Can meaningful clinical information be derived from triaxial accelerometers in adult hospital psychiatry?

McLean Hospital – Fellow

Philippe Beauchamp, MD

Scientific Abstract


Physical activity can be influenced by a variety of factors in the context of mental health care, notably by pharmacological drugs. This study aims at identifying which environmental and individual factors most saliently influence change in the level of activity, including sleep, as measured by a triaxial accelerometer in inpatient psychiatry settings, a highly structured environment. Here, we report on our initial data set from the GENEActiv wristband in adult inpatient units as well as on the potential use of these activity datasets when coupled with electronic health records data such as pharmacological interventions.


GENEActiv triaxial accelerometer was worn by 157 adult psychiatric inpatients at McLean Hospital during their admission. Open-source DPSleep (Deep Phenotyping of Sleep) pipeline is used to derive sleep onset, sleep offset, sleep duration, the sleep fragmentation index, as well as the power spectral percentile of activity intensity within individuals. EHR datasets are represented in real use case scenarios in the form of summary reports. It comprises information on activity summary, diagnosis, pharmacological interventions, mental state examinations, and sleep/activity metrics derived from DPSleep.


A total 44,375 hours of actigraphy data were processed with a mean of 282.6. (SD 311.1) hours per participant. DPSleep pipeline was used to process raw actigraphy data and yielded sleep quantitative measures such as sleep duration and sleep onset/offset for 1823-night events. Color coded power spectrum percentiles of activity intensity was also derived for all participants.


Gathering large sets of physical activity data via actigraphy wristbands in psychiatric inpatients settings is possible and yields quantifiable clinical information on activity level and sleep during hospitalization. Actigraphy and electronic health records represent rich datasets that can be leveraged and summarized to inform clinical states, as measured by activity intensity and sleep metrics estimates.

Live Zoom Session – March 9th

research Areas


Philippe Beauchamp, MD, Habiballah Rahimi-Eichi, PhD, Joshua D. Salvi, MD, PhD, Einat Liebenthal, DSc, Savannah D. Layfield, BA, Charles Bray, BA, Robert Patterson, MD, Justin T. Baker, MD, PhD

Principal Investigator

Justin T. Baker, MD, PhD