Fall prediction of the elderly with a logistic regression model based on instrumented timed up & go

• Published in: Journal of mechanical science and technology Vol. 33; no. 8; pp. 3813 - 3818
• Main Authors: Seo, Jeongwoo, Kim, Taeho, Lee, Jinsoo, Kim, Junggil, Choi, Jinseung, Tack, Gyerae
• Format: Journal Article
• Language: English
• Published: Seoul Korean Society of Mechanical Engineers 01.08.2019; Springer Nature B.V; 대한기계학회
• Subjects: Control; Dynamical Systems; Engineering; Gait; Industrial and Production Engineering; Mechanical Engineering; Prediction models; Regression analysis; Regression models; Vibration; 기계공학; Timed-up & go test; Fall prediction in the elderly; Logistic regression analysis; Inertia measurement unit sensor
• ISSN: 1738-494X; 1976-3824
• DOI: 10.1007/s12206-019-0724-0

An attempt has been made to use an instrumented TUG (iTUG) that complements the limitations of the traditional “timed up and go” (TUG). However, the variables that represent the characteristics of a faller have been reported to be different among preceding studies with iTUG. Thus, the purpose of this study was to develop a fall prediction model based on three years follow-up study with iTUG. Total 69 subjects participated in this experiment: 26 fallers (4 male and 22 female) who fell within 12 months from the first year measurement date and those newly fallen within 12 months from the second and third year measurement day were added up, and 43 non-fallers (11 male and 32 female) who had no falls. ITUG was performed once a year (two experiments per year) for three consecutive years using IMU sensor system (APDM Inc.). Among 30 variables, the final fall prediction model with logistic regression analysis consisted of five variables such as the duration of the total and the sit-to-stand phase, peak velocity of trunk sagittal plane and range of motion of trunk horizontal plane during gait phase and peak turn velocity during the turn-to-sit phase. Prediction accuracy using the receiver operation characteristic curve was 69.9 %. It is necessary to develop a more accurate fall prediction model by increasing the follow-up period and adding the numbers of the fallers. Further, it is important to identify meaningful variables by consecutive years rather than simple annual comparison.