Falls are a major problem for older adults. Balance impairment is one of several major fall risk factors. One of the best ways to measure balance is posturography; however, lack of standardization in testing and reporting first need to be overcome to make the tool clinically useful. This study comprehensively examined four testing conditions, traditional time-domain postural sway parameters, and newer, promising fractal measures to develop a clinical protocol to differentiate fallers and non-fallers.
One hundred-fifty individuals aged 65 through 97 participated in this study. Sixty second quiet-standing trials were taken in four testing conditions: eyes open, comfortable stance; eyes closed, comfortable stance; eyes open, narrow stance; and eyes closed, narrow stance. Eight traditional postural-sway parameters and six fractal dimensions were calculated. Stepwise logistic binary regression was performed to identify the group of the postural sway parameters and physical characteristics that best differentiated fallers from non-fallers for each condition. This analysis was performed twice with fallers defined using two different definitions: at least one fall in the past year and multiple falls in the past year.
Results found that individuals who had fallen at least once were not well differentiated from non-fallers. A single fall, or lack of, does not indicate the presence, or absence, of postural instability and should not be used clinically to identify fall risk. Multiple fallers were differentiated from individuals who had not fallen or had only fallen once. Medial-Lateral Sway Velocity was the most important postural sway parameter in differentiating the two groups in all conditions. Logistic regression promoted the use of the eyes closed, comfortable stance condition to best differentiate individuals based on fall history. The associated model included, in order or importance: Medial-Lateral Velocity, Anterior-Posterior Short-Term Fractal Dimension, Medial-Lateral Short-Term Fractal Dimension, Body Mass Index, and Age. This model demonstrated very good ability in identifying non-fallers as low-risk, and moderate accuracy in correctly identifying fallers as high-risk. Fractal analysis was an important inclusion and revealed novel findings of two distinct scaling regions. Future prospective work is necessary to extend findings to prediction of future falls.