Artificial neural networks based fall risk detection in nursing home residents and investigation of the effects of circuit exercises on physical fitness and balance

Title: Artificial neural networks based fall risk detection in nursing home residents and investigation of the effects of circuit exercises on physical fitness and balance Aim: The aim of this study is to investigate the effects of circuit exercises on physical fitness and balance in elderly people living in nursing homes and to develop an artificial neural network model for the prediction of fall risk. Method: The study was conducted in nursing homes affiliated with the Eskisehir Provincial Directorate of Family and Social Services. 41 elderly people were randomly divided into two groups as study and control groups. Circuit exercises were applied to the study group for six weeks, two days per week. The control group was given a written exercise program consisting of similar exercises. Participants' scores were evaluated both before and after the exercise intervention. Data was collected by using the Sociodemographic information form, Standardized mini-mental state examination, Senior fitness test, Performance oriented mobility assessment, Berg balance scale, EQ-5D Quality of life scale and the Friendship scale. An Artificial Neural Network model was developed to detect the risk of falls in elderly by predicting their balance scores. Results: After the exercise, improvements were observed in all subtests of the Senior Fitness Test in the study group, while only the two minutes's step test showed improvement in the control group (p<0,05). In the study group, significant improvements were also observed in balance and social participation scores after the exercise intervention (p<0,05), whereas no significant differences were found in the control group (p>0,05). Overall, the study group showed significantly better outcomes in terms of physical fitness, balance, and social participation compared to the control group. The performance criteria of the artificial neural network model designed to predict the risk of falls were found as R 2 0,97 and 0.88; RMSE 0,56 and 2,33; MAE 0,27 and 1,66; MAPE 0,01 and 0,07; MSE 0,35 and 0,36 for training and test data. Conclusion: It was observed that circuit exercises improved physical fitness, balance, and social participation in nursing home residents and the artificial neural network model performed effectively to predict fall risk in elderly. Keywords: Circuit Based Exercises, Aged, Neural Networks, Nursing Home, Physical Fitness, Falls, Balance