Algorithms for identification of Guillain-Barré Syndrome among adolescents in claims databases

Background: Health insurance claims databases can provide data for studies of vaccine-related Guillain-Barre' Syndrome (GBS), but not all patients with a diagnostic ICD-9-CM code for GBS have the disease. The objective of this study was to evaluate the positive predictive values (PPVs) of claims-based algorithms for identifying GBS cases in 4 claims database environments.

Methods: Potential cases were adolescents ages 11-21 with at least one claim for GBS (ICD-9-CM code 357.0). Medical record reviews by a panel of 3 neurologists were conducted for case confirmation. Claims data considered for inclusion in the case-ascertainment algorithm included coding position, physician specialty, visit type, diagnostic tests. PPVs were used to assess the contribution of study factors in predicting case status.

Results: Among 361 individuals with a GBS diagnosis code, 106 were confirmed overall (PPV=0.29), varying from 0.24 to 0.56 across the 4 sites. Requiring the GBS code to be associated with a neurologist visit (PPV=0.53) or to be in a primary position on an inpatient claim (0.56) improved the performance. A composite algorithm including a primary inpatient GBS code and a neurologist visit associated with any GBS code gave the highest PPV (0.70). Incorporating claims for diagnostic testing had little impact on the PPV. Findings were generally similar across study sites.

Conclusions: Algorithms were able to identify GBS cases better than the single occurrence of the diagnostic code for GBS, and these algorithms may perform similarly in different claims environments.