Algorithm for Identifying Drug-Resistant Epilepsy in Danish Health Registers

Overview

A novel algorithm was developed to identify drug-resistant epilepsy (DRE) using Danish national health registers, validated against a medical record gold standard. The algorithm demonstrated moderate sensitivity and high specificity, identifying 10.3% of a large epilepsy cohort as having DRE and revealing key risk factors including early epilepsy onset and substance abuse.

Background

Epilepsy affects approximately 1% of the population, with up to one-third developing drug-resistant epilepsy (DRE), which is linked to increased morbidity and mortality. Traditional clinical studies on DRE are limited by small sample sizes and resource-intensive data collection, while administrative databases offer large-scale data but often lack validated definitions. The International League Against Epilepsy defines DRE based on failure of two adequate antiseizure medication trials, a concept difficult to capture in administrative data without detailed clinical information. This study aimed to create and validate a computable phenotype algorithm for DRE using Danish prescription and hospital registers.

Data Highlights

Key Findings

• The best-performing algorithm defined DRE as either filling prescriptions for ≥3 distinct antiseizure medications within 3 years or an acute hospital visit for epilepsy/convulsions after filling prescriptions for two distinct ASMs.
• The algorithm showed moderate sensitivity (59%) and high specificity (93%) against a medical record-validated gold standard.
• Applying the algorithm to a large national cohort identified 10.3% of individuals with incident epilepsy as having DRE.
• Multivariable logistic regression revealed early epilepsy onset, focal or generalized epilepsy type, somatic comorbidity, and substance abuse as independent risk factors for DRE classification.
• The algorithm enables large-scale research on DRE using administrative data without requiring detailed clinical chart review.

Clinical Implications

This validated algorithm facilitates identification of patients with drug-resistant epilepsy in large administrative datasets, enabling epidemiological and health services research on DRE at scale. Clinicians and researchers can use this tool to better understand risk factors and outcomes associated with DRE, potentially guiding earlier interventions for high-risk groups such as those with early onset epilepsy or substance abuse. The approach may improve resource allocation and support development of targeted therapies.

Conclusion

The study successfully developed and validated a register-based algorithm to identify drug-resistant epilepsy with good specificity, enabling large-scale research and risk factor analysis. This tool represents a significant advance in leveraging administrative data for epilepsy research and may inform clinical and public health strategies.

References

1. International League Against Epilepsy (ILAE) 2010 -- Definition of Drug-Resistant Epilepsy
2. Danish National Patient Registry -- Data Source for Epilepsy Research
3. Original Study Authors 2024 -- A Methodology for Identifying Drug-Resistant Epilepsy in Danish National Health Databases