Investigating the usability of off-the-shelf sensors and using patient data to diagnose frailty

In recent years healthcare systems of countries worldwide have been overwhelmed with an increasing demand due to population aging (Care Quality Commission, 2017). The governments of these countries are focused on promoting remote/ in-home rehabilitation when possible as a potential solution to reduce costs and free up resources in critical hospitals (Lankila et al., 2016). This has been the case particularly with the elderly, whose population has increased exponentially within the past years. Remote home rehabilitation can empower patients to have control over their own rehabilitation process and makes them aware of their health-related habits and overall health situation (Tuntland, 2017). Another potential solutions to this problem are the avoidance of healthcare problems as well as educating patients to live healthier lifestyles. Wearable sensor technology is transforming rehabilitation processes in a positive way, providing valuable information which previously was not available, to health care staff - such as physiotherapists, nurses and GPs – who previously did not have access to patients own data (Patel et al., 2012). As a result, enhanced and better-informed decisions may be taken in situations where the patient does not demonstrate capacity. Currently, advances in technology have made “off-the-shelf activity trackers”, including advanced sensors, such as accelerometers, magnetometers, heartbeat and GPS sensors, available at a lower cost (Arriba-Pérez, Caeiro-Rodríguez and Santos-Gago, 2016; Tedesco, Barton, O'Flynn, 2017). However, there are currently only a few studies that investigate the usability of “off-the-shelf” wearable sensor technologies from an elderly person’s viewpoint, or their validity to assist in the diagnosis of frailty from a healthcare perspective.

In our SENDOC Northern Peripheries and Artic project (SENDoc Team, 2017), we are evaluating the effectiveness of off-the-shelf wearables for monitoring and rehabilitating remote and rural patients. Therefore, we are conducting demonstrations in 4 partner locations, where healthy participants aged over 60 years will wear a Mi Band activity tracker (Mi Global Home, 2018), a data logger and a smartphone to attain comparable data. The usability of this technology will be assessed from elders’ perspective. The data attained will then be analysed in combination with medical patient data to identify frailty. We hypothesise that off-the-shelf sensors can be used to automatically identify frailty. Statistical methods and qualitative usability questionnaires will be applied to validate or reject this hypothesis. Artificial Intelligence and machine learning methods will be employed to classify frail and pre-frail patients from non-frail patients. Cohen’s Kappa will be used to assess accuracy of classifications.

Results are not available at this stage. However, we expect that on-time therapeutic and medical advice can assist patients to recover full capacity, before frailty becomes irreversible.