Abstracts

Rationale: An epileptic seizure can be detected before clinical onset by monitoring alterations in heart rate variability (HRV). We have developed a wearable HRV-based epileptic seizure prediction system. The use of our system in hospitals could contribute to prevention of in-hospital injuries, reducing nurses' burden and enabling easier observation of epileptic seizures. Here, we report medical staff members' views on the in-hospital use of our system. Methods: This survey was conducted among medical staff members working in 8 medical facilities that specialize in epilepsy in Japan. These medical staff members included doctors, nurses, and medical technologists. They were asked to complete a self-administered questionnaire designed to assess their views with respect to our epileptic seizure prediction system. At the beginning of the questionnaire, a brief description of our system was provided. The questionnaire items were focused on the sensitivity of the system, false alarms, and clinical importance. Results: Two hundred sixty-four medical staff members completed our questionnaire. In total, 71% of the respondents indicated that a prediction rate of >=70% is necessary. Regarding the frequency of false alarms without seizures, 65% of the respondents reported that this rate would preferably be below several times per day. In addition, 27% of the respondents considered it to be acceptable for the false alarm rate to be less than the baseline seizure frequency in individual patients. Regarding the in-hospital usage of our system, all possible uses were considered to be of nearly equal importance; these included securing patient safety, calling for help, observing epileptic seizures easily, and reducing anxiety. Regarding the form of the alarm, 56% of the respondents considered an acoustic alarm signal to be suitable, whereas 20% of the respondents regarded a vibration alarm signal to be suitable. With respect to the timing of the alarm, 89% of the respondents wanted an alert at least 5 minutes before seizure onset. Regarding a notification of seizure probability―instead of predicting individual seizures, a notification of the probability of seizures could be provided as '10%, 50%, or 90%' or 'green, yellow, red'―79% of the respondents considered this approach to be useful. Conclusions: For an HRV-based seizure prediction system to be considered useful, the respondents reported that a prediction rate of >70% is needed. Although false alarms were acceptable to some extent, their frequencies must be reduced to a maximum of several times per day. Respondents indicated that it could be meaningful to use the phased notification of seizure probability, instead of predicting individual seizures. The in-hospital use of our system can provide a variety of benefits; it could prevent securing patient safety, calling for help, observing epileptic seizures easily, and reducing anxiety. In order to achieve such benefits, the findings of the present study suggest that an alert is needed at least 5 minutes before seizure onset. Funding: This work was supported by Japan Agency for Medical Research and Development.