Using wearable technology to identify activity and sleep patterns in inpatients with delirium and Parkinson's disease

People with Parkinson's disease may be at increased risk of delirium. Delirium is an acute neuropsychiatric syndrome associated with altered levels of consciousness, confusion and impaired attention. Delirium in Parkinson's has been poorly defined in previous studies and has overlapping symptoms with Parkinson's and Parkinson's dementia, such as experience sleep-wake disturbance, hallucinations and delusions; this may have underestimated the prevalence. Delirium is associated with poorer outcomes including falls, cognitive decline and mortality. As delirium is preventable in a third of cases, early diagnosis could lead to improved outcomes for patients. This project will be the first to use wearable technology (WT) to objectively and continuously quantify activity, mobility and sleep patterns in people with Parkinson's with delirium over prolonged periods. Delirium fluctuates throughout the day, with many patients experiencing day-night reversal.
Methods
100 people with Parkinson's who are admitted to hospital will be complete delirium assessments over five consecutive days. Participants will be assessed with WT for seven days in hospital and again at home to compare their activity levels in the two environments. Data will be recorded continuously for seven days with an accelerometry-based WT (Axivity AX3) placed on the lower back. Seven-day accelerometer-based data will be segmented using an automatic cloud-based analytical pipeline, which has been developed and validated by the secondary supervisor. Individual walking bouts (continuous length of time spent walking) will be extracted and for each bout 14 validated gait characteristics will be evaluated. Novel sleep metrics will be developed using night-time wearable sensor data. We will develop novel models under a nonlinear functional data analysis (FDA) framework such as wavelet-domain functional analysis of variance which maximises information from the original process data and characterises functional features of different scale and location.