Appling Decision Tree Technique To Predict Fall Risks Classified By Functional Fitness In Older Adults

Abstract

PURPOSE: To examine an accuracy of the fall risk classification predicted by senior functional fitness tests and health conditions in Korean older adults utilizing a decision tree technique, which is a machine learning algorithm METHODS: A total of 732 older adults (Males 45.9%, age: 73.51±6.20yrs, BMI: 24.07) participated in Korea Survey of National Physical Fitness (KSNPF). All participants performed the senior physical fitness test (SPFT) including body composition (FAT%), 6min walk, hand grip(HG), timed up-and-go (TUG), and chair sit-and-stand. Also, demographic variables (income, employment, etc.) and health markers were measured, which were including blood pressure & waist and hip circumferences as well as physical activity levels and health conditions (disease, medication, and fall experiences). All utilized measures were validated for the Korean elderly in 2014. To determine a classifier of a fall risk and to set cut-off scores of the fall risk classification, a decision tree technique with CHAID algorithm (Kass, 1980) was applied. To examine the accuracy of the fall risk classification predicted by selected variables, 80% of participants were randomly selected to derive the equation for a training group (GR) and the others were assigned for a cross-validation GR (20% for testing). RESULTS: Only TUG and HG were significant classifiers with 87.6% of accuracy in the testing GR and 86.6% of precision in the cross-validation GR, respectively. The cut-off score of TUG was 7.13sec. (chi-square=21.22, p<.001), in which 21.1% of participants were classified into the fall risk GR. Within the TUG GR (<7.13sec.), the cut-off score of HG were set as 23.6kg (chi-square=14.87, p<.001), and 52.8% (<23.6kg) was classified into the fall risk GR. CONCLUSIONS: Among SPFT tests and health related variables, TUG and HG were relatively important to predict the fall risk for the healthy elderly in Korea. Coordination and strength exercise are critical for fall prevention in older adults. *Corresponding author ([email protected])