Cross-validation of predictive models for functional recovery after post-stroke rehabilitation.
Silvia Campagnini, Piergiuseppe Liuzzi, Andrea Mannini and 4 others
PMID 36071452WHAT IT FOUND
A machine learning model predicted stroke rehabilitation outcomes with 79% accuracy using admission data.
Trunk control, communication level, and absence of pressure ulcers were the strongest predictors of functional improvement.
Key findings
01The model predicted functional class transition with 79.1% accuracy.
02Higher trunk control, absence of bedsores, and better communication levels were strong positive predictors of recovery.
03Severe communication limitations predicted a lack of functional class transition.
STILL TO COME
How it was doneWhat they foundWhat it means for PTsWhat it means for OTsWhat it means for SLPs
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 study was retrospective, using existing data from 2015 to 2017. The outcome measure was a class transition on the Modified Barthel Index, which assumes a linear relationship between score and clinical condition and treats all class changes as equal. The model was trained on data from two specific hospitals, so its accuracy may not generalise to other settings or populations. The study identifies predictors of outcome but does not prove that changing these factors (e.g., removing a catheter) will improve the result.
Declared interests
The study was funded by the Italian Ministry of Health and Regione Toscana. The authors declared no other conflicts of interest.
The easy way to misread this
Do not assume that the identified predictors cause the outcome. The study shows that patients with better trunk control and communication at admission are more likely to improve, but it does not prove that interventions targeting these specific factors will change the rehabilitation trajectory.