BACKGROUND
Critical Congenital Heart Disease (cCHD) globally occurs in 2-3 of every 1000 live births and often requires cardiac surgery in the first weeks of life for survival. In the critical peri-operative period, intensive multimodal monitoring at a Paediatric Intensive Care Unit (PICU) is warranted as their organs – especially the brain – may be severely injured due to hemodynamic- and respiratory events. These 24/7 clinical data streams yield large quantities of high-frequency data, which are challenging in terms of interpretation due to the varying and dynamic physiology innate to cCHD. Through advanced data science algorithms such as Machine Learning (ML), these dynamic data can be condensed into comprehensible information, reducing the cognitive load on the medical team and providing data-driven monitoring support through automated detection of clinical aberrations – which may facilitate timely intervention.
OBJECTIVE
To develop a clinical aberration detection algorithm for PICU patients with cCHD.
METHODS
Retrospectively, synchronous per-second data of cerebral regional oxygen saturation (rSO2) and four vital parameters (respiratory rate, heart rate, oximetry (SpO2) & invasive mean blood pressure (IBP)) in neonates with cCHD admitted to the University Medical Center Utrecht, the Netherlands between 2002 and 2018 were extracted. Patients were stratified based on mean SpO2 during admission to account for physiological differences between (a)cyanotic cCHD. Each subset was used to train our algorithm in classifying data as either stable, unstable, or sensor dysfunction. The algorithm was designed to detect combinations of parameters abnormal to the specific subpopulation through Support Vector Machine learning, as well as significant deviations from the analyzed patient’s unique baseline. These deviations were further analyzed for directional movement to distinguish clinical improvement from deterioration. Novel data were used for testing, visualized in detail, and evaluated by paediatric intensivists.
RESULTS
A retrospective query yielded 4600 hours and 209 hours of per-second data in 78 and 10 neonates for, respectively, training- and testing purposes. Overall, our algorithm provided accurate detection in 88% of stable- and 81% of unstable episodes. Twelve expert-confirmed unstable episodes were missed in testing. A total of 138 sensorial dysfunctions were detected, of which 130 (94%) were correct.
CONCLUSIONS
In this proof-of-concept study, an aberration detection algorithm was developed and retrospectively evaluated to classify clinical (in)stability, achieving a reasonable performance considering the heterogeneous population of neonates with cCHD. The combination of baseline deviations (i.e. patient-specific) and simultaneous parameter-shifting (i.e. population-specific) proofs to be promising with respect to enhancing applicability to heterogeneous critically ill paediatric populations. Although this model needs to be validated prospectively, advanced data science models such as the one presented here may in the future be used in the automated detection of clinical aberrations and eventually provide data-driven monitoring support to the medical team, allowing for timely intervention.