Abstract P405: A Time-Series Forecast Model to Assess Vital Sign Waveform Variability Prior to Vasospasm
Introduction: Symptomatic vasospasm (SV) is a complication of aneurysmal subarachnoid hemorrhage (aSAH) and can lead to cerebral infarction. Changes in vital trends, such as heart rate (HR) and mean arterial blood pressure (MAP), have been associated with SV in aSAH. Real-time assessment of instantaneous vital sign waveform data could improve detection of vital sign variability associated with vasospasm. However, no model using instantaneous waveform data exists to predict SV. We hypothesize that autoregressive integrated moving average (ARIMA) analysis, a time-series forecast model, is a useful approach to assess the variability of vital sign waveforms associated with SV. Methods: In this small case-control study, vital signs of patients admitted to the neuroICU with aSAH were obtained using a software-based analytics platform, Sickbay. HR and MAP from 15 aSAH patients were continuously obtained from ECG and arterial line waveforms. Ten patients developed neurologic deficits attributed to angiographically-confirmed SV (Det). Five controls (Con) without SV were matched based on age. 3 Det and 3 Con were randomly selected for further analysis. For Det, waveforms were analyzed at 5-second intervals for 48 hours prior to clinical deterioration. For Con, waveforms were analyzed at a random 48-hour interval. Results: Visually, MAP and not HR was more variable in Det than in Con patients (Figure). The ARIMA model plotted the forecasted-fit for each delta-variable waveform. The MAP confidence interval margins were significantly larger for Det patients compared to the Con patient. This trend was consistent across all other patients. Conclusion: ARIMA is a useful tool to assess HR and MAP waveform variations prior to SV in aSAH. Larger studies are required to solidify this concept and further explore the combination of data analytics platform and ARIMA to predict neurological deterioration in SV.