Measurement properties of the Chelsea Critical Care Physical Assessment tool (CPAx): a systematic review and meta-analysis.

Abstract

What are the measurement properties of the Chelsea Critical Care Physical Assessment tool (CPAx) for evaluating physical function and activity for rehabilitation of patients who are critically ill? A systematic search was conducted in five databases to 12 January 2026. Two reviewers independently screened studies, extracted data and assessed quality for each measurement property (COSMIN Risk of Bias checklist); quality of each available measurement property was independently evaluated by any two reviewers who judged the quality as very good, adequate, doubtful or inadequate. Data were synthesised according to COSMIN guidelines, with random-effects meta-analyses for intraclass correlation coefficients (ICCs). Pooled results were rated and the evidence graded as very low, low, moderate or high quality. 27 studies investigated 71 measurement properties (inadequate to very good). The CPAx was feasible, with small floor and ceiling effects across the trajectory of recovery. Content validity was high (pooled index 0.94; high-quality evidence). The CPAx demonstrated excellent inter-rater reliability (pooled ICC 0.99, 95% CI 0.98 to 1.00, six studies, high-quality evidence), intra-rater reliability (ICC 0.95, 95% CI 0.82 to 0.97, one study, low-quality evidence) and acceptable standard error of measurement (pooled SEM 1.36, five studies) corresponding to a smallest detectable change of 3.76 (sufficient rating, high-quality evidence). Construct validity was supported with 26 out of 34 predefined hypotheses accepted (76%, sufficient, 10 studies, moderate-quality evidence) and responsiveness with five out of six predefined hypotheses accepted (83%, sufficient, five studies, very low-quality evidence). Moderate- to high-quality evidence indicates that the CPAx is a valid and reliable instrument for evaluating physical function and activity of critically ill patients. Further research is needed to evaluate responsiveness. PROSPERO CRD420250655518.