Predicting intracranial pressure monitor placement in children with traumatic brain injury: a prospective cohort study to develop a clinical decision support tool

Abstract

Objective: Clinicians currently make decisions about placing an intracranial pressure (ICP) monitor in children with traumatic brain injury (TBI) without the benefit of an accurate clinical decision support tool. The goal of this study was to develop and validate a model that predicts placement of an ICP monitor and updates as new information becomes available.

Materials and methods: A prospective observational cohort study was conducted from September 2014 to January 2024. The setting included one US hospital designated as an American College of Surgeons Level 1 Pediatric Trauma Center. Participants were 389 children with acute TBI admitted to the ICU who had at least one Glasgow Coma Scale (GCS) score ≤ 8 or intubation with at least one GCS-Motor ≤ 5. We excluded children who received ICP monitors prior to arrival, those with GCS = 3 and bilateral fixed, dilated pupils, and those with a do not resuscitate order.

Results: Of the 389 participants, 138 received ICP monitoring. Several machine learning models, including a recurrent neural network (RNN), were developed and validated using 4 combinations of input data. The best performing model, an RNN, achieved an F1 of 0.71 within 720 minutes of hospital arrival. The cumulative F1 of the RNN from minute 0 to 720 was 0.61. The best performing non-neural network model, standard logistic regression, achieved an F1 of 0.36 within 720 minutes of hospital arrival.

Conclusions: These findings will contribute to design and implementation of a multidisciplinary clinical decision support tool for ICP monitor placement in children with TBI.

Keywords: clinical decision support; intracranial hypertension; intracranial pressure; machine learning; pediatrics; traumatic brain injury.

Figure 1.

Model performance: F1 and cumulative sum of F1, first 12 hours. First 12 hours for minimal and maximal variables models: Column 1—CT and invasive blood pressure variables as input, Column 2—no CT and no invasive blood pressure variables input to model. GAM, generalized additive models; GEE, generalized estimating equations; Lasso, Lasso penalized logistic regression; LR, standard logistic regression; Ridge, Ridge penalized logistic regression; RNN, Recurrent Neural Network; SVM, support vector machines. See Figure S5 for additional intermediate models with a mix of CT and Invasive Blood Pressure variables. See Figures S6 and S7 for first 24 hours of inpatient stay. See Table 3 for numeric values.

Figure 2.

Logistic regression coefficients for models that include CT and invasive blood pressure data. Mean rank ordered coefficients for linear models where both CT and Invasive Blood Pressure are inputs. Darker color indicates a larger coefficient. Yellow indicates negative coefficient, green indicates positive coefficient, and red indicates overall rank of absolute value of coefficient across the models.

Figure 3.

Logistic regression coefficients for models without CT or invasive blood pressure data. Mean rank ordered absolute value of coefficients for linear models where CT and Invasive Blood Pressure were not used as model inputs. Darker color indicates larger coefficient. Yellow indicates negative coefficient, green indicates positive coefficient, and red indicates overall rank of absolute value of coefficient across the models.