Prediction of upper limb recovery after stroke. Which stratifying algorithms can clinicians integrate into practice?

The aim of this review is to summarise the post-stroke upper limb recovery stratifying algorithms and identify the ones that can be implemented in daily practice by physiotherapists and occupational therapists. 

The search for prospective and retrospective observational studies including stroke survivors aged ≥18 with upper limb impairment took place between the 1^st July and 15^th August 2024 in the following databases: Pubmed, AMED via Ovid, Scopus, CIHNHAL via EBSCOhost (MEDLINE and Academic Search Complete). The results of the search were imported in Zotero first, and then in Rayyan.ai to identify duplicates, as well as assign and manage articles to be screened. The authors used the Critical Appraisal Skills Programme (CASP) tool for guidance in assessing the quality of included papers.

The authors screened 1762 papers, leading to the inclusion of eight articles and three algorithms (Predict REcovery Potential or PREP2, Stroke Arm Longitudinal study at the University of Gothenburg or SALGOT, Alternative prognosis of recovery assessment for the hemiparetic limb or APRAHL). The PREP2 algorithm, based on factors like shoulder abduction, finger extension, age, National Institute of Health Stroke Scale (NIHSS), and motor evoked potentials, reported an accuracy of 75% at 3 months and 80% at 2 years post-stroke. It was lower in the validation studies outside New Zealand (60%). The SALGOT study, based on the Action Research Arm Test (ARAT) cube task, the grip strength and shoulder elevation AND/OR abduction, showed high specificity (0.92) and sensitivity (0.96) for different ARAT scores at three months. The APRHAL algorithm, including Shoulder Abduction Finger Extension (SAFE), Mobilization and Stimulation of Neuromuscular Tissue (MASONT), and NIHSS, reported a prediction accuracy of 69%, increasing to 76% for subjects with a SAFE score of ≥ 8

The PREP2 is the only validated and the best tool available, despite the barriers to its clinical implementation. The APRAHL algorithm presents challenges with standardisation, notably the use of MASONT techniques. The SALGOT is the easiest for clinical implementation by therapists in terms of time, equipment and training necessary, making it the most promising for its implementation in clinical practice. Further research is necessary to validate the algorithms included this review.

Clinicians can effectively participate in upper limb recovery post-stroke prediction with simple procedures learn in physiotherapy school, helping decision making in clinical practice.