Behaviorally Assessed Sleep Disruptions in Memory Care Residents Using Passive Sensors: A Prospective Uncontrolled Cohort Pilot Study

Background:  This pilot study examined the feasibility of a passive sleep sensor system to assess sleep quality with a sample of memory care residents. The commercially available system is a ballistocardiography-based general sleep analysis tool developed to measure heart rate, breathing rate, and musculoskeletal movement. The system uses continuous cardiac and respiratory signals and movement data to infer sleep patterns and quality.  Although research with the general population indicates sleep monitoring using non-wearable sensors embedded in the environment is possible and provides meaningful longitudinal data that can potentially signal the emergence of health issues, little is known about the potential of the technology for people with dementia (PWD) with neuropsychiatric symptoms.  This gap in the literature is problematic as sleep disturbances and agitation are common with PWD, making established approaches to long-term sleep assessment impractical.   

Methods: Data were collected using a prospective cohort pilot design over a 12-week period.  Seventeen PWD residing in a single memory-care unit were recruited.  The validated commercially available system analyzed sleep patterns by detecting restlessness and human presence in bed with a single sensor pad.  Neuropsychiatric symptoms were assessed at baseline using the neuropsychiatric symptom inventory-nursing home version (NPI-NH), a valid scale used widely in dementia research.  Feasibilty issues related to our approach were tracked.  To evaluate high levels of neuropsychiatric symptoms, relative to low, weekly rates for sleep disruptions were modeled using generalized linear mixed models (GLMM) with the negative binomial distribution specified to allow for over- or under dispersion.  

Results:  The study revealed our approach to assessing sleep in PWD is feasible.  Fourteen of 17 (82%) memory care residents were enrolled in the study by their legal guardians and 10 of 14 residents participated in the study (71%).  The system continuously collected and securely transmitted data via Wi-Fi to a password protected website for analysis.  No systems were damaged or removed by participants and there were no participant, legal guardian, or staff injuries or complaints.  Five hundred and ninety sleep disruption measurements were obtained.  On average, there were four missing sleep disruption measurements for each participant.  Preliminary effect sizes were calculated to inform future studies.  Though data fit our GLMM and a trend emerged, there was no evidence of a significant association between neuropsychiatric symptoms and weekly sleep disruption (RR= 1.29, 95% CI: 0.68 – 2.44), our secondary outcome.   

Conclusions and Implications:  This is the first pilot study to demonstrate in a well-defined sample of PWD that passive sleep assessment is possible.  We anticipate given a larger sample, and a more robust clinical trial methodology, the trend would continue and a predictive model for sleep disruptions could be developed based on NPI scores.  

Advances in sensor, communication, internet-connectivity, and information technology over the past decade has created opportunities for social workers to test interventions in older adult living environments, and perhaps other environments.  Passive sleep assessment in PWD is an emerging area for social workers in both research and clinical contexts.  Future research in this emerging area is warranted.