Pilot Study of Real-Time Blood Pressure Decision Support During Cesarean Delivery Under Spinal Anesthesia

This single-center, open-label, single-group pilot study evaluated the prospective performance and operating-room feasibility of a clinician-facing autoregressive with exogenous input (ARX) decision-support algorithm for short-horizon maternal blood pressure forecasting during elective cesarean delivery under spinal anesthesia. Adult pregnant patients at term undergoing elective cesarean delivery with planned spinal anesthesia and prophylactic phenylephrine management were enrolled at a tertiary care academic hospital. The ARX system generated real-time forecasts of maternal mean arterial pressure (MAP) at 1-, 2-, and 3-minute horizons from the time of spinal anesthesia administration until delivery or 20 minutes, whichever occurred first. The model used prior MAP values together with time-stamped phenylephrine infusion and bolus dosing and intrathecal bupivacaine dose/time as inputs. Forecasts were displayed on a dedicated screen adjacent to the anesthesia workstation. The display was intended to support clinician awareness of short-horizon hemodynamic trends and did not provide dosing recommendations. Blood pressure management remained under usual clinical care, including phenylephrine infusion and bolus administration according to institutional practice, and treating anesthesiologists retained full autonomy over all management decisions. The primary objective was to assess prospective forecast accuracy at each prediction horizon using the difference between predicted and observed MAP during the study observation window. Additional analyses assessed other forecast-accuracy metrics, short-horizon identification of hypotension defined as MAP below 80% of baseline, and implementation feasibility based on uninterrupted display operation and clinician-reported overrides. Baseline MAP was defined as the mean of the last three pre-spinal noninvasive blood pressure measurements. Baseline MAP was defined as the mean of the last three pre-spinal noninvasive blood pressure measurements. Noninvasive blood pressure was recorded at 1-minute intervals, and model inputs were entered in real time by in-room research staff. Secondary outcomes assessed intraoperative hypotension burden and phenylephrine exposure during the observation window, including exploratory comparison with matched nonconcurrent controls; these analyses were not part of prospective assignment to the study intervention.