• Gorana Capkun, Raquel Lahoz, Elisabetta Verdun, Xue Song, Weston Chen, Jonathan R Korn, Frank Dahlke, Rita Freitas, Kathy Fraeman, Jason Simeone, Barbara H Johnson, and Beth Nordstrom.
• Novartis Pharma AG , Basel , Switzerland.
• Curr Med Res Opin. 2015 May 1; 31 (5): 1029-39.

ObjectiveAdministrative claims databases provide a wealth of data for assessing the effect of treatments in clinical practice. Our aim was to propose methodology for real-world studies in multiple sclerosis (MS) using these databases.Research Design And MethodsIn three large US administrative claims databases: MarketScan, PharMetrics Plus and Department of Defense (DoD), patients with MS were selected using an algorithm identified in the published literature and refined for accuracy. Algorithms for detecting newly diagnosed ('incident') MS cases were also refined and tested. Methodology based on resource and treatment use was developed to differentiate between relapses with and without hospitalization.ResultsWhen various patient selection criteria were applied to the MarketScan database, an algorithm requiring two MS diagnoses at least 30 days apart was identified as the preferred method of selecting patient cohorts. Attempts to detect incident MS cases were confounded by the limited continuous enrollment of patients in these databases. Relapse detection algorithms identified similar proportions of patients in the MarketScan and PharMetrics Plus databases experiencing relapses with (2% in both databases) and without (15-20%) hospitalization in the 1 year follow-up period, providing findings in the range of those in the published literature.LimitationAdditional validation of the algorithms proposed here would increase their credibility.ConclusionsThe methods suggested in this study offer a good foundation for performing real-world research in MS using administrative claims databases, potentially allowing evidence from different studies to be compared and combined more systematically than in current research practice.

20 keywords...

hide…