Objective: To investigate the effect of using Vibration Sensory Analyzer-3000 (VSA-3000) in patients with impaired vibration sensation caused by central nervous system injury. Design: Prospective observational study. Setting: A university hospital for the research and clinical practice of rehabilitation. Subjects: Sixty patients (30 stroke and 30 spinal cord injury) were recruited, aged between 20 and 71 years old, under stable medication. Interventions: Not applicable. Main Measure: VSA-3000 threshold test, tuning fork test and somatosensory evoked potential (SSEP) measurement. Results: Test-retest reliability was determined based on data collected from 60 subjects, and the intraclass correlation coefficient (ICC) for vibration perception thresholds (VPTs) was in the "substantial" range. The kappa value between VSA-3000 and SSEP was 0.877, which was higher than that of tuning fork (κ = 0.732). VSA-3000 had good diagnostic accuracy with a sensitivity of 94.8%, specificity of 92.9%, and positive-predictive value of 93.8% and negative-predictive value of 94.0%, each value was higher than that of tuning fork. The area under the receiver operating characteristic curve (AUC) of VSA-3000 was 0.95 (95% CI: 0.91 to 0.98) and that of tuning fork was 0.89 (95% CI: 0.85 to 0.95), and there was a significant difference between the two values (P = 0.0216). The types of injury and age were the independent correlates of the VPTs. Conclusion: The present study provides preliminary evidence that VSA-3000 is a non-invasive and convenient quantitative testing instrument with good diagnostic accuracy, and it may be useful as a screening tool for assessing impaired vibration sensation caused by central nerve injury.