Falls pose a serious threat to older people, since they may lead to severe injuries, reduced quality of life and increased health care costs. Every third person over 65 years old falls at least once each year, and the number of falls increases with age and frailty level. Falls are multifactorial by nature and a person can have several risk factors contributing to a fall. A variety of assessment scales have been developed for assessing fall risk factors and estimating the probability of future falls. These are typically administrated by a health care professional. However, selection of an assessment scale with high enough sensitivity and specificity and reasonable administration time can be difficult. The goal of this thesis was to develop new methods for fall risk assessment utilizing accelerometry-based movement sensing, which enables objective detection and assessment of a person's balance deficits. The first objective was to investigate the perceptions of prospective end-users of new technologies via focus group interviews. The analysis showed that familiarity, prior experience and self-efficacy presumably affect the acceptance of new solutions. The second objective was to investigate how an individual's fall risk is manifested through different assessment scales. The Disease State Fingerprint visualization method was examined for its potential in comparing different fall risk assessment scales. It was found useful in discovering the most relevant assessment scales for separating fallers from non-fallers in the study population, and for presenting how the overall fall risk of an individual is constituted. The third objective was to study how body-worn accelerometry could be utilized in the assessment of individual fall risk. For the third objective, three data sets were collected from a total of 111 subjects. The results showed that features derived from the body-worn accelerometer signals could be used for assessment of a person's balance. Furthermore, they seem to be able detect balance deficits even earlier than the traditionally used clinical assessment scales. The results provide a basis for studies validating these methods and further transferring them into practice.
|Award date||17 Nov 2017|
|Place of Publication||Espoo|
|Publication status||Published - 2017|
|MoE publication type||G5 Doctoral dissertation (article)|
- wearable sensors
- machine learning
- gait analysis