Objectives: To assess the extent of discrepancies between different vascular registries, at various levels of validation, and to investigate whether such differences might alter the morbidity and mortality rates obtained from the gold standard dataset for carotid endarterectomy (CEA).
Methods: All CEA operations in Helsinki University Central Hospital from 2000-2005 were retrieved from the local vascular registry (HUSVASC) and the Hospital Discharge Registry (HILMO). Both registries were validated at different levels to form the final dataset. Total and indication-specific perioperative morbidity and mortality rates were estimated from each level of validation and compared with those from the final dataset and with pooled rates from systematic reviews.
Results: Initial search provided 675 and 681 CEAs from HUSVASC and HILMO, respectively, decreasing to 636 (94%) and 614 (90%) when using the specific operative codes for thrombendarterectomy and patch angioplasty. Manual verification of initial HUSVASC results proved that 655 (97%) operations were true CEAs. 18 further proven CEAs, registered only in HILMO, were added to form the final CEA dataset (n=673). The peri-operative morbidity and combined morbidity and mortality rates were 2.23% and 2.67%, respectively. Comparable rates were obtained from both registries, irrespective the level of verification.
Conclusion: Registry data do not appear to be biased by random loss of some operations and thus they are reliable for decision-making. However, further research is still needed to estimate the permissible volume of omissions in a registry for the data-base to remain trustworthy.