO criteria or all neurological deficits lasting at least 1 h. Completeness was estimated using capture-recapture method.

Results: For 2008, 2009 and 2010, 851, 898, 823 patients were collected, respectively. The number of sources of identification per patient was as follows: one source: 30.8, 24.1 and 18.7%; two sources: 54.5, 42.9 and 31.0%; three sources: 13.4, 30.1 and 46%; four sources: 1.3, 3.0 and 3.8%. Capture-recapture analysis showed data completeness over 90%. Standardized cumulative first-ever stroke incidence using a world standard population was 87 in 2008, 87 in 2009 and 84 in 2010.

Conclusions: Case ascertainment by a neurologist, numerous sources, as well as 'hot' and 'cold' pursuit can provide a reliably large data set suitable for further epidemiological studies.