Accuracy of predicting arterial blood gases following mechanical ventilator adjustment

Introduction

Clinicians adjust mechanical ventilator (MV) in critical care units to optimise arterial blood gases (ABG). The present study evaluates their accuracy in predicting ABG values following MV adjustment.

Methods

Thirty-one sets of data were collected from 16 patients requiring mechanical ventilation on the ICU. Each set consisted of two subsets - before and after MV adjustment. Each of subset consisted of the following variables: FIO₂, respiratory rate (breaths/minute), tidal volume (ml/kg), body temperature (°C), haemoglobin concentration (g/dl), and ABG values (pH, PaCO₂, PaO₂, base excess). All data, with the exception of ABG values after MV adjustment, were presented to clinicians. Thirty-three clinicians (11 respiratory therapists (RTs), 11 nurses, and 11 medical staff) were asked to predict the new ABG values as they would in clinical settings. All data are presented as mean ± SD unless stated otherwise.

Results

All datasets include changes in FIO₂, while only 13 datasets include changes in minute ventilation. Increases in FIO₂ were 0.19 ± 0.04 and decreases were -0.12 ± 0.04. Size of change of minute ventilation was 1.38 ± 0.95 l/minute. The absolute magnitudes of change for measured pH, PaCO₂, and PaO₂ were 0.024 ± 0.027, 0.40 ± 0.37 kPa and 5.91 ± 3.35 kPa, respectively. The magnitudes of change for predicted pH, PaCO₂, and PaO₂, respectively, were 0.025 ± 0.033, 0.34 ± 0.23 kPa, and 3.45 ± 2.03 kPa for medical staff ; 0.024 ± 0.021, 0.40 ± 0.41 kPa, and 3.91 ± 1.75 kPa for RTs; and 0.022 ± 0.015, 0.32 ± 0.22 kPa, and 3.21 ± 2.1 kPa for nurses. The 95% limits of agreement between predicted and measured size of changes for pH, PaCO₂, and PaO₂, respectively, were 0.043 to -0.041, 0.724 to -0.596 kPa, and 3.22 to -8.14 kPa for medical staff ; 0.036 to -0.036, 0.66 to -0.66 kPa, and 2.64 to -6.64 KPa for RTs; and 0.036 to -0.076, 0.74 to -0.9 kPa, and 3.91 to -9.31 kPa for nurses. Correlation between duration of clinicians' clinical experience and errors in predicting size of change for pH, PaCO₂, and PaO₂ (-0.0225, -0.054, and 0.173, respectively) were weak and not statistically significant.

Conclusions

Clinicians show considerable variation in predicting ABG values. Medical staff and nurses underestimate the magnitude of PaCO₂ changes whereas RTs show some accuracy. All clinicians have a large bias in predicting the size of PaO₂ changes.