ACCURACY OF PROGNOSTIC MODELS TO PREDICT LONG-TERM RECOVERY OF INDEPENDENT WALKING AFTER STROKE: A SYSTEMATIC REVIEW

N.C. Wouda¹, B. Knijff², M. Punt², J. Visser-Meily^1,3, M. Pisters³

¹University Medical Center Utrecht, Utrecht University and De Hoogstraat Rehabilitation, Center of Excellence for Rehabilitation Medicine, UMC Utrecht Brain Center, Utrecht, Netherlands, ²University of Applied Sciences Utrecht, Research Group Lifestyle and Health, Utrecht, Netherlands, ³University Medical Center Utrecht, Department of Rehabilitation, Physical Therapy Science & Sports, UMC Utrecht Brain Center, Utrecht, Netherlands

Background: People recovering from stroke experience reduced participation especially when they are limited in daily activities involving walking, such as housekeeping, mobility and physical exercise. This is in accordance to Lord et al. (2004) who concluded that 75% of stroke patients considered ‘being able to get out of the home’ essential or very important. From this perspective, it is not surprising that independent walking is one of the main goals in stroke rehabilitation. Therefore, it is necessary to gain insight in predicting the extent to which one will recover after stroke. Many prognostic models for predicting motor outcome after stroke have been developed, but it lacks of an overview of all these models and it is unclear which model is most accurate in predicting independent walking after stroke. Therefore, this systematic review provides an overview of current evidence about prognostic models and its performance to predict recovery of independent walking (i.e. walking without physical help) after stroke.

Purpose: To find the most accurate model to predict independent walking after stroke, so that individualized stroke rehabilitation can be improved.

Methods: In MEDLINE, CINAHL and Embase we searched up to December 2021 for all relevant studies in English and Dutch. Keywords were ‘stroke’, ‘independent walking’, ‘prediction’, ‘model performance’ and all terms related to these. Based on the criteria, title and abstract were screened on relevance by two reviewers. They retrieved full article copies of the relevant studies and select them independently. A conflict of opinion was resolved by consensus after discussion. The QUIPS Tool was used to determine the methodological quality of each paper. Descriptive statistics, study methods and model performance in terms of discrimination (AUC value, overall accuracy, sensitivity and/or specificity) were extracted.

Results: From the initially identified 12058 studies, only sixteen papers fulfilled all the search criteria, in which 21 prediction models were developed. All of these papers showed good to high methodological quality. Six prediction models showed an excellent performance (AUC value and/or overall accuracy ≥.90). Nevertheless, the CART model of Smith et al. (2017) showed highest overall accuracy (100%) in predicting the ability of independent walking in the subacute phase after stroke. Their model included one prognostic factor: hip extension one week poststroke, with which an excellent prediction could be made for independent walking twelve weeks after stroke (sensitivity = .80 (95%CI: .28-.99), specificity = 1.0 (95%CI: .90-1.0).

Conclusions: Only six of the 21 prediction models showed an excellent performance. However, in most cases there is a lack of external validation, which also applies for the most accurate model of Smith et al. (2017).

Implications: Smith et al. (2017) determined a cut-off value of “hip extension against gravity (MRC≥3) one week poststroke” which predicts the ability of walking independently twelve weeks after stroke. Since only two measures should be tested in the first week poststroke (TCT and hip extension), the model is easy applicable in clinical practice. The ability to inform patients and their relatives about the recovery potential, supports the individualized stroke rehabilitation at stroke units and rehabilitation centers.

Funding acknowledgements: SIA RAAK (Grant number RAAK.PRO03.006)

Keywords:
Stroke
Prediction
Independent walking

Topics:
Neurology: stroke
Community based rehabilitation

Did this work require ethics approval? No

Reason: Ethics approval was not required, because of the study design (systematic review)