Predict Near Future Initiation of Bed Exit
| Status: | Completed |
|---|---|
| Conditions: | Insomnia Sleep Studies |
| Therapuetic Areas: | Psychiatry / Psychology |
| Healthy: | No |
| Age Range: | 18 - Any |
| Updated: | 1/1/2014 |
| Start Date: | December 2012 |
| End Date: | December 2013 |
| Contact: | Sean Halpin, MA |
| Email: | sean.halpin@va.gov |
| Phone: | 404-321-6111 |
Predict Near Future Initiation of Bed Exit to Prompt Effective Intervention to Avoid Nighttime Falls With Pattern-recognition Algorithms Using Unobtrusive Monitoring of Movement and Vital Signs
Presence/absence in bed along with heartbeat, respiration, and gross motion in bed will be
measured in 48 Budd Terrace residents, a long-term care facility of Emory Healthcare.
Measurement will be done using only pressure-sensitive mats that lie underneath the mattress
and never touch the patient. PHI information will be collected by Emory staff. This PHI
will be restricted to: age at time of participation; medical conditions; and medications.
The PHI will be stored in a locked file behind a locked door. Data management will provide
a unique identifier for each participant linked to a name that will be kept separately from
the aggregate data.
The data collected from the bed sensor will be processed offline and separately from the PHI
to do proof of concept evaluation for the use of machine learning technology to predict bed
exits 1 to 5 minutes ahead of time.
measured in 48 Budd Terrace residents, a long-term care facility of Emory Healthcare.
Measurement will be done using only pressure-sensitive mats that lie underneath the mattress
and never touch the patient. PHI information will be collected by Emory staff. This PHI
will be restricted to: age at time of participation; medical conditions; and medications.
The PHI will be stored in a locked file behind a locked door. Data management will provide
a unique identifier for each participant linked to a name that will be kept separately from
the aggregate data.
The data collected from the bed sensor will be processed offline and separately from the PHI
to do proof of concept evaluation for the use of machine learning technology to predict bed
exits 1 to 5 minutes ahead of time.
Falls and fall-related injuries are the leading cause of injury deaths among older adults.
This proposal will help prevent falls at night by developing a new alarm system. Current
bed-exit alarm systems sound when the patient is half way out of the bed or on the ground.
We need a warning for when a patient is about to try to exit the bed.
The investigators believe that patients' heart rate or breathing changes before they leave
bed. They may also start moving within the bed. This is a brief study with nursing home
patient participants. Our primary outcome of interest is bed-exits, and up to 10
participants at a time will be monitored for an average of 6 weeks (less than their
anticipated stay) until which time that 250 bed exits have been recorded. Nearly all
participants will have physical and/or mental impairments and will be at high risk for
falling.
The investigators will use an investigational device to watch over the patient using a pad
under the mattress. This monitor is called the "Early-Sense 5". The system works like a
microphone for very low sounds. It changes heart, lungs, and movement vibrations into tiny
electrical signals. A wire carries these signals to a control box.
The information collected in the box will be stored and checked later. We will use five
different math descriptions for recognizing patterns. One or more of these may be useful to
give a 1 - 5 minute early warning that the patient is about to exit the bed.
The plan is to determine whether patterns of differences in three areas (heart rate,
breathing rate, and body movement) can be recognized and depended on to warn us about
bed-exits or attempted bed-exits.
There are four study targets. The first is to develop five possible mathematical
descriptions. The second is to use the rest of the information to test which of the
descriptions have meaningful ability to predict that a patient is about to get out of bed.
The third is to show that warning times are one to five minutes. The fourth is to test the
best mathematical descriptions for false alarms and true fall prevention.
How doable Phase I is will depend on how well we can predict that a patient is about to get
out of bed. If we can identify a pattern easily, then Phase II research will be put
forward.
This study is supported by the National Institute on Aging (SBIR-I).
This proposal will help prevent falls at night by developing a new alarm system. Current
bed-exit alarm systems sound when the patient is half way out of the bed or on the ground.
We need a warning for when a patient is about to try to exit the bed.
The investigators believe that patients' heart rate or breathing changes before they leave
bed. They may also start moving within the bed. This is a brief study with nursing home
patient participants. Our primary outcome of interest is bed-exits, and up to 10
participants at a time will be monitored for an average of 6 weeks (less than their
anticipated stay) until which time that 250 bed exits have been recorded. Nearly all
participants will have physical and/or mental impairments and will be at high risk for
falling.
The investigators will use an investigational device to watch over the patient using a pad
under the mattress. This monitor is called the "Early-Sense 5". The system works like a
microphone for very low sounds. It changes heart, lungs, and movement vibrations into tiny
electrical signals. A wire carries these signals to a control box.
The information collected in the box will be stored and checked later. We will use five
different math descriptions for recognizing patterns. One or more of these may be useful to
give a 1 - 5 minute early warning that the patient is about to exit the bed.
The plan is to determine whether patterns of differences in three areas (heart rate,
breathing rate, and body movement) can be recognized and depended on to warn us about
bed-exits or attempted bed-exits.
There are four study targets. The first is to develop five possible mathematical
descriptions. The second is to use the rest of the information to test which of the
descriptions have meaningful ability to predict that a patient is about to get out of bed.
The third is to show that warning times are one to five minutes. The fourth is to test the
best mathematical descriptions for false alarms and true fall prevention.
How doable Phase I is will depend on how well we can predict that a patient is about to get
out of bed. If we can identify a pattern easily, then Phase II research will be put
forward.
This study is supported by the National Institute on Aging (SBIR-I).
Inclusion:
1. Ambulatory patient able to leave the bed.
2. Willingness to consent and participate in a 30-night study
Exclusion:
1. Lack of capacity to consent, without an identifiable surrogate.
2. Terminal Prognosis
3. Unstable health, as determined by the principal investigator, medical doctor, or
registered nurse.
We found this trial at
1
site
