AUTHORS: Ermer, Sean BS et al
METHODS: Volunteers received remifentanil and propofol infusions at selected target concentration pairs to induce depression of ventilation. Signals from each sensor were analyzed by an identical threshold-based detection algorithm to compute the breathing rate. Bland-Altman limits of agreement and error rate analyses were used to characterize the performance of each sensor compared to the reference sensor.
RESULTS: The analysis of the accelerometer and capnometer signals, using Bland-Altman and error rate analyses, showed the highest breath rate agreement (1.96 × standard deviation) of the 7 sensors with −2.1 to 2.2 and −2.5 to 2.7 breaths per minute, respectively. All other signals exhibited wider limits of agreement, with impedance being the widest at −7.8 to 7.4 breaths per minute. For the abdomen accelerometer, 95% of Bland-Altman data points were within ±2 breaths per minute. For the capnometer, 96% of data points were within ±2 breaths per minute. Nasal pressure, thermistor, and microphone all had >80% of data points within ±2 breaths per minute. Impedance and photoplethysmograph signals had 58% and 64%, respectively.
CONCLUSIONS: A unified approach can be applied to a variety of sensor signals to estimate respiratory rates in spontaneously breathing, nonintubated, sedated volunteers. However, detecting clinically relevant low respiratory rates (<6 breaths per minute) is a technical challenge. By our analysis, no single sensor was able to detect slow respiratory rates with adequate precision (<±2 breaths per minute of the reference signal). Of the sensors evaluated, capnometers and abdominal accelerometers may be the most reliable sensors for identifying hypopnea and central apnea.