Abstracts

Rationale: Epilepsy is the most commonly reported neurological condition in primary health care in most of the world, affecting people of all ages, gender and socioeconomic status. As the prevalence of epilepsy is expected to rise with the aging population it is important to develop surveillance programs to monitor and project future demands epilepsy will place on society and the healthcare system. Comprehensive prospective cohorts are extremely costly and time-consuming. Passive surveillance using administrative data is a cost-effective way to follow individuals who have chronic conditions such as epilepsy. However, administrative databases must first be validated. Inpatient facilities across Canada use the ICD-10 system to code diagnoses, comorbidities and procedures in their administrative databases. The purpose of this study was to assess the validity of ICD-10 data from an inpatient administrative database.Methods: All inpatient visits at a Canadian tertiary care center during the fiscal year 2004 with an ICD-10 code in the primary diagnostic position corresponding to epilepsy, transient ischemic attack (TIA), syncope, or classical migraine were identified. This enriched population which includes patients with diagnoses other than epilepsy was specifically chosen because these conditions can resemble seizure. A random sample of these charts (n=132) were reviewed by an epileptologist and neurology resident and independently given an ICD-10 code. Sensitivity, specificity, positive and negative predictive values were calculated for ICD-10 epilepsy codes, as was the level of agreement between the trained health records technologists who had originally coded the charts and the epileptologist. The agreement between the epileptologist and the neurology resident was also calculated. The ICD-10 code given by the epileptologist to the particular hospital visit was considered the “gold standard”.Results: Epilepsy coding in the primary diagnostic position using the ICD-10 system was very good, with a sensitivity of 96% (95% CI 87.9-99.0) and specificity of 99% (95% CI 92.9-99.8). The positive predictive value of an ICD-10 epilepsy code was 0.98 (0.90-1.00 and the negative predictive value was 0.97 (95% CI 0.91-0.99). Agreement between the epileptologist and trained coder was excellent with a Kappa value of 0.95 (95% CI 0.90-1.00). Agreement between the epileptologist and the neurology resident (inter-rater reliability) was also excellent at 1.00 (1.00-1.00).Conclusions: Epilepsy cases coded in the primary diagnostic position can be accurately identified using inpatient administrative databases utilizing the ICD-10 coding system. Ongoing research targeted at validating secondary diagnostic positions coding, emergency visits and outpatient visits are necessary for the development of a future comprehensive population-based epilepsy surveillance program.