Exploring physiotherapists' and occupational therapists' perceptions of the upper limb prediction algorithm PREP2 after stroke in a rehabilitation setting: a qualitative study

Biomarkers of Motor Outcomes After Stroke

Predicting motor outcomes after stroke based on clinical judgment alone is often inaccurate and can lead to inefficient and inequitable allocation of rehabilitation resources. Prediction tools are being developed so that clinicians can make evidence-based, accurate, and reproducible prognoses for individual patients. Biomarkers of corticospinal tract structure and function can improve prediction tool performance, particularly for patients with initially moderate to severe motor impairment. Being able to make accurate predictions for individual patients supports rehabilitation planning and communication with patients and families.

Predicting later categories of upper limb activity from earlier clinical assessments following stroke: an exploratory analysis

BACKGROUND

Accelerometers allow for direct measurement of upper limb (UL) activity. Recently, multi-dimensional categories of UL performance have been formed to provide a more complete measure of UL use in daily life. Prediction of motor outcomes after stroke have tremendous clinical utility and a next step is to explore what factors might predict someone's subsequent UL performance category.

PURPOSE

To explore how different machine learning techniques can be used to understand how clinical measures and participant demographics captured early after stroke are associated with the subsequent UL performance categories.

METHODS

This study analyzed data from two time points from a previous cohort (n = 54). Data used was participant characteristics and clinical measures from early after stroke and a previously established category of UL performance at a later post stroke time point. Different machine learning techniques (a single decision tree, bagged trees, and random forests) were used to build predictive models with different input variables. Model performance was quantified with the explanatory power (in-sample accuracy), predictive power (out-of-bag estimate of error), and variable importance.

RESULTS

A total of seven models were built, including one single decision tree, three bagged trees, and three random forests. Measures of UL impairment and capacity were the most important predictors of the subsequent UL performance category, regardless of the machine learning algorithm used. Other non-motor clinical measures emerged as key predictors, while participant demographics predictors (with the exception of age) were generally less important across the models. Models built with the bagging algorithms outperformed the single decision tree for in-sample accuracy (26-30% better classification) but had only modest cross-validation accuracy (48-55% out of bag classification).

CONCLUSIONS

UL clinical measures were the most important predictors of the subsequent UL performance category in this exploratory analysis regardless of the machine learning algorithm used. Interestingly, cognitive and affective measures emerged as important predictors when the number of input variables was expanded. These results reinforce that UL performance, in vivo, is not a simple product of body functions nor the capacity for movement, instead being a complex phenomenon dependent on many physiological and psychological factors. Utilizing machine learning, this exploratory analysis is a productive step toward the prediction of UL performance. Trial registration NA.

Prognosis of Individual-Level Mobility and Daily Activities Recovery From Acute Care to Community, Part 2: A Proof-of-Concept Single Group Prospective Cohort Study

OBJECTIVE

To demonstrate a proof-of-concept for prognostic models of post-stroke recovery on activity level outcomes.

DESIGN

Longitudinal cohort with repeated measures from acute care, inpatient rehabilitation, and post-discharge follow-up to 6 months post-stroke.

SETTING

Enrollment from a single Midwest USA inpatient rehabilitation facility with community follow-up.

PARTICIPANTS

One-hundred fifteen persons recovering from stroke admitted to an acute rehabilitation facility (N=115).

INTERVENTIONS

Not applicable.

MAIN OUTCOME MEASURE(S)

Activity Measure for Post-Acute Care Basic Mobility and Daily Activities domains administered as 6 Clicks and patient-reported short forms.

RESULTS

The final Basic Mobility model defined a group-averaged trajectory rising from a baseline (pseudo-intercept) T score of 35.5 (P<.001) to a plateau (asymptote) T score of 56.4 points (P<.001) at a negative exponential rate of -1.49 (P<.001). Individual baseline scores varied by age, acute care tissue plasminogen activator, and acute care length of stay. Individual plateau scores varied by walking speed, acute care tissue plasminogen activator, and lower extremity Motricity Index scores. The final Daily Activities model defined a group-averaged trajectory rising from a baseline T score of 24.5 (P<.001) to a plateau T score of 41.3 points (P<.001) at a negative exponential rate of -1.75 (P<.001). Individual baseline scores varied by acute care length of stay, and plateau scores varied by self-care, upper extremity Motricity Index, and Berg Balance Scale scores.

CONCLUSIONS

As a proof-of-concept, individual activity-level recovery can be predicted as patient-level trajectories generated from electronic medical record data, but models require attention to completeness and accuracy of data elements collected on a fully representative patient sample.

External validation and extension of the Early Prediction of Functional Outcome after Stroke (EPOS) prediction model for upper limb outcome 3 months after stroke

Objective

The ‘Early Prediction of Functional Outcome after Stroke’ (EPOS) model was developed to predict the presence of at least some upper limb capacity (Action Research Am Test [ARAT] ≥10/57) at 6 months based on assessments on days 2, 5 and 9 after stroke. External validation of the model is the next step towards clinical implementation. The objective here is to externally validate the EPOS model for upper limb outcome 3 months poststroke in Switzerland and extend the model using an ARAT cut-off at 32 points.

Methods

Data from two prospective longitudinal cohort studies including first-ever stroke patients admitted to a Swiss stroke center were analyzed. The presence of finger extension and shoulder abduction was measured on days 1 and 8 poststroke in Cohort 1, and on days 3 and 9 in Cohort 2. Upper limb capacity was measured 3 months poststroke. Discrimination (area under the curve; AUC) and calibration obtained with the model were determined.

Results

In Cohort 1 (N = 39, median age 74 years), the AUC on day 1 was 0.78 (95%CI 0.61, 0.95) and 0.96 (95%CI 0.90, 1.00) on day 8, using the model of day 5. In Cohort 2 (N = 85, median age 69 years), the AUC was 0.96 (95%CI 0.93, 0.99) on day 3 and 0.89 (95% CI 0.80, 0.98) on day 9. Applying a 32-point ARAT cut-off resulted in an AUC ranging from 0.82 (95%CI 0.68, 0.95; Cohort 1, day 1) to 0.95 (95%CI 0.87, 1.00; Cohort 1, day 8).

Conclusions

The EPOS model was successfully validated in first-ever stroke patients with mild-to-moderate neurological impairments, who were independent before their stroke. Now, its impact on clinical practice should be investigated in this population. Testing the model’s performance in severe (recurrent) strokes and stratification of patients using the ARAT 32-point cut-off is required to enhance the model’s generalizability and potential clinical impact.