Accuracy of remote continuous respiratory rate monitoring technologies intended for low care clinical settings: a...

Altered respiratory rate (RR) has been identified as an important predictor of serious adverse events during hospitalization. Introduction of a well-tolerated continuous RR monitor could potentially reduce serious adverse events such as opioid-induced respiratory depression. The purpose of this study was to investigate the ability of different monitor devices to detect RR in low care clinical settings.

This was a prospective method-comparison study with a cross-sectional design. Thoracic impedance pneumography (IPG), frequency modulated continuous wave radar, and an acoustic breath sounds monitor were compared with the gold standard of capnography for their ability to detect RR in breaths per minute (breaths·min^-1) in awake postoperative patients in the postanesthesia care unit. The Bland and Altman method for repeated measurements and mixed effect modelling was used to obtain bias and limits of agreement (LoA). Furthermore, the ability of the three devices to assist with correct treatment decisions was evaluated in Clarke Error Grids.

Twenty patients were monitored for 1,203 min, with a median [interquartile range] of 61 [60-63] min per patient. The bias (98.9% LoA) were 0.1 (-7.9 to 7.9) breaths·min^-1 for the acoustic monitor, -1.6 (-10.8 to 7.6) for the radar, and -1.9 (-13.1 to 9.2) for the IPG. The extent to which the monitors guided adequate or led to inadequate treatment decisions (determined by Clarke Error Grid analysis) differed significantly between the monitors (P = 0.011). Decisions were correct 96% of the time for acoustic, 95% of the time for radar, and 94% of the time for IPG monitoring devices.

None of the studied devices (acoustic, IPG, and radar monitor) had LoA that were within our predefined (based on clinical judgement) limits of ± 2 breaths·min^-1. The acoustic breath sound monitor predicted the correct treatment more often than the IPG and the radar device.