Content uploaded by Veronique M Boscart
Author content
All content in this area was uploaded by Veronique M Boscart on May 22, 2015
Content may be subject to copyright.
IEEE TRANSACTIONS ON AUTOMATION SCIENCE AND ENGINEERING, VOL. 7, NO. 3, JULY 2010 695
Embedded System for Hygiene Compliance Monitoring
Alexander I. Levchenko, Member, IEEE, Graham C. Hufton,
Veronique M. Boscart, and Geoff R. Fernie
Abstract—Different approaches to implementation of hygiene com-
pliance monitoring are presented. The architecture and operation of an
embedded distributed system for hygiene compliance monitoring are de-
scribed. The performance of the system does not depend on the number of
monitored areas, number of caregivers being monitored, and no network
infrastructure is required.
Note to Practitioners—The embedded system for hand hygiene moni-
toring system was designed to be used in healthcare institutions. The main
function of the system is to monitor hand hygiene compliance according to
the rules (based on the Ministry of Health and Long-Term Care guidelines)
implemented as a part of the system software. The main components of
the system are wearable electronic monitors, disinfectant dispensers,
and monitored zones installed in the areas essential for hand hygiene
compliance.
Index Terms—Data acquisition, infrared tracking, intelligent systems,
microcontrollers.
I. INTRODUCTION
Approximately one in ten people admitted to hospitals in Canada ac-
quire a new infection during their stay. Gould et al. [1] report that these
nosocomial infections result in an estimated 8000 deaths per year. In
addition, there is considerable waste of resources and staffing because
of the increased length of stay required. It is evident that this situation
contributes significantly to the overall stress on the healthcare system
and increases wait times.
It is estimated that approximately half of these nosocomial infec-
tions are the result of inadequate hand hygiene by healthcare staff.
Unfortunately, published studies have generally found that compliance
with hand hygiene requirements by healthcare workers averages about
40% [2] and that various educational and management interventions
can only improve this to about 50%. Further, sustainability of the im-
provement produced by interventions is a problem.
Pittet et al. [3] provide the most recent authoritative review of evi-
dence that hand hygiene is the primary means to reduce nosocomial in-
fections. This review describes how pathogens are present even on un-
broken skin and on surfaces surrounding the patient. These organisms
are then transferred to healthcare workers hands where they can survive
Manuscript received June 30, 2008; revised January 20, 2009; accepted Feb-
ruary 21, 2010. Date of publication March 29, 2010; date of current version July
02, 2010. This paper was recommended for publication by Associate Editor
A. Colombo and Editor D. Meldrum upon evaluation of the reviewers’ com-
ments. This work was supported in part by the Health Technology Exchange and
the Canadian Institutes of Health Research (ref. # HTX 80047, XHT 83447).
We acknowledge the support of Toronto Rehabilitation Institute who receives
funding under the Provincial Rehabilitation Research Program from the Min-
istry of Health and Long-Term Care in Ontario. The views expressed do not
necessarily reflect those of the Ministry. Equipment and space have been funded
with grants from the Canada Foundation for Innovation, Ontario Innovation
Trust and the Ministry of Research and Innovation. Patent applications have
been filed.
A. I. Levchenko, G. C. Hufton, and V. M. Boscart are with the Research and
Development Technology Team, Toronto Rehabilitation Institute, Toronto, ON
M5G 2A2 Canada (e-mail: Levchenko.Alexander@torontorehab.on.ca; Hufton.
Graham@torontorehab.on.ca; Boscart.Veronique@torontorehab.on.ca).
G. R. Fernie is with the TorontoRehabilitation Institute and the Department of
Surgery, University of Toronto, Toronto, ON M5G 2A2 Canada (e-mail: Fernie.
Geoff@torontorehab.on.ca).
Digital Object Identifier 10.1109/TASE.2010.2044882
for hours. The review outlines the evidence that an alcohol-based hand
rub is significantly more effective than washing with soap and water
and concludes that experimental studies have all shown that improved
hand hygiene reduces transmission of pathogens.
II. INVESTIGATED HYGIENE COMPLIANCE MONITORING SOLUTIONS
There have been some attempts to improve hand hygiene compliance
through the incorporation of electronic prompting systems in wearable
and stationary alcohol gel dispensers. Kinsella et al. [4] conducted a
study using gel and soap dispensers, equipped with a microcontroller-
based time stamping device, in an intensive care unit to collect data
about soap and gel use for hand-hygiene quality improvement and ed-
ucational initiatives. Unfortunately, this approach has a limited value
as the records of hand cleansing actions are provided unrelated to the
caregiver’s location and activity. Venkatesh et al. [5] and Swoboda et
al. [6] evaluated an electronic prompting system that used light beams
and motion detectors at the threshold of each room in a surgical unit.
The system had visual and voice prompting function instructing the in-
dividuals to wash their hands if they had not done so before entering
or after exiting the room The system could work only with stationary
dispensers, did not record individual performance data and could not
provide a prompting signal in a multipatient environment when a care-
giver moves from one patient to another.
We postulate that in order to be able to estimate hand hygiene
compliance and provide efficient prompting functionality, the system
should have information both on caregiver’s location and hand
cleansing activities. One technical approach that we considered was
to use the hygiene compliance algorithms [7] and control software in
conjunction with commercially available location tracking systems.
In this case, caregivers would have identification tags working in
combination with arrays of readers that allow wearers coordinates to
be determined. Depending on the tag’s positional coordinates the host
computer would estimate the probability of contact between patient
and caregiver and would decide when to issue a prompting signal.
For this application, the major requirements for the location tracking
system are the ability to operate in a hard real-time mode providing
guaranteed maximum reaction time and the ability to determine a tag’s
positional coordinates with high accuracy. The user tag should be ac-
tive and it should be small enough to be integrated into a wearable dis-
penser or to be able to communicate with it. It should also be able to
log and process hand cleansing events or to communicate with the host
computer every time a hand cleansing occurs.
Our clinical tests demonstrated that to perform efficiently the system
must guarantee a reaction time (time interval between entering the pa-
tient environment and the moment when the prompting signal is issued)
not exceeding one second. The system must also identify that the care-
giver is within the patient area with a 20–30 cm accuracy. To date,
RF-based location tracking solutions [8]–[12] do not meet these re-
quirements. Some vendors of ultra-wideband (UWB) tracking systems
claim that they can achieve about 30 cm accuracy [13] and control mul-
tiple tags simultaneously. Although the UWB technology is advancing,
the main limitation of this approach is considerable capital investments
in infrastructure and hardware. In a complex indoor environment, extra
expensive sensors might be required to improve coverage and provide
accurate tracking [14].
During our development program, we next built a prototype of the
key elements of another approach which used ultrasound ranging sen-
sors. The arrays of programmable ultrasound receivers controlled by
a wireless ZigBee network coordinator were installed to mark the pa-
tient areas. Portable units based on the PIC18F2620 microcontroller
and a CC2420 RF transceiver were configured as ZigBee™ Reduced
1545-5955/$26.00 © 2010 IEEE
Authorized licensed use limited to: The University of Toronto. Downloaded on August 16,2010 at 20:24:39 UTC from IEEE Xplore. Restrictions apply.
696 IEEE TRANSACTIONS ON AUTOMATION SCIENCE AND ENGINEERING, VOL. 7, NO. 3, JULY 2010
Fig. 1. Embedded system for hygiene compliance monitoring. In this distributed architecture, both zone controllers and personal wearable monitors work com-
pletely independently without any network or central coordinating unit. The system is easily scalable, its performance does not depend on the number of monitored
zones or the number of wearable monitors.
Function Devices (RFD). The ZigBee coordinator constantly polled the
wearable units having ultrasound transmitters. It sent a command to the
wearable unit being located to produce a short ultrasound pulse. Imme-
diately after receiving confirmation from a ZigBee™ RFD device the
coordinator sent a “Listen” strobe to all zone sensors that included the
ID of the wearable unit and the current status of the hand disinfection
status “flags” of this caregiver. If none of the zone sensors detected
anything within a certain time interval (set depending on the size of
controlled zone), the procedure was repeated for the next unit. If any
of the zone sensors detected the presence of a wearable unit it could
provide a prompt to the caregiver if the hand disinfection status flag
was not set to clean. The identity of the wearable unit (a surrogate for
the caregiver’s identity), the times of entering and leaving identified
patient zones marked by sensors and the times of hand disinfection ac-
tions on the wearable unit were all stored by the coordinator/PC for later
downloading and data analysis. Since the system worked in a polling
mode the number of devices that could be controlled by one central unit
without affecting system performance was limited. The communication
intervals between the RFD and the coordinator were dictated by the re-
quired reaction time and resulted in higher power requirements for the
portable devices. In addition, extra routers were needed to cover areas
larger than the range of ZigBee coordinator increasing complexity of
the system.
III. DEVELOPED ARCHITECTURE:EMBEDDED MONITORING SYSTEM
We subsequently developed an alternative low-cost distributed ar-
chitecture (shown in Fig. 1) which is free from the above mentioned
limitations. In this concept all monitoring and data processing func-
tions are embedded in the personal wearable monitors, which allows
them to operate completely independently without any network infra-
structure and any central control and processing unit.
The patient areas are instrumented with arrays of low intensity in-
frared emitters, constantly transmitting modulated unique patient area
identifiers. The main functions of the wearable monitors are to demod-
ulate and decode patient zone identity signals, record the real time of
entering/leaving and codes of visited zones, record the real time of dis-
penser activations, and to provide prompting to perform hand hygiene
when required.
One of the major advantages of this approach is that reaction time of
the system is defined entirely by embedded software and does not de-
pend on the number of units being monitored, their locations, number
of patient areas installed and any network characteristics. The perfor-
mance is not affected when two or more caregivers work simultane-
ously in the same patient area. The cost of the distributed system as
well as the hardware and software requirements is considerably lower
compared to centralized architectures. The wearable monitors are de-
signed so they are not obscured during use to maintain reliable infrared
communication. Each of these units communicates with its wearable
alcohol gel dispenser by a short range RF link so that the dispenser can
be worn or carried in a variety of ways without restriction.
Infrared emitters defining the patient areas are housed in specially
designed conical directional elements (35 15 mm). Symmetrical con-
ical directional elements are used to provide even coverage inside the
zone, while narrow elliptical directional elements are angled toward the
center of the zone to provide a crisp zone perimeter. The directional el-
ements are injection molded white polypropylene with a polished finish
and are housed inside two-part PVC extrusions in groups of three, with
a total of thirty six emitters defining patient areas (3.8 each) in the
test room. The PVC extrusions housing the directional elements are
prewired for ease of installation and are clipped, screwed or adhered to
the ceiling to create a zone of required shape and size, without inter-
fering with overhead patient lifts and other equipment used inside pa-
tient area. The zone emitters are controlled by a PIC24FJ16GA004 mi-
crocontroller and emit 36.7KHz modulated infrared code that includes
the type and identification number of the zone, with each zone being
powered by a 6V DC adapter with a current consumption of approxi-
mately 200 mA.
The zone microcontroller can automatically adjust the intensity of
the infrared signal depending on ambient light to maintain the required
accuracy of the zone boundaries. This function is required only to com-
pensate for significant changes of ambient light conditions. In our tests
with the constant signal radiant intensity of 7 mW/sr per emitter, the
Authorized licensed use limited to: The University of Toronto. Downloaded on August 16,2010 at 20:24:39 UTC from IEEE Xplore. Restrictions apply.
IEEE TRANSACTIONS ON AUTOMATION SCIENCE AND ENGINEERING, VOL. 7, NO. 3, JULY 2010 697
Fig. 2. Block diagram of the wearable dispenser operation in the distributed system. A number of software timers are used to adjust disinfection expiry intervals
for different clinical environments. – time elapsed since the previous detection of the zone; – elapsed time since last hand hygiene action; – estimated
duration of protection provided by a single hand hygiene event when outside zones; – time permitted outside a zone before returning without the need for
hand hygiene.
variations of zone boundaries did not exceed 15 cm, with the illumi-
nance of ambient light changing from 30 to 500 lux. Assuming that all
infrared emission is perfectly concentrated inside the zone, the com-
bined irradiance from all emitters inside the patient zone at typical op-
erating ranges is around 0.118 . This is more than 130 times
under ACGIH [15] safety limits for continuous exposure.
The main components of the wearable electronic monitor are
an STM32F102CBT6 ARM Cortex™–M3 based microcontroller,
MICRF211 433.92 MHz receiver, and a PNA4701 infrared detector.
EEPROM emulation algorithms are implemented in software to use
embedded flash memory for the data logging function. A special
protocol with a pulse width modulation encoding pattern has been
developed to provide reliable detection of the zone signal and elim-
inate possibilities of false prompting, with an STM32F102CBT6
decoding zone signals in sleep mode using external interrupts. Each
monitor (dimensions: 50 40 11 mm) is powered by a single AAA
NiMH rechargeable battery, with charging required once a week.
Wearable monitors and portable dispensers work in pairs with each
dispenser which contains a replaceable gel cartridge. Typically, one to
two cartridges are required per shift, depending on the care activities
carried out. Each dispenser is equipped with a PIC16F506 microcon-
troller and a MAX1472 transmitter, with lower than 2 dBm output
power setting, to inform its corresponding monitor about dispenser
activations. The hand disinfection status “flags” of the caregiver are
stored in the wearable monitor so it knows whether the wearer has
recently performed hand hygiene (this time interval can be set in
the software), or whether the wearer’s hands have been disinfected
since the previous patient zone was visited. The data recorded by the
monitors are later downloaded to a PC via a USB interface for further
analysis and reporting.
A simplified logic diagram illustrating the operation of the monitor
is shown in Fig. 2. The device is in sleep mode most of the time and
wakes up periodically to check the presence of the zone emitters. The
duration of the power saving intervals is controlledby a microcontroller
watchdog timer and defined by the maximum acceptable reaction time
when a caregiver enters the zone. The experiments were conducted in
a four-bed room on a chronic care unit of a larger complex continuing
care hospital. Nine registered nurses, employed on the unit, participated
in the tests. In multiple-bed rooms, a nurse often responds to different
patients in a short time by moving between patient beds. The minimum
distance between two beds in the four-bed room was 120 cm. The reac-
tion time of the monitor was determined experimentally to guarantee
that when a caregiver enters the patient environment, the prompting
signal is issued before any contact with a patient or other objects in-
side the protected area. In our test, this parameter was configured not
to exceed 800 ms. If the zone is detected the device checks that the
last disinfection occurred not earlier than the disinfection expiry time
outside of the zone which is programmable and may vary for dif-
ferent applications. If disinfection was not performed or the time has
already expired the device prompts the caregiver to activate the dis-
penser. When the monitor leaves the zone its disinfection status flag
remains set to clean for a certain period of time ( ), so the care-
Authorized licensed use limited to: The University of Toronto. Downloaded on August 16,2010 at 20:24:39 UTC from IEEE Xplore. Restrictions apply.
698 IEEE TRANSACTIONS ON AUTOMATION SCIENCE AND ENGINEERING, VOL. 7, NO. 3, JULY 2010
giver is allowed to leave the zone temporarily and come back without
prompting. In our tests, it was initially set at 30 seconds. When care-
givers performed procedures involving complex motion patterns, for
example, transferring a patient to a wheelchair, the monitors periodi-
cally lost communication with the zones for longer than periods
of time, as some procedures require a caregiver to frequently step in and
out of the zone. Nurses performing complex procedures sometimes re-
ported false prompts inside the patient areas, until was adjusted
to 60 seconds. In the system software, variable immediately ex-
pires if, after leaving the patient zone, a caregiver leaves the patient
room, goes to a shared bathroom, or goes to another patient area, as
in a full system configuration all these destinations are controlled by
specific zone types. This allows the monitors to filter the events when
communication with a patient zone is lost due to the complex maneu-
vering inside the patient area required by certain procedures. Another
timer implemented in software (not shown in Fig. 2) is used to con-
trol a hand hygiene status flag when the dispenser is activated inside
the patient zone, so that if a caregiver performs hand hygiene inside
the patient zone and then leaves the zone within short programmable
time interval going directly to another patient zone, the system does not
issue a prompting signal when the next zone is entered. If the device de-
tects a zone which is different from the previous one and the dispenser
was not activated between the zones, or within a short programmable
interval before leaving the previous zone, then the disinfection status
flag changes immediately issuing the prompting signal. Time intervals
and other parameters of the monitors are configured via USB without
changing their software and are adjusted depending on the clinical en-
vironment. Three types of monitored zones are defined for the patient
rooms: entrance zones, individual patient zones, and bathroom zones.
The zones can also be installed in utility rooms, laundries and other
areas where hand hygiene monitoring is required.
The logic described above applies to the patient zones, but is dif-
ferent for entrance and bathroom zones. When an entrance zone is de-
tected and hand hygiene has not been performed within an acceptable
time ( ) the device produces a longer audio signal prompting the
caregiver to perform cleansing after entering the patient room. If the
dispenser is activated just before entering the room and, after entering,
the caregiver goes directly to a patient zone the prompting is not is-
sued. When a caregiver enters a bathroom zone, the wearable monitor
does not produce a prompting signal but it is issued immediately after
exiting the bathroom zone.
Most clinical settings are equipped with a number of wall-mounted
alcohol gel dispensers installed in hallways close to the room entrances.
In our system, each of these wall mounted dispensers has its own small
infrared zone that is used to inform the monitor that its wearer has used
a stationary dispenser. These small zones are inactive most of the time
and transmit only for a few seconds after activation of the stationary
dispenser.
There are obvious privacy issues for the staff since the system can
be used to determine how long caregivers spend with specific patients,
not just whether they wash their hands. A degree of anonymity can be
provided by not associating the names of individuals with the identity
numbers of the monitors. Nevertheless, together with clinical mem-
bers of our team, we conducted a separate study [16] to evaluate ac-
ceptability of the new technology. Staff agreed that the system would
increase hand hygiene and found it was a convenient approach to re-
mind them about this task, which is often overlooked in the hectic pace
of competing clinical activities. The study demonstrated positive feed-
back from clinicians who felt comfortable with receiving individual
performance information and found the system acceptable with some
caveats.
IV. CONCLUSION
The distributed embedded system has a number of advantages com-
pared to other architectures. Embedding all monitoring and data pro-
cessing functions into the wearable monitors provides guaranteed reac-
tion time and makes the system seamlessly scalable without affecting
performance, with the number of wearable monitors being practically
unlimited. The proposed approach simplifies installation and allows the
wearable devices and monitored zones to be added or removed at run-
time. The zone controllers and wearable monitors work completely in-
dependently without any central control unit and no network infrastruc-
ture is required.
ACKNOWLEDGMENT
The authors wish to acknowledge the contribution of S. Gorski and
other members of hand hygiene project team.
REFERENCES
[1] G. D. J. Gould, J. H. Chudleigh, D. Moralejo, and N. Drey, “Interven-
tions to improve hand hygiene compliance in patient care,” Cochrane
Database of Systematic Reviews, vol. 2, no. Art. No. CD005186.
DOI:10.1002/14651858. CD005186. pub2, 2007.
[2] M. G. Lankford, T. R. Zembower, W. E. Trick, D. M. Hacel, G. A.
Noskin, and L. R. Peterson, “Influence of role models and hospital de-
sign on hand hygiene of health care workers,” Emerging Infectious Dis-
eases, vol. 9, pp. 217–223, 2003.
[3] D. Pittet, B. Allegranzi, H. Sax, S. Dharan, C. L. Pessoa-Silva, L. Don-
aldson, and J. M. Boyce, “On behalf of the WHO global patient safety
challenge, world alliance for patient safety. Evidence-based model for
hand transmission during patient care and the role of improved prac-
tices,” Lancet Infectious Diseases, vol. 6, pp. 641–652, 2006.
[4] G. Kinsella, A. N. Thomas, and R. J. Taylor, “Electronic surveillance
of wall-mounted soap and alcohol gel dispensers in an intensive care
unit,” J. Hospital Infection, vol. 66, pp. 34–39, 2007.
[5] A. K. Venkatesh, M. G. Lankford, D. Rooney, T. Blatchford, C. M.
Watts, and G. A. Noskin, “Use of electronic alerts to enhance hand hy-
giene compliance and decrease transmission of vanvomycin-resistant
enterococcus in a hematology unit,” Amer. J. Infection Control, vol.
36, no. 3, pp. 199–205, 2008.
[6] S. M. Swoboda, K. Earsing, K. Strauss, S. Lane, and P. A. Lipsett,
“Electronic monitoring and voice prompts improve hand hygiene and
decrease nosocomial infections in an intermediate care unit,” Critical
Care Medicine, vol. 32, no. 2, pp. 358–63, 2004.
[7] Hand Hygiene. Ontario Ministry of Health and Long Term Care. [On-
line]. Available: http://www.health.gov.on.ca/patient_safety/pro/hh/
hh_pro.html
[8] P. Bahl and V. Padmanabhan, “RADAR: An in-building RFbased
user location and tracking system,” in Proc. IEEE Infocom, 2000, pp.
775–784.
[9] K. Lorincz and M. Welsh, “MoteTrack: A robust, decentralized ap-
proach to RF-based location tracking,” in Proc. Int. Workshop on Lo-
cation and Context-Awareness, 2005, pp. 63–82, Pervasive.
[10] D. Pandya, R. Jain, and E. Lupu, “Indoor location using multiple wire-
less technologies,” in Proc. IEEE PIMRC, Beijing, China, 2003, vol. 3,
pp. 2208–2212.
[11] Y. Gwon, R. Jain, and T. Kawahara, “Robust indoor location tracking
estimation of stationary and mobile users,” in Proc. IEEE Infocom,
Hong Kong, 2004, pp. 1012–1022.
[12] S. Ray, D. Starobinski, A. Trachtenberg, and R. Ungrangsi, “Robust
location detection with sensor networks,” IEEE J. Sel. Areas Commun.,
vol. 22, pp. 1016–1025, 2004.
[13] L. Coyle, J. Ye, E. Loureiro, S. Knox, S. Dobson, and P. Nixon,“A pro-
posed approach to evaluate the accuracy of tag-based location systems,”
in Proc. Ubiquitous Systems Evaluation. Workshop at UbiComp2007,
2007, pp. 292–296.
[14] W. Kearns, D. Aglase, D. Moore, and S. Ahmed, “Ultra wideband
radio: A novel approach for measuring wandering in persons with de-
mentia,” Gerontology, vol. 7, pp. 48–57, 2008.
[15] American Conference of Governmental and Industrial Hygienists
(ACGIH) “Documentation of the threshold limit values for physical
agents,” 7th ed. 2001.
[16] V. M. Boscart, K. S. McGilton, A. I. Levchenko, G. C. Hufton, P. J.
Holliday, and G. R. Fernie, “Acceptability of a wearable hand wash
device with monitoring capabilities,” J. Hospital Infection, vol. 70, pp.
216–222, 2008.
Authorized licensed use limited to: The University of Toronto. Downloaded on August 16,2010 at 20:24:39 UTC from IEEE Xplore. Restrictions apply.