Inpatient stroke rehabilitation: prediction of clinical outcomes using a machine-learning approach.
Yaar Harari, Megan K O'Brien, Richard L Lieber and 1 others
PMID 32522242WHAT IT FOUND
Admission scores on the FIM, Ten-Meter Walk, Six-Minute Walk and Berg Balance were the strongest predictors of discharge scores, stronger than age, sex or stroke details.
The models were rough and built from 50 patients at one hospital.
Key findings
01Admission clinical test scores were the most important predictors of discharge scores, contributing 80 to 90 percent of relative importance.
02For a new patient, the average prediction error was 9.5 FIM points, 0.3 m per second walking speed, 80.8 m walking endurance and 7.4 Berg Balance points.
03Dysphagia was associated with lower discharge FIM and Berg Balance scores, and speech or language impairment was included as a predictor in some models.
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
Only 50 patients from one inpatient rehabilitation hospital were studied, so the models may not work well in other hospitals or patient groups. The only validation described was leaving one patient out at a time within the same 50 patients, not testing the models in a new sample. Patients received physical, occupational and speech-language therapy together, so the study cannot say how much any one discipline contributed to recovery. The authors warned that the percent of variance explained may be over-optimistic and based their claims on average error. The models did not identify patients who never achieved enough walking ability to complete the walking tests. The models did not include medical comorbidities or complications, and the study did not measure the specific amount of therapy each patient received from each discipline or patient engagement.
Declared interests
Funding is named as Shirley Ryan AbilityLab and the National Institutes of Health. The supplied text does not state author conflicts of interest.
The easy way to misread this
Do not treat these equations as exact forecasts for your patient. They were built from 50 patients at one hospital, and for a new patient the average errors were 9.5 FIM points, 0.3 m per second walking speed, 80.8 m walking endurance and 7.4 Berg Balance points.