Use this URL to cite or link to this record in EThOS:
Title: Semi-automatic falls risk estimation of elderly adults using single wrist worn accelerometer
Author: Sankar Pandi, Sathish Kumar
ISNI:       0000 0004 5353 011X
Awarding Body: University of Newcastle upon Tyne
Current Institution: University of Newcastle upon Tyne
Date of Award: 2014
Availability of Full Text:
Access from EThOS:
Access from Institution:
The population of the oldest old (aged 85 years and over) is growing. It is estimated that 30% of the adults over the age of 65 years experience falls at least once a year. This figure rises to 50% per annum for adults over 80 years living either at home or in care home. Currently older people are the fastest growing segment of the population. In the UK alone, the proportion of people aged 85 years old has increased from 2% to 4% in the past six decades. This marked increase in growth of population aged over 85 years is expected to have substantial impact on overall falls rate and pose serious issues to meet care needs for social and health care departments. In the light of such negative consequences for the faller and the associated costs to society, simple and quantitative techniques for falls risk screening can contribute significantly. This study describes a semi-automated technique to estimate falls risk of community dwelling elderly adults (aged 85 and over). This study presents the detailed analysis of tri-axial accelerometer movement data recorded from the right wrist of individuals undertaking the Timed Up and Go (TUG) test. The semi-automated assessment is evaluated here on 394 subjects’ data collected in their home environment. The study compares logistic regression models developed using accelerometer derived features against the traditional TUG measure ‘time taken to complete the test’. Gender based models were built separately across two groups of participants- with and without walking aid. The accelerometer derived feature model yielded a mean sensitivity of 63.95%, specificity of 63.51% and accuracy of 66.24% based on leave one-out cross validation compared to manually timed TUG (mean sensitivity of 52.64%, specificity of 45.41% and accuracy of 55.22%). Results show that accelerometer derived models offer improvement over traditional falls assessment. This automated method enables identification of older people at risk of falls residing both at home and in care homes and to monitor intervention effectiveness of falls management.
Supervisor: Not available Sponsor: Not available
Qualification Name: Thesis (Ph.D.) Qualification Level: Doctoral
EThOS ID:  DOI: Not available