Machine Learning Algorithms for Prediction of Ambulation and Wheelchair Transfer Ability in Spina Bifida.
Gina McKernan, Matt Mesoros, Brad E Dicianno
PMID 39631515WHAT IT FOUND
A spina bifida prediction model correctly classified 83.22% of ambulation categories, but missed most household and therapeutic ambulators.
Wheelchair transfer prediction also struggled to identify patients unable to transfer.
Key findings
01The ambulation model was 80.21% accurate with the unbalanced training dataset and 83.22% accurate with the balanced training dataset.
02In the balanced model, the network correctly identified 93.21% of community ambulators and 76.70% of nonambulators, but only 10.00% of household ambulators and 23.96% of therapeutic ambulators.
03For wheelchair transfers, the unbalanced model was 70% accurate overall, correctly classified 97.3% of patients able to transfer unassisted, but classified only 3.6% of patients unable to transfer; balancing raised correct classification of inability to transfer to 29.1%.
STILL TO COME
How it was doneWhat they foundWhat it means for PTs
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 used patients from large academic medical centers only, so it may not apply to community settings. Children younger than 5 years were excluded, so it does not address younger children. 76 cases were dropped because of missing input variables. The sample size was not large by machine learning standards. Ambulation categories were unbalanced, with most patients in community or nonambulator groups. The model did not predict household or therapeutic ambulators well. Wheelchair transfer prediction missed most patients unable to transfer unassisted.
Declared interests
No explicit funding or conflict-of-interest statement appears in the supplied text. Publication types include Research Support, U.S. Gov't, P.H.S.
The easy way to misread this
Do not treat the 83.22% overall accuracy as evidence that the model can guide individual treatment decisions. It correctly identified only 10.00% of household ambulators and 23.96% of therapeutic ambulators in the balanced model, and only 3.6% of patients unable to transfer in the unbalanced transfer model.