Predicting later categories of upper limb activity from earlier clinical assessments following stroke: an exploratory analysis.
Jessica Barth, Keith R Lohse, Marghuretta D Bland and 1 others
PMID 36810072WHAT IT FOUND
Early arm strength and function scores best predicted later daily arm use categories in people with stroke.
However, models were poor at accurately predicting these categories for new patients, with high error rates suggesting many other factors are involved.
Key findings
01Upper limb clinical measures, specifically the Shoulder Abduction Finger Extension (SAFE) test and the Action Research Arm Test (ARAT), were the most important predictors of later upper limb performance categories.
02While models explained the existing data well, their ability to predict outcomes for new individuals was poor, with out-of-bag error rates remaining high across all machine learning approaches.
03Adding non-motor clinical measures and demographic data to the models did not substantially reduce prediction errors, indicating that factors not captured in this study likely influence daily arm use.
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 sample size was small (n=54), preventing the use of separate training and testing sets. The study is exploratory and hypothesis-generating; results require validation in larger cohorts. Two of the five performance categories were under-represented in the sample. Important potential predictors like motor evoked potentials were not collected. Rehabilitation services received by participants were not controlled or measured.
Declared interests
The study was funded by the National Institutes of Health. The authors declared no conflicts of interest.
The easy way to misread this
Do not interpret the high in-sample accuracy (up to 1.00) as evidence that these models work well for new patients. The out-of-bag error rates were high (0.46 to 0.55), meaning the models failed to predict the correct category for nearly half of the cases when tested against data they had not seen during training.