An analysis of the errors in recorded heart rate and blood pressure in the ICU using a complex set of signal quality metrics

Abstract

We analyzed over 95,000 individual values of heart rate and blood pressure derived from 118,000 hours of electrocardiogram (ECG) and 71,000 hours of Arterial Blood Pressure (ABP) data from 1,071 patients using two methods. One method was a nursing-staff verified automatic measurement transmitted from the bedside monitor to central nursing station at intervals of 5 to 60 minutes. The other method involved re-deriving the estimates from continuous ECG and ABP waveforms using independent algorithms and a set of previously described signal quality metrics to reject noisy and untrustworthy data. Results demonstrate that after the removal of obvious artifactual derived HR and ABP estimates, the two measurement sources disagree, on average, by a clinically insignificant amount. Furthermore, after rejection of data using signal quality metrics, the error distribution curve significantly tightens. The clinically-verified BP values exhibit a small but significant bias towards overestimation, both as a function of time of day and as a function of day of the week. Differences in values between time of day and day of week were small but statistically significant. Inter-nurse differences are also described.