ABSTRACT
Background: Acute neurological injury is a leading cause of permanent disability and death in the pediatric intensive care unit (PICU). No predictive model has been validated for critically ill children with acute neurological injury. Objectives: We hypothesized that PICU patients with concern for acute neurological injury are at higher risk for morbidity and mortality, and advanced analytics would derive robust, explainable subgroup models. Methods: We performed a secondary subgroup analysis of the Trichotomous Outcomes in Pediatric Critical Care (TOPICC) study (2011-2013), predicting mortality and morbidity from admission physiology (lab values and vital signs in 6â h surrounding admission). We analyzed patients with suspected acute neurological injury using standard machine learning algorithms. Feature importance was analyzed using SHapley Additive exPlanations (SHAP). We created a Fast Healthcare Interoperability Resources (FHIR) application to demonstrate potential for interoperability using pragmatic data. Results: 1,860 patients had suspected acute neurological injury at PICU admission, with higher morbidity (8.2 vs. 3.4%) and mortality (6.2 vs. 1.9%) than those without similar concern. The ensemble regressor (containing Random Forest, Gradient Boosting, and Support Vector Machine learners) produced the best model, with Area Under the Receiver Operating Characteristic Curve (AUROC) of 0.91 [95% CI (0.88, 0.94)] and Average Precision (AP) of 0.59 [0.51, 0.69] for mortality, and decreased performance predicting simultaneous mortality and morbidity (0.83 [0.80, 0.86] and 0.59 [0.51, 0.64]); at a set specificity of 0.995, positive predictive value (PPV) was 0.79 for mortality, and 0.88 for mortality and morbidity. By comparison, for mortality, the TOPICC logistic regression had AUROC of 0.90 [0.84, 0.93], but substantially inferior AP of 0.49 [0.35, 0.56] and PPV of 0.60 at specificity 0.995. Feature importance analysis showed that pupillary non-reactivity, Glasgow Coma Scale, and temperature were the most contributory vital signs, and acidosis and coagulopathy the most important laboratory values. The FHIR application provided a simulated demonstration of real-time health record query and model deployment. Conclusions: PICU patients with suspected acute neurological injury have higher mortality and morbidity. Our machine learning approach independently identified previously-known causes of secondary brain injury. Advanced modeling achieves improved positive predictive value in this important population compared to published models, providing a stepping stone in the path to deploying explainable models as interoperable bedside decision-support tools.
ABSTRACT
Introduction: Continuous neurologic assessment in the pediatric intensive care unit is challenging. Current electroencephalography (EEG) guidelines support monitoring status epilepticus, vasospasm detection, and cardiac arrest prognostication, but the scope of brain dysfunction in critically ill patients is larger. We explore quantitative EEG in pediatric intensive care unit patients with neurologic emergencies to identify quantitative EEG changes preceding clinical detection. Methods: From 2017 to 2020, we identified pediatric intensive care unit patients at a single quaternary children's hospital with EEG recording near or during acute neurologic deterioration. Quantitative EEG analysis was performed using Persyst P14 (Persyst Development Corporation). Included features were fast Fourier transform, asymmetry, and rhythmicity spectrograms, "from-baseline" patient-specific versions of the above features, and quantitative suppression ratio. Timing of quantitative EEG changes was determined by expert review and prespecified quantitative EEG alert thresholds. Clinical detection of neurologic deterioration was defined pre hoc and determined through electronic medical record documentation of examination change or intervention. Results: Ten patients were identified, age 23 months to 27 years, and 50% were female. Of 10 patients, 6 died, 1 had new morbidity, and 3 had good recovery; the most common cause of death was cerebral edema and herniation. The fastest changes were on "from-baseline" fast Fourier transform spectrograms, whereas persistent changes on asymmetry spectrograms and suppression ratio were most associated with morbidity and mortality. Median time from first quantitative EEG change to clinical detection was 332â minutes (interquartile range: 201-456â minutes). Conclusion: Quantitative EEG is potentially useful in earlier detection of neurologic deterioration in critically ill pediatric intensive care unit patients. Further work is required to quantify the predictive value, measure improvement in outcome, and automate the process.
