How to improve specific databases for clinical data in rare diseases? The example of hereditary haemorrhagic telangiectasia.

INTRODUCTION: The objectives of reference centres for rare diseases are multiple and mainly concern disease management and coordination between specialties, but they also have to improve knowledge through epidemiological studies and biomedical research. A first database was created by the hereditary haemorrhagic telangiectasia network to achieve these objectives, but facing a lack of data entered in the first database, we established a new database, named CIROCO (Clinical Investigation for the Rendu-Osler Cohort). This new database was constructed after the first database assessment.

METHODS: We listed all difficulties encountered in the first database. We focused on three themes: database technical characteristics, database design (the number and characteristics of variables and the overall structure) and data entry. Based on this expertise, we defined the characteristics that should lead to an optimization of the database. We then compared the performance of these databases in terms of available data.

RESULTS: The first database had 1273 fields spread out on 14 forms. A total of 838 patients was entered from 2001 to 2005 and among the 1273 fields, 205 (16%) had no data at all. The new database gathered 362 fields from the first database and 173 new fields. These fields were rearranged into 443 variables. A total of 1410 patients directly seen by the clinicians were entered. For patients seen by clinicians each variable used for defining the disease was available for at least 93% of patients.

CONCLUSION: Items to be included in a specific database should be carefully selected in order to achieve good results and an epidemiological utility.