Perme Score and Machine Learning for Detecting ICU-acquired Weakness: A Prospective Observational Cohort Study.

Background and aims: Intensive care unit-acquired weakness (ICUAW) is a common complication in critically ill patients on prolonged mechanical ventilation (MV), impairing recovery and prolonging intensive care unit (ICU) stays. Standard diagnostic tools like the handgrip strength test (HGST) require patient cooperation and may be limited in critical settings. This study evaluated whether the Perme ICU Mobility Score, a multidimensional functional assessment, is associated with ICUAW and can serve as a screening tool using HGST as the diagnostic reference.

Patients and methods: We conducted a prospective observational study in a Brazilian tertiary ICU from May 2021 to October 2023. We assessed adults (≥18 years) undergoing MV for ≥7 days who were clinically stable (pressure support 7 cm H₂O, PEEP 5 cm H₂O, RASS -1 to +1). ICUAW was defined using sex-specific HGST cutoffs (<11 kg men, <7 kg women). Logistic regression, least absolute shrinkage and selection operator (LASSO), and Random Forest models assessed the association between ICUAW and Perme Score. ROC curves and the Youden index determined the optimal cutoff.

Results: Among 97 patients, ICUAW was identified in 78.4%. Lower Perme Scores were significantly associated with ICUAW (p < 0.001). A cutoff ≤9 showed 76.3% sensitivity, 71.4% specificity, and OR = 8.06 (95% CI: 2.72-23.8). In multivariate analysis, the Perme Score (OR = 0.86; p = 0.0004) and SAPS 3 remained independent predictors. Machine learning models confirmed Perme Score as the most significant variable.

Conclusions: The Perme Score is a feasible, complementary screening tool for ICUAW. A cutoff ≤9 supports early functional risk stratification but should be interpreted alongside clinical context due to limited specificity and negative predictive value (NPV).

How to cite this article: Delazari LEB, Ratti LSR, da Silva AC, Sibinelli M, Heidemann AM, Montedioca HLM, et al. Perme Score and Machine Learning for Detecting ICU-acquired Weakness: A Prospective Observational Cohort Study. Indian J Crit Care Med 2025;29(7):562-568.