This RUBIN_readme20230519.txt file was generated on 20230519 by Michael Rubin. Edited on 102030530 by Tina Kirkham. ------------------- GENERAL INFORMATION ------------------- 1. Title of Dataset: Raw Wireless Sensor Data from the University of Utah from the CDC-funded Granular Modeling Project 2. Author Information Principal Investigator Contact Information Name: Michael Rubin, MD, PhD Institution: University of Utah/Salt Lake City Veterans Affairs Address: Division of Epidemiology, 295 Wakara Way, Salt Lake City, UT 81408 Email: michael.rubin@hsc.utah.edu Associate or Co-investigator Contact Information Name: Molly Leecaster, PhD Institution: University of Utah/Salt Lake City Veterans Affairs Address: Division of Epidemiology, 295 Wakara Way, Salt Lake City, UT 81408 Email: molly.leecaster@utah.edu Alternate Contact Information Name: Candace Harolden, MSPH Institution: University of Utah/Salt Lake City Veterans Affairs Address: Division of Epidemiology, 295 Wakara Way, Salt Lake City, UT 81408 Email: candace.haroldsen@hsc.utah.edu 3. Date of data collection (single date, range, approximate date) 20180301--20180428 4. Geographic location of data collection (where was data collected?): University of Utah Hospital Acute Care Ward, Salt Lake City, Utah, USA 5. Information about funding sources that supported the collection of the data: Centers for Disease Control and Prevention (CDC) RFTOP #2015-006, "Granular Modeling–Simulating the Transmission of Healthcare Associated Infections in Hospitals" -------------------------- SHARING/ACCESS INFORMATION -------------------------- 1. Licenses/restrictions placed on the data: Open Access 2. Links to publications that cite or use the data: N/A 3. Links to other publicly accessible locations of the data: None 4. Links/relationships to ancillary data sets: None 5. Was data derived from another source? No If yes, list source(s): 6. Recommended citation for the data: Rubin, M., Leecaster, M., and Haroldsen, C. 2023. Raw Wireless Sensor Data from the University of Utah from the CDC-funded Granular Modeling Project. The Hive: University of Utah Research Data Repository. https://doi.org/10.7278/S50d-acm8-epwg --------------------- DATA & FILE OVERVIEW --------------------- 1. File List A. Filename: SensorDataRaw05192023.csv Short description: This dataset contains raw sensor level pings of RFD reads between healthcare worker worn sensors and environmental sensors placed in facility. 2. Relationship between files: N/A, only one file. 3. Additional related data collected that was not included in the current data package: N/A 4. Are there multiple versions of the dataset? NO If yes, list versions: N/A -------------------------- METHODOLOGICAL INFORMATION -------------------------- 1. Description of methods used for collection/generation of data: The objective of using the wireless sensors was to improve understanding of the heterogeneity of healthcare worker (HCW) contact with patients and the physical environment in patients’ rooms. The framework and design were based on contact networks with a) nodes defined by HCW’s, rooms, and items in the room and b) edges defined by HCW’s in the room, near the bed, and touching items. Nodes had characteristics of HCW role and room number. Edges had characteristics of day, start time, and duration. Thus, patterns and heterogeneity could be understood within contexts of time, space, roles, and patient characteristics. At the University of Utah Hospital Cardiovascular ICU (CVICU), a 20-bed unit, we collected data for 54 days. HCW contact with patients was measured using wireless sensors to capture time spent in patient rooms as well as time spent near the patient bed. HCW contact with the physical environment was measured using wireless sensors on the following items in patient rooms: door, sink, toilet, over-bed table, keyboard, vital signs monitor touchscreen, and cart. HCW’s clipped a sensor to their clothing or lanyard. A sensor sent radio signals that other sensors recorded. They record the signal strength received and the loss (difference) between sending and receiving signal strengths indicated the distance between them. The sensors also had an accelerometer; it sent a signal when vibration was detected. Non-accelerometer sensors sent signals at a pre-determined interval and recorded signals continuously. The signal included the sensor identifier and time stamp. Sensors were programmed with specifications tested during pilot work to balance collection of accurate data with battery and memory capacity. The specifications were sending strength, receiving filter, accelerometer sensitivity, and sending interval. There were four categories of sensors we programmed: 1) HCW sensors send & receive, 2) wall sensors send on an interval, 3) item sensors send on vibration, and 4) select item sensors receive only. Low sending strength was recorded within 2-3 feet, medium was recorded within 6 feet, and high was recorded within 10 feet. The filter was specified to limit memory use; the lower the value the lower the signal strength they recorded. Accelerometer sensitivity was set at a low value for items that vibrated a lot so that slight vibrations did not trigger a signal. The sending interval for linked motes was set very high to reduce noise from that sensor, which recorded signals but the signals it sent were not useful. Maintenance was performed to download data and charge batteries. This process resulted in clumps of data, many records among sensors in a short period of time. Maintenance also resulted in sporadic, but few, records during that time. For interpretation, these time periods were indicated in the data. 2. Methods for processing the data: Sensors were collected after deployment and downloaded to a local computer, creating hundreds of raw data files. Files were then combined into one dataset where they were cleaned to remove obvious unreliable data. Several metadata components were incorporated directly into this analysis dataset including: 1) timesync files used to calculate time of sensor event, 2) id dataset used to characterize each sensor and associated role (e.g. healthcare worker, near-bed sensor, sink, gel, etc.), 3) and event log files used to create indicator variables for whether a charging/swapping event took place during that particular sensor event (charging/swapping events could create noise in data that hinders its use/interpretation). 3. Instrument- or software-specific information needed to interpret the data: SAS 9.4 was used for importing and processing data. R 4.3.0 was used for validation and analysis. No special software is required for use of current dataset (csv format). 4. Standards and calibration information, if appropriate: Several pilot studies were done in order to calibrate the devices. 5. Environmental/experimental conditions: This study was performed in vivo in an acute care hospital 6. Describe any quality-assurance procedures performed on the data: Initial data processing steps: Import raw sensor data files Join time sync file to calculate sensor date time Join IDs file to identify sensor role and location Assign event categories (EventCat) Contact between HCWs Item use recorded by HCW: sink, soap, gel, door, table, toilet, keyboard, touchscreen, cart, ventilator Patient room location recorded by HCW (room and near-bed) Item use recorded by environmental sensor: sink, door, table, toilet, ventilator Combine all sensor files into one dataset Define sublocation for gel/soap/sink in HA using ‘placement’ (provides location along hallway wrt patient room numbers) Define limitingEvent: Charge for times when sensors could have been taken away, charged, and replaced Restrict for times when the patient or family requested no one enter the room Define location as Patient Room #, HA (hallway), NS (nursing station), or unknown (Contacts) Data cleaning steps: Delete HCW IDs that never interact with environmental sensors Delete HCW IDs that either never record signals or are never recorded by others Delete Reboot > 0 [occurred in Sensor Lab Office] Delete IsSync > 0 [occurred in Sensor Lab Office] Delete Ego_global=ego_local [occurred in Sensor Lab Office] Delete Battery < 3.9 [too low for reliable sending/recording] Keep: Ego roles: nurse (na), nurse assistant (cn), physician (ph), physical therapist (pt), respiratory therapist (rt), study personnel (sa) Remove if rssi<15 Keep only event category pings: Keep: Ego and alter are linked (same and in: ventilator, door, sink, toilet, table, sinkpipe) so ego is Mote and alter is Accelerometer, in the same location and, rssi>29. Delete sensor events that occurred while sensors were in transport, indicated by: rssi>=30 and signal internal <= every 20 seconds for 5+minutes. 7. People involved with sample collection, processing, analysis and/or submission: Data Manager: Candace Haroldsen, MSPH Study Coordinators: Kristina Stratford, Tavis Huber ----------------------------------------- DATA-SPECIFIC INFORMATION FOR: SensorData_Phase1.csv ----------------------------------------- 1. Number of variables: 20 2. Number of cases/rows: 15,941,136 3. Variable List 1. row number 2. Name: ego_role Description: indicator of mote vs accelerometer VALUES: M=mote A=accelerometer 3. Name: ego_location Description: type of healthcare worker or if environmental sensor, location of that sensor. VALUES: nu=nurse, rt=repiratory therapist, pt=physical therapist, ph=physician, sa=sampler, cn=certified nurse assistant. For locations: NS=nurses station, HA=hallway, other locations are numeric and correspond to a patient room. 4. Name: ego_subloc Description: 3 letter identifier for hcw. Used in conjuction with ego, will identify a unique HCW. Non-HCW sensors identified by name of item, e.g., wall, table, sink. 5. Name: alter_role Description: indicator of mote vs accelerometer VALUES: M=mote A=accelerometer 6. Name: alter_location Description: type of healthcare worker or if environmental sensor, location of that sensor. VALUES: nu=nurse, rt=repiratory therapist, pt=physical therapist, ph=physician, sa=sampler, cn=certified nurse assistant. For locations: NS=nurses station, HA=hallway, other locations are numeric and correspond to a patient room. 7. Name: alter_subloc Description: 3 letter identifier for hcw. Used in conjuction with ego, will identify a unique HCW. Non-HCW sensors identified by name of item, e.g., wall, table, sink. 8. Name: alter_placement Description: letter indicating placement within the subloc. b for second gel dispenser within patient room. For wall, C = cabinet, W = window, B = nearbed 9. Name: ego Description: id number of sensor Value labels if appropriate 10. Name: alter Description: id number of sensor Value labels if appropriate 11. Name: rssi Description: strength of signal recorded by alter sensor Value labels if appropriate 12. Name: downloaddate Description: date sensor data was downloaded Value labels if appropriate 13. Name: hours Description: hours from start of data collection for study 14. Name: SensorDate Description: date of sensor event 15. Name: SensorTime Description: time of sensor event 16. Name: EventCat Description: type of event VALUES: Cart=sensor on cart triggered, Contact=hcw to hcw event, Door=sensor on door triggered, Gel=sensor on gel triggered, Keyboard=sensor on keyboard triggered, Near-bed=sensor signal from wall of patient room behind bed, Room=sensor signal from patient room wall, Sink=sensor on sink triggered, Soap=sensor on soap triggered, Table=sensor on table triggered, Toilet=sensor on toilet triggered, Touchscreen=sensor on touchscreen triggered. 17. Name: location Description: location of sensing event VALUES: NS=nurses station, HA=hall, unknown=hcw contact event - all locations unknown, the rest are integers corresponding to a room number. 18. Name: LM Description: this is 1 to indicate accelerometer and mote are linked on one subloc to measure use without being recorded by a HCW 19. Name: SwapRestrict Description: sensors were swapped out and/or recharged during that sensing event or patient/family requested no one come into the room. Could indicate unreliable record. VALUES: 0=no, 1=yes 20. Name: shiftnum Description: sequential shift numbering, from 1, day shift of first day of data collection. 4. Missing data codes: NA 5. Specialized formats of other abbreviations used: NA