Prediction of emergence from prolonged disorders of consciousness from measures within the UK rehabilitation outcomes collaborative database: a multicentre analysis using machine learning.
Richard J Siegert, Ajit Narayanan, Lynne Turner-Stokes
PMID 36031885WHAT IT FOUND
Routine rehab scores identified patients likely to stay unconscious better than patients likely to regain consciousness.
Models predicted staying unconscious more reliably than regaining consciousness. When tested on unseen patients, accuracy fell, so these scores are not yet dependable prognosis.
Key findings
01Motor impairment and communication dependence were the strongest admission predictors.
02When tested on withheld data, the neural network correctly identified 78% of patients who stayed unconscious but only 57% of patients who emerged.
03A rule predicted non-emergence with 91% capture when motor score was greater than 13, complex disability score was greater than 2, and cause was anoxia, stroke or toxic.
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 did not use a direct consciousness measure; PDOC and emergence were inferred from total UK FIM + FAM scores, and the authors say these scores still need validation against a gold standard. Patients with borderline scores of 32 to 35 were excluded, so the models do not show how to judge those cases. The sample was relatively young and admitted to specialist services, so it may not apply to older adults or patients in non-specialist services. Many patients were admitted an average of 231 days after onset, so early emergence may have been missed. Prediction of emergence was poor on withheld data, especially for the neural network and linear discriminant models. Database reporting changed over time and some services reported only a reduced dataset.
Declared interests
The original development of the UKROC database was funded by the UK National Institute for Health Research, and ongoing commissioning is funded by NHS England. The paper does not report a commercial sponsor.
The easy way to misread this
Do not use these models to tell a family whether a patient will regain consciousness. The study defined emergence with a disability score rather than a direct consciousness test, and prediction of emergence was much weaker than prediction of staying unconscious.