Maryam Alavi
, Steven L. Johnson
and Youngjin Yoo
Goizueta Business School, Emory University, 1300 Clifton Road, Atlanta, GA, U.S.A.
Fox School of Business and Management, Temple University, Philadelphia, PA, U.S.A.
Keywords: Knowledge Management Systems (KMS) Applications and Use, Organizational Impacts of KMS, KMS in
Medical Settings.
Abstract: Use of Information Technology (IT) to facilitate storage, search and access to codified knowledge can
enhance organizational performance. Considering the knowledge intensive nature of medical decision
making, use of knowledge management systems (KMS) can potentially enhance delivery of patient care.
This paper reports on a field study that investigated the impact of use of a KMS by emergency physicians on
their decision making behaviors (admission of patients for in-hospital care) and efficiency of care. The
findings indicate that in the emergency room (ER) under study, the use of a KMS for placing diagnostic
orders resulted in lower hospital admissions, cost savings, and a shorter ER stay.
An extensive body of research based on knowledge-
based view of organizations (for example, Alavi and
Leidner, 2001; Spender, 1996; Nonaka and
Takeuchi, 1995) suggests that Information
Technology (IT) can enhance organizational
performance through consistent and effective
application of knowledge in organizational practices.
Considering the knowledge-intensive nature of the
healthcare domain, the high rate of change and
innovation, and the wide scope of knowledge needed
by the professionals in this field, patient care can
greatly benefit from advanced information
technologies that facilitate storage, search, and
timely access to the best knowledge available.
In the U.S., IT is viewed as integral to achieving
substantial quality and efficiency improvements in
healthcare delivery and management. A key idea
underlying this view is the use of IT to support
knowledge management (KM) to enhance and
facilitate clinical decision-making. The goal is to
apply best available evidence gained from scientific
method to delivery of patient care. It seeks the ideal
in which all medical professionals should make
explicit, conscious and judicious use of all available
best knowledge in making clinical decisions.
Proponents argue that application of evidence-based
practice would reduce the inconsistencies and
uncertainties in patient care. Many institutions try to
implement evidence-based clinical decision-making
in the form of policy, recommendations, guidelines
and regulations. However, translating such
institutional behaviors into actual behaviors
practiced by healthcare professionals is a
challenging task. IT is seen by many as an integral
solution to close the gap in implementing evidence-
based clinical decision-making practices. By
codifying best available knowledge and integrating
it into the electronic medical record (EMR) systems
that healthcare professionals interact with to manage
various clinical information, hospitals and medical
practices might be able to implement the evidence-
based practice more consistently and improve their
The literature, however, is clear that mere
stocking of knowledge through codification is not
enough to improve organizational performance
(Alavi and Tiwana, 2002). Knowledge needs to be
applied in practice in order to produce intended
outcomes (Cook and Brown, 1999; Orlikowski,
2002). Thus, the potential benefits KMS cannot be
realized through the codification and accumulation
of medical knowledge in IT systems alone. Rather, it
is in the ability to take effective action by applying
knowledge. Thus, it is not clear to what extent KMS
will be effective for Physicians who make complex
Alavi M., L. Johnson S. and Yoo Y..
DOI: 10.5220/0003669902360240
In Proceedings of the International Conference on Knowledge Management and Information Sharing (KMIS-2011), pages 236-240
ISBN: 978-989-8425-81-2
2011 SCITEPRESS (Science and Technology Publications, Lda.)
clinical decisions in constrained timeframes and
under high degree of uncertainty.
In order to address this issue, we have conducted
a field study to examine the impact of KMS in an
emergency room (ER) at a major hospital in the
United States. Specifically; we investigated the
following research questions:
Research Question 1: How does the use
knowledge management system change the clinical
decision-making behaviors by medical
Research Question 2: What are the impacts of the
use of knowledge management system on cost and
efficiency in clinical decision-making?
Lately, the popularity and deployment of EMR
(electronic medical records) have been on the rise in
the U.S. EMR systems are integrated IT systems for
healthcare information management and process
support. We anticipate that the eventual ubiquitous
availability of EMR systems in the U.S. will create
the necessary “backbone” that serves as the
infrastructure for knowledge codification, storage,
search, and delivery for clinical decision-making.
This in turn creates a need for conducting studies
that provide the necessary insights and
understanding for development and deployment of
knowledge management systems (KMS) for
effective and efficient delivery of patient care.
The effectiveness and outcomes of the decisions
made in an emergency room depend on timely and
accurate diagnosis and delivery of appropriate
treatments to patients. These decisions are in turn
impacted by the timely availability of the requisite
knowledge such as the knowledge of diagnostic tests
and their outcomes, treatment protocols, and
accurate presentation of the patient’s condition and
symptoms. As such, clinical decisions in emergency
rooms can benefit from the codification,
accumulation, and delivery of knowledge to
augment physicians’ judgment and know-how.
We investigated the impact of the use of a KMS
by emergency room physicians on the rate of
admission of emergency room patients to the
hospital, emergency room charges, and patient time
spent in the ER. The emergency room under
investigation serves the 573-bed university hospital
and is staffed by 38 physicians, working between
one to fifteen 8-hour shifts per month. The ER
treated approximately 93 patients per day and in
2009, a total of about 34,000 patients were treated at
this facility.
The emergency room in our field study routinely
collects and maintains patient records in a data
warehouse. The records of all adult patients (18
years and older) who visited the ER in a 321 day
period between January 2009 and November 2009
were used in this study (the precise dates were
masked to protect patient anonymity). For the
purpose of this study, we analyzed the records of ER
patients who complained of abdominal pain as their
primary symptom. We chose to focus on abdominal
pain complaints because they represent relatively
ambiguous cases and can potentially benefit the
most from the use of a KMS. Considering patient
privacy issues, we used a de-identified data sample.
This was accomplished by creating new data sets
from the warehouse patient records by excluding all
identifying fields, assigning appropriate aliases, and
copying the remaining data fields. The data fields
used in our analyses are described in Table 1.
The IT infrastructure in the ER consisted of an
EMR system, FirstNet, by Cerner Corporation in
Kansas City, Missouri. FirstNet functionalities
include electronic records and notes, results
management (e.g., lab and radiology reports),
clinical provider order entry (e.g., test orders), and a
KMS for decision support (e.g., standard diagnostic
orders and clinical guidelines and protocols). The
KMS module provides physicians with
recommended diagnostic tests and medication order
sets, based on the patient’s symptoms. As such, this
module can enhance the accuracy of diagnosis and
delivery of appropriate patient care. The standard
order sets (the codified knowledge embedded in the
system) are developed by expert physicians based on
the best available evidence of their efficacy.
Physicians and nurses, however, can choose not to
use the KMS recommendation and place their own
We focused our investigation on the impact of
the use of diagnostic order sets placed through the
KMS on patient care outcomes.
In order to control for the exogenous impact of the
severity of the conditions that affect the outcome of
the ER visit, we only focused on two most prevalent
acuity levels (urgent and emergent), resulting in a
sample of 2238 emergency department visits. A
small number of immediate, stable, and non-urgent
cases were dropped. Table 2 shows basic descriptive
statistics and the correlations between variables. All
correlations with an absolute value of 0.07 or greater
are significant at the p <0.05 levels.
Table 1: Emergency Room (ER) Patient Record Fields
Included in the Study.
Field Description
Admitted to
Outcome of ER stay resulted in
admission to hospital for in-
patient care
Duration of ER Stay Length of stay in ER
Total Charges
Total of doctor charges and
hospital charges for this ER stay
(available only for a subset of
Level of acuity of patient
condition as assessed at time of
arrival to ER
Age of patient on date of
Gender Gender of patient
First Diagnostic
Order through KMS
During this ER stay the first
diagnostic order was placed via
the KMS (a standard care set)
Total Number of
Diagnostic Orders
Total number of diagnostic orders
placed during this ER stay
Number of KMS
Diagnostic Orders
Of the total number of diagnostic
orders placed during this ER stay,
the number placed through KMS
Rate of orders
placed through
Number of diagnostic orders
through KMS divided by ER stay
Average Diagnostic
Order Placement as
% of ER Duration
Considering all diagnostic orders
placed during this ER stay, the
average time of order placement
as a percentage of stay duration
We ran separate analyses to answer the research
questions. In the first analysis, we examined if the
use of recommendations from KMS affect hospital
admissions. In particular, we focused on the impact
of the use of the system (represented by Ordering
Rate through KMS) on specific clinical decision-
making behaviors (represented by Total Number of
Diagnostic Orders, Average Diagnostic Order, and
First Diagnostic Order through KMS). Since hospital
admission is a binary variable, we used the logistic
regression. Table 3 presents the logistic regression
results for the predictors of hospital admissions.
Model 1 contains three significant control variables
related to the inherent complexity of patient care:
acuity, age, and gender. The parameter estimates for
acuity and age have the expected signs and
magnitude: higher acuity patients are far more likely
(~ 3x as likely) to be admitted to the hospital than
lower acuity patients; each additional year of patient
age leads to a very slight (~ 1.02x) increase in
hospital admission. Model 1 has an AIC value of
2739 and a log-likelihood of -1365 (df=4). Model 2
is a full test with our explanatory variables. It has an
AIC value of 2294 and a log-likelihood of -1138
(df=9). This is a statistically significant
improvement over model 1 in model fit (p<0.001).
Our results show that the use of KMS in clinical
decision making contributes to the reduction of
hospital admission. Furthermore, if the first
diagnostic order follows the recommendation from
the system, the reduction of the admission rate is
even larger.
In the second analysis, we explored if the use of
recommendations from KMS affected patient care
cost. Since the total charge does not follow the
normal distribution, we used a log transformation.
The results of an OLS regression model of
predictors of total charges (log transformed) are
presented in Table 4. This model includes only the
788 emergency room encounters for which both
physician charges and emergency department
hospital charges are available. For those patients
who were eventually admitted to the hospital,
hospital charges were not available as the charges
were rolled into other hospital charges incurred
during the in-patient care. Because a portion of total
charges are directly associated with the duration of a
patient’s emergency room stay, duration is included
as an additional control variable. As expected, our
results show that the duration of stay and total
number of orders are the main drivers of the total
charges. Our results also show that if the physician
starts with a diagnostic order recommended by the
KMS, it leads to a significant reduction in total
Given that duration is one of the key variables
that affect the total cost, in the third analysis, we
examined if the use of recommendations from the
KMS can contribute to the reduction of the duration.
The results of an OLS regression model of
predictors of duration of stay in the emergency
department are presented in Table 5.
Taken together, our analyses show several
statistically significant relationships for emergency
room encounters with patients. The variables of
acuity, age, and total number of diagnostic orders all
increase the likelihood of an encounter resulting in a
hospital admission. Also, female patients are less
likely to result in hospital admission. However, we
found that when a physician places a higher rate of
diagnostic orders through the KMS, it significantly
reduces the chance that the patient gets admitted in
the hospital. Furthermore, this relationship is
strengthened when the first diagnostic order is
placed through the KMS.
KMIS 2011 - International Conference on Knowledge Management and Information Sharing
Table 2: Correlations (n-2238).
mean s.d. 1 2 3 4 5 6 7 8 9
Admitted to Hospital
0.40 0.49
2 Duration of ER Stay 0.38 0.23 0.43
3 Acuity (1=Higher) 0.34 0.47 0.29 0.07
4 Age 44.6 18.2 0.23 0.09 0.16
5 Gender (1=Female) 0.67 0.47 -0.13 0.02 -0.17 -0.09
First Diag. Order through
KMS (1=Yes)
0.69 0.46 0.05 0.03 0.04 0.03 -0.01
Total Number of Diagnostic
11.0 4.6 0.45 0.29 0.20 0.20 -0.02 0.23
Number of KMS Diagnostic
4.00 4.56 0.08 0.07 0.02 0.09 -0.04 0.59 0.20
KMS Diagnostic Ordering
13.1 14.3 -0.12 -0.31 -0.02 0.01 -0.02 0.47 0.02 0.76
Average Diagnostic Order
Placement as % of ER
0.22 0.14 0.08 -0.13 -0.04 -0.00 0.03 -0.15 0.19 -0.09 -0.06
Table 3: Logistic Regression for (In-Patient) Hospital Admission (n=2238).
Model 1: Controls Model 2: Full Model
Sig. Est.
(Intercept) -1.58 0.21 0.18 *** -3.72 0.02 0.25 ***
Acuity 1.10 3.02 0.10 *** 0.91 2.49 0.11 ***
Age 0.02 1.02 0.00 *** 0.02 1.02 0.00 ***
Gender -0.37 0.69 0.10 *** -0.55 0.58 0.11 ***
First Diagnostic Order through KMS -0.12 0.89 0.15
Total Number of Diagnostic Orders 0.26 1.30 0.02 ***
KMS Diagnostic Ordering Rate -0.08 0.92 0.02 ***
Average Diagnostic Order Placement as % of
ER Duration
0.23 1.26 0.41
Interaction: First Order through KMS
KMS Diagnostic Ordering Rate
0.05 1.06 0.02 **
** p < 0.01; *** p < 0.001
In the subset of patient encounters where
hospital charge data are available, patient age, the
duration of emergency department stay, and the
total number of diagnostic orders are all associated
with higher charges. Female patients and
encounters where the diagnostic orders are placed
earlier in the stay are all associated with lower
charges. If the physician starts the diagnostic
orders following the recommendations from the
system, it leads to a significant reduction of total
In the full data set the measures associated with
a longer duration of emergency department stay
were gender (female) and the total number of
diagnostic orders. Encounters with higher acuity
and encounters where the diagnostic orders were
placed earlier in the stay had lower durations.
Again, if the physician starts the diagnostic orders
following the recommendations from the system, it
leads to a significant reduction in the duration of
patient stay in ER.
Table 4: OLS Regression for Total Charges (Log
Transformed) (n=788).
(Intercept) 6.18 0.10 ***
Acuity 0.04 0.05
Age 0.00 0.00 **
Gender -0.09 0.05 *
Duration 2.02 0.19 ***
First Diagnostic Order
through KMS
-0.22 0.07 ***
Total Number of
Diagnostic Orders
0.12 0.01 ***
KMS Diagnostic
Ordering Rate
0.00 0.00
Average Diagnostic
Order Placement as % of
ER Duration
-0.87 0.17 ***
Interaction: First Order
through KMS * KMS
Diagnostic Ordering Rate
0.00 0.00
*p < 0.05; ** p < 0.01; *** p < 0.001; Adj. R
= 0.47
Table 5: OLS Regression for ER Duration (n=2238).
(Intercept) 0.30 0.02 ***
Acuity -0.03 0.01 ***
Age 0.00 0.00
Gender 0.04 0.01 ***
Admitted to Hospital 0.19 0.01 ***
First Diagnostic Order
through KMS
-0.03 0.01 **
Total Number of
Diagnostic Orders
0.01 0.00 ***
Average Diagnostic
Order Placement as %
of ER Duration
-0.36 0.03 ***
** p < 0.01; *** p < 0.001; Adj. R
= 0.24
An open question—and one that Information
Systems (IS) researchers are well-situated to help
address—is if and when knowledge management
systems may actually impact practice. It is merely
not enough to identify best practices; it is a greater
challenge to consistently enact them. Our study
provides preliminary evidence in how the
implementation of KMS for diagnostic testing in
an ER is leading to positive patient outcomes.
Alavi, M. and Leidner, D. (2001). Knowledge
Management and Knowledge Management Systems:
Conceptual Foundations and Research Issues, MIS
Quarterly, 25, 107-136.
Alavi, M. and Tiwana, A. (2002). Knowledge
Integration in Virtual Teams: The Potential Role of
KMS, Journal of the American Society for
Information Science and Technology, 53, 1029-
Cook, S. D. N. and Brown, J. S. (1999). Bridging
Epistemologies: The Generative Dance between
Organizational Knowledge and Organizational
Knowing, Organization Science, 10, 381-400.
Nonaka, I. and Takeuchi, H. (1995). The Knowledge
Creating Company: How Japanese Companies
Create the Dynamic Innovation. New York: Oxford
University Press.
Orlikowski, W. J. (2002). Knowing in Practice: Enacting
a Collective Capability in Distributed Organizing,
Organization Science, 13, 249-273.
Spender, J. C. (1996). Making Knowledge the Basis of a
Dynamic Theory of the Firm, Strategic Management
Journal, 17, 45-62.
KMIS 2011 - International Conference on Knowledge Management and Information Sharing