Reliable Real-time Clinical Monitoring Using Sensor Network Technology

Octav Chipara, Christopher Brooks, Sangeeta Bhattacharya, Chenyang Lu,

Roger D. Chamberlain, Gruia-Catalin Roman, Thomas C. Bailey
Washington University in St. Louis, St. Louis, MO
Abstract 24 hours following an operation4. This may lead to
either missing, or prolonged delays in the detection of
We propose wireless sensor networks composed of
clinical deterioration. Therefore, a real-time and
nodes using low-power 802.15.4 radios as an ena-
reliable clinical monitoring system for patients in
bling technology for patient monitoring in general
general wards is critical for effective early detection
hospital wards. A key challenge for such applications
of clinical deterioration. The integration of real-time
is to reliably deliver sensor data from mobile pa-
clinical data automatically collected from patients
tients. We propose a monitoring system with two
into electronic record systems will enable advanced
types of nodes: patient nodes equipped with wireless
clinical applications. For example, an electronic
pulse oximeters and relays nodes used to route data
scoring system can automatically generate alerts to a
to a base station. A reliability analysis of data collec-
rapid response team in the event of patient
tion from mobile users shows that mobility leads to
deterioration, thereby reducing the delay of response
packet losses exceeding 30%. The majority of packet
from hours to minutes.
losses occur between the mobile subjects and the
first-hop relays. Based on this insight we developed Emerging wireless technologies hold the promise to
the Dynamic Relay Association Protocol (DRAP), an meet the challenge of real-time clinical monitoring.
effective mechanism for discovering the right relays Commercial medical device vendors (e.g., Phillips,
for patient nodes. DRAP enables highly reliable data GE, Cisco) are moving toward solutions that use
collection from mobile subjects. Empirical evaluation IEEE 802.11 in their wireless patient monitoring
showed that DRAP delivered at least 96% of data systems. These systems require the deployment of
from multiple users. Our results demonstrate the fea- wireless routers connected to a wired infrastructure.
sibility of wireless sensor networks for real-time As a result, the equipment and deployment costs of
clinical monitoring. such systems are high, limiting their adoption in
specialized departments such as cardiology. In
Introduction
contrast, our goal is to develop an inexpensive real-
Early detection of clinical deterioration in patients is time wireless clinical monitoring system for patients
a key factor in reducing mortality rates and length of in general hospital wards.
stay in hospitals. The prevalence of clinical
The system we developed employs emerging wireless
deterioration resulting in cardiopulmonary or
sensor network (WSN) technology based on IEEE
respiratory arrests for patients admitted to hospitals is
802.15.4 standard. Such a solution is attractive for
between 4-17%1. An automated scoring system that
patient monitoring for the following reasons. First, in
uses vital signs and other commonly recorded clinical
contrast to 802.11 radios, 802.15.4 radios may be
data has been shown to identify patients at risk of
integrated with inexpensive microcontrollers and
clinical deterioration2. Similarly, early detection and
hardware platforms. In fact, the retail price of
treatment of patients with sepsis, an infection induced
802.15.4 radios is approaching $1, and single-
syndrome resulting in systemic inflammatory
package radio, processor, and memory are about $313.
response complicated by dysfunction of at least one
Second, we reduce the deployment cost of the system
organ, resulted in statistically significant lower
through mesh networking technologies, which
mortality rates3.
require minimal infrastructure. In contrast to current
While the benefits of early detection are real, the systems in which a patient must be within the radio
sensitivity and accuracy of such systems hinges upon range of an access point connected to a wired local
having timely clinical data1. This is not a major area network, our system may route data to a base
problem for patients in Intensive Care Units (ICUs) station over multiple hops. Thus, we remove the
since their vital signs are monitored by wired expensive task of wiring access points. Third,
electronic monitoring systems. However, due to the 802.15.4 radios consume significantly less energy
significant cost of monitoring systems, the vital signs (e.g., in idle state a typical 802.15.4 draws 17.4 mA,
of patients in general wards are most often collected compared to 200 mA for 802.11). The small form
manually, and are obtained at relatively long factor and weight coupled with proper power
intervals. For example, even in postoperative care, management techniques that allow operating the
nurses measure vital signs only 10 times in the first device for the duration of hospitalization without
 Figure 1. Circles denote relay nodes that are part of testbed. Diamonds indicate relay nodes plugged into wall
outlets. A square denotes the base station. Positions A and B are used in the mobility study.

replacing batteries thereby reducing the the delivery of data from the mobile node to a relay.
inconvenience of wearing a monitoring device. This enables us to avoid reconstructing end-to-end
Therefore, even in hospitals where 802.11 routes, facilitating a simple and efficient mechanism
infrastructure has already been deployed, 802.15.4 for reliable data collection from resource-constrained
technology still delivers important advantages in sensors attached to mobile users. DRAP features a
terms of power consumption and patient mechanism that enables a mobile node to
convenience. dynamically associate with relay nodes based on
channel conditions or in response to mobility.
For a wireless clinical monitoring system to be
widely used in hospitals, it is essential to reliably As part of the CodeBlue project5 a publish/subscribe
collect data from patients. We identified two key system that uses the TinyAMDR protocol was
challenges in developing such a system: (1) In proposed. TinyAMDR is an adaptation of the AMDR
contrast to 802.11 radios, low-power 802.15.4 radios (Adaptive Demand-Driven Multicast Routing)
are unreliable due to dynamic channel conditions as protocol11 to WSNs. TinyAMDR handles mobility by
well as susceptible to interference9. (2) Unlike restarting the route discovery process every 15
patients in ICUs, patients in the general hospital seconds. Thus, TinyAMDR suffers from the same
wards may be ambulatory. Patient mobility may lead issues as the routing protocols for ad hoc networks.
to significant packet losses. This work focuses on the
A number of projects propose to use body sensor
challenge of reliable clinical monitoring of mobile
networks for different medical applications such as:
subjects. Specifically, the contributions of this paper
assisted living12, disaster response8, and wireless
are: (1) Through systematic empirical studies on a
clinical monitoring5. These systems emphasize
prototype system we show that the Collection Tree
architectural concerns, hardware and medical sensors
Protocol (CTP)9, the standard data collection protocol
choice, and specific application requirements.
on WSNs, performed poorly under mobility. We
Unfortunately, those projects did not provide
identify the root cause of these problems to be the
systematic study of their networking performance
first hop; that is, the changing connectivity between
and reliability in the presence of mobile users, which
the mobile node and the relay nodes. (2) We propose
is the focus of this paper.
a robust data collection mechanism called Dynamic
Relay Association Protocol (DRAP) which allows a System Description
patient node to dynamically associate with a reliable
Our wireless clinical monitoring system consists of
relay node as the user moves in the environment. (3)
three types of nodes: a base station node, relay nodes,
Empirical results show that DRAP achieves at least
and patient nodes. The base station node is the end-
96% of the sensor readings to the base station and has
point of the data collection system. The base station
a radio duty-cycle (the fraction of time the radio is
turned on) of 2.2%, thus conserving power. node is the only node that requires access to the
wired network. The base station stores the received
Related Work data in a PostgreSQL relational database.
The mobile ad hoc network community has proposed The relay nodes are used to route packets from
numerous routing protocols with support for mobile patient nodes to the base station when there is no
entities10. It is common for routing protocols to direct wireless link between a patient node and the
introduce significant overhead by reconstructing a base station. Sufficient stationary relay nodes are
portion or an entire route in response to mobility- deployed to cover the area in which users may move.
triggered routing failures. Due to reliability concerns Relay nodes are connected to power outlets to reduce
and the characteristics of clinical environments, we maintenance cost and enhance reliability. In contrast
deployed sufficient relay nodes to cover the area in to existing commercial telemetry systems, we do not
which a user may move. As a result, a mobile node is require that the relay nodes have access to the wired
always one hop away from a relay node. In this case, network, thus removing the high wiring cost. Instead,
the impact of mobility is limited to the first hop, i.e.,
the relay nodes form a multi-hop mesh network that section we consider the case of a single mobile user
covers the area of interest. carrying a patient node in a pouch around his neck.
Patients carry the patient nodes (Figure 2). A patient To characterize the network reliability we introduce
node integrates an embedded platform that provides the following metrics. End-to-end reliability is the
computation, storage, radio, and sensors. The patient fraction of data packets received by the base station
nodes are battery operated. out of the total number of packets generated by the
patient node. First-hop reliability is the reliability of
the packet transmission over the first-hop, i.e., from
the patient node to a relay node. First-hop reliability
allows us to directly characterize the impact of user
mobility on routing reliability.
Normal activity: In the first experiment, we asked the
volunteers to (1) go about their normal routine and
(2) record the times when they leave their offices and
Figure 2. Hardware prototype of a patient node. when they return in order to correlate network
Baseline Software Prototype: The goal of the reliability and mobility. The experiment lasted for 2.5
baseline software prototype was to reuse components hours during which a pulse and oxygenation reading
provided by the TinyOS operating system to develop was taken every 30 seconds. We chose this data rate
a clinical monitoring system. The prototype uses the because the pulse and oxygenation typically do not
Collection Tree Protocol (CTP)9 which is the de facto vary significantly within 30 seconds.
data collection protocol for multi-hop WSNs. CTP is Figure 3 shows the vital signs collected during the
designed to be robust and efficient: it provides close experiment. The background of the figures is color-
to 90% packet reliability for stationary networks9. coded as follows: a white background color indicates
However, as shown in the next section, CTP that the volunteer was in his office; the gray
performs poorly in the presence of mobility. background indicates the volunteer outside his office.
Empirical Study The end-to-end reliability was 82.39%. A careful
analysis indicates a strong correlation between user
The empirical study is designed to compare the mobility and packet drops, as consecutive packet
reliability of the baseline prototype in an indoor drops often occurred following periods of mobility.
environment that emulates a clinical environment The first-hop reliability in the two experiments was
with mobile patients. The experiments were 85.38%. This indicates that most of the packet drops
conducted on a WSN testbed deployed in Jolley and occur during the first-hop transmission and that once
Bryan Halls at Washington University in St. Louis a packet was delivered to the first relay, CTP was
(Figure 1). The testbed consists of 30 TelosB motes able to route the packets through the remaining hops
located in Jolley Hall. All sensor data collected from with high reliability.
sensor nodes are transmitted over the wireless mesh
network. It is important to note that the scale of the Mobility Stress Test: We hypothesized that the
experiments matches our target application: observed packet drops were due to poor routing table
monitoring a hospital unit. The formed network is management. The patient node running CTP
highly redundant, many relays receiving packets from discovers nodes that are within its communication
the patient node. range. When the subject moves sufficiently to break
the link to its current parent in the routing tree, CTP
For instrumentation, each node is connected to a will attempt to transmit to the next best node in its
central server using a wired USB and Ethernet routing table. However, since the subject has moved,
backbone used to collect experimental results without many of the entries in the routing table may have
interfering with ongoing wireless transmissions. In been invalidated by the subject's movement and, as a
some experiments, an additional 7 TelosB motes result, CTP will attempt to transmit to nodes outside
were added to provide coverage in Bryan Hall. These the patient node's communication range.
nodes are not connected to the USB backbone and
logged their data in flash memory, which is inserted To test this hypothesis we asked a subject to move
in a database upon completion of each experiment. according to the following pattern: stay at position A
One of the motes was selected as the base station for 3 minutes, move to position B, stay at position B
while the remaining nodes acted as relays. To better for 3 minutes, and move back to position A (Figure
characterize the reliability of the system, in this 1). This mobility pattern was repeated five times.
Figure 4 shows the reliability of CTP under mobility.
 Figure 3. CTP under normal activity. Figure 4. CTP under stress mobility

Figure 5. DRAP under normal activity. Figure 6. DRAP under stress mobility.

The white background indicates the user at position easily integrated with existing data collection routing
A, the dark the user moving, and grey the user at protocols such as CTP.
position B. The pattern of packet drops indicates that
The key design challenge in DRAP is how to
our hypothesis is correct: packet drops tend to occur
optimize for the common case when there is no
following changes in subject's location.
mobility and to adapt to user mobility. DRAP
CTP performed poorly under mobility, having an accomplishes this by associating a mobile node with
end-to-end reliability of 71.18% even when a packet the relay node that has the best link quality. DRAP
was transmitted up to 30 times if it remained detects when its relay table becomes outdated due to
unacknowledged. The long sequences of packet drops mobility by keeping track of the number of broken
are particularly disconcerting. The first-hop links. We consider a link to be broken when five
reliability was 72.88%, indicating that 97.67% of the packet transmissions remain unacknowledged. When
packet drops occurred in the first hop. In contrast, if the number of consecutive broken links exceeds a
the packets from the mobile node are delivered threshold set to three in our experiments, the entire
reliably to the first relay node, then CTP's reliability relay table is flushed and new relay is discovered to
is 97.67%. This suggests that the key research avoid the previously discussed problem. A detailed
question is to develop a more reliable data collection description of the DRAP protocol may be found in10.
mechanism from the mobile nodes to appropriate
Empirical Evaluation: To evaluate the reliability of
relays, which we call the dynamic relay association
DRAP we reran the mobility experiments presented
problem in this paper.
in the Empirical Study section. Figure 6 shows the
Dynamic Relay Association reliability of DRAP under mobility stress. During the
one-hour experiment, we did not observe any packet
We developed the Dynamic Relay Association
drops. In contrast, CTP dropped 28.82% of the data
Protocol (DRAP) to achieve reliable data collection
in a similar experiment (Figure 4). Figure 5 shows
from mobile users to their first relay nodes. Based on
the reliability of DRAP under normal activity. During
the insights from the empirical study presented in
the 2.5-hour experiment, DRAP dropped only 3
Empirical Study, DRAP is not designed to be an end-
packets resulting in an end-to-end reliability of
to-end routing protocol for replacing CTP. Instead, it
99.33% at a duty cycle of 2.09%.
dynamically discovers and selects the right relay for
the patient node to send its data after the user moves. Multiple User Reliability Study: To test the
DRAP complements routing protocols in that the performance of the system in the presence of multiple
former focuses on reliable data delivery over the first users, we monitored 7 healthy volunteers for an hour.
hop (from the patient node to the first relay), while The volunteers followed their normal activities while
the latter focuses on delivering data from the first carrying our patient nodes. The end-to-end reliability
relay to the base station. DRAP can therefore be of our system is shown in Table 1. The end-to-end
reliability for each user was between 96-100%. We
also measured the duty cycle of two of the References
volunteers, which were 2% and 0.2%, respectively.
1. The Joint Commission. 2008 National Patient
The difference in the duty cycles may be explained
Safety Goals. 2008.
by the difference in the mobility patterns of the two
2. Kho A, Rotz D, Alrahi K, Cárdenas W, Ramsey
users. The user with more frequent movement had a
K, Liebovitz D, Noskin G, Watts C. Utility of
higher duty cycle because DRAP kept the node active
commonly captured data from an EHR to iden-
to discover new relays.
tify hospitalized patients at risk for clinical dete-
User End-to-End Reliability rioration. AMIA Annu Symp Proc. 2007 Oct
1 100.00% 11:404-8.
2 98.33%
3. Jones AE, Brown MD, Trzeciak S, et. al. The
3 100.00%
4 100.00% effect of a quantitative resuscitation strategy on
5 100.00% mortality in patients with sepsis: a meta-analysis.
6 96.15% Crit Care Med. 2008 Oct;36(10):2734-9.
7 98.59% 4. Zeitz K, McCutcheon H. Policies that drive the
Table 1. End-to-end reliability for multiple users. nursing practice of postoperative observations.
An important concern is the scalability of the system Int J Nurs Stud. 2002 Nov;39(8):831-9.
as the number of patients is increased. While the 5. Lorincz K, Malan DJ, Fulford-Jones TRF, Na-
experiments are limited to 7 users, the low data rates woj A, Clavel A, Shnayder V, Mainland G,
make it feasible to monitor numerous patients. Welsh M. Sensor Networks for Emergency Re-
Typical data rates achieved by 802.15.4 are 25-50 sponse: Challenges and Opportunities. IEEE
kbps. Assuming a 25kbps data rate and excluding Pervasive Computing. 2004 Oct;3(4):16-23.
other traffic overhead (e.g., due to routing), the 6. Hong X, Xu K, Gerla M. Scalable routing proto-
proposed system may monitor up to 4100 patients cols for mobile ad hoc networks. IEEE Network.
when the pulse and oxygenation are sampled every 2002 Jul;16(4):11-21.
30s. The large margin gives us confidence that we 7. Gao T, Greenspan D, Welsh M, Juang R, Alm A.
may monitor all patients in a hospital unit. Vital signs monitoring and patient tracking over
a wireless network. Conf Proc IEEE Eng Med
Conclusion Biol Soc. 2005;1:102-5.
This paper proposes wireless patient monitoring as an 8. Curtis DW, Pino EJ, Bailey JM, Shih EI, Wa-
enabling technology for early detection of clinical terman J, Vinterbo SA, Stair TO, Guttag JV,
deterioration in general hospital wards. We present Greenes RA, Ohno-Machado L. SMART: an in-
an empirical study on the reliability of real-time data tegrated wireless system for monitoring unat-
collection from mobile users wearing wireless pulse- tended patients. J Am Med Inform Assoc. 2008
oximeters. We observe that the key challenge in such Jan;15(1):44-53
systems is to prevent packet loss from mobile users to 9. Gnawali O, Fonseca R, Jamieson K, Levis P.
their first-hop relays, while the standard data CTP: Robust and Efficient Collection through
collection routing protocol called CTP can provide Control and Data Plane Integration. Stanford
reliable data transport over stationary wireless relay Univ. Technical Report SING-08-02.
networks. Based on this insight we developed the 10. Chipara O, Brooks C, Bhattacharya S, Lu C,
Dynamic Relay Association Protocol (DRAP), an Chamberlain R, Roman GC, Bailey TC. Reliable
effective mechanism for dynamically discovering Data Collection from Mobile Users for Real-
right relays for mobile users. The key advantage of Time Clinical Monitoring. Washington Univer-
DRAP is that it enables reliable data collection from sity in St. Louis Technical Report.
mobile. Empirical evaluation showed DRAP 11. Jetcheva JG, Johnson DB. Adaptive demand-
delivered at least 96% of the data from multiple driven multicast routing in multi-hop wireless ad
users, while maintaining a radio duty cycle below hoc networks. Conf. Proc. MobiHoc.2002;33-44.
2.8%. Our results demonstrate the feasibility of 12. Wood A, Virone G, Doan T, Cao Q, Selavo L,
WSNs for clinical monitoring. Wu Y, Fang L, He Z, Lin S, Stankovic J.
ALARM-NET: Wireless Sensor Networks for
This publication was made possible by Grant Number UL1 Assisted-Living and Residential Monitoring.
RR024992 from the National Center for Research Resources
(NCRR), part of the National Institutes of Health (NIH) and NIH Technical Report. Univ. of Virginia CS-2006-11
Roadmap for Medical Research. Its contents are solely the 13. Jon Adams, Bob Heile. Busy as a ZigBee. IEEE
responsibility of the authors. Additional, funding provided by NSF Spetrum.http://www.spectrum.ieee.org/computin
NeTS-NOSS Grant CNS-0627126 and CRI Grant CNS-0708460. g/networks/busy-as-a-zigbee