A decision model to predict the risk of the first fall onset
| Autor: | Jean-Baptiste Mignardot, Christophe Cornu, Thibault Deschamps, Camille G. Le Goff, Gilles Berrut |
|---|---|
| Rok vydání: | 2016 |
| Předmět: |
Male
Aging medicine.medical_specialty Decision tree Poison control Risk Assessment Biochemistry Suicide prevention Machine Learning 03 medical and health sciences 0302 clinical medicine Endocrinology Physical medicine and rehabilitation Risk Factors 030502 gerontology Injury prevention Genetics medicine Humans Prospective Studies Postural Balance Molecular Biology Aged Human factors and ergonomics Cell Biology Test set Cohort Accidental Falls Female France 0305 other medical science Psychology Decision model 030217 neurology & neurosurgery Follow-Up Studies |
| Zdroj: | Experimental Gerontology. 81:51-55 |
| ISSN: | 0531-5565 |
| DOI: | 10.1016/j.exger.2016.04.016 |
| Popis: | BACKGROUND: Miscellaneous features from various domains are accepted to be associated with the risk of falling in the elderly. However, only few studies have focused on establishing clinical tools to predict the risk of the first fall onset. A model that would objectively and easily evaluate the risk of a first fall occurrence in the coming year still needs to be built. OBJECTIVES: We developed a model based on machine learning, which might help the medical staff predict the risk of the first fall onset in a one-year time window. PARTICIPANTS/MEASUREMENTS: Overall, 426 older adults who had never fallen were assessed on 73 variables, comprising medical, social and physical outcomes, at t0. Each fall was recorded at a prospective 1-year follow-up. A decision tree was built on a randomly selected training subset of the cohort (80% of the full-set) and validated on an independent test set. RESULTS: 82 participants experienced a first fall during the follow-up. The machine learning process independently extracted 13 powerful parameters and built a model showing 89% of accuracy for the overall classification with 83%-82% of true positive fallers and 96%-61% of true negative non-fallers (training set vs. independent test set). CONCLUSION: This study provides a pilot tool that could easily help the gerontologists refine the evaluation of the risk of the first fall onset and prioritize the effective prevention strategies. The study also offers a transparent framework for future, related investigation that would validate the clinical relevance of the established model by independently testing its accuracy on larger cohort.Copyright © 2015. Published by Elsevier Inc. Language: en |
| Databáze: | OpenAIRE |
| Externí odkaz: |
|
Full Text from ScienceDirect