Decoding accelerometry for classification and prediction of critically ill patients with severe brain injury
The goal of this research is to explore quantitative motor features in critically ill patients with severe brain injury (SBI). We hypothesized that computational decoding of these features would yield important information on underlying neurological states and clinical outcomes. Using wearable microsensors placed on all extremities, we recorded 1,701 hours of continuous, high-frequency accelerometry data from a prospective cohort of patients (n = 69) admitted to the ICU with SBI. Models were trained using time-, frequency-, and wavelet-domain motion features and levels of responsiveness and outcome as labels. The two primary tasks were detection of levels of responsiveness, assessed by motor sub-score of the Glasgow Coma Scale (GCSm), and prediction of functional outcome at hospital discharge, measured with the Glasgow Outcome Scale—Extended (GOSE). Detection models achieved significant (AUC: 0.70 [95% CI: 0.53—0.85]) and consistent (observation windows: 12 min — 9 hours) discrimination of SBI patients capable of purposeful movement (GCSm > 4). Prediction models accurately discriminated SBI patients of upper moderate disability or better (GOSE > 5) with 2—6 hours of observation (AUC: 0.82 [95% CI: 0.75—0.90]). Results suggest that computational analysis of time series motor activity in patients with SBI yields clinically important insights on underlying neurologic states and short-term clinical outcomes.