Building an inter-organizational communication network and challenges for preserving interoperability

ArticleinInternational Journal of Medical Informatics 77(12):818-27 · July 2008with139 Reads
Impact Factor: 2.00 · DOI: 10.1016/j.ijmedinf.2008.05.001 · Source: PubMed

The ideal scenario for information technology to bridge information gaps between primary and secondary healthcare and to improve the quality of healthcare in the medication process is to build an interoperable communication network. This type of undertaking requires diverse information systems to be integrated, and central to this are the preservation of data integrity and the integration of different pieces of patient data. OBJECTIVES AND METHODOLOGY: In this study, we focused on sources of challenges to the integration process and to the building of an interoperable communication network. Interviews, document analysis, and observations were conducted to evaluate the integration process in a project that involved medication data communication between primary healthcare providers (i.e., general practitioners and community pharmacists) and secondary healthcare providers (i.e., hospital pharmacists and specialist physicians). The project encountered numerous integration problems, many of which persisted even after extensive technical intervention. An analysis of the problems revealed that they were mostly rooted either in problematic integration of work processes or in the way the system was used. Despite the project's ideal technical condition, the integration could be accomplished only by applying human interfaces. The main challenge to building interoperable communication network does not lie in technical integration. The real problem occurs when the technical linkage is implemented without the work processes being aligned and integrated.


Available from: Habibollah Pirnejad
international journal of medical informatics 77 (2008) 818–827
journal homepage:
Building an inter-organizational communication network
and challenges for preserving interoperability
H. Pirnejad
, R. Bal, M. Berg
Healthcare Governance, Institute of Health Policy and Management, Erasmus University Medical Center, P.O. Box 1738, 3000 DR,
Rotterdam, The Netherlands
article info
Article history:
Received 30 July 2007
Received in revised form
28 March 2008
Accepted 12 May 2008
Medication error
Computer communication network
Information system integration
Background: The ideal scenario for information technology to bridge information g aps
between primary and secondary healthcare and to improve the quality of healthcare in
the medication process is to build an interoperable communication network. This type of
undertaking requires diverse information systems to be integrated, and central to this are
the preservation of data integrity and the integration of different pieces of patient data.
Objectives and methodology: In this study, we focused on sources of challenges to the integra-
tion process and to the building of an interoperable communication network. Interviews,
document analysis, and observations were conducted to evaluate the integration process
in a project that involved medication data communication between primary healthcare
providers (i.e., general practitioners and community pharmacists) and secondary healthcare
providers (i.e., hospital pharmacists and specialist physicians).
Results: The project encountered numerous integration problems, many of which persisted
even after extensive technical intervention. An analysis of the problems revealed that they
were mostly rooted either in problematic integration of work processes or in the way the
system was used. Despite the project’s ideal technical condition, the integration could be
accomplished only by applying human interfaces.
Conclusion: The main challenge to building interoperable communication network does not
lie in technical integration. The real problem occurs when the technical linkage is imple-
mented without the work processes being aligned and integrated.
© 2008 Elsevier Ireland Ltd. All rights reserved.
1. Introduction
In the last decade, concern has grown with regards to under-
standing, identifying, and preventing medical errors before
they harm patients [1–3]. Many of the errors are “medication
errors” [1,4] and cause considerable morbidity and mortality in
different healthcare systems [5–7]. Limited or impaired access
to patients’ medication-related data is the frequent source
of medication errors [8]. This is especially relevant when a
patient is shifted from one level of healthcare to another due
Corresponding author. Tel.: +31 10 408 8811; fax: +31 10 408 9094.
E-mail address: (H. Pirnejad).
to problematic communication between different healthcare
levels. A recent systematic review by Canadian researchers of
22 studies involving 3755 patients showed that errors made in
current medication histories during hospital admission were
“disturbingly common and potentially harmful to patients”.
Mistakes of this kind were seen in up to 67% of the studies,
which were published between 1966 and April 2005 [9].
One of the possibilities that recent ICT developments have
raised is to enhance the quality of healthcare by improving
communication, especially across healthcare boundaries. The
1386-5056/$ see front matter © 2008 Elsevier Ireland Ltd. All rights reserved.
Page 1
international journal of medical informatics 77 (2008) 818–827 819
ideal scenario for ICT to do this and to reduce medication
errors lies in building interoperable
communication net-
works among different care providers, whereby they can work
on the same set of patient data [10–14]. To accomplish this,
the information systems would have to be integrated. Thus
far, however, many complications (e.g., inability to integrate
different parts of patient data and problems in synchroniza-
tion between communicating systems) have been reported
in the integration of diverse information systems and have
resulted in costly but underutilized or failed projects [12,13,15].
Evaluations are necessary to understand more about these
complications and to discover efficient and less costly inte-
gration methods.
For an integration process to succeed, it is necessary to
combine diverse items of patient data stored in a variety of
information systems (data integration) and to prevent data
loss or distortion (preserve data integrity). Many studies thus
far have evaluated the challenges inherent in the replace-
ment of paper-based communication with IT communication
networks [16,17] or in the technical integration of diverse
information systems or different standards for incorporat-
ing patient data [18]. However, as a recent systematic review
revealed, the quality of data integration and the types of error
detection constitute a key point that is missing from most
project publications [15].
We studied an inter-org anizational communication project
in The Netherlands in which primary care providers, the gen-
eral practitioners (GPs) and community pharmacists, shared
medication data with secondary care providers, the hospi-
tal pharmacists and specialist physicians. By evaluating the
attending challenges for data integrity and data integration
throughout the communication network, we sought to answer
the question: How are data integration and data integrity, as
practical achievements of technical integration, attained in
this communication network? Qualitative research methods
were used to evaluate the preservation of data integrity and
the integration of medication data, and to answer the research
question. The study enabled us to extend our knowledge about
building an interoperable communication network between
different healthcare organizations and about the role of tech-
nical integration in its attainment.
2. Study environment
The study environment was Almere, a city near Amsterdam in
the northwestern part of The Netherlands. As in the rest of The
Netherlands, every patient in Almere has a GP as a family doc-
tor who acts as gatekeeper between primary and secondary
healthcare. Every patient also has his/her own community
pharmacist who fills his/her prescriptions. At the primary
care level, together with GPs, pharmacists are responsible for
ensuring the safety of their patients with regard to medica-
In this article, we define “interoperability” as the ability of
parties, either human or machine, to exchange and use data or
information [10].
Fig. 1 Schematic representation of the technical
integration in TUMA.
2.1. Organizational setting
An IT project, named TUMA
, was launched in 2004 in the
Almere region. Its purpose was to exchange patient medica-
tion records
between primary and secondary care providers.
Almost all of the 115 GPs and the 17 community pharma-
cists from the primary care side were involved. Representing
the only regional hospital, the Flevo Hospital, the pharmacy
department was the main participant in the project.
2.2. Network details
In Almere’s primary healthcare setting, nearly all GPs use
an information system, and all community pharmacists use
information systems from the same vendor (Medicom
, respectively). There is a common server
the “Local Health Server” for all Medicom and Pharma-
com systems in the region. By sharing the server, GPs and
community pharmacists have built an application-specific
communication network through which they can easily share
and work on elements of their patients’ data, including med-
ication records. TUMA is intended to connect this regional
primary care network to the only general hospital in the
region. The hospital pharmacy department uses an informa-
tion system (Zamicom
) similar to Pharmacom, again from
the same vendor. TUMA, therefore, is building a communi-
cation network in what could be called an ideal situation,
as compared to other regions in The Netherlands, which are
characterized by their “patchwork” of information systems
TUMAs network is a “Virtual Private Network” (VPN) con-
necting the Local Health Server to the Zamicom server. At
the center of this communication network (eHealthNet) is a
Central-Patient-Index system to ensure a one-by-one match of
patients’ records between primary and secondary healthcare
(Fig. 1). Through this network, patients’ medication records,
including current medications and a summary of medical
TUMA stands for trans-mural exchange of medication data in
This includes patient medication data as well as a summary of
patient medical records.
Page 2
820 international journal of medical informatics 77 (2008) 818–827
Fig. 2 The timeline of the project and research in TUMA. The blocks with diagonal lines show the dates when the system
was tested by hospital pharmacists.
records, is exchanged by an EDIFACT
-based communication
protocol named OZIS-DWA1.0. Communication by way of this
protocol is rendered operational by an intermediate system
(the OZIS sever) from another major local vendor.
Each time a patient is admitted to hospital an enquiry
is sent to primary healthcare. The reply is returned through
the network and contains the patient’s medication record(s).
Data from primary healthcare is integrated into Zamicom.
To create the message, two main ‘drug-related’ and ‘disease-
related’ record structures are considered in Pharmacom for
the period of the last 15 months. From drug-related records,
‘delivered medication records’ are listed into the message if
they are indicated by code “C” (continuous medication) or “P”
medication) in front of them; all the “current medica-
tion” (coded with “*”) are listed unless they have been stopped
manually (active stop) in the system. Drug allergies and con-
traindications are also picked up from here into the message.
From disease-related records, patient chronic conditions and
co-morbidities such as diabetes are listed. Moreover, patient
characteristics, including date of birth and gender are also
included into the message.
3. Methods
In this study, building a communication network was consid-
ered as developing a “loop” within which care providers try to
communicate in order to link and complement each other’s
medication work [20,21]. If this communication network has
to be interoperable, this loop has to contain communicative
in its core in order to help care providers to bind
together (i.e., articulate) their work and to build mutual under-
standing upon the exchanged information [21,22]. The word
“loop” emphasizes the end-to-end closure of the medica-
tion data communication within the network. It emphasizes
that patients’ medication records have to be circulated and
updated by different care providers (in a timely manner) with-
out suffering any loss or distortion. Evaluating the medication
data transaction within this frame of reference helped us to
consider and to evaluate the factors either related to the
system or to its users or to the implementation environment
The electronic data interchange for administration, communi-
cation and transport.
A PRN medication is a drug used by a patient whenever symp-
toms of the disease occur.
Here, norm means a principle of a right action binding upon
the members of a group and serving to guide, control, or regulate
proper and acceptable behavior (Merriam Webster).
that hamper the preservation of data integrity or the inte-
gration of different pieces of data.
Launched in 2004, TUMA went live in March 2005. We were
involved in its evaluation as an external research group; this
evaluation was in line with our research interest in under-
standing the development of local communication networks
in the Dutch healthcare system [14,19]. The evaluation took
place in two stages; the first stage started in January 2005,
before the project became operational. During this stage our
evaluation was mostly focused on project level. The second
stage started in July 2005 during which we mostly focused on
work-floor (Fig. 2). Almost 1 month later the project ran into
unforeseen problems, which are briefly discussed in Section 4.
We also had to stop data collection from work-floor and shift
the evaluation focus to project level in order to find out the
reasons of this halt. Almost 8 months later, in April 2006, after
the problems had been dealt with, we were able return to the
field and continue our data collection. The study finished in
July 2006 (Fig. 2).
Qualitative research methods were used to evaluate the
attainment of data integration and saving data integrity in
TUMA. We conducted interviews in both stages of the study,
and in the second, we also analyzed documents and made
observations. Among the care providers in the Almere region,
community and hospital pharmacists played the main role in
implementing and testing the TUMA network. Together with
the project team, pharmacists were the main stakeholders in
TUMA and thus were the focus of our interviews and observa-
The first author interviewed the project leader, two project
managers, two community pharmacists, and four hospital
pharmacists. In total, 10 interviews were conducted, each last-
ing 1.5–2h; four of the interviews were during the first stage
and six interviews during the second. The semi-structured,
in-depth interviews were one-on-one and face-to-face. The
interviews were audio taped, transcribed, and coded according
to Bowling [23]. They were then integrally analyzed for emerg-
ing trends based on Grounded Theory. Interviews conducted
in the first stage of the research provided useful information
about the study context, medication data communication and
the information gaps prior to TUMA, the baseline measure-
ments. In the second stage, we asked the interviewees about
the changes in their work, the improvement, and the prob-
lems; after the network was tested, we also asked their opinion
about reasons for the problems in the test results. These inter-
views helped us to discover and to deepen our understanding
of changes brought about by TUMA, to interpret the results of
the network’s tests, and to recognize the challenges faced in
the effort to create and maintain interoperability.
Page 3
international journal of medical informatics 77 (2008) 818–827 821
To understand how community and hospital pharmacists
deal with medication data, the first author observed their work
for approximately 6 h during the second stage of the project. At
the primary healthcare level, the observation focused on how
community pharmacy technicians entered data into infor-
mation systems; at the hospital, it focused on data entry
in Zamicom and data acquisition from primary healthcare.
During the observations, pharmacists were asked about their
experiences with the system and the reasons for commonly
encountered problems. Notes were made and analyzed in the
same way as the data from the interviews.
After the network was set up and tested technically, hos-
pital pharmacists evaluated the functionality of the network
as part of the implementation improvement process. The first
test was done in November 2005 by requesting from one com-
munity pharmacist the medication records for 100 randomly
chosen primary care patients. The results of the enquiry in
Zamicom were then compared with the original data in Phar-
macom. The results were evaluated and commented upon by
a hospital pharmacist, the community pharmacist, and the
system vendor with respect to reasons for message incon-
sistencies with the original data as well as other problems.
After necessary interventions were carried out to improve the
network performance, a second test was conducted in April
2006, again by requesting medication records for 100 randomly
selected primary care patients (excluding those that had been
chosen for the first test) from the same community pharma-
cist. As the second test was performed on the medication
records from the same community pharmacist, the result of
the test could reflect the effect of improvement measures. We
analyzed the results of the two tests both qualitatively and
quantitatively (descriptive analysis), drawing upon the com-
ments of the parties on the test results and the pre- and
post-implementation interviews with pharmacists. Four main
sources of problems in the test results were defined and the
problematic items of both tests were then distributed among
the authors for final classification.
4. Results
First, results of the pre-implementation study are presented to
depict the pitfalls and information gaps in the old medication
data communication. Second, TUMA and its effect on bridg-
ing the information gaps and improving the communication
are presented, focusing on the test results and their analysis.
Third, important unforeseen problems and conflicts related to
the articulation work and responsibility distribution between
the involved parties are presented, focusing on their impact
on TUMA.
4.1. Information gaps in the medication data
communication loop prior to TUMA
At the primary care level of the medication process loop,
GPs were entering patient data into their information sys-
tem based on “episodes”.
Except for a diagnosis, which was
For GPs, the unit of observation is called “episode” and refers
to a patient with a specific medical problem over time.
coded by ICPC
-2, and medication data, which was coded
by ATC
-classification, most data entries into Medicom were
made in free text format. Community pharmacists also used
the same ATC-classification to code data. GPs and commu-
nity pharmacists shared patients’ medication records through
their information systems; however, direct communication
(e.g., phone calls) was also common between them. GPs and
community pharmacists in The Netherlands already commu-
nicated to a certain extent, but Almere is unique with respect
to the extent of communication and collaboration that exists
among GPs and community pharmacists (please also see Sec-
tion 2).
“We in primary healthcare always check each other’s work
[on patient medication]. This is normally done both by
our information system and by direct observation of the
prescriptions. If we see there is something wrong in the
prescriptions, we just pick up the phone and call the GP for
further clarification. Every time a prescription is filled, the
information system generates an automatic message that
updates the records of the prescribing GP. [A community
Hospital pharmacists stated that, in most cases, GP refer-
ral letters did not contain medication data for hospital care
providers. This created the first information gap in the commu-
nication loop between primary and secondary patient care. To
fill this gap, patients themselves were frequently the source
of their medication-related information when they arrived at
the hospital. However, it was common that patients did not
remember all the types of medication, or became confused by
look-alike drug names. It was therefore possible for hospital
care providers to fail to obtain an accurate medication his-
tory from patients. Some part of the patient’s data might be
missed when it was handed over or when it was transferred
from paper-based forms into Zamicom. A hospital pharmacist
reported the following:
A nurse failed to register a drug name (Methoteroxate)
while she was taking the drug history from a patient, only
because the drug had been used at intervals. The patient
then got cystitis during his hospitalization and a physician
prescribed Cotrimoxazole. When Cotrimoxazole treatment
started for the patient, his condition suddenly worsened
with leucopoenia and other signs of Methoteroxate toxicity.
Such a dangerous condition happened because the nurse
failed to take an appropriate drug history from the patient.
Our information system failed to react to this drug interac-
tion because Methoteroxate had not been registered in it.
[A hospital pharmacist]
When discharged from the hospital, a patient was given
prescriptions that had to be filled by a community pharmacist.
In addition to this early contact between secondary and pri-
mary healthcare, additional information including diagnosis
and medication was sent to primary care providers by means
of a discharge letter. Community pharmacists claimed that
this process usually took a long time. Moreover, in most cases,
International classification for primary care.
Anatomical therapeutic chemical classification system.
Page 4
822 international journal of medical informatics 77 (2008) 818–827
Table 1 Different categories and frequency of problematic items in the two tests
Type of problematic items Frequency in the 1st test Frequency in the 2nd test
Missing items
Currently used chronic medication did not appear in the
7 (12.3%) 4 (7.3%)
Stopped chronic medication did not appear in the message 4 (7%) 4 (7.3%)
Administration data did not appear in the message 2 (3.5%) 0 (0%)
Potential (PRN) medication did not appear in the message 1 (1.7%) 6 (10.9%)
Temporary medication did not appear in the message 2 (3.5%) 3 (5.4%)
Medication appeared in double form (both in generic and
commercial forms)
3 (5.3%) 0 (0%)
Stopped chronic medication appeared without the stop date 3 (5.3%) 13 (23.6%)
Wrongly appeared items
Patient current or old medical condition records appeared
in the wrong form
4 (7%) 0 (0%)
Non-chronic medication from past medication history
appeared in the message
17 (29.8%) 10 (18.1%)
Non-medication-related information appeared in the
1 (1.7%) 10 (18.1%)
Unknown medication appeared in the message 5 (8.7%) 2 (3.6%)
Actively stopped medication in Pharmacom appeared in the
8 (14%) 3 (5.4%)
Total 57 (100%) 55 (100%)
when patients contacted their GPs after discharge, the GPs
wereunaware of the most recent changes in the patients’ med-
ication. This delay created the second information gap in the
medication data process loop.
After discharge [from the hospital], most patients fre-
quently don’t know what to do with the drugs they were
using before hospitalization. They don’t know whether
they have to take them together with their discharge med-
ication or to stop using them. They usually ask us because,
as community pharmacists, we are supposed to be respon-
sible for their medication safety. But we cannot help them,
because we do not know the reasons for the changes. [A
community pharmacist who was also project manager]
Therefore, patients were considered a link, filling the
information gaps between primary and secondary healthcare
levels. TUMA, as its authorities claimed, replaces this weak-
est link in the medication data process loop and builds an
interoperable network.
4.2. Medication data integration in TUMA
Despite the fact that similar information systems were applied
to both sides of the network, the project ran into trou-
ble mainly due to software compatibility problems. The
project stopped, and it took time until the necessary soft-
ware patches were developed and tested successfully. As
soon as software patches were considered working prop-
erly, the project started to run again. The first test was
performed to evaluate the network operation before the
application went live again. However, the results were sur-
prising for project authorities; several problems were detected
in the data transacted to Zamicom, and the number of
these problems indicated the scale of the integration difficul-
“One main reason that the project fell behind in its
timetable lay in the problems we had in integrating the
medication data from primary care to Zamicom; it was a
software functionality problem. To solve it, we consulted
with people from other projects, who have already worked
with the same method of data transaction. To our surprise,
we learned that they only use the system to acquire data
from primary care and then transfer it manually to Zami-
com. [Project leader]
In total, 59 problematic items from 32 medication records
were identified in the first test by comparing the data that
emerged through the network with the original data in
Pharmacom. These problematic items were evaluated and
commented upon by one hospital pharmacist, by the com-
munity pharmacist whose records were involved, and the
software vendors. We excluded two of the detected problems
from our study because the community pharmacist, the hos-
pital pharmacists, and the system vendor could not agree as to
whether they should be considered communication problems.
We then analyzed the results and comments for the remain-
ing 57 items and allocated them into 11 groups. Table 1 shows
that the problems were of different types, including data that
did not make it through the network, discrepancies between
the transferred data and its origin in Pharmacom, and the
transference of inappropriate data.
Extensive measures were taken to reduce the prob-
lems: software patches were developed, coding system was
improved, and patient medication records in Pharmacom were
revised. In addition, it was decided that a free text form of
patients’ records should accompany every message through
the system. After these changes were implemented, the net-
work was supposed to work properly, so its performance
was tested again by requesting medication records for 100
randomly chosen patients. This time, despite all the above-
mentioned improvements, the total number of problematic
Page 5
international journal of medical informatics 77 (2008) 818–827 823
Table2–Categories and frequency of the problems’ sources in the first and the second tests
Sources of
problematic items
Number of the
problems in the 1st test
Frequency among the total
problematic items (%)
Number of the
problems in the 2nd test
Frequency among the total
problematic items (%)
Faulty application of the coding system 11 19.3 19 34.5
Faulty application of software 22 38.6 30 54.5
Coding system deficiency 4 7 1 1.8
Software problem 20 35.1 5 9.1
Total 57 100 55 100
Note: one hundred records were evaluated in every test and some of the records have more than one problem.
items detected was 55 in 14 medication records. Table 1
shows the problematic items distribution among the differ-
ent categories as well as the changes that occurred after the
improvement measures.
To determine the sources of the problematic items in both
tests, we subsequently analyzed the interviews and the com-
ments made by interested parties on the test results. With
respect to their sources, we were able to allocate the prob-
lems detected in the two tests into four general categories:
those due to coding system deficiencies; those due to soft-
ware failures; those related to the faulty application of the
coding system; and those related to the faulty application of
Problems due to coding system deficiencies included items
that were a result of code mismatches or due to different gran-
ularity levels of the codes in the two information systems.
For example, items such as bandages, syringes, and catheters
were frequently registered in Pharamcom, while Zamicom had
no registration code for them. When such data was received
by Zamicom, it produced an error of “unknown medication”, a
problem that required the attention of a pharmacist. In Phar-
macom, on the other hand, both types of diabetes (types I and
II) were given the same code, while in Zamicom they were
represented by different codes. The opposite scenario held
for “intention to become pregnant” and “being pregnant”. In
Pharmacom, these conditions were each coded differently; in
Zamicom, only one code represented both of them.
Problems due to software failures were considered solved
by improving the current or the subsequent version of the
software. Many software failures were related to problems in
creating the message. For example, the program was picking
up temporary medications from Pharmacom that did not have
an end-date registered in the system but that were supposed
to be stopped after 14 days (theoretical end-date). Since by the
first test Pharmacom was not able to calculate and register
this theoretical end-date, the message was picking up those
medications even if they were supposed to have been stopped
by the date the message was created. As another example, if
a medication was registered once by its trade name and later
by its generic name, both were incorporated in the message.
Problems due to the faulty application of the coding system
included: applying a wrong code, a failure in coding while
entering patient data into Medicom or Pharmacom, or a failure
to update the coding status. In such instances, the neces-
sary data would be missed, while unnecessary data would
be appeared in the enquiry response to Zamicom. For exam-
ple, in the test results, items were missing from the enquiry
responses because GPs failed to code the episodes or coded
them wrongly.
“One important issue [in filling the message] is the end-
date of usage for medications. This is especially important
for temporary and PRN medications and cannot be done
by our information system if they were not coded properly.
If GPs do not code temporary medication properly, there
will be no clue that that medication is to be stopped after
a certain time. As a result, hospital pharmacists receive a
long list of different drugs in their enquiry and will become
confused as to which one is still used and which one has
already been stopped. [A community pharmacist]
Page 6
824 international journal of medical informatics 77 (2008) 818–827
Finally, many problematic items in the test results could
be considered a result of the faulty application of software.
Those parts of medication data that had to do with delivering
and administration were managed by community pharma-
cists. However, in Zamicom it was possible to enter something
else in the space where a drug’s name had to be registered.
Hospital pharmacists experienced communication problems
especially after the second test because, for instance, com-
munity pharmacy assistants used this information system’s
possibility inappropriately: for example, writing “the status is
OK” instead of inserting the name of the medication. Another
example concerned instances of pharmacy technicians fail-
ing to enter an end-date for non-current items; this led to the
appearance of this medication in the enquiry response.
“The chronic medications are labeled with code C at the
beginning of their listing in the information system. How-
ever, we know that this code does not mean that a drug
has to be continued forever. Some chronic medications are
discontinued or switched to other ones after a while. If
community pharmacists do not update the drug’s status
in their information system, the discontinued drugs will
be presented in response to our enquiry. This is something
that the community pharmacists usually forget to do. [A
hospital pharmacist]
The sources of the problematic items were also analyzed
quantitatively, and the result is shown in Table 2. The analy-
sis demonstrated that the most common sources of problems
in the first test were either related to software (35.1%) or to
faulty application of software (38.6%). In the second test, the
total number of the problematic items changed slightly (57 vs.
55), while the reasons for the problems shifted considerably.
Problems due to software or due to coding system decreased
considerably; ho w e ver, the rate of the problems concerning
faulty applications of the coding system (34.5% vs. 19.3%) and
software (54.5% vs. 38.6%) increased tremendously.
4.3. Unforeseen problems and conflicts
An important responsibility of community pharmacists in
TUMA was to code the administration of drugs in their infor-
mation system precisely. However, this could not be done
unless GPs coded medications rightly in their information
systems. Therefore, the medication coding work became a
joint work and responsibility between GPs and community
pharmacists. The community pharmacists in TUMA ran into
trouble with respect to this coding responsibility since it was
an unforeseen and unmanaged conflict between GPs and com-
munity pharmacists.
“For example, if a GP prescribes a medication that has to be
used at half an ordinary dose and does not code it properly
in his information system, our system will calculate it, for
example, for one and half months and after that time, the
system will show that the drug is stopped by the patient.
Yet, the patient has the drug at home and will use it for
a further one and half months. This is something that the
system cannot do automatically; it has to be done man-
ually by GPs. However, there is a problem at the moment
with respect to convincing GPs to accept this responsibil-
ity. They argue that they are not interested in doing this
work, and excuse themselves for being busy. [A commu-
nity pharmacist after the second test]
One responsibility of the hospital pharmacy department
in TUMA is to keep the data transmission line from primary
healthcare to the hospital wards operational. In order to do
this, and to integrate primary with secondary healthcare data
as well as the communication between the information sys-
tems in TUMA, specialist attention and manual steps are
required by hospital pharmacists. As practically it was not
possible to correct all old medication records in the primary
healthcare, appearing problematic items were expected to per-
sist. There was a great concern that if hospital pharmacists
did not correct the problematic items in the messages, spe-
cialist physicians in the wards would not appreciate using the
system. On the other hand, there was no exact idea what por-
tion of the medication data had to be observed, checked, and
improved by hospital pharmacists. And hospital pharmacists
were concerned about the time and effort they were forced to
spend on the process, especially as this type of contribution
and the role of hospital pharmacists was not anticipated by
the project team.
One ambition of the TUMA project team was to replace
the patient the weakest link in the medication data transac-
tion loop with ICT. Nevertheless, the contribution of patients
in saving the integrity of data and in integrating medication
data is valuable. We discovered at least three reasons in TUMA
why this is still the case. First, some patients have to fill their
prescriptions at a pharmacy other than their designated com-
munity pharmacy. Second, over-the-counter drugs (OTCs) are
not registered in any information system and thus their names
are requested by medical specialists. Last but not least, the
adherence of patients to the medication administration plan
from primary care needs to be verified. These aspects of a
patient’s medication history are important and can only be
obtained through a patient’s involvement in the medication
data communication loop.
5. Discussion
The results of our study demonstrate that the technical inte-
gration of information systems is necessary but it is not
enough to save data integrity and to integrate various pieces of
patient data during the communication process. Other factors
are important. The Almere situation was ideal for the techni-
cal integration of a healthcare information system, since there
was a well-integrated communication network at the primary
healthcare level and all participants use same-vendor infor-
mation systems. Only one general hospital was in the region,
and it also used a very similar information system and the
same standards for data coding. Nevertheless, a number of
communication problems arose in the course of testing the
TUMA network. The persistence of the problems even after
extensive technical improvements made it clear that the cre-
ation of technically integrated information systems is not a
straightforward solution to achieving data integration and to
preserving data integrity. Moreover, results of the second test
demonstrated that while the total number of communication
problems did not change in comparison to the first, most of
Page 7
international journal of medical informatics 77 (2008) 818–827 825
the problems shifted toward faulty application of software and
coding system. This implies that technical integration is not to
blame for the problems encountered in TUMA; the work pro-
cesses had to be integrated and the work routines and habits of
users had to be improved if data integration and the preserving
of data integrity were to be accomplished.
In many studies, the heterogeneity of information systems
and standards are referred to as main impediments to building
interoperable communication networks [13,18,24]. Our study,
however, shows that social and organizational factors are also
paramount. Lack of attention to how the technological arti-
fact will affect and be affected by the organization in which
it becomes embedded lies at the core of many technologi-
cal failures [11]. Hanseth et al. [25] argue that practices and
technologies co-develop over time and adapt to each other,
creating a socio-technical network. In the mutual effect of
technology and work practices, one changes the other. This
means that “technology changes work practices, which in turn
changes how the technology is used, which leads to changes in
the technology, which induces new changes in work practices,
and so on” [11]. It means that alterations in either the tech-
nical or the social realm will somehow require alterations in
the other. Building an interoperable communication network
through the integration of information systems, therefore,
requires changes in the organization of care practices and the
way people use the system. It is clear that the registering and
coding of medication-related data in primary care have to be
adapted to facilitate the retrieval and use of this data by a med-
ical specialist at the hospital. In TUMA, however, we found
idiosyncratic uses of the information and coding systems by
GPs and community pharmacists. Although those uses were
appropriate for their purposes and saved considerable time,
they were considered inappropriate for hospital pharmacists,
since they resulted in communication problems that required
the pharmacists to pay special attention and to try to solve the
problems. The second test showed that the amount of “non-
medication-related information” was increased considerably.
This, we think, was due to persisting inappropriate application
of the information systems, which led the community phar-
macist’s staff to make more mistakes and which, for example,
resulted in information (e.g., the status is OK) being inserted
in inappropriate places. Moreover, since TUMA-required type
of articulation work was not performed to integrate the work
of different care providers across the network, the commu-
nity pharmacy staff could not fill in all the missing data in
their information system without the GPs’ cooperation. For
example, they inserted the missing code for chronic medica-
tions but could not enter the end-date of those medications.
As a result, the number of “missing end-date of an old chronic
medication” was increased dramatically in the second test.
Whereas most technical (e.g., software) problems seemed
solvable, the use of the system implied the need to create
extra checks in the socio-technical network. In the second
test, despite all the improvement measures, the total num-
ber of problematic items remained almost the same, mainly
because the number of problems due to non-technical rea-
sons increased. This was an unexpected result. We believe the
improvement measures after the first test could improve some
of the software functionality problems and the coding system
deficiencies; however, at the same time they did not change
or they even added to the non-technical problems. Interest-
ing is that data integration as a practical accomplishment in
the TUMA project was only achieved by hospital pharmacists
checking and improving primary care medication data. Our
study therefore confirms that it is not possible to build an
interoperable communication network and to fill the informa-
tion gaps merely by the technical integration of information
systems; the work processes of communicating care providers
in the network also have to be integrated. Thus, the technical
linkage is not the real problem in integration; it is that the
technical linkage is implemented without the work processes
being aligned and interconnected.
An increasing number of publications describe projects
that integrate data from multiple information systems. How-
ever, as Cruz-Correia et al. [15] argue, one key omission in
most of these publications is the mention of any type of error
detection; this leaves the effect of integration processes on
data quality inadequately researched. In this paper we elab-
orated upon the problematic items that TUMA encountered
in the course of its information system integration. In con-
trast to many of the projects that build inter-organizational
communication networks upon diverse information systems,
in this study we reported on a condition of similar infor-
mation and coding systems. Hence, the problems discussed
in this paper are most likely to occur in many communica-
tion projects in which there are less ideal a priori conditions.
Though our study involved a situation quite different from
that of Ellingsen and Monteiro [13] and Monteiro et al. [26],
there are many similarities in the kinds of problems encoun-
tered. While some of these problems might be resolved over
time (e.g., GPs become accustomed to a coding routine that
is appropriate for other users), some of them (e.g., the work
load of hospital pharmacists) are likely to persist or to trans-
form into other difficulties that may compromise instead of
improve patient safety [27]. This is not to say that building an
interoperable communication network should not be strived
for, but the socio-technical links that exist within the pro-
cess of integrating information systems in healthcare must
be taken into account. As a fully automated process is still
far from realization in healthcare settings, human interfer-
ence may be reduced or transformed, but it is still necessary
in many instances of information communication processes:
for instance, for the validation of exchanged data or the clari-
fication of ambiguous information.
One important issue concerns the role and the position
of patients. We saw that patients, prior to TUMA, played an
important part in transferring their medication data from one
level of healthcare to the other; their role was that of a mes-
senger [28]. However, although patients were able to provide
information to care providers in a timely fashion, it was known
that their information was not always reliable. Studies have
shown that patients get it wrong 28–38% of the time [28,29].
Nevertheless, after the implementation of TUMA their role
was still dominant in verifying and updating their medica-
tion data and thus in preventing medication errors. Patients
are an integral part of the medication data communication
loop and their verification role has to be considered in every
project [29]. For example, in a study by van der Kam et al. [16]
on medication data exchange between GPs and pharmacists,
there was no difference between electronic and paper-based
Page 8
826 international journal of medical informatics 77 (2008) 818–827
communication with respect to the drugs “reported only by
patients”. Therefore, ignoring the role of a patient in complet-
ing and updating medication data can lead to the integrity of
the data being damaged. However, further studies are required
to conceptualize an appropriate place for the contribution of
patients in inter-organizational communication networks.
This study had several limitations. First, the study on TUMA
only involved taking patient medication records from primary
healthcare to the hospital. However, it is clear from our anal-
ysis that similar problems, only on a larger scale, will be
encountered when communication from the hospital to pri-
mary healthcare becomes operational. Second, the hospital
pharmacist information system was an application shared
with the nearby hospital in the city of Lelystad. This posed
problems for the project; for example, there was resistance to
carrying out the required adaptations to the system’s server at
Lelystad’s hospital. We also observed that logging into Zam-
icom from the Flevo hospital was sometimes difficult and
time consuming. Third, defining categories for the problem
sources was a challenging issue. The problems were in many
instances socio-technical, while the categories drew a line
between social and technical issues. For example, the inap-
propriate use of a data field in Pharmacom can be considered
a user problem or a software problem that does not prevent
this user mistake. To cope with this challenge, we carefully
defined different categories and remained with these defi-
nitions in allocating the different communication problems.
Any interpretation of the test results is limited to the defini-
tion of the categories. Fourth, although the study brought up
the potential sources of medication errors during the commu-
nication process, quantifying and determining their clinical
importance were not of immediate relevance here.
6. Conclusion
With regard to the exchange of medication information, the
safeguarding of data integrity and the integration of different
pieces of medication data are crucial to create and main-
tain the interoperability of healthcare providers. Our study
shows that technical integration is not the real problem in
an interoperable communication; the problem emerges when
the technical linkage is implemented without the work pro-
cesses being aligned and integrated and the work routines
being improved. Moreover, a thorough communication solu-
tion must address a way to combine the role of patients with
that of other care providers in the communication network.
The authors would like to thank Arjen P. Stoop (our research
colleague in the first stage of the study), and Jelle Dijkstra and
Rob Linde (TUMA project managers) for their cooperation dur-
ing our study. Our special thanks go to Peter de Bruijn, the
former project leader, for his positive attitude and help while
we conducted this research.
Contribution: The first and the second authors were respon-
sible for designing and implementing the research. The first
author also had the main role in data collection and analy-
sis, and drafting and revising the paper. The second and the
Summary points
What was already known
In healthcare, building inter-organizational commu-
nication network through integrating information
systems has been costly but underutilized or failed.
Problems with technical integration of diverse infor-
mation systems especially those related to data
integrity and data integration have been considered as
the main impediment.
Lack of attention to how the technological artifact will
affect and be affected by the organization in which it
becomes embedded lies at the core of many techno-
logical failures.
What is learned from this study
Even in a technically ideal condition, problems with
data integrity and data integration will persist.
In building an interoperable communication network,
integrating work processes of the communicators
throughout the network and improving the way they
apply information systems precede technical integra-
Patients’ central role in data integration and preserving
data integrity need to be addressed in any IT configu-
rations for building medication data communication
third authors were both involved in supervising the research
and managing the cooperation with project authorities. They
were also involved in analysis of the data; drafting, revising,
and scientific approval of the paper.
[1] S.N. Weingart, R.M. Wilson, R.W. Gibberd, B. Harrison,
Epidemiology of medical error, Br. Med. J. 320 (2000) 774–777.
[2] Committee on the Quality of Health Care in America,
Crossing the Quality Chasm: A New Health System for the
21st Century, National Academy Press, Washington, DC,
[3] Committee on the Quality of Health Care in America, To Err
is Human, National Academy Press, Washington, DC, 2000.
[4] M. Lisby, L.P. Nielsen, J. Mainz, Errors in the medication
process: frequency, type, and potential clinical
consequences, Int. J. Qual. Health Care 17 (1) (2005) 15–22.
[5] N. Barber, M. Rawlins, B.D. Franklin, Reducing prescribing
error: competence, control, and culture, Qual. Saf. Health
Care 12 (Suppl. 1) (2003) i29–i32.
[6] B. Dean, M. Schachter, C. Vincent, N. Barber, Prescribing
errors in hospital inpatients: their incidence and clinical
significance, Qual. Saf. Health Care 11 (4) (2002) 340–
[7] P. van den Bemt, A.C.G. E. National Medication Safety Study
in the Netherlands, Research report, Utricht: Netherlands
Association Hospital Pharmacists (NAZA) and the Order of
Medical Specialists (OMA), accessible from, 27th November 2006.
Page 9
international journal of medical informatics 77 (2008) 818–827 827
[8] L.L. Leape, A.I. Kabcenell, T.K. Gandhi, P. Carver, T.W. Nolan,
D.M. Berwick, Reducing adverse drug events: lessons from a
breakthrough series collaborative, Jt. Comm. J. Qual. Improv.
26 (6) (2000) 321–331.
[9] V. Tam, R. Knowles, P. Cornish, N. Fine, R. Marchesano, E.
Etchells, Frequency, type and clinical importance of
medication history errors at admission to hospital: a
systematic review, CMAJ 173 (2005) 510–515.
[10] C.N. Mead, Data interchange standards in healthcare
IT—computable semantic interoperability: now possible but
still difficult, do we really need a better mousetrap? J. Health
Care Inf. Manage. 20 (1) (2006) 71–78.
[11] R.L. Wears, M. Berg, Computer technology and clinical work:
still waiting for Godot, JAMA 293 (10) (2005) 1261–1263.
[12] G. Ellingsen, E. Monteiro, A patchwork planet: integration
and cooperation in hospitals, Computer Supported
Collaborative Work 12 (2003) 71–95.
[13] G. Ellingsen, E. Monteiro, The slight surprise of integration,
in: C. Sørensen, Y. Yoo, K. Lyytinen, J.I. deGross (Eds.),
Designing Ubiquitous Information Environments:
Socio-technical Issues and Challenges, Springer, New York,
2005, pp. 261–274.
[14] H. Pirnejad, R. Bal, A.P. Stoop, M. Berg, Inter-organisational
communication networks in healthcare: centralised versus
decentralised approaches, Int. J. Integr. Care 7 (2007) e14.
[15] R.J. Cruz-Correia, P.M. Vieira-Marques, A.M. Ferreira, F.C.
Almeida, J.C. Wyatt, A.M. Costa-Pereira, Reviewing the
integration of patient data: how systems are evolving in
practice to meet patient needs, BMC Med. Inform. Decis.
Making 7 (2007) 14.
[16] W.J. van der Kam, P.W. Moorman, M.J. Koppejan-Mulder,
Effects of electronic communication in general practice, Int.
J. Med. Inform. 60 (1) (2000) 59–70.
[17] T. Schabetsberger, E. Ammenwerth, S. Andreatta, G. Gratl, R.
Haux, G. Lechleitner, et al., From a paper-based transmission
of discharge summaries to electronic communication in
health care regions, Int. J. Med. Inform. 75 (3–4) (2006)
[18] R. Lenz, M. Beyer, K.A. Kuhn, Semantic integration in
healthcare networks, Int. J. Med. Inform. 76 (2–3) (2007)
[19] R. Bal, F. Mastboom, H.P. Spiers, H. Rutten, The product and
process of referral optimizing general practitioner–medical
specialist interaction through information technology, Int. J.
Med. Inform. 76 (Suppl. 1) (2007) 28–34.
[20] R. Medina-Mora, T. Winograd, R. Flores, F. Flores. The Action
Workflow Approach To Workflow Management Technology,
CSCW, ACM Conference on Computer-Supported
Cooperative Work, Toronto, 1992, pp. 281–288.
[21] H. Weigand, A. de Moor, Workflow analysis with
communication norms, Data Knowl. Eng. 47 (3) (2003)
[22] T. Windgrad, Designing a new foundation for design,
Commun. AMC 49 (5) (2006) 71–73.
[23] A. Bowling, Research Methods in Health, Open University
Press, Philadelphia, 2002.
[24] S. Garde, P. Knaup, E.J.S. Hovenga, S. Heard, Towards
semantic interoperability for electronic health records,
Methods Inf. Med. 46 (2007) 332–343.
[25] O. Hanseth, E. Monteiro, M. Hatling, Developing information
infrastructure: the tension between standardization and
flexibility, Int. J. Sci., Technol., Human 11 (4) (1996) 407–426.
[26] E. Monteiro, Integrating health information systems: a
critical appraisal, Methods Inf. Med. 42 (4) (2003) 428–432.
[27] E. Balka, N. Kahnamoui, K. Nutland, Who is in charge of
patient safety? Work practice, work processes and utopian
views of automatic drug dispensing systems, Int. J. Med.
Inform. 76S1 (2007) 48–57.
[28] W. van der Kam, B. Meyboom de Jong, T. Tromp, P. Moorman,
J. van der Lei, Effects of electronic communication between
the GP and the pharmacist: the quality of medication data
on admission and after discharge, Family Practice 18 (6)
(2001) 605–609.
[29] Committee on the Quality of Health Care in America,
Preventing Medication Errors, Institute Of Medicine (IOM) of
the National Academies, Washington, DC, July 2006.
Page 10
    • "physician communication (Aarts and Koppel 2009, Berner et al. 2005, Harrison et al. 2007, Niazkhani 2007, Pirnejad et al., 2008b, Vaaste 2007). And so, the pragmatics of how to manage the highly cognitive, collective and ad hoc nature of medical work remains a persistent concern. "
    [Show abstract] [Hide abstract] ABSTRACT: The design and implementation of an electronic medical record system pose significant epistemological and practical complexities. Despite optimistic assessments of their potential contribution to the quality of care, their implementation has been problematic, and their actual employment in various clinical settings remains controversial. Little is known about how their use actually mediates knowing. Employing a variety of qualitative research methods, this article attempts an answer by illustrating how omitting, editing and excessive reporting were employed as part of nurses' and physicians' political efforts to shape knowledge production and knowledge sharing in a technologically mediated healthcare setting.
    Full-text · Article · Mar 2015 · Sociology of Health & Illness
    • "One of the major design challenges identified related to lack of consistency in information layouts and limited considerations of information semantics in the system design. Striving for information consistency in health care settings is not only a problem of technical integration, but also of aligning inter-organisational work processes343536. Achieving the right degree of information consistency therefore requires input from all stakeholders about residents' medication information to negotiate how to stabilise the content and meaning of coordinative artefacts like medication charts [35,37,38]. "
    [Show abstract] [Hide abstract] ABSTRACT: Electronic medication administration record (eMAR) systems are promoted as a potential intervention to enhance medication safety in residential aged care facilities (RACFs). The purpose of this study was to conduct an in-practice evaluation of an eMAR being piloted in one Australian RACF before its roll out, and to provide recommendations for system improvements. A multidisciplinary team conducted direct observations of workflow (n=34 hours) in the RACF site and the community pharmacy. Semi-structured interviews (n=5) with RACF staff and the community pharmacist were conducted to investigate their views of the eMAR system. Data were analysed using a grounded theory approach to identify challenges associated with the design of the eMAR system. The current eMAR system does not offer an end-to-end solution for medication management. Many steps, including prescribing by doctors and communication with the community pharmacist, are still performed manually using paper charts and fax machines. Five major challenges associated with the design of eMAR system were identified: limited interactivity; inadequate flexibility; problems related to information layout and semantics; the lack of relevant decision support; and system maintenance issues. We suggest recommendations to improve the design of the eMAR system and to optimize existing workflows. Immediate value can be achieved by improving the system interactivity, reducing inconsistencies in data entry design and offering dedicated organisational support to minimise connectivity issues. Longer-term benefits can be achieved by adding decision support features and establishing system interoperability requirements with stakeholder groups (e.g. community pharmacies) prior to system roll out. In-practice evaluations of technologies like eMAR system have great value in identifying design weaknesses which inhibit optimal system use.
    Full-text · Article · Nov 2014 · Applied Clinical Informatics
    A Tariq A Tariq E Lehnbom E Lehnbom K Oliver K Oliver +3 more authors... A Georgiou A Georgiou
    • "Healthcare IT systems, especially mobile tools, can enhance such environments. However, many healthcare professionals remain reluctant to embed these tools in their medical practices (e.g. Pirnejad et al., 2008; Protti et al., 2007). Much of their work must be carried out quickly, and is completed away from their offices, so the need to frequently interact with technology, especially fixed workstations, can be resisted. "
    [Show abstract] [Hide abstract] ABSTRACT: This study formulates and empirically tests a theoretical model involving factors in the use of mobile devices for knowledge sharing in hospitals. The research model is derived from two important studies, Kankanhalli et al.'s (2005a, 2005b) studies on electronic knowledge repositories, and is adapted to the healthcare context. We conduct an exploratory case study of three mobile devices in two units of a hospital. The preliminary results reveal that factors such as image, privacy, reciprocity, quality of output, resource availability and portability influence the use of these devices for knowledge sharing.
    Full-text · Conference Paper · Aug 2013
Show more