Reliability and validity of inertial sensor–derived scores for the MDS-UPDRS motor examination in Parkinson’s disease: a systematic review and meta-analysis

Abstract Background Motor assessment in Parkinson’s disease (PD) mainly relies on the Movement Disorder Society–Unified Parkinson’s Disease Rating Scale (MDS-UPDRS) Part III. Inertial measurement units (IMUs) are promising tools for objective scoring the standardized motor tasks included in this scale. Objectives To synthesize evidence on the metric properties (reliability and criterion validity) of IMU-derived scores for different items of the MDS-UPDRS in the assessment of individuals with PD. Methods This systematic review followed PRISMA-COSMIN guidelines. A literature search was conducted on Nov 6, 2025, in PubMed, Embase, CINAHL, Scopus, and Web of Science databases. Studies including adults with PD using IMUs during at least one MDS-UPDRS Part III task were eligible if sensor-derived metrics were linked to clinician-rated scores. Risk of bias was assessed using QAREL (reliability) and COSMIN checklists (validity). Random-effects meta-analyses were performed when appropriate. Results Of 1,794 records, 22 studies were included, most of which investigated validity. Quality appraisal showed that the risk of bias varied across studies and was generally moderate or high. Data from three studies consistently found good-to-excellent intra-rater reliability for a few items of MDS-UPDRS and for a composite index designed to measure the overall response to drug therapy (pooled ICC = 0.867; 95% CI:0.79–0.915; N = 84; I 2 = 0.00%). No data on inter-rater reliability and on measurement error were available. The criterion validity was assessed in terms of agreement/correlation with clinicians’ scores and showed high variability across the items, with pooled coefficients ranging from 0.487 (Kinetic tremor) to 0.908 (Gait), and high heterogeneity in most of the estimates. Conclusion At present, there is a low level of evidence of adequate reliability of IMU-derived scores for a very limited number of MDS-UPDRS items as measured by IMUs. Although for some items the agreement/correlation with clinician-rated scores was high, the large inconsistency and the small sample of most estimates also question the validity of these scores. These findings indicate that the integration of digital measures into evidence-based clinical decision-making requires further research.