Accelerometer-derived movement features as predictive biomarkers for muscle atrophy in neurocritical care: a prospective cohort study

Abstract

Background: Physical inactivity and subsequent muscle atrophy are highly prevalent in neurocritical care and are recognized as key mechanisms underlying intensive care unit acquired weakness (ICUAW). The lack of quantifiable biomarkers for inactivity complicates the assessment of its relative importance compared to other conditions under the syndromic diagnosis of ICUAW. We hypothesize that active movement, as opposed to passive movement without active patient participation, can serve as a valid proxy for activity and may help predict muscle atrophy. To test this hypothesis, we utilized non-invasive, body-fixed accelerometers to compute measures of active movement and subsequently developed a machine learning model to predict muscle atrophy.

Methods: This study was conducted as a single-center, prospective, observational cohort study as part of the MINCE registry (metabolism and nutrition in neurointensive care, DRKS-ID: DRKS00031472). Atrophy of rectus femoris muscle (RFM) relative to baseline (day 0) was evaluated at days 3, 7 and 10 after intensive care unit (ICU) admission and served as the dependent variable in a generalized linear mixed model with Least Absolute Shrinkage and Selection Operator regularization and nested-cross validation.

Results: Out of 407 patients screened, 53 patients (age: 59.2 years (SD 15.9), 31 (58.5%) male) with a total of 91 available accelerometer datasets were enrolled. RFM thickness changed - 19.5% (SD 12.0) by day 10. Out of 12 demographic, clinical, nutritional and accelerometer-derived variables, baseline RFM muscle mass (beta - 5.1, 95% CI - 7.9 to - 3.8) and proportion of active movement (% activity) (beta 1.6, 95% CI 0.1 to 4.9) were selected as significant predictors of muscle atrophy. Including movement features into the prediction model substantially improved performance on an unseen test data set (including movement features: R² = 79%; excluding movement features: R² = 55%).

Conclusion: Active movement, as measured with thigh-fixed accelerometers, is a key risk factor for muscle atrophy in neurocritical care patients. Quantifiable biomarkers reflecting the level of activity can support more precise phenotyping of ICUAW and may direct tailored interventions to support activity in the ICU. Studies addressing the external validity of these findings beyond the neurointensive care unit are warranted.

Trial registration: DRKS00031472, retrospectively registered on 13.03.2023.

Keywords: Accelerometer; ICU; ICUAW; Machine learning; Muscle atrophy; Sarcopenia.

Fig. 1

Accelerometer-derived features. Body motion was monitored using tri-axial accelerometers bilaterally attached to the upper thigh (1). Raw triaxial accelerometer recordings were first offset eliminated, and the time series were segmented into non-overlapping 5 s windows. The signal magnitude area was computed for every window (2). Bouts of dynamic activity were identified based on the threshold ≥ 0.135 g (3) and a set of motion features was computed for every bout of activity (4). Finally, the average and distribution of motion features across all bouts of activity were computed (5). acc = acceleration

Fig. 2

Nested-cross validation of a regularized GLMM model. After standardization and a stratified 80/20 split, the training data set was partitioned into 4 folds (outer loop). Within each outer fold, an inner loop of 2 folds was used for hyperparameter tuning. The hyperparameter (lambda) that minimized the mean squared error in the inner loop was selected. The model with this optimal lambda was then evaluated on the validation fold of the outer loop. This process was repeated for all 4 outer folds, resulting in an optimal lambda for each fold. The final model was chosen using the average of the optimal lambdas from all outer folds. Finally, the performance of this final model was assessed using the unseen test set. GLMM = generalized mixed effects model; lasso = least absolute shrinkage and selection operator

Fig. 3

Screening and study inclusion. ICU = Intensive Care Unit;

Fig. 4

Active movement and muscle atrophy during ICU treatment. Muscle atrophy at days 3, 5 7 and 10 relative to day 0 for RFM and TM (A). Proportion of active movement (%active) over time (B). ICU = Intensive Care Unit;

Fig. 5

Prediction of MRF muscle atrophy with and without movement features. Standardized coefficients and 95% confidence intervals (asterisks indicate significant predictors) of the regularized regression models with (model_movement+, A) and without movement features (model_movement−, C). Out of all demographic (age, sex), clinical (baseline RFM muscle mass, mSOFA), nutritional (calorie deficit, protein deficit) and movement variables (intensity, %active, AB-intensity_log-mean, AB-duration_log-mean, AB-intensity_log-SD, AB-duration_log-SD), the depicted 10/12 independent variables for model model_movement+ and 4/6 independent variables for model_movement− were selected for the final models, respectively. Significant predictors in model_movement+ included baseline RFM muscle mass (beta − 5.1, 95% confidence interval (95% CI) − 7.9 to − 3.8) and %active (beta 1.6, 95% CI 0.1 to 4.9). For model_movement-, only baseline RFM muscle was found as a statistically significant predictor (beta − 4.6, 95% CI − 7.6 to − 3.9). Scatter plots with regression line of predicted versus actual muscle wasting (grey dots: training data; black dots: unseen test data) for model_movement+ (B) and model_movement- (D), respectively (R²: 0.79 vs. 0.55, RMSE: vs. 8.4 vs. 10.7 mm; MAE: 6.2 vs. 8.0 mm). mSOFA = SOFA without GCS;