Evaluatie van motorisch functioneren na beroerte - psychometrische eigenschappen van klinische en sensor-gebaseerde metingen
Year: 2024
Type: dissertation
Abstract: Stroke remains a leading cause of disability globally, with motor impairments of the upper and lower limbs being prevalent within the first six months post-stroke and persisting for years in about half of the cases. This prolonged impairment limits daily activities and social participation, challenging healthcare providers to optimise motor recovery through evidence-based treatments and community reintegration. Given the crucial role of motor function in regaining and maintaining independence, regular motor assessments are essential for capturing functional health status and guiding the recovery process within and beyond clinical rehabilitation settings. Clinical assessments typically provide insights from an observer's perspective or through subjective, patient-reported experiences. In contrast, wearable sensors can elucidate an objective view, providing valuable insights into patients' real-world motor performance. Despite their potential, these devices are not widely utilised in clinical settings due to the current lack of standardised methods for data extraction and interpretation. This dissertation aims to guide clinical professionals in regularly integrating clinical motor assessment in the stroke care continuum and to facilitate the interpretation of clinical and sensor-based measures, particularly for upper limb functions. Chapter 1 of the dissertation addresses the need for a specific guideline on which assessments to choose and when to administer them. A three-round e-Delphi study involving European stroke rehabilitation professionals found a consensus on nine prioritised outcome measures and a measurement framework with seven time points. This core set of outcome measures addresses a wide range of motor impairments. It includes assessments for ambulatory and non-ambulatory patients, covering the ICF domains of body function, activities, and participation. This framework spans the entire stroke care continuum, incorporating recurrent assessments every six months, ensuring comprehensive motor recovery monitoring and timely intervention adjustments. In Chapter 2, we followed the World Health Organisation's call to develop tools allowing monitoring of precise physical activity performance outcomes. We conducted a valid study video-recording fourteen stroke survivors' daily physical activities in their home environment while they wore movement sensors on their wrists, ankles, and one on the chest. Using ground truth video and sensor data, computational algorithms were trained and validated to detect gait (walking, stair walking) and posture types (lying, sitting standing) for five sensor setups. Superior machine learning approaches achieved high accuracy 229 (85%-93%) in distinguishing gait from non-gait activities, with the optimal setup involving bilateral wrist and ankle sensors. Detection accuracy among all gait and posture types remained above 80%. Chapter 3 addressed the need to specify sensor-based arm use performance measures to improve their interpretability. Using ground truth video labels and acceleration features from wrist sensors, we developed machine learning and thresholding methods to classify functional versus non-functional upper limb movements during lying, sitting and standing activity types. Classification performance was evaluated across developed methods and the conventional approach in this research field. Our machine learning and thresholding-based approach achieved 80% accuracy in distinguishing between functional and nonfunctional movements, outperforming conventional methods by approximately 10%. Optimal thresholds for distinguishing functional movements were identified, indicating the new methods' superiority for classifying functional upper limb activities. These findings provide information on the validity of arm use metrics when different data processing methods are used. The findings in Chapters 2 and 3 provide information on the criterion validity of physical activity and arm use performance outcome. We highlight the potential of wearable sensors to provide an objective layer of information on physical activity types, which is rarely applied in clinical settings due to a lack of accessibility to similar tools. Thus, we open-sourced our methods to grant access to future applications in research and clinical environments. Chapter 4 provides information on measurement properties of clinical upper limb measures across stroke recovery phases, which is important for interpreting scores and score changes. The research project collected data in a longitudinal observation study that included ninety-three participants to determine the convergent validity and responsiveness of clinical and sensor-based measures. The interpretability of change scores was determined using patient-perceived change between time points to estimate minimal important change (MIC) values. High convergent validity was found between capacity measures and arm use performance of the affected upper limb across recovery phases, indicating that participants used their upper limb ability during daily life to a large extent. Strong negative correlations between capacity and arm use of the non-affected side indicated adaptive compensation in participants with more severe impairments. Responsiveness across all investigated measures was consistent within a five-week period of the early subacute phase (day 10-day 28), whereas upper limb capacity measures showed good responsiveness across recovery phases. Eight cut-off values indicating MIC could be estimated for clinical assessments, and nine for arm use measures. Distinguishing the importance of change through MICs, accuracies varied from 66-87%. These findings suggest that while certain measures are valid indicators of motor function, their responsiveness to changes over time varies, possibly due to changes in response behaviour when assessing the relevance of observed change. These findings on upper limb measurement properties refer to a conventional rehabilitation setting and, therefore, be used as a reference when assessing individual patients. The arm use performance outcome is valid and responsive, essential for integrating regular sensor-based monitoring. This dissertation guides clinical professionals in integrating motor assessments into the stroke care continuum. It facilitates the interpretation of clinical and sensor-based upper limb outcome measures, particularly for upper limb functions.
Source: Lirias
Author: Pohl, Johannes
Language: English
Subfield: Rehabilitation
Field: Medicine
Domain: Health Sciences
Open Access status: green
OpenAlex ID: W7009717606
Cites:
Cited by:
Related to: