Cookies on this website

We use cookies to ensure that we give you the best experience on our website. If you click 'Accept all cookies' we'll assume that you are happy to receive all cookies and you won't see this message again. If you click 'Reject all non-essential cookies' only necessary cookies providing core functionality such as security, network management, and accessibility will be enabled. Click 'Find out more' for information on how to change your cookie settings.

Data smoothing of vital signs has been reported in the anesthesia literature, suggesting that clinical staff are biased toward measurements of normal physiology. However, these findings may be partially explained by clinicians interpolating spurious values from noisy signals and by the undersampling of physiological changes by infrequent manual observations. We explored the phenomenon of data smoothing using a method robust to these effects in a large postoperative dataset including respiratory rate, heart rate, and oxygen saturation (SpO2). We also assessed whether the presence of the vital sign taker creates an arousal effect.Study data came from a UK upper gastrointestinal postoperative ward (May 2009 to December 2013). We compared manually recorded vital sign data with contemporaneous continuous data recorded from monitoring equipment. We proposed that data smoothing increases differences between vital sign sources as vital sign abnormality increases. The primary assessment method was a mixed-effects model relating continuous-manual differences to vital sign values, adjusting for repeated measurements. We tested the regression slope significance and predicted the continuous-manual difference at clinically important vital sign values. We calculated limits of agreement (LoA) between vital sign sources using the Bland-Altman method, adjusting for repeated measures. Similarly, we assessed whether the vital sign taker affected vital signs, comparing continuous data before and during manual recording.From 407 study patients, 271 had contemporaneous continuous and manual recordings, allowing 3740 respiratory rate, 3844 heart rate, and 3896 SpO2 paired measurements for analysis. For the model relating continuous-manual differences to continuous-manual average vital sign values, the regression slope (95% confidence interval) was 0.04 (-0.01 to 0.10; P = .11) for respiratory rate, 0.04 (-0.01 to 0.09; P = .11) for heart rate, and 0.10 (0.07-0.14; P < .001) for SpO2. For SpO2 measurements of 91%, the model predicted a continuous-manual difference (95% confidence interval) of -0.88% (-1.17% to -0.60%). The bias (LoA) between measurement sources was -0.74 (-7.80 to 6.32) breaths/min for respiratory rate, -1.13 (-17.4 to 15.1) beats/min for heart rate, and -0.25% (-3.35% to 2.84%) for SpO2. The bias (LoA) between continuous data before and during manual observation was -0.57 (-5.63 to 4.48) breaths/min for respiratory rate, -0.71 (-10.2 to 8.73) beats/min for heart rate, and -0.07% (-2.67% to 2.54%) for SpO2.We found no evidence of data smoothing for heart rate and respiratory rate measurements. Although an effect exists for SpO2 measurements, it was not clinically significant. The wide LoAs between continuous and manually recorded vital signs would commonly result in different early warning scores, impacting clinical care. There was no evidence of an arousal effect caused by the vital sign taker.

Original publication

DOI

10.1213/ANE.0000000000003694

Type

Journal article

Journal

Anesthesia and analgesia

Publication Date

10/2018

Volume

127

Pages

960 - 966

Addresses

From the Department of Engineering Science, Institute of Biomedical Engineering, University of Oxford, Oxford, United Kingdom.