Purpose Develop a risk prediction model for epilepsy-related death.

Methods In this age/sex-matched case-control study, we compared adults (aged ≥16 years) suffering epilepsy-related death between 2009-2016 to living adults with epilepsy in Scotland. Cases were captured from national mortality records, and controls from a research database or epilepsy clinics. Medical record data were used in univariable and multivariable conditional logistic regression to develop a risk predic- tion model consisting of four variables chosen a priori. A sum of the factors present was taken to create a risk index – the SEDS Score. Odds ratios (OR) with 95% CIs were estimated.

Results 224 cases and 224 controls were compared (mean age 48 years). Univariables predicting epi- lepsy-related death were recent epilepsy-related A&E attendance (OR 5.1, CI 3.2–8.3), living in deprived areas (OR 2.5, CI 1.6–4.0), developmental epilepsy (OR 3.1, CI 1.7–5.7), alcohol abuse (OR 4.4, CI 2.2–9.2), absent recent neurology review (OR 3.8, CI 2.4–6.1), generalised epilepsy (OR 1.9, CI 1.2–3.0), and mental health problems (OR 1.6, CI 1.0–2.6). SEDS Score model variables consisted of the first three listed above, alongside the number of comorbidities (adjusting variable). Compared to having a SEDS Score of 0, those with a SEDS Score of 1, 2, and 3, had 3.6x (CI 1.9–6.8), 17.2x (CI 7.4–39.6), and 19.8x (CI 5.1–76.6) increased odds of death, respectively.

Conclusion SEDS Scoring may help predict epilepsy-related death and requires external validation.