Increasing epilepsy-related mortality: A multiple causes of death study in Northern Italy

Epilepsy-related mortality during the COVID-19 pandemic: A nationwide study of routine Scottish data

“… [41] However, the latter strategy is likely to offer a more representative picture of a population as it is likely to include data from areas where resources may have been limited. Epilepsy-related death: We defined epilepsy-related death using the same ICD-10 coding strategy as other authors of epilepsy-related death studies that use routine data, [46] , [47] , [48] which is also the same strategy used by NRS and the ONS to capture epilepsy-related death when reporting the UK's mandatory national mortality statistics, [49] , [50] , [51] and the same coding strategy used by the WHO to identify epilepsy in their mortality database. [52] This strategy uses codes G40–41 (epilepsy and/or status epilepticus) listed as an underlying or contributory cause of death within death records to denote epilepsy-related death.…”

Developing and validating a clinical prediction model to predict epilepsy-related hospital admission or death within the next year using administrative healthcare data: a population-based cohort study protocol

“…We performed a sample size calculation for each outcome based on our previous methodology [33]. We used prevalence figures for each outcome from prior literature [9,18], supported by our pre-submission CPRD feasibility search. Logistic regression modelling assumptions (i) The models have a predictive performance of either Nagelkerke's R 2 of 15% or achieve at least a c-statistic of 0.6 (whichever leads to higher sample size to give a conservative estimate).…”

“…A Based on a seizure-related ED or hospital admissions prevalence of 8.8% [9], developing a logistic regression model for this outcome will require a minimum sample size of 44,750 if the model achieves at least a c-statistic of 0.6, or 6,430 if the model achieves a Nagelkerke's R 2 of 15%. B Based on an epilepsy-related deaths prevalence of 15.4% [9,18], developing a logistic regression model for this outcome will require a minimum sample size of 27,528 if the model achieves at least a c-statistic of 0.6, or 4,951 if the model achieves a Nagelkerke's R 2 of 15%.…”

“…However, a substantial proportion are epilepsy-related [16, 17]. These are operationally defined as any death listing epilepsy as an underlying or contributory cause within death records [11, 18, 19], in line with national guidance [20]. Although sudden unexpected death in epilepsy (SUDEP) is a common epilepsy-related death, we have shown it is equally common for epilepsy-related death to occur through other mechanisms including aspiration pneumonia, cardiac arrest, antiseizure medication (ASM) poisoning, drowning, and alcohol dependence [11, 16].…”

“…When we looked at epilepsy-related deaths as a group, nearly 80% were potentially avoidable [11]. Therefore, it is becoming increasingly pragmatic to investigate epilepsy-related deaths as a group rather than focusing on each cause individually [17, 18, 21].…”