AbstractObjectiveSudden unexpected death in epilepsy (SUDEP) is a significant cause of mortality in epilepsy. The aim of this study is to evaluate the validity of the SUDEP‐7 inventory and its components as tools for predicting SUDEP risk, and to develop and validate an improved inventory.MethodsThe study included 28 patients who underwent video–electroencephalography (EEG) monitoring and later died of SUDEP, and 56 age‐ and sex‐matched control patients with epilepsy. The SUDEP‐7 score, its individual components, and an alternative inventory were examined as predictors of SUDEP.ResultsSUDEP‐7 scores were significantly higher among SUDEP patients compared with controls, both at time of admission (p = 0.024) and most recent follow‐up (p = 0.016). SUDEP‐7 scores declined only among controls, who demonstrated reduced seizure frequency. Seizure freedom after epilepsy surgery was also associated with survival. Several components of the SUDEP‐7 inventory were independently associated with higher risk of SUDEP, including more than three generalized tonic‐clonic (GTC) seizures (p = 0.002), one or more GTC seizures (p = 0.001), or one or more seizures of any type within the last year (p = 0.013), and intellectual disability (p = 0.031). In stepwise regression models, SUDEP‐7 scores did not enhance the prediction of SUDEP over either GTC seizure frequency or seizure frequency alone. A novel SUDEP‐3 inventory comprising GTC seizure frequency, seizure frequency, and intellectual disability (p < 0.001) outperformed the SUDEP‐7 inventory (p = 0.010) in predicting SUDEP.SignificanceOur findings demonstrate the limitations of the SUDEP‐7 inventory. We propose a new three‐item SUDEP‐3 inventory, which predicts SUDEP better than the SUDEP‐7.