Reliability, validity, and clinical feasibility of a rapid and objective assessment of post-stroke deficits in hand proprioception

Abstract

Background: Proprioceptive function can be affected after neurological injuries such as stroke. Severe and persistent proprioceptive impairments may be associated with a poor functional recovery after stroke. To better understand their role in the recovery process, and to improve diagnostics, prognostics, and the design of therapeutic interventions, it is essential to quantify proprioceptive deficits accurately and sensitively. However, current clinical assessments lack sensitivity due to ordinal scales and suffer from poor reliability and ceiling effects. Robotic technology offers new possibilities to address some of these limitations. Nevertheless, it is important to investigate the psychometric and clinimetric properties of technology-assisted assessments.

Methods: We present an automated robot-assisted assessment of proprioception at the level of the metacarpophalangeal joint, and evaluate its reliability, validity, and clinical feasibility in a study with 23 participants with stroke and an age-matched group of 29 neurologically intact controls. The assessment uses a two-alternative forced choice paradigm and an adaptive sampling procedure to identify objectively the difference threshold of angular joint position.

Results: Results revealed a good reliability (ICC(2,1) = 0.73) for assessing proprioception of the impaired hand of participants with stroke. Assessments showed similar task execution characteristics (e.g., number of trials and duration per trial) between participants with stroke and controls and a short administration time of approximately 12 min. A difference in proprioceptive function could be found between participants with a right hemisphere stroke and control subjects (p<0.001). Furthermore, we observed larger proprioceptive deficits in participants with a right hemisphere stroke compared to a left hemisphere stroke (p=0.028), despite the exclusion of participants with neglect. No meaningful correlation could be established with clinical scales for different modalities of somatosensation. We hypothesize that this is due to their low resolution and ceiling effects.

Conclusions: This study has demonstrated the assessment's applicability in the impaired population and promising integration into clinical routine. In conclusion, the proposed assessment has the potential to become a powerful tool to investigate proprioceptive deficits in longitudinal studies as well as to inform and adjust sensorimotor rehabilitation to the patient's deficits.

Keywords: Difference threshold; MCP; Metacarpophalangeal joint; Parameter Estimation by Sequential Testing; Psychophysics; Quantitative measurements; Robot-assisted assessment; Somatosensory function.

Fig. 1

Robotic Sensory Trainer. a Rest position of the index finger. Side view on the remote-center-of-motion (RCM) mechanism of the apparatus used to apply passive movements around the metacarpophalangeal (MCP) joint. b Flexed position of the index finger. c Experimental setup with a touchscreen, covering the tested hand, for instructions and post-trial subject feedback on perceived stimuli

Fig. 2

Comparison of the difference limen (DL) of both hands in neurologically intact control (NIC) subjects, participants with left hemisphere stroke (LHS) and right hemisphere stroke (RHS). For the patients, test and retest were averaged for a better DL estimate. The dashed bracket indicates that the statistical test was conducted on baseline-removed data (i.e., using the median for the corresponding hand of the NIC)

Fig. 3

Bland-Altman plot of the test-retest of the robotic assessment in participants with stroke. The bars indicate the mean difference $\overline{d}$ between the difference limen (DL) of the two test sessions (solid black line) and its 95% CI (gray bar) for both impaired and unimpaired hand separately