Long-term forecasting of a motor outcome following rehabilitation in chronic stroke via a hierarchical bayesian dynamic model.
Nicolas Schweighofer, Dongze Ye, Haipeng Luo and 2 others
PMID 37386512WHAT IT FOUND
A computer model predicted upper-limb recovery in chronic stroke patients by combining their initial scores with repeated progress checks.
Accuracy improved significantly as more data arrived. The model worked best after the second week of therapy, offering reliable forecasts for six months ahead.
Key findings
01The model's ability to predict recovery six months ahead improved as more therapy data became available, reaching a level considered clinically useful after the second week of training.
02Using information from previous patients (hierarchical structure) significantly helped predict outcomes for new patients when only initial data was available, but this advantage disappeared once three weeks of therapy data were collected.
03The model successfully replicated known clinical patterns, such as how higher doses of therapy lead to greater initial gains but also faster forgetting, and how self-practice outside of therapy sustains progress.
STILL TO COME
How it was doneWhat they foundWhat it means for PTsWhat it means for OTs
Read the rest of this summary
You get three full summaries a month, free, and we do not ask for a card. Search, the TL;DRs and your library stay unlimited either way.
What it does not show
The model was tested only on two specific trials with chronic stroke patients who had mild-to-moderate arm impairment. It has not been validated in acute stroke or in patients with severe impairment. The outcome measure (MAL) is based on patient self-report, which can be influenced by mood or memory, rather than objective performance tests. The study used existing data; the model has not yet been used prospectively to guide actual clinical decisions. The model did not include biological factors like brain imaging or genetic data, which might improve predictions. The benefit of using population data (hierarchy) vanishes after three weeks, meaning the model's value for new patients is limited to the very early phase of therapy.
Declared interests
The study was supported by the National Institute of Neurological Disorders and Stroke, the National Institute of Biomedical Imaging and Bioengineering, and the Alfred E. Mann Institute. The authors declared no competing interests.
The easy way to misread this
Do not use this model to predict a patient's final outcome after only one or two assessments. The study shows that predictions are unreliable until after the second week of therapy, and the model is not yet available for clinical use.