220. Development of a risk score to predict accidental falls in an elderly cohort
Contributed abstract in session FB-4: Artificial Intelligence, stream Regular talks.
Friday, 11:00-12:30Room: Room S3
Authors (first author is the speaker)
| 1. | Rachda Naila MEKHALDI
|
| Centre Ingénierie et Santé, Ecole des Mines de Saint Etienne | |
| 2. | Julia Fleck
|
| Ecole des Mines de Saint-Etienne | |
| 3. | PHAN Raksmey
|
| Centre Ingénierie et Santé , Ecole des Mines de Saint Etienne | |
| 4. | Xiaolan Xie
|
| Mine de Saint-Etienne |
Abstract
Accidental falls are the primary cause of accidental death among the elderly and a significant barrier to active and healthy aging. Early fall risk prediction can assist caregivers in defining interventions that reduce the risk and prevent a fall, thus significantly impacting healthcare services. In this study, our goal is to construct a risk score that accounts for the probability that an elderly individual will fall within six months. We use an open source database containing information on fall incidents, fall history, and fall risk assessment scores, among others, from a cohort of 301 patients aged 65 and over. To preprocess the data, we perform numerical data standardization and categorical data encoding, and use Boruta algorithms for feature selection. We assess the performance of three classification algorithms (Random Forest, Extreme Gradient Boosting, and Falling Rule List) whose hyper-parameters are optimized through Bayesian methods. We report our results using classification metrics including recall, precision, and F1-score. Our findings will guide future algorithmic developments in the context of an ongoing project.
Keywords
- Artificial Intelligence
- Data analysis and risk management
- Analytics
Status: accepted
Back to the list of papers