Coordinating Heterogeneous Work: Information and Representation in Medical Care

Abstract

Medical care involves intense collaboration amongst a number of practitioners including physicians, nurses, and pharmacists Their work is concentrated on a single patient, and yet their activities, motivations, and concerns are very different. We explore the use of a shared information system in helping these individuals coordinate their work. In particular, we use the idea of a common information space to explore how the shared information is incorporated into the diverse work practices of an intensive care unit. In addition to physical co-location, we found that providing information in many specialised representations is critical to managing their coordination. Unlike paper records, computer systems offer the ability to decouple information from its representations. This decoupling opens up a rich design space for systems that allow people with different interests, concerns and work practices to work together effectively.

Full text

W Pnnz, M Jarke, Y. Rogers, K. Schmidt, and V Wulf(eds.),
Proceedings
of the
Seventh European
Conference
on
Computer-Supported Cooperative
Work,
16-20 September
2001,
Bonn,
Germany,
pp. 239-258
© 2001 Kluwer Academic Publishers Printed in the Netherlands.
Coordinating Heterogeneous Work:
Information and Representation in
Medical Care
Madhu C. Reddy, Paul Dourish, and Wanda Pratt
University of California, Irvine, USA
mreddy@ics.uci.edu,
jpd@ics.uci.edu,
pratt@ics.uci.edu
Medical care involves intense collaboration amongst a number of practitioners including
physicians, nurses, and pharmacists Their work is concentrated on a single patient, and
yet their activities, motivations, and concerns are very different We explore the use of a
shared information system in helping these individuals coordinate their work In
particular, we use the idea of a common information space to explore how the shared
information is incorporated into the diverse work practices of an intensive care unit In
addition to physical co-location, we found that providing information in many specialised
representations is critical to managing their coordination. Unlike paper records, computer
systems offer the ability to decouple information from its representations. This decoupling
opens up a rich design space for systems that allow people with different interests,
concerns and work practices to work together effectively.
Introduction
The concept of a common information space, or CIS, has become an influential
way to think about the use of shared information in collaboration. Originating in
the work of Schmidt and Bannon (1992), and further explored by Bannon and
B0dker (1997), it was designed to extend then-current notions about the role of
technology and shared information.
At the time this was originally proposed, a great deal of technical attention was
being paid to the development of "shared workspace" systems (e.g. Lu and

240
Mantel 1991; Ishn et al. 1992) These systems attempted to extend the
workspaces of conventional single-user applications such as word processors and
drawing tools, allowing synchronous or asynchronous collaboration across digital
networks. Designing effective shared workspace systems presented a range of
technical challenges concerning appropriate network protocols, synchronisation,
concurrency control mechanisms, and user interface design. Still, over time
considerable progress was made, resulting today in the widespread use of systems
such as Microsoft NetMeeting that emerge directly out of the "shared workspace"
tradition
However, by introducing the concept of common information space, Schmidt
and Bannon sounded a note of caution about the technological conception of
shared information. They pointed out that information is not shared
unproblematically. It has to be explicitly placed in common - extracted from one
person's work context? and reformulated in some way that displays its relevance
to others (by being related to some common conceptual scheme, for instance).
Similarly, when individuals come to examine shared information, they need to
recontextuahze it, making it relevant for their immediate needs. Further, the same
information may be relevant to two people in quite different ways, for instance, a
purchase order has different consequences for the person who must process the
shipment and the person who must balance the budget. A common information
space according to Schmidt and Bannon incorporates not only a repository of
information held in common amongst different parties, but also the work practices
surrounding that information - how it is used, managed and integrated into the
work of those who share it. The practices by which information is placed in
common, and then made relevant to individuals' activities, make the information
meaningful in the context of their work.
The value of the common information space concept, then, is that it relates
shared information to the activities that are conducted over and through the
information. While the precise formulation has, lately, been subject to a certain
amount of critical scrutiny (see, for example, Bannon 2000; Randall 2000), there
is still considerable value in the perspective that it offers on how shared
information is incorporated into daily work practices.
In this paper, we report on a field study of the use of a shared information
repository in medical work. In particular, we describe the challenges to using a
computer-based patient record system as a CIS in an intensive care unit. The
paper is structured as follows: in the following section we discuss in greater detail
the CIS concept and related work. Next, we present our field study: the research
site,
staff,
and technology as well as examples of daily work activities in the unit
We then discuss the implications of our findings for the construction and use of a
CIS,
and finally, present some design considerations for CIS systems followed by
some concluding remarks. .

241
Background
Schmidt and Bannon (1992) introduced the concept of common information space
by contrasting it with technical conceptions of shared information:
Cooperative work is not facilitated simply by the provisioning of
a
shared database, but
rather requires the active construction by the participants of
a
common information space
where the meanings of the shared objects are debated and resolved, at least locally and
temporarily (Schmidt and Bannon,
p
22)
A CIS, then, encompasses not only the information but also the practices by
which actors establish its meaning for their collective work. These negotiated
understandings of the information are as important as the availability of the
information itself
The actors must attempt
to
jointly construct a common information space which goes beyond
their individual personal information spaces The common information space is negotiated
and established
by
the actors involved (Schmidt and Bannon, p 28)
This is not to suggest that actors' understandings of the information are identical;
they are simply "common" enough to coordinate the work. People understand
how the information is relevant for their own work. Therefore, individuals
engaged in different activities will have different perspectives on the same
information. The work of maintaining the common information space is the work
that it takes to balance and accommodate these different perspectives A "bug"
report in software development is a simple example. Software developers and
quality assurance personnel have access to the same bug report information.
However, access to information is not sufficient to coordinate their work. Instead,
it is their more or less shared understanding of the record's organizational
structure that allows developers and quality assurance personnel to coordinate
their activities They know where to find certain information, what it means if the
information is not present, and what implications this information carries for their
own work
The distinction between access and practical understanding is at the heart of
the CIS concept. Moving from one to the other is not straight-forward. Schmidt
and Bannon discuss potential problems actprs face in interpreting information
when the information's creator, the context of its creation, or politics of its use is
unknown to the actors involved. They provide examples of what occurs when this
contextuahzing information is not present and discuss how common information
spaces are created in different work situations.
In many work settings, a CIS involves not only local work practices but also
crosses group boundaries. The information artifacts at the heart of the space are
the focus of heterogeneous workgroups and have characteristics of "boundary,
objects" (Star and Gnesemer 1989). Boundary objects are informatidn artifacts
flexible enough to fit local work practices but also stable enough to convey
information across group boundaries, enabling them to act as coordinating
mechanisms for interactions between diverse workgroups. For example, Berg and

242
Bowker (1997) examine the medical record as "an organizational
infrastructure...[that] affords the interplay and coordination between divergent
worlds." They argue that the patient record is both a representation of the patient
as well as a representation of the work being carried out on the patient. The
record is used by different groups (e g physicians, nurses, administrators, etc.) in
their own local work context. To each group, the record has a localized meaning,
but it also serves to coordinate the different activities of these groups. The patient
record functions as a boundary object, spanning the borders of a number of
different groups.
Bannon and B0dker (1997) use boundary objects as a lens for viewing
common information space. They contend that, as with a boundary object, the
dialectical nature of the common information space is an important characteristic:
It is this tension between the need for openness and malleability of information on the one
hand, and on
the
other, the need for some form of closure, to allow for forms of translation and
portability between communities, that we believe characterizes the nature of common
information space (Bannon and B0dker,
p
86)
Resolving the tension between the need for both openness (supporting diverse
work practices) and closure (supporting coordination) depends on features of the
work and work setting. Bannon and B0dker use a variety of examples to discuss
CIS construction in different settings, and suggest that physical co-presence plays
an important role in making it easier to construct a common information space:
In the case of physically shared workspace, due to the common work setting and exposure to
the same work environment, actors are able to co-operate with each other, both in the
production and reception of utterance and information, without having to resort to extended
descriptions or elaborated
codes,
due to their understanding of
the
shared context within which
they work (Bannon and B0dker, p 83)
The physical co-location of the workgroup members provides a number of
benefits. First, the work related to the information is highly visible. Participants
can see not only what other individuals are doing, but also when and how they are
doing it. Second, in a physically shared space, individuals can easily ask other
individuals for explanation of something they do not understand - "popping your
head over the cubicle wall." Participants can ask their neighbors questions about
the work before looking elsewhere for the information. Finally, a shared
workspace allows human mediators to play a more effective role in sharing and
communicating knowledge about the artifact. Blomberg and colleagues' (1997)
study of attorneys in a law firm highlight the role of human mediation of an
artifact. Over a period of time, M, a firm attorney, had amassed a large collection
of legal documents that he deemed potentially useful or reusable. These
documents were available to all, but stored in a filing cabinet in his office When
other attorneys wanted to find a document in the cabinet, M acted as gatekeeper,
helping them locate and interpret the needed documents:
The utility of M's file for other attorneys depends on his knowledge of its contents and
organization, derived in turn from his creation, maintenance, and regular use of
the
file Other

243
attorneys rely on M to help determine whether the form file contains documents relevant to the
transaction on which they are working, to point them to likely places in the file where relevant
documents might be found, and to justify the choice of particular documents (Blomberg et al ,
P 195)
M mediated the other attorneys' search for documents in the filing cabinet. This
was successful because he was physically available along with the files.
Although physical co-location provides a number of benefits, how much of a
role it actually plays in the creation of a common information space remains
unclear. We know that when the work of the participants is similar (e g. as with
the attorneys using M's filing cabinet), the physical co-location of actors helps
them to create a usable common information space This depends, though, on
some mutual intelligibility of action, so that when participants observe each other
at work with the artifacts, they have some understanding of what work is being
earned out. However, what if the work practices are so heterogeneous that the
work of different actors is no longer intelligible to others? What role does
physical co-location play then? And what other elements can be brought to bear
in order to resolve the tensions of openness and closure?
A Common Information Space in Medical Work
We explore these issues by looking in detail at an example of a common
information space supporting divergent forms of work. We focus on medical
work in an intensive care unit (ICU) supported by a shared patient record system
called HealthStat.1
Research Site and Methods
The surgical intensive care unit (SICU), where we conducted our fieldwork, is
one of nine ICUs of a large urban teaching hospital. The majority of participants
in this study worked in the SICU. The research team had access to the SICU
staff.
In addition, we observed and interviewed the HealthStat technical team members
from the hospital's information systems department. The first author observed
work in the SICU for approximately three months during summer 2000. He
collected data through 30 formal interviews, as well as a number of informal
interviews, and observations. The formal interviews were taped and transcribed.
The research team had access to the HealthStat application and internal
communications, including written policies, procedures, and meeting notes.
Each ICU provides rigorous invasive and non-invasive care-monitonng for
patients requiring special attention due to a critical medical condition.
Specifically, the SICU is a 20-bed unit that treats the most seriously ill surgical
A pseudonym

244
patients, including those who have undergone liver transplant, major trauma, or
major elective surgery It is equipped with sophisticated equipment including
digital physiological monitors and a fully computerized patient record system.
The SICU is an extremely busy unit with 15 out of 20 beds occupied on a daily
basis Patients usually stay in the unit for 5-6 days and are the focus of a team of
health-care workers. In most cases, patients are in such critical condition that any
minor change in their condition could have rapid and severe implications.
Therefore, the specialised equipment and staff in the SICU allows even small
changes in a patient's condition to be detected early, thus permitting rapid
changes in treatment to prevent problems from developing.
SICU Staff
The SICU staff includes surgical critical care nurses, physical therapists, social
workers, respiratory therapists, surgical residents, critical care fellows and
faculty. We focus on three ,SICU work groups physicians, nurses, and
pharmacists because these groups interact with each other and the computerized
patient record system on a daily basis. We will now briefly discuss each group.
Physician Staff
The physician staff is organized hierarchically and consists of three rotating
surgical residents, two critical care fellows, and four attending physicians. At the
bottom of the hierarchy are residents. They are considered physicians-in-training
and provide the most hours of patient care in the unit. The fellows are in the
middle of the hierarchy. They have completed their residency and are undergoing
specialized training in intensive care. Fellows supervise and monitor residents'
activities on a day-to-day basis. They resolve a majority of problems that
residents cannot handle. If a fellow cannot resolve a problem, an attending
physician is notified. The SICU has four attending physicians, each with many
years of experience in intensive care The attending physicians supervise fellows
and residents to ensure that they receive proper training as well as maintain a high
standard of patient care The ultimate responsibility for success or failure in the
unit lies with the attending physicians.
Nursing Staff
The nursing staff has more than fifty registered nurses certified in critical care,
supervised by a SICU nurse manager. Depending on the number of patients, there
are 10-12 nurses on each 12-hour shift. The nursing experience in the unit vanes;
some nurses have more than 20 years of experience but the majority of nurses
have been in the unit less than five years. The nursing staff has experienced high
turnover due to the stress of ICU work. The nurse's responsibilities range from
patient assessment and monitoring to medication administration. Because of the

245
serious condition of the patients in the unit, each nurse is responsible for only 1-2
patients per shift. (In the non-ICUs, the nurse-to-patient ratio is 1:6.) Therefore,
SICU nurses can provide more focused care for their patients.
Pharmacist
A pharmacist is assigned to the SICU on a regular basis. She spends 3-4 hours per
day in the unit. The medical
staff,
especially the residents, rely heavily on her
knowledge to help them make the appropriate medication decisions. She is also
the pnmary resource for the nurses on any questions concerning drug dosage or
usage. The pharmacist participates in the SICU team's rounds each morning and
is familiar with the conditions and medications of all the SICU patients.
Diverse Work Practices
The patient is the center of the work activities in the SICU. The pnmary goal of
the SICU staff is to stabilise patients so they can be safely transferred out of the
unit. On a daily basis, physicians, nurses, and pharmacists successfully coordinate
their patient care activities. However, although patient care is the central focus,
the vanous groups have their own work to do; their motivations, concerns, and
activities are quite different (Strauss et al. 1985).
Consider one case that we observed. A nurse noticed that her patient's fingers
were turning blue. She knew that blue fingers were an indication of blood vessel
constnction and correctly attnbuted the condition to the medication. The patient
was in obvious discomfort. The nurse did not understand why this medication was
being administered since it clearly caused so much distress to the patient. She
asked the physician if they could stop the medication. However, he insisted that,
despite the discomfort that it might cause, the medication was necessary to
improve the patient's overall medical condition.
This example highlights the distinct roles and concerns of the physician and
nurse. To the physician, the patient's discomfort was not as important as treating
the other medical problems. On the other hand, the nurse was pnmarily concerned
with the patient's comfort and well-being. The different emphases of their work
continually feature in the life of the unit; with their different concerns, physicians
and nurses frequently do not understand the details of each other's work. As one
SICU physician stated,
There is a scope of practice for nurses There are certain nursing actions.lbut] they are not the
same as my actions They are involved with patient care and they make patient care decisions
on a routine basis. It is true that they cannot do what I do They cannot order medications, [but]
when I order it they administer it That is her job But they make nursing care decisions. I am
not minimizing them. It is a different sphere of things For example, patient comfort measures.
I don't prescribe that I don't tell them when to clean a patient, when to put a pillow here or
there, and yet they are important to the patients Patients remember that Nurses are the ones

246
who make those decisions and decide that care I think that they have a very specific sphere of
care,
just different from mine
If physicians view their activities as distinct from nurses' activities, then nurses
too view their work as differing from that of other groups (e.g. physician,
pharmacist). Berg and Bowker (1997) and Bowker and Star (1999) discuss the
creation of the Nursing Interventions Classification, a classification of nursing
work that was developed by nurses as a way to describe their activities
independently of other groups' work. Nurses created this classification as a means
to legitimate their activities and make them visible to physicians and other
hospital staff who otherwise neither recognized nor understood their work
While their work practices may be quite diverse, effective and timely
coordination between physicians, nurses, and pharmacists is critical otherwise the
patient will suffer In one example we observed, a nurse failed to notify the
physician that the patient's sodium was raising to dangerous levels. If the
physician had been notified quickly, he would have been able to give the patient
medication to lower the sodium. However, the physician only found out about the
sodium levels six hours later, by which time the patient's condition had
detenorated so far that the physician had to intubate the patient to protect her
airways. As the example highlights, these groups work under constant time
pressure that can effect patient care They do not have the luxury of waiting an
extended period of time for important patient information.
HealthStat: A Common Information Repository
Information technology plays a crucial role in the SICU. A computerized patient
record system, HealthStat, mediates much of the work among the physicians,
nurses, and pharmacists. The staff has used HealthStat for more than nine years
and is well acquainted with its functionality. Originally implemented in the SICU,
the system is now in use in eight of the other nine ICUs in the hospital.
Almost all patient information ts in the computerized record Since the
patient's bedside monitoring systems are linked to HealthStat, physiological data
such as temperature, blood pressure, heart rate, and fluid levels are downloaded
automatically into the patient's HealthStat record However, before the
information is permanently entered into the record, the patient's nurse ensures the
validity of the data by cross-checking the data in the record with the displays on
the bedside monitoring systems. The record also contains medication information,
progress notes, and laboratory results.
Most of the data that is' not automatically downloaded into HealthStat is
entered by nurses. They can spend up to 15 minutes every hour entering data into
the system. In a busy ICU, this is a great deal of time but is still shorter than the
time that would be spent entering the same information on a paper chart.
Physicians, by contrast, do minimal data entry; they largely use HealthStat to
monitor the patient's status and to find needed patient information. Finally,

247
pharmacists are interested in ensuring that the patient is receiving the appropriate
medication and that all the information related to the patient's medication is
correct; the SICU pharmacist spends a couple of hours each day using HealthStat
The Work of the SICU
We present three examples of activities related to patient care in the SICU. These
activities highlight the collaboration required for successful completion of work
tasks.
SICU Morning Rounds
SICU morning rounds play an important role in the unit's patient care process A
multi-disciplinary team led by a fellow and consisting of three residents, attending
physician, pharmacist, and nurse visits each patient. The goal of morning rounds
is to discuss and decide upon a plan of care for that day for each patient. The team
uses HealthStat workstations outside patient rooms to find patient information.
The team begins by viewing x-rays of all the SICU patients. After examining the
x-rays, the team "rounds" on each patient. Each of the three residents are
responsible for a certain number of patients in the unit. During rounds, the
residents "present" their patients to the team As a resident outlines the patient's
current condition, vitals and other information, the fellow and other team
members view the patient's record on the HealthStat workstation. They do this
both to verify the resident's information and to gather other pertinent information.
As one fellow stated, "It is much easier for me to find the information in the
system than to wait for them [residents] to give it to me." After the resident
presents, the fellow examines the patient. The team then discusses the patient's
condition and decides on the plan of care for the day. After all the decisions are
made, a resident writes a progress note in the patient's HealthStat record. The
following vignette presents a typical patient round.
MC,
a resident, presents the patient to the team The patient has recently undergone a male to
female gender change operation She was admitted to the unit because of complications from
the administration of high levels of progesterone and estrogen TK, a fellow, suggests that the
hormones be discontinued However, MC argues that the patient needs them for the gender
change WK, another fellow, looks at HealthStat and asks MC whether the patient is getting
both Heparin and TPA (both drugs prevent blood clotting) MC tells WK that the patient is
only receiving Heparin AL, an attending, asks whether estrogen and progesterone have a dose
response level None of the residents know the answer to this question. Later, TK asks the JC,
the pharmacist about dose-related complications for estrogen and the relationship between
estrogen and progesterone JC tells TK and WK that the drugs are dose independent of each
other After a discussion, the team decides not to discontinue the progesterone and estrogen
As the example illustrates, rounding involves a collaborative dialogue among
physicians, nurses, pharmacists and the patient record system. Different questions

248
were raised during the interaction: Should the estrogen and progesterone be
stopped? What other medications is the patient receiving? Are there dose level
concerns between estrogen and progesterone? HealthStat provided some
information, but the different team members brought their individual perspectives
to understanding the information so the questions could be answered. MC gave
the context of the case and explained the need for the high level of estrogen and
progesterone. She also answered WK's question about the patient medication. JC
answered the question about the drug interaction. HealthStat played a role in
answering questions, but only as a component of the entire collaborative process.
For instance, WK raised the medication question after looking at the patient's
physiological data in HealthStat. Instead of asking the resident, WK could have
also looked up the medications in HealthStat. However, she was interested in not
only the patient's medication but also the rationale for giving the medication. MC
was in the best position to provide that information. The information
itself,
in
HealthStat, does not tell the complete story. During rounds, team members
actively collaborate to integrate that information into the context of their work
Medication Administration
Ordering and administering medication requires collaboration between
physicians, nurses, and pharmacists In routine situations, most surgeons use a
standard set of drugs. However, for complex cases, nurses and pharmacists often
provide information that help physicians tailor the medication prescription. Since
nurses are constantly by the bedside, they can inform physicians about the
patient's physical and mental state. This information can help physicians to decide
whether a current drug and dosage are appropriate If physicians need to prescribe
a drug for a problem with which they are not familiar, pharmacists can provide a
list of appropriate medications
Nurses must collaborate directly with both physicians and pharmacists When
ordered to give an unfamiliar drug, nurses commonly ask the physician why it is
being given, especially when the drug causes discomfort or pain to the patient.
Most physicians want the nurse to understand the plan of care and will answer
such questions readily. The nurses also ask the pharmacist questions concerning
the medication and dosage administration. For certain kinds of drugs, such as pain
relievers, it is the nurse who observes the patient's response most directly, and
whose opinion is usually given high regard by physicians for subsequent pain
medication orders.
HealthStat plays an important role in supporting the collaborative process of
medication administration. The central element that HealthStat provides in this

249
Figure 1 Different Representations of Medication Information: (a) Pharmacists use the
Medication Administration Record (MAR) to provide them with the more detailed information on
each medication (b) Nurses use the Medication Worklist to keep track of their medication
administration work activities
process is the Medication Administration Record, or MAR (Figure la). The MAR
coordinates both the prescnption and administration of medication. When the
physician writes a medication order, a nurse or pharmacist enters the order into
the MAR, recording the details of the prescribed medication. Although the MAR
provides the detailed information necessary for the pharmacists, it provides too
much detail for the nurses to allow them to plan their medication administration
activities for a shift. Consequently, to administer medications effectively and on-
time,
nurses use another "view" of the MAR, the Medication Worklist (Figure
lb),
which provides a time-ordered list of dosages, and administration times for
all drugs due to be administered on the current nursing shift. The nurses use the
Worklist to plan their medication administration activities for each of their
patients.
Each group uses the system to view a patient's medication information,
although in different ways. For example, pharmacists check the appropriateness
of the medication based on the patient's condition. If they do not believe that the
drug is appropriate, they will offer the physician advice about alternative
medications. Physicians may consider the pharmacists' recommendations when
making their final medication decision, based on the information that HealthStat
provides them concerning the patient's response to previous treatments.

250
Figure 2 HealthStat Flowsheet's MEDS Section: The ICU staff especially physicians use the
MEDS section to quickly check on patient medications
Configuration Group
HealthStat is implemented in eight ICUs in the hospital. Due to technical
constraints, any changes to the various interfaces to the system are replicated to
only seven of the eight ICUs (the eighth ICU uses HealthStat on an independent
platform). Still, coordinating interface changes for seven ICUs is a difficult
process. Although all ICUs have some information in common, much is particular
to each ICU and its specialized work. Terms used in one unit may not be used in
another. To prevent any misunderstandings between the different ICUs, a
committee called the Configuration Group was created. The group consists of
nurses from the different ICUs, HealthStat programming team members, and the
HealthStat director. Any interface changes must be discussed in the Configuration
Group meeting. The group then decides whether changes will be implemented.
An important aspect of these meetings is dealing with changes to the
HealthStat Flowsheet (Figure 2). In our observations and interviews, nurses and
physicians described the Flowsheet as the most widely used interface of the
computerized record. Each of its fourteen subsections contains information about
the patient. For example, the MEDS subsection contains brief information about
the patient medication. ICU staff use the Flowsheet to get a quick overview of the
patient's condition.
Since interface changes made to the Flowsheet for one ICU are propagated to
the Flowsheets in the other ICUs, the Configuration Group has to mediate the
differing requirements of the various ICUs. For example, both the medical and the
surgical cardiac units are interested in the section of the Flowsheet dealing with

251
cardiac data. Because HealthStat was implemented first in the medical unit, the
nurses there had the data visually arranged in the Flowsheet to fit their work
activities. However, when the nurses in the surgical unit began using the system,
they complained about this arrangement, arguing that they could not easily find
needed information because the data was not arranged according to their work
activities.
The responsibility for resolving these problems falls to the Configuration
Group. The group plays an important role in minimizing friction between
different units concerning changes to the system. The Configuration Group
meetings also provide a rare opportunity for practitioners to cross organizational
boundaries and discuss their work with others from different organizational
groups. As such, the Configuration Group engages in an explicit negotiation of
the meaning and role of the information in HealthStat.
HealthStat as a CIS
As we have outlined, HealthStat is a shared repository of information used to
coordinate the different aspects of medical care in the SICU. However, looking at
HealthStat as a CIS leads us to focus less on the idea of shared information, and
concentrate more on the practices by which that information is put to use and is
made meaningful for the different sets of people who use it. For example,
although much information is automatically logged from systems that monitor the
patient's vital signs, that information is not accepted into HealthStat until it has
been reviewed and approved by a member of the SICU nursing
staff.
In other
words, the information needs to be explicitly "vetted" according to a set of SICU
expectations in order to determine its acceptability. In tum, this vetting allows the
SICU staff to maintain a common understanding about the appropriateness and
accuracy of the information contained in the system. Thus, the notion of
"information" here is not uncontested; rather, HealthStat is a repository for
approved and trusted information.
Prospective and Retrospective Use
The patient record in HealthStat incorporates a broad set of concerns and a wide
range of information about all aspects of a patient's current treatment regime and
medical history. The amount of information that it contains about a given patient
is potentially overwhelming. One way that the system's design deals with this
problem is by providing a range of interfaces tailored to the needs of either the
different practitioners who may deal with the patient or the different activities that
make up the patient's care.
For example, physicians interact with the system primarily through the
Flowsheet Since they do not have a great deal of time, the Flowsheet provides

252
them with quick information about the patient's condition. Pharmacists have a
different set of goals. They are less concerned with the overall medical treatment
of each individual patient and more concerned with ensuring proper medication
administration. Their interactions with the system consist largely of checking on
patient medications through the MAR On the other hand, the nursing
staff,
who
are primarily responsible for the moment-by-moment care of the patients, interact
with the system through a number of screens, depending on their particular tasks
One of the primary interfaces that they use to coordinate their work is the
Medication Worklist.
HealthStat stores information about the administration of medical care.
However, the different screens reflect very different aspects of that care. The
physicians' primary concern is with diagnosis and monitoring of the effectiveness
of a treatment regime. The process of rounding, for example, is about describing
how the patient has responded to treatment since the previous round, and on the
basis of that, deciding what path should be taken next. So, the physicians' primary
use of HealthStat's information is retrospective; they want to know what has
happened over the last 24-hour period. In contrast, the nurses, who must arrange
their activities in such a way as to ensure that each patient receives appropriate
attention at relevant points in the shift, look to HealthStat for prospective
information about the activities which will need to be earned out in order to effect
the prescribed regime of care. HealthStat sits at the nexus of these two concerns -
retrospective and prospective - both detailing what has gone before and
projecting what will come next (See Berg (1999) for a similar argument
concerning "reading" and "writing" a computerized patient record).
There are two consequences to this use of HealthStat. The first is the issue of
temporal coordination in a CIS. Previous investigations of CIS have pointed out
how the activities that surround an information store, and the practices by which
information is explicitly transformed in order to "place it in common" are
frequently oriented towards the anticipated lifetime of the information. At the
most banal level; when information is placed into some form of storage, it is with
an expectation that the information may need to be retrieved at a later date.
Participants record information in such a way as to anticipate the circumstances
under which it might be found again at some time in the future. Discussing the
case of the Danish National Labour Inspection Service, Bannon and B0dker
(1997) point out that records about encounters with companies may be reused
years later, and that inspectors need to be sensitive to the potential future uses of
the matenal they create; while Dounsh et al. (1999) discuss how the evolution of
a "common" classification scheme presented problems for the long-term storage
of engineering documents. In contrast, in the SICU, temporal coordination
through the CIS is both much more explicit and much finer-grained. The system
not only stores information but also transforms the information into a hour-by-
hour schedule'by which work activities can be coordinated. This mediation

253
between retrospective and prospective information is a key feature of how the
different groups within the SICU make use of the CIS.
Our second point concerns the coordination of these multiple representations
of information.
Information and Representations
One of the motivating concerns that we have been pursuing through this work is
the following problem, if the work practices of the different groups whose work is
coordinated through HealthStat are sufficiently diverse that many of the benefits
of co-location, as discussed earlier, are effectively lost, then what compensates for
this loss? How can coordination be re-established?
The issue of prospective and retrospective uses of information by nurses and
physicians offers a clue. Through its multiple screens, HealthStat offers different
views of the same information, and these different views are attuned to the needs
of the different groups who use them. For the nursing
staff,
HealthStat transforms
information about a treatment regime into a schedule of tasks and activities that
will need to be earned out. The information that HealthStat records gives rise to
many different representations (Flowsheet, MAR, Medication Worklist),
according to how the information is used by the different groups.
This decoupling that HealthStat allows between the information and its
representations is unusual amongst the CIS scenarios explored in the research
literature. In the case of a CIS that is based on physical records, there is, clearly,
only a single representation or physical form for each information artifact.
However, even in cases where the information is recorded electronically, a single
representation is still the norm (e.g. Tngg et al. 1999). HealthStat, however,
expresses the same information through different representations. As we have
seen, these representations are crucially integrated into the different working
styles and practices of the groups who collectively carry out the work of the
SICU. However, this need for different representations is balanced by the need
for shared information. It is not enough that the representations be different, as
would be afforded by translation-based approaches (Simone et al. 1999), but that
these be different representations of identical underlying information, since it is
through the sharing of this information that coordination is achieved.
The tension between the need for diverse representations (matching diverse
work needs) and common information (for stable coordination) is reflected in the
need to coordinate over the forms of the representations themselves. The staff
needs to coordinate their activities through more than simply the information;
they need to be able to discuss, to exchange, and to compare representations. The
work of the Configuration Group reflects this concern. The seven ICUs have
varied work practices requiring not only disparate information but also different
arrangements of the same information This diversity is a common feature of a

254
CIS,
but one interesting element here is that the diverse needs of the units are
explicitly negotiated through the Configuration Group. Here, the issue is not
different representations of the same information, but compatibility between the
representations used in different places. Since, the ICUs all use the same system,
there has to be clear understanding of the representations' meaning to each unit
Unlike most CIS negotiations which are informal in nature and carried out during
the course of the actual work, the Configuration Group allows nurses from the
ICUs to meet and exchange information about their different work practices. The
Configuration Group meeting is an opportunity for the group members to find out
in explicit detail how the same information might be differently used in the
various ICUs By exchanging information about each unit's local work practices,
Configuration Group members have a better understanding of how making
changes to representations can effect each unit.
Discussions in the Configuration Group meetings help overcome the problem
that each group has understanding the other's work This problem is manifest in
the SICU
itself.
While previous investigations described in Bannon and B0dker
(1997) have suggested that physical proximity is a key feature in allowing
different groups to coordinate their work in and around a common information
space, our field data suggests that this is true only in the cases where the work of
the different groups is sufficiently integrated (or, at least, mutually
comprehensible) that the information can have some general relevance. In these
circumstances, then, the ability to see how the work of others is being earned out
with and through the information allows participants to coordinate their actions
In cases where the work is more disparate, though, physical proximity is of less
immediate value. Even though they work in the same environment, the different
groups in the SICU do not feel that their work is understood by the others. The
role of a single information representation as a site of work coordination breaks
down. However, electronic information systems allow us to present multiple,
coordinated representations of information. When the system can present the
same information in ways that are differently attuned to the information needs of
different groups, participants see other's work transformed in ways that make
sense from their own perspective.
Design Considerations for CIS Systems
Our exploration of the use of HealthStat in the SICU has highlighted a number of
interesting issues concerning the role of information in coordinating work. In
particular, we have seen that work coordination through HealthStat depends on
the separation it offers between the information and its representation (how that
information is configured for particular uses). Although the observational material
presented in the paper has been very specific, our findings suggest a number of
broader implications.

255
First, the work of the SICU suggests that we should reconsider the role of
physical proximity and accessibility in coordinating cooperative work. As we
have already noted, previous studies have observed that physical proximity is
critically important in a range of collaborative settings, affording participants
visual and auditory access to each other's activities and facilitating easy
communication. Clearly, this is true, but it rests on a more fundamental
assumption that the activity going on in the physical space is intelligible to those
who witness it. For example, Heath and Luff's (1992) classic analysis shows how
the London Underground control room operators achieve a remarkably smooth
and intricate coordination between their activities through a combination of, first,
continually monitoring the actions of others in the room, and, second, explicitly
organising their actions so as to disclose what is happening to others nearby.
However, this depends not only on their proximity, but on their ability to interpret
what is going on around them, through their familiarity with the work of the
control room and the practices by which their colleagues organise that work It is
precisely this in-depth familiarity with the detail, motives and consequences of
each other's work that is absent in the case of the SICU. The physicians and the
nursing staff have only a limited and superficial understanding of each other's
work - certainly not enough to achieve the delicate choreography that Heath and
Luff observe So, the observation that physical proximity and accessibility
support the coordination of group work glosses over an important detail. More
accurately, physical proximity and accessibility afford the mutual interpretation of
working activities to those who share a sufficiently detailed understanding of
those activities in the first place. In cases such as the SICU, where this
understanding is not present, physical proximity is not, by
itself,
sufficient
Second, the case of the SICU shows us that although participants interact with
the information through different representations, coordinating their activities
depends on these representations reflecting the same underlying information.
Because it is the same underlying information, the different representations are
always sychronised; any changes in the underlying information will be
immediately reflected in all the different representations. The alternative would
be to maintain two different systems in parallel - perhaps an information store
that describes medication information, and a separate schedule that outlines
nursing tasks, such as that observed by Bardram (1997). However, the possibility
for inconsistency and the difficulty of moving information back and forth would
compromise the SICU's ability to coordinate activities around the patient; it is
important that the underlying information be shared The role of shared
information in promoting coordination has been explored extensively in CSCW,
particularly in the form of technologies promoting awareness (e.g Dounsh and
Bellotti 1992; Gutwin et al. 1996; Mark et al. 1997). The effectiveness of most of
these approaches, however, depends on a common representation of the
underlying information: a common information structure in the case of Dounsh

256
and Bellotti or Mark et al. and a common set of spatial arrangements in the case
of Gutwin et al. There has been much less exploration of uses of awareness
techniques in a coordinated fashion across multiple distinct representations,
although an exploration by Greenberg and his colleagues (1996) provides an
interesting example of the opportunities. Cases where the different forms of work
are highly diverse, such as in the SICU, may require this sort of approach. In turn,
our attention to the ways in which information representations can be designed to
naturally convey a sense of the activities in which they are involved. This is not,
in
itself,
a new observation (see Nygren et al. (1992) for an exploration of this
issue, also in the medical domain), but the separation between information and
representation implied in our study suggests that this meta-information must also
be coordinated with multiple representations of the information.
Finally, the separation between information and representation also highlights
how the same information is enmeshed in a variety of work processes. It serves
multiple purposes and enables multiple individuals to carry out their own work.
Traditional software architectures, however, typically provide no direct support
for this feature of information work. Often, the information is embedded in a
structure (such as a schema or hierarchy) that makes it tractable and manipulable
by software systems. However, these information structures make it harder to
pick information up and move it from place to place, decontexualising it and
recontextualising it according to the situation of need; and similarly, they only
reflect a single point of view on the role of the information rather than the many
different points of view that we see at work in situations such as the SICU. Our
fieldwork, then, provides support for approaches to information architecture that
separates the information from the structures that surround and describe it
(Dounsh et al. 1999; Parsons and Wand 2000). By decoupling information from
its structure and supporting diverse representations of the same information, these
approaches can facilitate better coordination of heterogeneous work.
Conclusion
Common information spaces exist in diverse work environments. In our study of
medical work in an intensive care unit, we focus on the use of a CIS in which the
actors are physically co-located. The work of the SICU, like that of many other
workplaces, is detailed, demanding, time-cntical, and involves interaction among
many different groups. At the center of the SICU work is the patient whose health
is dependent on the effective coordination among physicians, nurses, and
pharmacists. However, in many ways, each group's work practices are opaque to
others. Although being physically co-located does help coordinate their activities,
the diverse work practices of these groups prevent them from receiving the full
benefits of co-location. Under these circumstances, our observations in the SICU
point to the important role played by specific information representations in

257
coordinating diverse work activities. For example, the system's ability to present
both retrospective and prospective representations of the same information is
important for coordinating physician and nursing activities. Unlike paper records,
computer systems offer the ability to decouple information from its
representations to help smooth coordination. This decoupling opens up a rich
design space for systems that allow people with different interests, concerns and
work practices to work together effectively.
Acknowledgments
We thank the SICU physicians, nurses, and pharmacists and the HealthStat technical team for
allowing us to observe and interview them We also thank Erin Bradner, Ulnk Chnstensen, Beki
Gnnter, and Suzanne Schaefer for their helpful comments on this paper This work was supported
by the University of California's Life Science Informatics grant #L98-05
References
Bannon, L (2000). "Understanding Common Information Spaces in' CSCW," Paper
presented at the Workshop on Cooperative Organisation of Common Information
Spaces, Technical University of Denmark, August 2000.
Bannon, L. and Bodker, S. (1997). "Constructing Common Information Space,"
Proceedings of the European Conference on Computer-Supported Cooperative
Work £CSCW97.(Lancaster, UK), Dordrecht Kluwer, pp 81-96.
Bardram, J E. (1997)" "Plans as Situated Action: An Activity Theory Approach to
Workflow Systems," Proceedings of the European Conference on Computer-
Supported Cooperative Work ECSClV97.(Lancaster, UK), Dodrecht- Kluwer, pp.
17-32.
Berg, M (1999) "Accumulating and Coordinating: Occasions for Information
Technologies in Medical Work," Computer-Supported Cooperative Work 8 pp.
373-401.
Berg, M. and Bowker, G. (1997). "The Multiple Bodies of the Medical Record. Toward a
Sociology of an Artifact," The Sociological Quarterly 38. pp. 511-535.
Blomberg, J , Suchman, L. and Trigg, R H. (1997) "Reflections on a Work-Oriented
Design Project" in Bowker, G C, Starr, S.L., Turner, W. and Gasser, L. (eds.)
Social Science technical Systems, and Cooperative Work: Beyond the Great Divide.
Mahwah, New Jersey Lawrence Erlbaum Associates.
Bowker, G.C. and Star, S L. (1999). Sorting things out: classification and its
consequences Cambridge, Mass/ MIT Press.
Dounsh, P. and Bellotti, V. (1992). "Awareness and Coordination in Shared
Workspaces," Proceedings of the ACM Conference on Computer-Supported
Cooperative Work CSCW V2.{ Toronto, Canada), New York ACM Press, pp. 107 -
114

258
Dounsh, P, Edwards, K., LaMarca, A and Salisbury, M. (1999). "Presto An
Experimental Architecture for Fluid Information Spaces," ACM Transactions on
Computer-Human Interaction 6 (2)' pp. 133-161
Greenberg, S, Gutwin, C. and Roseman, M (1996) "Semantic Telepointers for
Groupware," OzCHI'96. (Hamilton, NZ), IEEE Press, pp 54-61
Gutwin, C , Roseman, M and Greenberg, S (1996) "A Usability Study of Awareness
Widgets in a Shared Workspace Groupware System," ACM Computer-Supported
Cooperative Work CSCW96. (Cambridge, MA), New York- ACM Press, pp. 258-
267
Heath, C and
Luff,
P (1992). "Collaboration and Control Crisis Management and
Multimedia Technology in London Underground Line Control Rooms," Computer-
Supported Cooperative Work 1 (1-2): pp. 69-118.
Ishii, I, Kobayashi, M and Grudin, J. (1992) "Integration of Inter-Personal Space and
Shared Workspace: Clearboard Design and Experiments," Proceedings of the ACM
Conference on Computer-Supported Cooperative Work CSCW'92. (Toronto,
Canada), New York: ACM Press, pp 33-42.
Lu, I. and Mantei, M. (1991). "Idea Management in a Shared Drawing Tool ,"
Proceedings of the European Conference on Computer-Supported Cooperative
Work ECSCW9I. Dordrecht Kluwer, pp. 97-112
Mark, G, Fuchs, L. and Sohlenkamp,,M. (1997). "Supporting Groupware Conventions
through Contextual Analysis," Proceedings of the European Conference on
Computer-Supported Cooperative Work ECSCW97. (Lancaster, UK), Dordrecht-
Kluwer, pp. 253-268.
Nygren, E., Lind, M , Johnson, M and Sandblad, B (1992). "The Art of the Obvious,"
Proceedings of the ACM Conference on Human Factors in Computing Systems CHI
'92 (Monterey, California), New York: ACM Press, pp 235-239
Parsons, J and Wand, Y. (2000). "Emancipating Instances from the Tyranny of Classes
in Information Modelling,!' ACM Transactions on Database Systems 25 (2)- pp
228-268.
Randall, D. (2000). "What's 'Common' about Common Information Spaces9," Paper
presented at the Workshop on Cooperative Organisation of Common Information
Spaces, Technical University of Denmark, August 2000.
Schmidt, K. and Bannon, L. (1992). "Taking CSCW Seriously. Supporting Articulation
Work," Computer-Supported Cooperative Work 1: pp 7-40.
Simone, C, Mark, G. and Giubbilei, D. (1999) "Interoperability as a Means of
Articulation Work," Work Activities Coordination and Collaboration (WACC99).
(San Francisco, California), pp. 39-48.
Star, S L. and Gnesemer, J R (1989) "Institutional Ecology, Translations and Boundary
Objects Amatuers and Professionals in Berkeley's Museum of Vertebrate Zoology,
1907-1939," Social Studies of Science 19. pp 387-420
Strauss, A, Fagerhaugh, S., Suczek, B. and Wiener, C. (1985) Social Organization of
Medical Work. Chicago: University of Chicago.
Trigg, RH., Blomberg, J. and Suchman, L. (1999). "Moving document collections
online. The evolution of a shared repository," Proceedings of the European
Conference on Computer-Supported Cooperative Work ECSCW99. (Copenhagen,
Denmark), Dordrecht Kluwer, pp. 331-50.