Abstracts

Rationale: Adjustment for the confounding effects of comorbid conditions is vital in observational studies with disease or mortality outcomes. A number of comorbidity indices have been developed to serve as standardized tools for this purpose. We compared the performance of the Charlson Comorbidity Index and the more recently developed Epilepsy Specific Comorbidity Index, in predicting mortality in a cohort of epilepsy patients.

Methods: In a nation-wide register-based cohort study, we identified all individuals with a first diagnosis of epilepsy in the Danish National Patient Registry from 1 Jan 1980 to 31 Dec 2016. The day of the first diagnosis with epilepsy was defined as the index date. Information on comorbid conditions was obtained from diagnoses recorded in the same registry, as well as the Danish Central Psychiatric Registry, with an admission date on or before the epilepsy index date. We calculated total prognostic scores from the weighted scores assigned to each diagnosis in the Charlson and the Epilepsy Specific Comorbidity index. Total scores were stratified into five categories: ‘0 points’, ‘1 point’, ‘2 points’, ‘3 points’ and ‘4 points or more’. We followed the cohort from date of epilepsy diagnosis (index date) and until death, censoring or end of follow-up in 2016, whichever came first. We assessed the ability of the indices to predict mortality and discriminate between risk strata by Kaplan Meier-survival curves and Cox Proportional Hazards regression analyses, adjusted for sex and age at epilepsy diagnosis.

Results: In a study population of 8,730,911 individuals, we identified 143,478 with a first diagnosis of epilepsy (54% males) with a median age at diagnosis of 42 years (interquartile range 17-65 years). The cohort was followed for a mean of 11.6 years (standard deviation 9.8 years). The Charlson Comorbidity Index produced steeper Kaplan Meier-survival curves with clearer discrimination between risk scores (Figure 1), compared to the Epilepsy Specific Comorbidity Index, in which the mid-range stratified curves for the total scores crossed (Figure 2). There was a stronger mortality gradient across risk scores in the adjusted Cox model for the Charlson Comorbidity Index (HRs for total scores 1, 2, 3, and 4+ points vs 0 points: 1.45 (95% CI 1.42-1.48), 2.12 (95% CI 2.06-2.17), 2.32 (95% CI 2.24-2.39), and 3.47 (95% CI 3.37-3.58)), than for the Epilepsy Specific Comorbidity Index (HRs for total scores 1, 2, 3, and 4+ points vs 0 points: 1.14 (95% CI 1.10-1.17), 1.64 (95% CI 1.59-1.68), 2.23 (95% CI 2.17-2.29), and 2.52 (95% CI 2.45-2.60)).

Conclusions: The results suggest that in a large cohort of epilepsy patients from the Danish National Patient Registry, the Charlson Comorbidity Index is closer associated with mortality than the Epilepsy Specific Comorbidity Index.

Funding: Please list any funding that was received in support of this abstract.: This research projected was funded by the Lundbeck Foundation and the Novo Nordisk Foundation (NNF16OC0019126).