Related Experiment Video
Updated: Nov 6, 2025
05:08
Gathering Self-Initiated Rat Behavioral Data to Characterize Post-Stroke Deficits
Published on: March 15, 2024
1.3K
Using random forests to model 90-day hometime in people with stroke
Jessalyn K Holodinsky1, Amy Y X Yu2,3, Moira K Kapral2,4,5
1Department of Clinical Neurosciences, Cumming School of Medicine, University of Calgary, 3330 Hospital Drive NW, Calgary, AB, T2N4N1, Canada. jkholodi@ucalgary.ca.
BMC Medical Research Methodology
|May 11, 2021
Summary
Random forests regression effectively analyzed 90-day hometime post-stroke, despite data challenges. This method captured complex relationships, though extreme values require further accuracy improvements.
Area of Science:
- Stroke research
- Health services research
- Biostatistics
Background:
- Ninety-day hometime, a key post-stroke metric, has a challenging non-normal distribution.
- Analyzing hometime is difficult due to data constraints.
- This study explores random forests for hometime analysis.
Purpose of the Study:
- Evaluate random forests regression for analyzing 90-day post-stroke hometime.
- Assess the performance of random forests in predicting hometime.
- Understand the relationships between covariates and hometime.
Main Methods:
- Utilized administrative data from Ontario, Canada (2010-2017) for stroke hospitalizations.
- Employed random forests regression to predict 90-day hometime using 15 covariates.
- Assessed model accuracy with R-squared and explored variable importance and marginal effects.
Main Results:
- Analyzed 75,745 stroke patients; median 90-day hometime was 59 days.
- Random forests achieved reasonable prediction accuracy (adjusted R-squared: 0.3462).
- Identified inverse non-linear relationships for frailty, stroke severity, and age; ambulance arrivals reduced hometime.
Conclusions:
- Random forests show promise for analyzing 90-day hometime and its complex predictor relationships.
- Further research should compare random forests with other models.
- Improving prediction accuracy for extreme hometime values is a future goal.
