Early Prediction of Poststroke Rehabilitation Outcomes Using Wearable Sensors.
Megan K O'Brien, Francesco Lanotte, Rushmin Khazanchi and 5 others
PMID 38169444WHAT IT FOUND
Wearable sensors during a brief walk improved prediction of discharge ambulation and fall risk in ambulatory stroke patients, but did not help predict independence.
For nonambulatory patients, standard clinical records alone remained as accurate as sensor data.
Key findings
01For ambulatory patients, adding sensor data from a 10-meter walk improved the prediction of discharge ambulation and fall risk compared to standard clinical records, but not independence.
02For nonambulatory patients, sensor data from simple balance tasks did not add value to predicting discharge ambulation, and standard clinical records were as accurate as sensor models for independence and fall risk.
03A simple model using patient information and sensor data from a 10-meter walk correctly identified 27 of 29 ambulatory patients who did not change their ambulation category from admission to discharge.
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 study was conducted at a single site with a small sample size, particularly for nonambulatory patients (only 8 analyzed), which limits generalizability and increases the risk of overfitting. The models classified patients into binary categories (e.g., household vs. community ambulators) based on cutoff scores, so patients with scores near these boundaries were frequently misclassified. The study did not compare the sensor models against other machine learning algorithms that might perform better. For nonambulatory patients, the discharge outcomes were very similar, making it difficult for the models to learn meaningful patterns from the sensor data.
Declared interests
The funders played no role in the design, conduct, or reporting of this study.
The easy way to misread this
Do not assume that wearable sensors are a superior tool for all stroke predictions. They only improved accuracy for ambulatory patients predicting walking and fall risk, and failed to help predict independence or assist with nonambulatory patients. Standard clinical assessments remain essential.