Decision Support Tool to Judiciously Assign High-Frequency Neurologic Examinations in Traumatic Brain Injury

Traumatic brain injury (TBI) management includes serial neurologic examinations to assess for changes dictating neurosurgical interventions. We hypothesized hourly examinations are overassigned. We conducted a decision tree analysis to determine an algorithm to judiciously assign hourly examinations.

A retrospective cohort study of 1022 patients with TBI admitted to a Level 1 trauma center from January 1, 2019, to December 31, 2019, was conducted. Patients with penetrating TBI or immediate or planned interventions and those with nonsurvivable injuries were excluded. Patients were stratified by whether they underwent an unplanned intervention (e.g., craniotomy or invasive intracranial monitoring). Univariate analysis identified factors for inclusion in chi-square automatic interaction detection technique, classifying those at risk for unplanned procedures.

A total of 830 patients were included, 287 (35%) were assigned hourly (Q1) examinations, and 17 (2%) had unplanned procedures, with 16 of 17 (94%) on Q1 examinations. Patients requiring unplanned procedures were more likely to have mixed intracranial hemorrhage pattern (82% versus 39%; P = 0.001), midline shift (35% versus 14%; P = 0.023), an initial poor neurologic examination (Glasgow Comas Scale ≤8, 77% versus 14%; P < 0.001), and be intubated (88% versus 17%; P < 0.001). Using chi-square automatic interaction detection, the decision tree demonstrated low-risk (2% misclassification) and excellent discrimination (area under the curve = 0.915, 95% confidence interval 0.844-0.986; P < 0.001) of patients at risk of an unplanned procedure. By following the algorithm, 167 fewer patients could have been assigned Q1 examinations, resulting in an estimated 6012 fewer examinations.

Using a 4-factor algorithm can optimize the assignment of neuro examinations and substantially reduce neuro examination burden without sacrificing patient safety.