Abstract
BACKGROUND
Intermittent vital sign (VS) documentation is the standard of care in pediatric wards and pediatric intensive care units (PICUs). However, intermittent capture of VS values may miss important trajectories and extreme values, particularly for patients at risk of organ dysfunction.
OBJECTIVE
Determine the agreement between intermittently charted and high-frequency (continuous) bedside monitor VS measurements, as well as the performance of the 2 types of VS measurements to predict cardiorespiratory dysfunction.
METHODS
This was a single-center, retrospective study of children admitted to the PICU. We examined agreement by comparing the range of high-frequency VS measurements with the corresponding intermittently charted VS measurements. To assess the performance for predicting cardiorespiratory dysfunction, both sets of VS measurements were summarized as the worst values at an hourly level. The dataset was temporally divided into training and test sets, and XGBoost models were trained using either the intermittent or high-frequency values. The area under the receiver operating characteristic (AUROC) curve of both models was compared using the DeLong methods.
RESULTS
There were 12 189 PICU encounters, with over 1.5 million paired VS observations. Agreement varied based on the VS, with 21% to 33% of all intermittently charted VS measurements out of range from the high-frequency VS measurements. Cardiorespiratory dysfunction occurred in 36.4% of encounters. High-frequency VS measurements had a small but significantly higher performance at predicting cardiorespiratory dysfunction compared with intermittently charted VS measurements (AUROC, 0.64 vs 0.63; P
CONCLUSIONS
About 20% to 30% of intermittently charted VS measurements do not fall into the expected high-frequency, bedside monitor range. High-frequency VS measurements are weakly but significantly better than intermittently charted VS measurements in predicting cardiorespiratory dysfunction.