Abstracts

Rationale: A TDSS is a computer-based algorithm that can potentially help clinicians select optimal treatment for their patients based on patient- and drug-related data. To guide the development of a TDSS in epilepsy, a pilot study was conducted to explore whether information in electronic medical records (EMRs) and claims data are of sufficient quality to allow evaluation of outcomes among patients treated with antiepileptic drugs (AEDs). Methods: Anonymized EMRs combined with aggregated multichannel US claims data were obtained from HealthVerity. Combined records were included if patient was aged ≥18 years with an epilepsy diagnosis and contained clinical history ≥1 year before and after AED initiation, >1-month treatment period and >1 physician note in that period. Five US-based epileptologists reviewed records to assess the adequacy and quality of the reporting of outcomes, and to assess the success/failure of AED using the Physician Impression Index (PII; Table). The PII consisted of 11 questions pertaining to reviewers’ ability to assess AED efficacy and tolerability, and overall clinical outcome based on information in records. Questions had discrete multiple choice-based (including ‘unable to assess’) and free-text answers where reviewers provided explanations supporting their choice. Assessable evaluations were defined as any answer other than ‘unable to assess’. Discrete and free-text answers were analyzed descriptively using quantitative and qualitative methods. Results: Of 414 records identified, 149 were randomly selected and allocated to reviewers (29–30 each). Assessment of overall clinical outcome (successful/unsuccessful) was achieved in 103 (69%). Reviewers were able to assess AED tolerability and treatment-emergent adverse events in 70% and 89% of records, respectively, and AED’s impact on seizure frequency and severity in 62% and 39%, respectively. Using thematic analysis, reviewers’ free-text responses were found to be consistent with answers corresponding to multiple-choice questions, and provided detailed assessments of the underlying diagnosis, observations in AED dose change, and explanation of complexities. The analysis also indicated that reviewers considered both structured (with assigned codes) and unstructured (eg, physician notes) data in records to inform decisions on outcome. Conclusions: Combined EMR/claims data were of sufficient quality to allow evaluation of clinical outcome related to treatment of patients with epilepsy. AED safety/tolerability and impact on seizure frequency were assessable in most records; assessment of seizure severity presented challenges. Current datasets are therefore unlikely to contain detailed information about this important outcome. Further research on how best to assess and document the impact of AEDs on seizure severity is needed. Results should be interpreted with caution due to the exploratory nature of the study and small sample size. Funding: UCB Pharma-sponsored.