Key result
A logistic regression model identified key risk factors for falls and achieved a prediction accuracy of 76.6%, compared to 73% for an artificial neural network model.
Why the study?
Does an artificial neural network model provide better prediction accuracy for falls compared to a logistic regression model in community-dwelling older adults?
Observational
Does an artificial neural network model provide better prediction accuracy for falls compared to a logistic regression model in community-dwelling older adults?
Absolute Event Rate: 76.6% vs 73%
Logistic regression outperformed artificial neural networks in predicting falls among community-dwelling older adults, identifying past balance issues as the strongest predictor.
Logistic regression modestly outperformed neural networks for fall prediction; hypothesis-generating and should not yet alter clinical risk assessment in older adults.
The objective of this study is to identify the significant risk factors of 'fall' among elderly people living in community centres. Different factors that may associate with falls are analysed via logistic regression (LR) and artificial neural networks (ANNs) models. According to the analysis, the most significant predictor in LR analysis is found to be the variable 'lost balance within the past year'. Other significant predictors are stroke, Hx of dizziness/fainting, fall within past 30 days, and the number of supplements/vitamins used. The LR model gives a prediction accuracy of 76.6% while an ANN model gives 73%. ANN is expected to provide a more robust computational model than LR to analyse past falls and predict future ones. However our research results are contrary to those expectations. These results suggest that the ANN model need a larger sample size to reach its full potential and to give more accurate predictions.
No takes yet. Share an insight, caveat, or question.
Onel et al. (2010) conducted an observational in Falls. Logistic regression (LR) model vs. Artificial neural networks (ANNs) model was evaluated on Prediction accuracy for falls. A logistic regression model identified key risk factors for falls and achieved a prediction accuracy of 76.6%, compared to 73% for an artificial neural network model.
Synapse has enriched 5 closely related papers on similar clinical questions. Consider them for comparative context: