Feasibility of a Wireless Vital Signal Monitoring System in the NICU

In the Neonatal Intensive Care Unit (NICU), vital signs are monitored continuously via skin sensors connected to bedside monitors using wires and cables. This may interfere with patient care and increase the risks of skin damage, and infection. Wires may also tangle around the body. Recently, a wireless system called ANNE (R) One (Sibel Health, Chicago, USA) was developed, and investigated in a pilot study at the Montreal Children's Hospital. 25 neonates with a median gestational age at birth of 28 weeks (IQR: 26-31 weeks) were monitored simultaneously by this wireless system and a wired reference monitor (Intellivue MX450, Philips Healthcare, Best, Netherlands) for 8 hours a day, on 4 consecutive days. To assess the feasibility of using this wireless monitoring system in the NICU, we examined the coverage (defined as the percentage of time for which signal values were available with respect to total record length) of three vitals signs: heart rate (HR), oxygen saturation (SpO(2)) and respiratory rate (RR). The coverage of HR and SpO(2) were further compared to that of corresponding electrocardiograms (ECGs) and photoplethysmograms (PPGs), respectively. We found the coverage of wired vital recordings to be 99% for HR, 94% for SpO(2), and 97% for RR. In the wireless system, coverage was significantly lower for all three vitals: 86% for HR (compared to 94% in ECG), 66% for SpO(2) (compared to 99% in PPG), and 47% for RR. To investigate the reason for wireless gaps - defined as continuous periods with no signal for 1.25 seconds (equivalent to one missed sample) or more - we correlated them with automatically generated alert signals and manually recorded annotations. The alerts tracked wireless sensor functionality, including Bluetooth disconnection and drops in signal intensity; annotations tracked interactions between the infant and parents or healthcare professionals during recordings. We found: (1) 91% of gaps in wireless HR recordings were correlated with alert flags, with 56% linked to Bluetooth disconnections (67% of which occurred during kangaroo care); (2) only 10% of SpO(2) gaps and 25% of RR gaps were correlated with alerts; and (3) only 49% of gaps in SpO(2) and 34% in RR coincided with annotated activities. More investigations to explain the widespread occurrence of gaps in SpO(2) and RR signals are necessary. For this reason, future work will examine the relations of: (1) movement artifact, (2) signal-to-noise ratio, and (3) the algorithms used to display SpO(2) and RR, on gap classification with the use of the ANNE (R) system. At present, the coverage issues, especially in SpO(2) and RR, limit the feasibility of wireless vital sign monitoring in the NICU, but our results pinpoint key areas of improvement to be addressed.