Administrative databases present a viable alternative to disease registries when it comes to conducting research on multiple sclerosis. However, it’s important to note that administrative databases are not initially designed with research in mind. Therefore, it becomes imperative to thoroughly evaluate their performance in this context.
In this study, our primary objective was to assess the performance of the French administrative database, which includes hospital discharge records and national health insurance data, in its ability to accurately identify individuals with multiple sclerosis. We conducted this assessment by comparing it to a meticulously maintained registry that comprehensively compiles cases of multiple sclerosis among residents in the Lorraine region of northeastern France, serving as our reference point.
The study encompassed the period from 2011 to 2016 and involved the thorough examination of all individuals residing in the Lorraine region who were identified as having multiple sclerosis, either through the administrative database or the exhaustive registry. In our analysis, we employed various statistical measures, including the Matthews correlation coefficient, to gauge the level of agreement between the two data sources.
Remarkably, the Matthews correlation coefficient for the administrative database stood at 0.79 (95% CI 0.78–0.80), signifying a moderate level of performance in identifying individuals with multiple sclerosis. Intriguingly, our findings revealed a substantial time gap in the identification process. The registry consistently identified cases a noteworthy 5.5 years earlier than the administrative database.
These results underscore the need for caution when utilizing administrative databases for multiple sclerosis research. While they do offer utility in such studies, it is essential to be mindful of potential biases that may arise from their use. The findings from our study accentuate the value of regional registries, which offer a more exhaustive and expedited means of identifying cases, ensuring greater accuracy and reliability in research outcomes.