Using a Real-Time Locating System to Measure Walking Activity Associated with Wandering Behaviors Among Institutionalized Older Adults

All methods described here have been approved by the Institutional Review Board at the Corporal Michael J. Crescenz VA Medical Center in Philadelphia, PA.

1. Installation and Set-Up of a Real-Time Locating System (RTLS)

1. Review facility policies, safety, and personal information protections for residents with facility stakeholders. Determine whether written or verbal support on the use of RTLS in the facility is required. In discussions with stakeholders include local protocols and procedures (e.g., local facility technology waivers, union sign-off, etc.) and a project timeline.¹²
   NOTE: Update protocols, procedures, and timeline as they change over the course of the project, meeting with stakeholders and acquiring sign-off from interested parties.
2. Obtain institutional review board approval including a HIPAA waiver to review medical charts prior to obtaining consent from eligible residents.
3. Equip the desired area of study with a RTLS (see Figure 1). Mount sensors in the upper corners of all common rooms and hallways to triangulate resident location and movement in real-time.
   1. Point sensors toward the middle of the area to utilize their antenna pattern which is +/- 90 degrees in the azimuth (Horizontal) and +/- 45 degrees in the elevation (Vertical). Tilt the face of the sensor downwards so that if a laser beam came out of the face of the sensor if would hit the opposite corner of the space about 5-6ft off the ground. Ensure the sensor is level by placing a level on the two plastic pegs on the top back of the sensor.
      NOTE: For a typical communal area in a long-term care facility (about 10m x 13m or 1,000 square feet), four sensors are needed. These sensors will cover a larger area but this is dependent on the surrounding environment – e.g., walls and glass partitions in the area which may have an impact on transmissions.
   2. Each sensor needs a network cable running from the lower left port on the back of the sensor to the switch that the server is connected to; this cable is a Cat5e cable. With one sensor as the master, run the timing cable from the master to each other sensor, thus a star topology.
      1. To do so, plug a shielded cat5e cable into any of the 6 available ports on the master and run it to each other sensor where it will be plugged into the upper right port of the 6 ports. Run cables above ceiling tiles.
         NOTE: The number of sensors in the area determine the number of ports required for the power over Ethernet (POE) switch. Each sensor will require two ports. Multiple POE switches can be connected if needed.
   3. Measure where the sensors are located in the area and choose an origin point on the sensor (e.g., the lower left corner so that moving north is the positive y axis and moving east is the positive x axis). Measure the x, y, and z of each sensor (with a laser distance measurer) in relation to this origin. Record the MAC address off the back of the sensor and keep to enter into the graphical user interface (GUI;a specialized software developed to manage the RTLS).
4. In the GUI, open Platform Control and click on Core Server to highlight it and then click start. Repeat this for the Service Controller. Click Apply and then OK.
   1. Open the Service Installer and press next. Browse to C:Ubisense Software and go into the Location Engine folder and highlight the “packages” folder. Click OK and next. Install all the services listed. Repeat this process again but go into the Platform folder and highlight the “packages” folder. Install all the services listed. Click on Service Manager and ensure all services appear as “running.”
5. Open the Site Manager and go to the Areas tab. Create a floor plan by opening notepad and specify the start and stop point of each wall by giving the x, y coordinates of the starting point followed by the ending point. Save the file as a .dat file. After the last set of points (0,0) press enter.
   1. In the Areas tab go to Walls>Load Walls and load the .dat file. Go to Regions>Set Origin and choose the lower left corner. Click the draw wall mode button and add a dummy wall anywhere inside the square. This tells the system where to compute the regions (inside vs. outside the region).
   2. Click Regions>Compute Regions; this highlights the square blue. Delete the wall – by selecting the Wall Mode button and pressing delete. Go to Area>Save Area and save the area. Go to the Cells tab and load the area by choosing it from the drop down Area box.
   3. Click the Add Extent button in the lower left corner. To use the defaults click Save. Right click on Site in the left column and choose New Geometry Cell. Click the Add Extent button and again use the defaults. Right click on the Geometry Cell and choose New Location Cell. Click the Add Extent button and use the defaults.
6. Open the Location Engine Configuration and load the area by going to MAP>Load Area. Add a Location Engine cell which will be used to set up a cell of sensors by going to Cell>New. There are no available sensors in left column; go to File>Import Sensors and locate the .xsc file that located at: C:Ubisense Software.
   1. Viewing all of the sensors, click on a sensor and drag it anywhere onto the map. It will also be under Location Cell 0001; right click on it and go to Properties. Enter the x, y, and z for that particular sensor and its MAC address. Do not enter in anything for yaw, pitch, or roll. Repeat this process for all other sensors and ensure the system places them correctly on the map.
   2. Click the Sensor Status tab; sensors are running – if not unplug and plug back in to the power source. Use the log tab to monitor the boot up process. Each sensor will download packets in groups of 100 and will eventually report sensor running. Refer back to the Sensor Status tab to make sure the sensors have booted and are running.
   3. Click on the incident power plot tab to examine the background noise level on each sensor. Let the graphs run. After a break, press the Set Thresholds button. This will set the Activity threshold on each sensor which will filter out background noise. Background noise is recommended below 1000.
   4. Right click on Location Engine Cell 00001>Properties. On the Radio tab set the RF Power to 255, which is the max radio level. On the Geometry tab set the Ceiling to 5, the Floor to 0 and the Max Standard Error to 0.05. The ceiling is the max height of the space, the floor is min, and the max standard error is for filtering poor readings.
7. Pick up a compact or hang tag and go to the Tags tab in Location Engine Config and click Tag>New. Enter tag number and set the upper Qos to 32, the same value as the lower Qos. These rates are the beaconing rates of the tag. Choose default information filter as the filter.
   1. Click on the Sensors and Cells tab. Right click on Location Engine Cell 0001 and click Monitor. This sets the tags in the cell to transmit and never fall asleep. Press and hold the bottom middle of the compact tag and the middle of the hang tag for three seconds to turn on the tag. It is on when the light in the upper right hand corner is steady and begins to blink.
   2. Put the tag in the middle of the area where all sensors have it in the line of sight. Measure the x, y, and z of that spot in relation to the same origin on the sensor used before. Right click on any of three other sensors and choose Dual Calibration. Use the master as the reference, type in the tag number, type in the location measured and choose Next. After the calibration finishes save the values for all sensors.
   3. Run this above step again to ensure the values are +/- 2. Repeat this process for all the other sensors but do not save the master sensor values. If using a hang tag, rotate it so that the face of the tag is pointing between the master and another sensor being calibrated and make sure the tag is in a vertical position. The compact tag needs to be in one spot lying flat.
   4. Ensure the sensors are pointing correctly toward the center of the area and view the green angle of arrival lines converging on the tag. Click the TDOA box near the bottom of the window and view the time delay of arrival curves converging on the tag. Note that these lines and curves will not be perfect. Repeat the calibration if necessary. Then, following instructions from step 5.1, click the monitor flag off.
   5. Open the Map and load the Area under Area>Load Area and view the tag on the map.

2. Use the RTLS Tags to Locate and Track Residents in Real-Time

1. Review medical charts to identify ambulatory (with/out an assistance device) residents or residents who can propel with their feet age 55 or over with CI/dementia. Obtain consent. Or, if the resident is unable to consent on their own, use contact information provided in the medical chart to contact their legally authorized representative (LAR) or next of kin (NOK).
2. Outfit consenting residents with wrist or hang-tag (see Figure 1). To turn on the tag, place a magnet under the bottom right of the tag and wait for the light  to blink continuously. Ensure the hang-tag it is not on backward or the signal will be attenuated. Attach the wrist tag to an area of the body with a small cross sectional area and more limited absorption of radio frequency energy and provides better tracking accuracy.
3. Develop a protocol for health care staff to remove a resident tag during bathing and showering and train health care staff on these steps. Communicate a pre-determined location to health care staff where they may leave tags they find in the unit (e.g., behind the nurse’s station) in the event research staff are not there to retrieve them.
4. Prior to putting the tag on the resident, in the Tag Association tab GUI (see Figure 2), assign each resident a random and unique "patient ID" number and input into the GUI. Using the number provided on the tag, input the Tag ID number associating it with the “patient ID.” The tag will be wirelessly tracked once assigned in the GUI. Keep position at "origin" but in "allow tag swaps", select "true," and then click save.
   NOTE: If data are compromised the privacy and security of the residents is maintained as only a random identification number and x, y coordinates are available; these coordinates do not correspond to any home/institution, city/town, etc.
5. Create a separate document saved on a secured server behind a firewall and a password-protected computer linking the residents' personal information with their patient ID and tag ID.
6. In Smart Space Config click view trace messages. Click the "get trace messages. Examine the events for tag/resident location and movement. Click the log tab to ensure there are no error messages.
7. Click the sensor status tab and view that all sensors are "running" (see Figure 4). If not, right click on the sensor and reboot. If timing or other statuses are noted after the reboot, check physical cables running to the problematic sensor.
   1. Ensure all cables are plugged in to the POE switch and that timing and power cables are working on the specific sensor. For example, if the power cable is not working, there will be no light on the sensor and a new power cable is needed. If there is power, a new timing cable is needed.
8. In C: Ubisense Software system files, set up a folder on the server to access the raw daily CSV data files.
9. Set up an automatic data backup system (external hard drive) and secure so it cannot be unplugged or moved from the server.
10. In a data management program, smooth RTLS raw data using a 5-second moving average time window (based on time provided in x and y raw data coordinates) and a threshold of 0.7 m of movement (based on location provided in x and y raw data coordinates).
    NOTE: This creates a stable series of coordinates, resembling the observed resident walking activity. To manage the jumps in data, when computing a day’s motion, only accrue distance and time (and path data) when time between points is less than 30 seconds.

3. Measuring Walking Activity and Wandering

1. Download daily csv files into a data management/analysis program.
   NOTE: Based on project aim, RTLS data can be reduced to hourly, daily, weekly, bi-weekly, and so forth. For the purposes of this project, data are averaged weekly (summed daily/7) to examine intra-individual changes in ambulation by week. Note that the number of daily samples available for each resident will vary based on their level of activity. Residents who are largely sedentary will have several hundred data points/day or less; residents who are more active will have more like several thousand data points/day.
2. Calculate average walking distance, sustained walking distance, and sustained gait speed, and calculate the extent of changes in these measures over time using the raw location data provided (weekly averages of x, y coordinates).
3. NOTE: Walking distance = average total number of meters walked per week [e.g., to calculate between each point: √(x2-x1)^2 + (y2-y1)^2], sustained walking distance = average number of continuous meters walked per week calculated only when the resident travels for at least 60 seconds with a stop not exceeding 30 seconds, gait speed = the average meters per second/week calculated during sustained walking only [to calculate between each point: √(x2-x1)^2 + (y2-y1)^2 and then t2-t1 to determine the time it takes to go this distance].
4. Visually check all sensor light indicators on the RTLS sensors and tags once a day. Check all supplementary equipment provided (e.g., POE switches and timing boxes) for lights.
   1. In the GUI, under "map" check to ensure all tagged residents are visible and being tracked each day (see Figure 5). If there is a resident missing on the map, click report to determine the last time the resident was seen by the system. Click hourly, daily, or weekly reports, which can also be filtered by Patient ID (see Figure 6).
      NOTE: This can also be accomplished by reviewing daily CSV files for Patient ID numbers.
   2. When a tag is not working, replace the tag and/or check battery. When batteries are replaced, click on the associated tag and the “tag battery replaced” button in the right hand corner of the SmartSpaceConfig.
   3. Some residents with CI may take off their tag (thrown away by mistake) when they forget about their participation in the project. If so, remind the resident of the project, ask if they wish to continue, and where applicable, replace the wrist tag. In meetings with health care staff remind the stakeholders to talk with residents and remind them of their participation in the project.
5. Daily check that no wrist/hang tags have been submerged or otherwise damaed by water (resident takes a bath instead of shower); if water damage if visible, replace the tag.

4. Measuring Cognitive Impairment, Gait and Balance

1. Register, download and assess the cognitive status of residents consenting to participate in the study at baseline and every 6 months over the course of the study using the Montreal Cognitive Assessment (MoCA).¹⁸
   1. Input resident MoCA scores in a dataset which can be merged with the RTLS data through a data management program.
2. Recode raw MocA scores such that a MoCA score of ≥24 indicates no CI, a score ranging between 10-23 indicates mild/moderate CI, and a score ranging between 0-9 indicates severe CI.¹⁹
3. Using the Tinetti Performance Oriented Mobility Assessment (POMA) and associated description,²⁰assess the gait and balance of residents consenting to participate in the study each week over the course of the study period.²⁰
   NOTE: There are two subscales in the POMA with gait quality ranging from 0-12 and balance ability ranging from 0-16. Higher scores suggest fewer gait and balance impairments. These subscales gauge a variety of associated abilities and include tasks such as getting up from a chair, sitting and standing balance, balance while turning,step length, step height, deviation from a pathand stance. Frail or institutionalized older adults, consistent with the population utilized in this project, have a mean score of 11-12 (SD=3.3-5.7) on the balance ability subscale and a mean score of 8.1-8.6 (SD=3.2-4.6) on the gait quality subscale.^1,21
   1. Input gait and balance subscale and total scores in the database with other variables along with demographic characteristics of interest (age, race/ethnicity, gender).
4. Analyze the relationship between CI, gait, balance and ambulation activity in a data management/analysis program. Click crosstabs and input variables to examine bivariate relationships. Click chi square to examine the strength of the association between these variables of interest.