Accurate assessment of oxygen delivery relative to oxygen demand is crucial in the care of a critically ill patient. The central venous oxygen saturation (Svo2) enables an estimate of cardiac output yet obtaining these clinical data requires invasive procedures and repeated blood sampling. Interpretation remains subjective and vulnerable to error. Recognition of patient's evolving clinical status as well as the impact of therapeutic interventions may be delayed. The predictive analytics algorithm, inadequate delivery of oxygen (IDo2) index, was developed to noninvasively estimate the probability of a patient's Svo2 to fall below a preselected threshold. A retrospective multicenter cohort study was conducted using data temporally independent from the design and development phase of the IDo2 index. A total of 20,424 Svo2 measurements from 3,018 critically ill neonates, infants, and children were retrospectively analyzed. Collected data included vital signs, ventilator data, laboratory data, and demographics. The ability of the IDo2 index to predict Svo2 below a preselected threshold (30%, 40%, or 50%) was evaluated for discriminatory power, range utilization, and robustness. Area under the receiver operating characteristic curve (AUC) was calculated for each index threshold. Datasets with greater amounts of available data had larger AUC scores. This was observed across each configuration. For the majority of thresholds, Svo2 values were observed to be significantly lower as the IDo2 index increased. The IDo2 index may inform decision-making in pediatric cardiac critical care settings by providing a continuous, noninvasive assessment of oxygen delivery relative to oxygen demand in a specific patient. Leveraging predictive analytics to guide timely patient care, including support for escalation or de-escalation of treatments, may improve care delivery for patients and clinicians.
Read full abstract