Clinical and kinesiological analysis of post-fall recovery strategies

The immediate consequences of falls in older persons are an insufficiently studied area. The recovery reactions are poorly understood. Annually, thousands of older persons are not capable of getting up independently after a fall. This often leads to psychological and physiological stress due to long time periods without external support. It should be a priority in healthcare to detect and prevent these so-called “long-lies” after falls as they can even be fatal. Motion sensor based monitoring technologies to detect and prevent falls and long lies seem very promising, but current detection algorithms lack sensitivity and specificity. Existing fall models and definitions ignore the recovery component completely and therefore need to be further developed. One valuable approach to assist the development of a more comprehensive method to detect and prevent long lies may be to identify the activity of standing up from a lying position, thereby describing recovery. However, at present, methodological approaches to the systematic objective description and assessment of kinematics of standing up are lacking. Therefore, the aim of this thesis was to analyse movement patterns during lie-to-stand transfers to help develop a reference motion- sequence model for further kinematic analyses of recovery after falls using wearable inertial sensors. The sensor-based analysis would help to understand characteristics of successful and non-successful recovery patterns after real-world falls and contribute knowledge for the development of automated analyses of falls and consequences of falls.
Common methodologies for fall-detection using wearable sensor technology were evaluated through a systematic literature review. The review identified the use of heterogeneous fall definitions and the use of simplified fall models that commonly lacked information on the resting period and recovery activities after the impact of a fall. The results led to a consensus process on the definition of a modified fall model involving the resting phase and recovery activities.
As a next step, a video analysis was conducted to define common movement patterns during lie-to-stand transfers of healthy younger and non-frail older adults. Based on this information, a motion-sequence model was developed. This model and the underlying movement patterns were used to enable automated analyses to assess different recovery strategies using wearable motion sensors.
Along with the video analysis, lie-to-stand transfer performance was assessed using body- worn motion sensors. The results showed the feasibility of obtaining relevant temporal and kinematic measures, such as transfer duration, angular velocity and vertical acceleration to analyse lie-to-stand movements and describe how older people stand up from the floor.
Finally, the kinematic analysis was applied on recovery patterns after real-world falls. Temporal and kinematic measures were analysed from accelerometer and orientation signals from the recovery phase of 77 validated real-world fall events from a large fall- repository. The characterisation of falls with successful recovery with and without resting after the impact as well as non-recovered falls with long-lies was feasible. Non-recovered and self-recovered falls showed differing movement patterns, e.g. the trunk pitch angle differed significantly. A temporal threshold that could differentiate between successful and unsuccessful recovery reactions was detected at 25 seconds. This information might help to guide fall-detection algorithms and develop automated personal fall emergency response systems with a better performance and higher acceptability. In summary, the thesis has developed a basis for an improved management of falls leading to long lies.