Pressure injuries, also known as bed sores or pressure ulcers, are largely preventable yet they affect one in four Canadians across all healthcare settings. The current practice for treating pressure injuries is to reposition patients every two hours to allow the compressed tissues to return to their normal state; however, adherence to repositioning schedules are poor.
Our team recently developed a system that can predict patient position in bed using data from load cells under each bed leg and a machine learning algorithm. Our pilot data using healthy participants in a simulated home environment showed that the system was able to detect position (left-side, right-side, or supine) with 94.2% accuracy. However, when tested in a real homecare setting with older individuals, the system’s prediction accuracy decreased to 70%. This loss in accuracy may be due to the difficulty of labelling the wide variability of side-lying positions that fall between side-lying and supine.
The primary objective of this project is to develop an improved machine learning algorithm to predict an individual’s position more precisely. The new algorithms will predict the patient angle with respect to the surface of the bed in the transverse plane as a continuous value between -90° (left-side) and 90° (right-side). Secondary objectives include determining the range of trunk angles for which each greater trochanter and coccyx are offloaded.
The ultimate goal of this project is to design a prompting system to trigger an alert when a patient requires repositioning to improve healing rates.
This research is ongoing and results are not yet available.