The Timed Up and Go (TUG) test is widely used for assessing mobility and falls risk of elderly individuals. In this study, we aim to utilize TUG test to estimate disability level of community dwelling elderly participants. Forty features are extracted from single wrist mounted accelerometer signals which are recorded in home environment from the 321 participants performing TUG test. As an initial exploratory analysis, linear discriminant classifier is used to estimate the disability levels. The study compares models built using features extracted from accelerometer signals with the standard measure which is the time taken to complete the test. The developed accelerometer model has yielded a mean accuracy of 62.16% outperforming the standard measure with a mean accuracy of 39.10%. The obtained results show that TUG test has an ability to classify disability levels and accelerometer has an added value in evaluations and monitoring progression of disability levels.