Bioengineering technologies to monitor movements in supported postures: a potential strategy to prevent pressure ulcers

Abstract

There are many clinical situations in which skin and soft tissues are subjected to sustained mechanical loading, particularly in individuals with restricted mobility. This can result in the breakdown of soft tissues in vulnerable areas, leading to the development of pressure ulcers (PUs). For several decades, interface pressure measuring systems have been employed to assess the magnitudes of pressures at the support surface interface of individuals at risk of developing PUs, typically to evaluate the short-term performance of pressure relieving systems (e.g. mattresses) and promote optimal postures. These technologies have recently been adapted to monitor over extended periods, providing the opportunity to estimate clinically-relevant temporal trends in posture and mobility. However, their ability to detect individual postural movements has not been established. Therefore, the present research was designed to assess the combination of pressure monitoring and intelligent data processing for the detection of postural changes during prolonged lying.A series of experimental studies utilised biomechanical parameters derived from pressure distribution and signals representative of body segmental movements using actimetry systems. Continuous measures were taken in cohorts of healthy individuals during evoked lying postures involving a raised head of the bed (HOB) and automated lateral tilt. The sensitivity and specificity of parameters for detecting changes in defined lying postures were examined. Data optimisation with Receiver Operating Characteristics (ROC) and Principal Component Analyses were performed to establish the most robust parameters, thus reducing the large volume of data associated with long-term monitoring. In particular, contact area and centre of pressure signals at specific body regions i.e. whole body and buttock, proved the most accurate of the interface pressure parameters, with ROC curve values (AUC) exceeding 0.5 for the majority of evoked postures. Signals derived from actimetry at the sternum also proved accurate in detecting postural movements, with the majority of postures revealing high AUC values. These parameters were combined with an automated detection method and machine learning algorithms to develop a robust methodology capable of predicting the frequency and magnitude of postural changes.The methodology was refined to accommodate a random sequence of postures on different support surfaces. The final automated methodology was then tested on pressure monitoring data from a small cohort of spinal cord injured subjects, who are vulnerable to PU development.Prediction of lying postural changes was achieved with a derivative threshold – based method, which yielded an accuracy of 100% when pressure signals were combined with body angles, and >85% for pressure signals in isolation. Prediction of lying postures was achieved by applying machine learning classifiers to either a combination of actimetry and pressure data or the pressure parameters in isolation. The most accurate combination of clinically relevant parameters involved pressure signals and body angles, achieving an average accuracy of ≥88%.The series of experiments and analytical approaches undertaken in this project contributed to the development of a semi-automated methodology based on robust biomechanical parameters for prediction of posture and mobility during prolonged lying. This was translated to a clinical data set, where long-term pressure monitoring was employed to evaluate previously unknown postures and provide the objective means to evaluate whether repositioning for pressure ulcer prevention adhered to international guidelines. Although further improvements are required for the analysis and visualisation of pressure data in clinical settings, this novel methodology has the potential to provide objective indication of posture and mobility which will inform effective personalised PU prevention.

More information

Identifiers

ORCID for Silvia Caggiari: orcid.org/0000-0002-8928-2141

ORCID for Peter Worsley: orcid.org/0000-0003-0145-5042

ORCID for Daniel Bader: orcid.org/0000-0002-1208-3507

Catalogue record

Export record