This article was downloaded by: [University of Saskatchewan Library]

On: 11 January 2015, At: 09:45

Publisher: Taylor & Francis
Informa Ltd Registered in England and Wales Registered Number: 1072954 Registered office: Mortimer House,
37-41 Mortimer Street, London W1T 3JH, UK

IIE Transactions on Healthcare Systems Engineering

Publication details, including instructions for authors and subscription information:
http://www.tandfonline.com/loi/uhse20

An alternative outpatient scheduling system: Improving

the outpatient experience
a b b
Yu-Li Huang , Walton M. Hancock & Gary D. Herrin
a
Department of Industrial Engineering , New Mexico State University , Las Cruces , NM ,
88003 , USA
b
Industrial and Operations Engineering Department , University of Michigan , Ann Arbor ,
MI , USA
Accepted author version posted online: 04 Apr 2012.Published online: 21 Jun 2012.

To cite this article: Yu-Li Huang , Walton M. Hancock & Gary D. Herrin (2012) An alternative outpatient scheduling
system: Improving the outpatient experience, IIE Transactions on Healthcare Systems Engineering, 2:2, 97-111, DOI:
10.1080/19488300.2012.680003

To link to this article: http://dx.doi.org/10.1080/19488300.2012.680003

PLEASE SCROLL DOWN FOR ARTICLE

Taylor & Francis makes every effort to ensure the accuracy of all the information (the “Content”) contained
in the publications on our platform. However, Taylor & Francis, our agents, and our licensors make no
representations or warranties whatsoever as to the accuracy, completeness, or suitability for any purpose of the
Content. Any opinions and views expressed in this publication are the opinions and views of the authors, and
are not the views of or endorsed by Taylor & Francis. The accuracy of the Content should not be relied upon and
should be independently verified with primary sources of information. Taylor and Francis shall not be liable for
any losses, actions, claims, proceedings, demands, costs, expenses, damages, and other liabilities whatsoever
or howsoever caused arising directly or indirectly in connection with, in relation to or arising out of the use of
the Content.
This article may be used for research, teaching, and private study purposes. Any substantial or systematic
reproduction, redistribution, reselling, loan, sub-licensing, systematic supply, or distribution in any
form to anyone is expressly forbidden. Terms & Conditions of access and use can be found at http://
www.tandfonline.com/page/terms-and-conditions
 IIE Transactions on Healthcare Systems Engineering (2012) 2, 97–111
Copyright

C “IIE”

ISSN: 1948-8300 print / 1948-8319 online

DOI: 10.1080/19488300.2012.680003

An alternative outpatient scheduling system: Improving the

outpatient experience

YU-LI HUANG1,∗, WALTON M. HANCOCK2 and GARY D. HERRIN2

1
Department of Industrial Engineering, New Mexico State University, Las Cruces, NM 88003, USA
E-mail: yhuang@nmsu.edu
2
Industrial and Operations Engineering Department, University of Michigan, Ann Arbor, MI, USA
Downloaded by [University of Saskatchewan Library] at 09:45 11 January 2015

Received June 2010 and accepted March 2012.

Patient wait time has long been a recognized problem in modern outpatient health care delivery systems. Despite all the efforts to
develop appointment rules and solutions, the problem of long patient waits persists. Regardless of the reasons for this problem, the
fact remains that there are few implemented models for effective scheduling that consider patient wait times as well as physician idle
time and are generalized sufficiently to accommodate a variety of outpatient clinic settings. This paper presents a solution of designing
appointment slots for scheduling appointments in outpatient facilities that both patient wait time and physician idle time meet the
declared scheduling policies without overbooking and double-booking. Furthermore, this paper provides the implementation results
from three case studies to support the approach. The results confirm that the system can effectively reduce patient wait time as
much as 56% without significantly increasing physician idle time per patient and still allow physician to see and schedule exactly
same number of patients per clinic session. Consequently, this research improves the outpatient experience for both patients and the
medical professions, changes the perception of long waits in a physician’s office and ultimately enhances the quality of care.
Keywords: Outpatient scheduling, patient flow, patient wait time, physician idle time, process improvement

1. Introduction fective scheduling that consider patient wait time as well

Patient wait time has long been a recognized problem in
modern outpatient health care delivery systems. As compe-
tition has increased for limited health care dollars, efforts
have been made to increase efficiency and reduce costs.
One of the main strategies to decrease overall healthcare
cost has been to shift traditionally inpatient services to
an outpatient setting, effectively increasing the burden on
outpatient facilities to efficiently manage health care deliv-
ery (Carr, 2006). One possible explanation for the limited
progress made in the systematic reduction of patient wait
may be that facilities have focused more on the efficient
scheduling of provider time, perceived to be a more easily
controlled variable than patient behavior (e.g., late arrivals,
no-shows). Another possible explanation is that physicians
and clinic management do not adequately define when a
patient appointment actually commences, i.e., at time of
arrival or at time of examination, or do not identify fac-
tors associated with a patient encounter or ‘visit’ that can
increase patient wait time. Regardless of the reason, the
fact remains that there are few implemented models for ef-
∗
Corresponding author
as physician idle time and are generalized sufficiently to
accommodate a variety of outpatient health care settings.
This paper demonstrates an alternative patient scheduling
approach that accounts for patient wait time, physician idle
time, and clinic capacity. The approach was actually imple-
mented in three outpatient clinics and the results indicate
patient wait time is reduced as much as 56%.
1.1. Background and literature survey
The 1950s saw the burgeoning of appointment systems for
outpatient clinics, particularly public and government facil-
ities, most likely in response to the growing demand placed
on these institutions by the return of veterans and the heavy
migration from rural America to wartime industry. These
appointment systems generally operated on a fixed set of
rules or policies, which were expanded over time. In re-
sponse to the establishment of Medicare in the mid-sixties,
the American Medical Association (AMA) developed a
universal system of codes, current procedural terminology
(CPT), defining patient visits, medical and surgical proce-
dures, and other healthcare services, to replace the plethora
of individual, local, and state procedural codes and service

1948-8300

C 2012 “IIE”
 98
definitions; reported by AMA (2009). However, the CPT
did not finally become used nationwide by all carriers un-
til the end of the 1980s when the Resource Based Value
System (RBVS), developed by Harvard University under
the leadership of William Hsiao, was institutionalized in
the Omnibus Budget Reconciliation Act of 1989. With this
Act, visits defined under the AMA’s different CPT codes
were assigned RBVS units, which in turn dictated Medicare
reimbursement of outpatient services, setting the precedent
for all insurers and indelibly tying physician time to cost.
Together with the introduction of the RBVS, there was a
concerted effort on the part of government carriers to re-
duce reimbursement for more expensive procedures, while
increasing reimbursement for outpatient visits. Not sur-
Downloaded by [University of Saskatchewan Library] at 09:45 11 January 2015
prisingly, there was a resurgence of interest in scheduling
efficiency, particularly in terms of cost.
Shortly after the implementation of the Omnibus Bud-
get Reconciliation Act, a number of studies developed
and evaluated appointment rules for efficient scheduling.
Ho and Lau (1992) formulated the patient wait time and
physician idle time categories associated with cost. They
evaluated various appointment rules using the “frontier”
approach and selected the eight best appointment rules ac-
counting for the major environmental factors which impact
scheduling, such as the probability of no-shows, the coeffi-
cient of variation of service time and the number of patients
per session. Ho, Lau, and Li (1995) evaluated appointment
scheduling rules and introduced the concept of variable-
interval rules to reduce cost and improve service quality.
Klassen and Rohleder (1996) evaluated the appointment
rules using a design of experiment approach for differ-
ing patient characteristics and concluded that scheduling
a ‘low-variation patient type’ in the beginning of a ses-
sion was the best solution among all appointment rules.
Yang, Lau, and Quek (1998) developed a new appoint-
ment rule taking into account variations in service time,
the percentage of no-shows, the number of appointments
per session, and the cost ratio of physician and patient
wait time. Harper and Gamlin (2003) identified and evalu-
ated 10 different scheduling policies for a specific clinic and
found that mixing different appointment intervals over the
duration of a clinic session is better in terms of reducing pa-
tient delay. Cayirli, Veral, and Rosen (2006) used a simula-
tion approach to build a two-factor experiment, sequencing
rules and appointment rules, including a number of unpre-
dictable factors such as walk-ins and no-shows and found
the impact of the choice of sequencing rules is greater than
the choice of appointment rules. Wijewickrama (2006) used
a simulation approach to evaluate four appointment rules
and their possible combinations and concluded that a hy-
brid combination of appointment rules works better than a
single rule. Despite the insights generated by research, the
appointment rules developed did not seem to be broadly
adopted by outpatient clinics because they are difficult to
follow and include no clear explanation as to which rules are
most appropriate in which type of outpatient environments.
Huang et al.
Also, actual case studies with implementation results were
not available to support the feasibility of the theoretical ap-
pointment rules. This leaves a significant gap between the
theoretical constructs and the realities of practical appli-
cations. Furthermore, regardless of how sophisticated the
proposed appointment rules are or how significantly they
have impacted the current scheduling systems, the vari-
ability in physician treatment time has not been addressed
from the patient’s standpoint under clinical conditions. In
essence, this simply means that patients are scheduled to
accommodate the physician’s schedule.
The basic model, employed by most of the approaches
and models presented in the literature (Ho and Lau,
1992; Ho, Lau, and Li, 1995; Klassen and Rohleder, 1996;
Yang, Lau, and Quek 1998; Vanden Bosch and Dietz,
2000; Denton and Gupta, 2003; Cayirli, Veral, and Rosen,
2006; Wijewickrama, 2006; Kaandorp and Koole, 2007)
focuses on minimizing total cost using the cost ratio
approach while still trying to accommodate variation.
The cost relative to physician idle time tends to be much
higher than to patient wait time due to the perception
of higher physician compensation and cost of medical
facility (Keller and Laughhunn, 1973; Yang, Lau, and
Quek, 1998). Hence, this basic model inherently favors the
reduction of physician idle time. Moreover, to the best of
our knowledge, clinical policy or constraints have been
discussed by many researchers on developing a schedule,
but limited, such as the time at which the last patient
should be seen in a session or the session should end.
Some approaches presented in the literature are case-
specific, which precludes generalization. One group used
simulation approaches to design for a specific clinic such as
Swisher et al. (2001), Guo, Wagner, and West (2003), and
Wijewickrama and Takakuwa (2005). Those simulation
studies, regardless of their focus, arrived at similar conclu-
sions regarding the factors that influence patient flow and
admission policies. Another group tended to rely on trial
and error to select the best schedule for a particular clinic
from a range of possible solutions such as Meza (1998),
Harper and Gamlin (2003), Klassen and Rohleder (2004).
One other group proposed the Open Access approach such
as Murray and Tantau (1999), O’Hare and Corlett (2004),
Green and Savin (2008). This approach focused on reduc-
ing no-shows and improving patient access to services.
Similarly, some research such as Kim and Giachetti (2006),
LaGanga and Lawrence (2007), Muthuraman and Lawley
(2008) suggested that overbooking and double-booking
would reduce physician idle time and increase profitability
and patient access. These approaches focus on maximizing
the utilization of the clinic capacity especially physicians’
time to reduce health care cost.
Cayirli and Veral (2003) conducted a fairly inclusive sur-
vey of the literature on the topic of outpatient schedul-
ing appointments, which provides an extensive review
of the problems associated with the definitions and for-
mulations of outpatient appointments, the performance
 Improving outpatient experience
measurements and evaluations, and the historically used
analysis methods. However, Cayirli and Veral concluded
that the studies they reviewed had one or more major limi-
tations: 1) most studies are case-specific and, therefore, dif-
ficult to generalize; 2) the models of patient flow are pred-
icated on unrealistic assumptions; 3) the studies focus on
developing appointment rules to accommodate additional
clinical scenarios as opposed to exploring the variation of
patient flow; 4) studies do not successfully implement ap-
pointment rules, presumably due to a lack of understanding
of how rules are implemented in practice and how decisions
are made regarding which rule to follow.
To develop a more generalized approach to effective
scheduling that could be easily implemented to reduce pa-
Downloaded by [University of Saskatchewan Library] at 09:45 11 January 2015
tient wait time without significantly impacting physician
idle time, we felt we needed to address a number of out-
standing questions: How do treatment times that are not
based on actual data impact the success of a schedule? How
can the actual data be most appropriately utilized to esti-
mate the treatment time distribution, to take into account
the variation of the treatment time, and to finalize the best
scheduled treatment time interval?
1.2. Problem statement
The problems faced in designing a model are how to, first,
best model the scheduling of procedures and appointments
so that both patient and physician wait times meet the
declared policy of a facility, such as clinic duration or start
and finish times, and, then, second, generalize that model
so that it can be effectively used in a variety of clinic settings
with differing policies.
In designing a model, our specific objectives were to enlist
a number of outpatient clinics, collect accurate data about
the actual time physicians spend with patients for each type
of visit, then assign a realistic time interval to each type of
visit under clinic policy or constraints without overbook-
ing, and finally implement the model in the participating
clinics to validate performance in realistic settings. From
the outset, our primary goal was to develop a practical and
easy-to-use approach to scheduling that would equitably
reduce, rather than try to eliminate, systemic wait times
for both physician and patient. Given the large number of
variables that can affect scheduling, we felt it was more im-
portant to provide a tool for managing the schedule rather
than controlling it. Furthermore, since the true benefit of
any model or approach can only be realized if the principals
are willing to use it, it was imperative that the approach be
neither difficult nor too complex to implement.
We believe that much of the wait time experienced by pa-
tients is the direct result of unrealistic estimations of treat-
ment time or unrealistic solutions to chronic problems. Un-
derestimating visit times, overbooking or double booking
all represent a ‘physician centric’ solution to the problem
of ‘no-shows’ or late patients: events that can unexpectedly
affect any daily schedule and create sudden unplanned idle
99
time for the physician. The perception is that by scheduling
15 minutes for a service that takes 25 minutes, there will
always be enough patients to keep the physician occupied,
even in the event of no-shows or late arrivals. In addition,
time estimations are often based on a physician’s perception
of the ‘ideal’ visit or what should be accomplished in a given
type of visit, without considering the actual variations that
occur. This again reflects a ‘physician-centric’ approach to
time estimation. Yet, whatever the underlying cause of the
unrealistic estimation, a simple remedy is to conduct a de-
tailed time study to provide accurate data about the actual
time involved in each visit type. The impact of inaccurate
estimates of treatment times is discussed in section 3.
In addition to determining the actual time that should
be allocated to a particular physician service, it is also im-
portant to define what is typically included in a particular
service besides strict face-to-face physician-patient contact.
For our purposes, we decided that to establish a realis-
tic time estimation for a visit type, it is critical to include
any ancillary services, such as reviewing charts, dictating
findings, studying labs and x-ray reports, that may be per-
formed outside the exam room, but which are ‘triggered’ by
or directly related to the service for a particular patient.
However, to counter physician-centric scheduling, we
needed to incorporate a patient arrival schedule into the
model together with the physician service time. The dis-
tinction between patient arrival time and the actual ap-
pointment time, which we defined as the encounter time or
service start time, is often blurred in practice. Therefore, it
is not uncommon for a patient to arrive on time for the
scheduled appointment, but be delayed at the front desk
completing necessary paperwork. Even though the patient
may not perceive this as wait time, physicians often regard
this unplanned delay as idle time. It is not uncommon for
physicians to fill this idle time with activities which, in turn,
may often extend beyond the initial delay causing a cas-
cade that compounds patient wait time throughout a given
session. Hence, to eliminate the discrepancy between what
patients understand as arrival time and clinic staff regard
as encounter time, we made a distinction between arrival
time and appointment time in our model.
In short, our research objective was to develop a method-
ology to best schedule procedures and appointments in
accordance with clinical policy or constraints so that pa-
tient and physician idle times are limited or, on occasion,
even eliminated. Specifically, our aim was to provide an ap-
proach based on simulation that could be easily adapted
to any clinical setting and that would reliably provide best
scheduling practices responsive to both physician and pa-
tient interests.
2. Proposed solution
The goal is to provide a solution of designing a prede-
termined scheduling template that accounts for patient
 100
wait time and physician idle time without overbooking and
double-booking. There are steps for the proposed solution.
1. We first collected data on the treatment times for all
services rendered by the clinic providers, including the
physician, residents, registered nurses (RNs), and medi-
cal assistants (MAs) to estimate treatment time param-
eters and distribution for each visit type.
2. Then, we determined the best (i.e., maximum) sched-
uled time interval for each visit type under any declared
clinic policy or constraints. Simulation was chosen as the
solution technique since Cayirli and Veral (2003) sum-
marized from numerous literature that the advantage of
simulation modeling is the ability to present the complic-
ity and the difference of outpatient clinic environments.
Downloaded by [University of Saskatchewan Library] at 09:45 11 January 2015
Two simulation models were developed. First, a model
for a single visit type was built to determine the best
scheduled time intervals for any given ‘wait ratio’ (the
ratio between patient wait time and physician idle time)
for each visit type. Second, an extended model sequenc-
ing the various visit types was constructed for the clinic
to finalize the optimal best scheduled time intervals un-
der clinical policy or constraints, assuming perfect clin-
ical conditions such as punctual patients, no physician
delay, no conflict in availability of staff or equipment
(e.g., for x-rays, EKGs, or vital signs) for the purpose of
designing scheduling template.
3. The provider schedule was subsequently developed from
the optimal best scheduled time interval established for
each visit type.
4. Once the provider schedule had been determined, the
patient arrival schedule was constructed based on any
ancillary patient activities associated with a given visit.
2.1. Determination of the best scheduled time interval
A successful appointment system should minimize patient
delays while fully utilizing medical resources. However,
there is a tradeoff in that reducing patient wait time may
increase physician idle time and vice versa. Hence, this re-
Fig. 1. The graphical representation of patient flow model in physician service process.
Huang et al.
search balances patient wait time and physician idle time
by determining the Wait Ratio between them. This in turn
allows us to determine the best treatment or service time
interval for each type of visit for a given Wait Ratio; dis-
cussed in 2.1.1. Determining the best treatment time in-
tervals across different visit types by the same wait ratio
is to insure every patient is treated equally in terms of
waiting.
The best scheduled time interval is defined as the max-
imum scheduled time interval allowed for each visit type
that satisfies medical and clinic constraints. This section
provides the mathematical model of how the best sched-
uled time intervals are generated and discusses policy or
constraints that impact the decision.
2.1.1. Definitions and formulations
For our purpose, treatment or service time is defined
as the time from which a physician enters the exam
room to the time at which the physician finally exits the
exam room and includes any ancillary physician service
for that patient such as reading charts or dictating. The
model is focused on patient flow in the physicians’ treat-
ment process; see Figure 1 for the graphical representa-
tion of the model. The parameters defined in the model
are:
Ti = physician’s service time to treat patient i where i =
1, 2, 3, . . . , n. Ti ∼ D(µ, σ 2 ), D is a probability distribution
with mean µ and standard deviation σ .
n = the number of patients scheduled per session. A
session could be a day, a morning or an afternoon.
d = number of standard deviations away from the mean
µ.
X = the scheduled time interval for a patient to see the
physician in minutes
X = µ + dσ (1)
Si = the scheduled time to start patient i where i =
1, 2, 3, . . . , n and let S1 = 0
Si = Si −1 + X = (i − 1) X (2)
 Improving outpatient experience 101
Fi = the finish time for patient i where i = 1, 2, 3, . . . , n Table 1. The average scheduled time intervals for each visit type
Ai = the actual time to start patient i where i = with various wait ratios
1, 2, 3, . . . , n. Assuming a punctual patient, from (2), Ai
Visit type (k)
and Fi are:
Wait ratio (R) NP RV FU ... HP
Fi = Ai + Ti (3)


Si if Fi −1 < Si 1 X1NP X1RV X1FU ... X1HP
Ai = (4) 2 X2NP X2RV X2FU ... X2HP
Fi −1 Otherwise
3 X3NP X3RV X3FU ... X3HP
Wi = the wait time for patient i in minutes, where i = .. .. .. .. ..
.
..
. . . . .
1, 2, 3, . . . , n. From (2) and (4), Wi is:
/ XlNP XlRV XlFU ... XlHP
Wi = Ai − Si (5)
W̄ = the average patient wait time based on (5): k = visit type, where k = NP (new patient), RV (return
Downloaded by [University of Saskatchewan Library] at 09:45 11 January 2015


n visit), FU (follow up), HP (pre-operation) . . .
Wi Therefore, XRk is the average scheduled time for visit
i =1
W̄ = (6) type k given Wait Ratio R and this is the end of the first
n
simulation; see Table 1.
Pi = the physician idle time waiting for patient i in min- Once the average scheduled time interval of each visit
utes, where i = 1, 2, 3, . . . , n. From (3) and (4), type for each wait ratio are established, the second simula-
Pi = Ai − Fi −1 (7) tion is run with an extended simulation model sequencing
the various visit types for a clinic. The sequence is decided
P̄ = the average physician idle time based on (7): based on the clinic preference. However, the suggested se-

n quence is to distribute each visit type evenly throughout a
Pi clinic session. This provides options for patients to accom-
i =1
P̄ = (8) modate their personal schedule. With the extended simu-
n lation model, the average scheduled time interval for each
R = the Wait Ratio, which is the degree to which patient visit type at any given wait ratio can be inputted and evalu-
wait time (6) exceeds R times of physician idle time (8), ated to determine the best wait ratio (R ∗ ) that satisfies the
where R = 1, 2, 3, . . . , l and l ∈ R+ , assuming patient wait clinic policy or constraints (discussed in 2.1.2) assuming
time is normally longer than physician idle time on average: perfect clinical conditions, which gives the optimal sched-
uled time interval ( X̄ ∗ ) for each visit type. An example is
W̄ presented in 2.4 to demonstrate the approach.
= R ⇒ P̄ × R = W̄ (9)
P̄
Use (9) as the objective function to find the value for the 2.1.2. Clinic policy and underlying constraint
decision variable (d) and scheduled time interval (X) at any To assure that our model does reflect actual clinic opera-
given R. tion, certain common clinic policies and underlying con-
Then a number of simulation runs are needed to conclude straints should be accounted for in order to generate the
average schedule time interval X̄ at any given R. Let: best treatment time interval:
m = the number of simulation runs (based on desired
1. Clinic or session finish time: Most clinics have either
error bounds or confidence interval widths)
4-hour session (half day) or 8-hour session (full day).
d j = the decision variable of simulation run j where j =
The normal finish time could be at noon or at 5:00 p.m.
1, 2, 3, . . . , m
However, some physicians would like to finish earlier, for
X j = the scheduled time interval of simulation run
example, finishing at 11:30 a.m. for a 4-hour morning
j where j = 1, 2, 3, . . . , m
session.
Equation (1) can be rewritten to:
2. Time of last appointment: Most clinics have their own
Xj = µ + d j σ (10) preference as to when the last scheduled patient should
be in for the appointment. For example, the last patient is
Therefore,
scheduled to come in at 10:30 a.m. for a 4-hour morning

m
session.
Xj
j =1
3. Number of patients to be seen (or number of desired ap-
X̄ = (11) pointment slots) in a given session: This decision should
m
be made by a clinic management team to best serve
Then, we need to find X̄ for each visit type with various patients’ needs based on the seasonal demands. This
wait ratios (R). Let: decision also highly depends on specialty.
 102
4. Average patient wait time and maximum patient wait
time: Some clinics use this as a measurement to evalu-
ate their service quality. As mentioned in Huang (1994),
the survey results that patient tolerance for delay dimin-
ishes after approximately 30 minutes. Clinics could use
this 30-minute as a benchmark for their performance
measurement.
5. Average physician idle time and maximum physician idle
time: Physicians are the most costly element in the entire
clinic visit process. Most clinics would like to utilize
their physicians at their full capacity or minimize their
physicians’ idle time.
6. The sequence of the various visit types: Some studies
concluded that the scheduling sequence of the various
Downloaded by [University of Saskatchewan Library] at 09:45 11 January 2015
visit types in a clinic session has significant impact on pa-
tient waiting, such as Klassen and Rohleder (1996) and
Harper and Gamlin (2003). However, an optimal se-
quence may not provide clinics a flexible enough sched-
ule to accommodate patients’ needs. Each clinic man-
agement team should be able to decide its own sequence
for individual physician without any restriction.
7. The underlying constraint for deciding the best treat-
ment time interval is to ensure that the probability of
patient waiting given this time interval is less than 50%.
This constraint defines the minimum value of what the
best treatment time interval should be for a good patient
flow. From the equation (4), the current patient delays
are highly influenced by the wait time (Wi −1 ) and treat-
ment time (Ti −1 ) from the previous patient. Since the
treatment time (Ti ) for each patient is a random vari-
able, the only controllable factor here is the scheduled
time interval ( X̄ ∗ ). It is desirable that the probability of
Ti − X̄ ∗ ≤ 0 is higher than at least 50%.
In other words, given Ti ∼ D(µ, σ 2 ) and best time in-
terval X̄ ∗ , the objective should be Pr(Ti ≤ X̄ ∗ ) ≥ 0.5; that
is the probability of a wait less than or equal to 50% (see
Figure 2 where the area under the curve less than best time
interval X̄ ∗ is greater than or equal to 0.5).
In short, our model does not aim at forcing the clinic
setting to fit the solution, but rather aims at fully utilizing
the available resources and capacity to achieve the best
solution. Although clinic policy and constraints are limited
in the literature, they are used here to determine a wait ratio
Fig. 2. The constraint of the probability less than the best time
interval is at least 0.5.
Huang et al.
to replace conventional cost ratios, in large part because
physicians tend to overestimate the cost of their time as
opposed to patients’ time, most likely due to their lack of
criteria for accurately evaluating time cost for the patient.
This method effectively eliminates the cost of patient time
and the bias inherent in cost ratios from the model in favor
of well defined policy and constraints. In the case where a
solution cannot be found based on these constraints, clinics
should either consider loosening up clinic constraints or
evaluate their capacity whether clinics have scheduled more
patients than they can possibly handle on a regular basis.
2.2. Provider schedule
Based on the optimal scheduled time interval ( X̄ ∗ ) for each
visit type, the provider schedule from equation (2),Si =
Si −1 + X̄ ∗ , is constructed so that appointment slots are
consecutive, without overbooking or double-booking. For
the physician, this schedule represents the actual time at
which each patient encounter should begin. This schedule
can also provide a timeline for physicians to best utilize their
‘idle’ time in between patients. For example, a single follow-
up visit type with X̄ ∗ = 7.4 minutes, the physician schedule
starts at 8:00 a.m. (S1 = 0), 8:07 a.m. (S2 = S1 + X̄ ∗ = 7.4),
8:15 a.m. (S3 = S2 + X̄ ∗ = 14.8), and so on.
2.3. Patient arrival schedule
Once a physician schedule is established, then the cor-
responding patient arrival schedule must be determined.
Again, the main concept behind the arrival schedule is
to provide sufficient time between the patient arrival at
the clinic and the actual examination time for the pa-
tient to complete activities such as signing in, filling out
paperwork, having vitals taken, having an x-ray taken,
providing a specimen, and moving between lab or x-ray
room and exam room. Let Yi = arrival time for patient
i and Bi = time scheduled for pre-visit activities for pa-
tient i; therefore, Yi = Si − Bi . For example, assuming an
8-minute x-ray (Bi = 8) is needed for some patients and
a patient is scheduled to see the physician at 8:13 a.m.
(S3 = 12.8), if he/she is required to have an x-ray taken
before seeing the physician, the arrival time is 8:05 a.m.
(Yi = Si − Bi = 12.8 − 8 = 4.8), otherwise the arrival time
is 8:13 a.m. (Yi = Si − Bi = 12.8 − 0 = 12.8). The time as-
signed to pre-visit activities will differ from clinic to clinic
and between specialties. However, if the time needed for
these activities is not well defined, wait time will be com-
pounded for either physician or patient. Ideally, the physi-
cians should be able to maintain their schedules without
contributing significantly to patient wait.
2.4. An example
The example of a case study from an Orthopedic Surgery
Clinic is presented here to demonstrate how this approach
 Improving outpatient experience 103
Table 2. Average scheduled time intervals for various wait ratios that the physician service times (Ti ) of all visit types can
be best fitted by Gamma distribution with averages of 10.6,
Average scheduled time intervals (in minutes)
7.3, 5.5 minutes and standard deviations of 4.5, 4.0, 3.4 min-
Wait ratio FU NP XR utes for NP, FU, and XR, respectively. Therefore, the aver-
age scheduled time interval ( X̄) for each visit type given wait
1 9.2 12.7 6.8 ratios (R) from 1 to 20 after 50 simulation runs (m = 50) are
2 8.6 12.1 6.4 shown in Table 2. The number of simulation runs taken was
3 8.4 11.8 6.1 based on the error rate to be within ±0.1 with probability
4 8.2 11.5 6.0 of 0.95 and the initial estimation of population standard
5 8.0 11.4 5.9
deviation to be 0.355 for the FU patient type. These as-
6 7.9 11.3 5.8
7 7.8 11.2 5.7 sumptions generated the final number of simulation runs
8 7.8 11.1 5.7 must be at least as large as 48.5, which 50 was chosen.
9 7.7 11.0 5.6 Then, after simulating the extended clinic model, the
10 7.6 11.0 5.6 results after 50 simulation runs are in Table 3.
Downloaded by [University of Saskatchewan Library] at 09:45 11 January 2015

11 7.6 10.9 5.5 Given clinical policy constraints of finishing by 11:30

12 7.6 10.8 5.5 a.m. (210 minutes), and last patient scheduled to arrive
13 7.5 10.8 5.4 before 11:00 a.m. (180 minutes), the best wait ratio (R ∗ )
14 7.5 10.8 5.4 that satisfies both constraints is 18, which gives the optimal
15 7.4 10.7 5.4 scheduled time intervals for follow-up (FU), new patient
16 7.4 10.7 5.4 ∗ ∗
(NP) and x-ray patient (XR) to be X18FU = 7.4, X18NP =
17 7.4 10.7 5.3 ∗
18 7.4 10.6 5.3 10.6, and X18XR = 5.3 minutes (see Figure 3). These opti-
19 7.3 10.6 5.3 mal scheduled time intervals have the probability of a wait,
20 7.3 10.6 5.3 Pr(Ti > X̄ ∗ ), of 44%, 44%, and 42%, respectively, which
satisfy the underlying constraint, less than 50%.
Then, the physician schedule is constructed accord-
works. A physician from the participating Orthopedic ing to the optimal best scheduled time intervals, as-
Surgery clinic has three different visit types: NP (new pa- suming a physician starts at 8:00 a.m. (S1 = 0) and
tients), FU (follow-up patients), and XR (patients needing the visit type sequences are FU, XR, NP, FU, . . .,
x-ray before being seen) and schedules 25 patients in a ses- and so on. Therefore, the physician schedule starts at
∗
sion (n = 25). From a 3-month data collection, we found 8:00 a.m. (S1 = 0), 8:07 a.m. (S2 = S1 + X18FU = 7.4),

Table 3. Simulation results for wait ratios (1–20) for case of orthopedic surgery clinic

Patient wait time (min) Physician idle time (min)

Wait ratio Average Maximum Average Maximum Finish time Last patient scheduled at

1 2.3 10.8 1.9 7.3 236 222

2 3.2 12.4 1.4 6.5 224 209
3 3.7 13.3 1.2 6.0 219 203
4 4.2 14.0 1.0 5.6 215 198
5 4.7 14.6 0.9 5.2 212 195
6 5.1 15.1 0.8 5.0 210 192
7 5.5 15.7 0.8 4.9 209 190
8 5.6 15.8 0.8 4.8 208 189
9 6.1 16.4 0.7 4.6 207 187
10 6.4 16.7 0.7 4.5 206 186
11 6.6 17.0 0.6 4.4 205 185
12 6.7 17.2 0.6 4.3 205 184
13 7.1 17.8 0.6 4.2 204 182
14 7.1 17.8 0.6 4.2 204 182
15 7.6 18.4 0.5 4.0 203 180
16 7.6 18.4 0.5 4.0 203 180
17 7.7 18.6 0.5 4.0 202 180
18 7.9 18.8 0.5 3.9 202 179
19 8.2 19.2 0.5 3.8 201 178
20 8.2 19.2 0.5 3.8 201 178
 104 Huang et al.

Finish Time with Various Wait Ratios Last Patient Scheduled Time with Various Wait
240 230 Ratios

220
230

Scheduled Time (min)

210
Finish Time (min)

220
Clinical Contraint Finish Time < 210 minutes (11:30 am) 200

210 190

180
200
Clinical Contraint Last Patient Scheduled
170 Time < 180 minutes (11:00 am)
190
160
Downloaded by [University of Saskatchewan Library] at 09:45 11 January 2015

180 150
1:1
2:1
3:1
4:1
5:1
6:1
7:1
8:1
9:1
10:1
11:1
12:1
13:1
14:1
15:1
16:1
17:1
18:1
19:1
20:1

1:1
2:1
3:1
4:1
5:1
6:1
7:1
8:1
9:1
10:1
11:1
12:1
13:1
14:1
15:1
16:1
17:1
18:1
19:1
20:1
Wait Ratios Wait Ratios

Fig. 3. The results of finish time with various wait ratios for the morning session shows that the best wait ratio is 7:1 for a finish time
of 11:30 a.m. or 210 minutes (left). For the last patient scheduled arrival time with various wait ratios, the results show that the best
wait ratio is 18:1 for a last patient visit before 11:00 a.m. or 180 minutes (right).

∗
8:13 a.m. (S3 = S2 + X18XR = 7.4 + 5.3 = 12.7), 8:23 a.m. uled at 8:03 a.m. (Y3 = 12.7 − 10 = 2.7), which rounds to
∗ 8:05 a.m. in reality.
(S4 = S3 + X18NP = 12.7 + 10.6 = 23.3), and so on; see
Table 4.
Finally, assuming 8 minutes for x-ray and 10 minutes
for other pre-visit activities, the patient arrival schedule is 3. Simulation result supporting importance of data
constructed as in Table 4. The first visit is FU and sched- collection
uled to see physician at 8:00 a.m. (S1 = 0), so this patient
should come in 10 minutes (B1 = 10) early for pre-visit This section is to demonstrate the impact of designing an
activities. Hence, this patient should be scheduled to ar- appointment template without considering the role of data
rive at 7:50 a.m. (Y1 = −10). The second visit is XR and collection. We actually demonstrated through our simu-
scheduled to see physician at 8:07 a.m. (S2 = 7.4), so this lation how critical even small errors in time estimation,
patient should come in 18 minutes early (10 minutes for such as one minute, can actually have on patient wait time,
pre-visit activities and 8 minutes for taking x-ray,B2 = 18). physician idle time and finish time in the course of a single
Hence this patient should be scheduled to arrive at 7:49 session. We believe that long wait times in current outpa-
a.m. (Y2 = 7.4 − 18 = −10.6), which rounds to the near- tient scheduling systems come primarily from inaccurate
est 5 minute increment to be 7:50 a.m. For the third visit, estimates of treatment time.
NP, scheduled to see physician at 8:13 a.m. (S3 = 12.7), A simulation was developed to demonstrate the im-
this patient should arrive 10 minutes (B3 = 10) early for pact of treatment time estimations on patient wait time,
pre-visit activities. Hence, the arrival time should be sched- physician idle time and total treatment time per day. The

Table 4. Patient arrival and physician schedule for case of orthopedic surgery clinic

Physician schedule The optimal scheduled Pre-visit activity scheduled Patient arrival
i Visit type (Si = Si −1 + X̄ ∗ , S1 = 0) time interval ( X̄ ∗ ) time (Bi ) (Yi = Si − Bi )

1 FU 8 :00 (0) 7.4 10 7:50 (0–10)

2 XR 8:07 (0 + 7.4) 5.3 18 (10 + 8) 7:50 (7.4–18)
3 NP 8:13 (7.4 + 5.3) 10.6 10 8:05 (12.7–10)
4 FU 8:23 (12.7 + 10.6) 7.4 10 8:15 (23.3–10)
5 XR 8:31 (23.3 + 7.4) 5.3 18 (10 + 8) 8:15 (30.7–18)
6 NP 8:36 (30.7 + 5.3) 10.6 10 8:25 (36–10)
.. .. .. .. .. ..
. . . . . .
 Improving outpatient experience
Downloaded by [University of Saskatchewan Library] at 09:45 11 January 2015
Fig. 4. A 15-minute schedule with standard deviation of 4 minutes
in actual treatment times shows significant increase on patient
wait time and a decrease in physician idle time after actual average
treatment time is at or above 15 minutes.
simulation model was based on the assumptions of esti-
mated treatment time of 15 minutes on average, that is,
a patient is scheduled every 15 minutes (X = 15), over-
all treatment time is 8 hours a day (32 patients per day
(n = 32)), and actual treatment time (Ti ) follows Gamma
Distribution with a standard deviation of 4 minutes.
The simulation results after 50 simulation runs are shown
in Figures 4 and 5. The x-axis represents the actual average
physician treatment time (µ), whereas the y-axis represents
average wait time per clinic day in minutes. As Figure 4
shows, patient wait time increases close-to-linearly, espe-
cially when the actual average treatment time is greater than
or equal to 15 minutes. In other words, under this condition
Fig. 5. Comparison of total treatment and scheduled times with
variation in actual treatment times shows overtime increases
close-to-linearly at or above 15 minutes.
105
patient wait time starts to accumulate. On the other hand,
as wait time starts to accumulate, the physician idle on aver-
age begins to drop. In short, there is an inverse relationship
between patient wait time and physician idle time.
Another major concern is whether physicians can see all
scheduled patients without working overtime or reschedul-
ing patients. Figure 5 indicates that if the true average treat-
ment time on average is 15 minutes or greater, a physician
will encounter overtime. This is compounded by the normal
variations in treatment times that can occur over a given
session. Needless to say, a physician may actually finish ear-
lier than scheduled if the actual treatment time on average
is less than the scheduled time.
The conclusion from this simulation is that failure to
determine the time interval based on actual treatment time
can significantly impact patient wait time, physician idle
time and total treatment time.
4. Case study summaries
In order to prove the efficiency of the proposed approach
in terms of average patient wait time, we implemented the
solution to three different physicians from three different
clinics. Some general details are included and discussed
for the purpose of implementing the proposed scheduling
approach. Table 5 summarized the difference of these
three clinics, the approaches taken, the clinical constraints
considered, the input parameters for simulations, no-show
rates, the proposed scheduled time interval for each
visit type, and the comparison of average patient wait
time and average physician idle time before and after
implementation.
4.1. General descriptions
Three clinics participated in this study are Orthopedic
Surgery, Plastic Surgery, and Vascular Surgery clinics. The
first clinic is a private clinic and the last two are teaching
clinics. They are located in the Southeast region of Michi-
gan. The Orthopedic Surgery clinic schedules two physi-
cians at any given clinic session along with two medical
assistants and a nurse. This clinic also offers x-ray services.
Two technicians are available at any time. This clinic is cur-
rently using a scheduling system that allows them to enter
any scheduling template to be used in three months. The
Plastic Surgery and the Vascular Surgery clinics both sched-
ule two physicians at a time with two medical assistants,
two nurses and one or two residents. Both clinics are using
a grid system for scheduling. The grid can be set as small as
one minute. All three clinics are using systems that allow us
to easily implement our solution without redesigning their
current systems. Since they are all surgery clinics, the pri-
mary functions for medical assistants including nurses are
assisting physicians, suture removal, changing dressings,
injection, and filling in prescriptions.
 Downloaded by [University of Saskatchewan Library] at 09:45 11 January 2015

106
Table 5. Detail summary results for three case studies
Cases Case 1: Orthopedic Surgery Case 2: Plastic Surgery Case 3: Vascular Surgery
Clinic Type Private Teaching Teaching
Number of Residents None One Two
Scheduling Approach Scheduled by Physician Scheduled by Scheduled by Physician
Resident Physician

Clinical Constraints 1. Last patient schedule at 11 am 1. Finish clinic at 4:45 pm 1. Finish clinic at 5 pm
2. Finish clinic at 11:30 am 2. Schedule 26 patients a day 2. Schedule 32 patients a day
3. Schedule 25 patients a session (6 NP, 13 FU, 6 XR) 3. Resident sees all NPs (8 NP, 24 RV).
6 NP, 3 POP, 4 HP, 5 RV 3 POP, 5 RV
Parameters Ti ∼ Gamma(µ, σ 2 ): (10.6, 4.52) for NP, (7.3, 4.02) for FU, (5.5, n = 26; Bi = 10; m = 50; Ti ∼ Gamma(µ, σ 2 ): Ti ∼ Gamma(µ, σ 2 ): (19.7, 9.22) for NP,
3.42) for XR; n = 25; Bi = 10 or 18 (with x-ray); m = 50. (30.6, 12.22) for NP, (15.9, 6.52) for POP, (11.1, 7.22) for POP, (12.7, (14.1, 6.12) for RV; n = 32; Bi = 20 (with
(15.9, 8.42) for RV, (18.6, 7.32) for HP 7.42) for RV resident); m = 50.
No-show Rate 9% 10% 4%
∗ ∗
Scheduled Time Interval (X) (min) Visit Type Original X(R) Proposed Visit Type Original X(R) Proposed X̄ (R = 3) Visit Type Original Proposed
and Associated Wait Ratio (R) X̄ ∗ (R ∗ = 18) X(R) X̄ ∗ (R ∗ = 9)
NP 10(58) 10.6 NP 30(25) 33.3 NP 30(1/9) 20.7
FU 5(2000) 7.4 RV 15(33) 17.5 13.8 RV 15(6) 14.6
∗ 12.7
XR None 5.3 POP 15(53) 17.4
HP 15(1500) 20.4
Average Patient Wait Time (min): Before After Before After Before After
(95% C.I.) (95% C.I.) % Reduction (95% C.I.) (95% C.I.) % Reduction (95% C.I.) (95% C.I.) % Reduction
Data Collection (3 months, 3 weeks) 27.8 13.1 53% 15.0 7.5 50% 27.8 12.4 56%
(26.9, 28.7) (10.9, 15.3) (13.1, 16.9) (4.2, 10.8) (23.8, 31.8) (8.1, 16.7)
Simulation (50 runs) 26.9 13.0 52% 15.8 7.4 53% 25.2 12.6 50%
(23.7, 30.1) (9.4, 16.6) (12.7, 18.9) (5.9, 8.9) (18.1, 32.3) (7.9, 17.3)
Simulated Average Physician Idle Before After Before After Before After
Time (min)
0.2 0.8 5.4 5.3 2.0 1.1
∗
Physician considered XR as squeeze-in patient, hence no time was assigned.
 Improving outpatient experience
4.2. Clinic conditions and constraints
Certain common clinic conditions are considered in our
simulation model for the performance evaluation such as
no-show rate (including actual no-shows, cancellation and
open slots), add-ins, patient and physician lateness (espe-
cially in the beginning of a session), x-ray conflicts, over-
write one patient type to another, and the original schedul-
ing templates (sequences) and times for these three clinics
(see Table 6). The most critical concern from these clinics
for redesigning the scheduling templates is the templates
must have the same number of slots and a similar number
of slots for each patient type to ensure they are able to
meet the current demands. Two of the clinics, Orthopedic
and Plastic Surgery, are using a four-hour session and the
Downloaded by [University of Saskatchewan Library] at 09:45 11 January 2015
107
pedic Surgery prefers to finish each session in three and a
half hours, i.e., 11:30 a.m., if possible and the last patient
should be scheduled at, for example, 11:00 a.m., to im-
pose the 11:30 a.m. finish time. The participated physician
prefers to use this last 30 minutes to dictate his patients. The
Plastic Surgery prefers to finish 15 minutes earlier if possi-
ble for wrapping up a clinic session. The Vascular Surgery
would prefer to complete a nine-hour session around 5 p.m.
4.3. Prior implementation performance and simulation
model development
Three months of data collection indicated that patient wait
times are 28 minutes on average for both Orthopedic and

Vascular Surgery is using a nine-hour session. The Ortho- Vascular Surgery and 15 minutes on average for Plastic

Table 6. Appointment scheduling templates for three clinics (a) original (b) proposed

(a) Original scheduling template

Orthopedic surgery Plastic surgery Vascular surgery

Patient Patient Patient Patient Patient Patient
type arrival type arrival type arrival

FU 8:00 POP 8:00 NP 8:00

XR 8:00 RV 8:00 RV 8:15
NP 8:05 HP 8:00 NP 8:15
FU 8:15 POP 8:15 RV 8:45
XR 8:15 RV 8:15 RV 9:00
NP 8:20 RV 8:30 RV 9:00
XR 8:25 RV 8:45 RV 9:15
FU 8:30 HP 8:45 NP 9:30
FU 8:30 POP 9:00 RV 9:30
NP 8:45 POP 9:15 RV 10:00
FU 8:55 RV 9:30 NP 10:00
FU 9:00 HP 9:30 RV 10:30
FU 9:00 RV 9:45 RV 10:30
FU 9:15 POP 10:00 RV 10:45
XR 9:15 POP 10:15 RV 11:00
NP 9:20 HP 10:15 RV 11:15
FU 9:30 RV 10:30 RV 11:30
XR 9:30 RV 10:45 RV 11:45
FU 9:45 RV 11:15 RV 11:45
NP 9:45 RV 11:30 NP 12:00
FU 10:00 NP 1:00 RV 12:00
XR 10:00 NP 1:30 NP 12:30
NP 10:15 NP 2:00 RV 12:30
FU 10:20 NP 2:30 NP 1:00
FU 10:30 NP 3:00 RV 1:15
NP 3:30 RV 1:30
NP 1:45
RV 2:00
RV 2:15
RV 2:30
RV 2:45
RV 3:00
(Continued on next page)
 108 Huang et al.
Table 6. Appointment scheduling templates for three clinics (a) original (b) proposed (Continued)

(b) Proposed scheduling template

Orthopedic surgery Plastic surgery Vascular surgery

Patient Physician Patient Patient Physician Patient Patient Physician Patient
type schedule arrival type schedule arrival type schedule arrival

FU 8:00 7:50 HP 8:00 7:50 RV 8:00 7:40

XR 8:07 7:50 RV 8:00 7:50 RV 8:15 8:00
NP 8:13 8:05 RV 8:14 8:05 NP 8:29 8:10
FU 8:23 8:15 NP 8:20 8:10 RV 8:50 8:30
XR 8:31 8:15 POP 8:54 8:45 RV 9:04 8:40
NP 8:36 8:25 RV 9:11 9:00 NP 9:19 9:00
FU 8:47 8:35 HP 9:29 9:20 RV 9:40 9:20
Downloaded by [University of Saskatchewan Library] at 09:45 11 January 2015

XR 8:54 8:35 RV 9:29 9:20 RV 9:54 9:30

FU 8:59 8:50 POP 9:43 9:35 RV 10:09 9:50
NP 9:07 8:55 NP 9:49 9:40 RV 10:23 10:00
FU 9:17 9:05 RV 10:22 10:10 NP 10:38 10:20
XR 9:25 9:05 NP 10:40 10:30 RV 10:59 10:40
FU 9:30 9:20 RV 11:13 11:05 RV 11:13 10:50
NP 9:37 9:25 HP 1:00 12:50 NP 11:28 11:10
FU 9:48 9:40 POP 1:00 12:50 RV 11:48 11:30
XR 9:55 9:40 RV 1:13 1:05 RV 12:03 11:40
FU 10:01 9:50 NP 1:20 1:10 RV 12:18 12:00
FU 10:08 10:00 RV 1:54 1:45 RV 12:32 12:10
NP 10:16 10:05 POP 2:11 2:00 NP 12:47 12:30
XR 10:26 10:10 HP 2:29 2:20 RV 13:07 12:50
FU 10:31 10:20 RV 2:29 2:20 RV 13:22 13:00
FU 10:39 10:30 POP 2:43 2:35 NP 13:36 13:20
NP 10:46 10:35 NP 2:49 2:40 RV 13:57 13:40
FU 10:57 10:45 RV 3:22 3:10 RV 14:12 13:50
FU 11:04 10:55 NP 3:40 3:30 RV 14:26 14:10
POP 4:13 4:05 RV 14:41 14:20
NP 14:55 14:40
RV 15:16 15:00
RV 15:31 15:10
NP 15:45 15:30
RV 16:06 15:50
RV 16:20 16:00
Boldface values indicate the patients seen by the physician only.

Surgery. In general, medical staff addressed their stress lev-
els are very high due to patients’ complaint about waiting
and service quality and the pressure from the physician. In
addition, when physicians are behind their schedule, they
tend to move patients along quicker, which reduces service
quality. In addition, some patients complained about the
flexibility of time slots, especially for the Plastic Surgery.
The clinic only scheduled new patients in the afternoon
session, which supports our concerns about the optimal se-
quencing of patient slots. The optimal sequencing may be
preferred by clinics but may not be flexible for patients to
schedule their appointments. Furthermore, in order to un-
derstand the current clinic situation, we built a simulation
model for each clinic that reflects current clinic conditions.
For the Orthopedic Surgery Clinic, the following condi-
tions were considered: 1) The probability of no-show rate
including cancellation and open slots is about 9%. 2) The
physician tends to be late for his first appointment by 18
minutes on average. The physician explained that since the
first patients were not generally ready to be seen at the
assigned appointment time, he simply got in the habit of
starting late. More than likely the physician’s perception
results from a discrepancy between what he assumes to be
the actual appointment time and what is in fact the arrival
time of the patient. 3) There are two physicians working
at the same time and there is only one available x-ray ma-
chine. The conflict of using x-ray has been an issue for this
clinic. 4) Patients’ lateness for appointment during the first
30 minutes is about 9 minutes on average. 5) Three patient
types are used: FU (follow-up), XR (x-ray), and NP (new).
6) Overwriting the designated slots is commonly occurred;
for example, schedule a FU patient in a NP slots.
 Improving outpatient experience
In the case of the Plastic Surgery Clinic, the conditions
considered were: 1) The probability of no-show rate
including cancellation and open slots is about 10%. 2) One
or two new patients on average are added in the afternoon
session outside the regular template. 3) Four patient types
are used: HP (pre-operation), POP (post-operation), RV
(return visit), and NP (new). 4) Overwriting the designated
slots. 5) There is normally one resident available and the
physician prefers to have the resident evaluate patients
first, especially new patients (NP). Therefore, we proposed
a schedule template based on resident’s time. Since all
pre-operation patients (HP) are only seen by the resident,
so we placed a couple of scheduling slots, especially return
visit patients (RV) and post-operation patients (POP), in
Downloaded by [University of Saskatchewan Library] at 09:45 11 January 2015
conjunction with HP for the physician to see without the
resident. Hence, we proposed a scheduling template with
6 NPs, 3 POPs, 4 HPs, and 5 RVs for the resident with the
physician and 3 POPs and 5 RVs (bolded in Table 6(b))
for the physician only. The associated best scheduled time
intervals are also different (see Table 5). It is ideal to keep
the physician busy while the resident is seeing HP patients.
Some variations in the proposed scheduling template due
to resident’s involvement are justified in this case.
As for the Vascular Surgery Clinic, the conditions con-
sidered are: 1) The probability of no-show rate is about
4%. 2) Two patient types are used: RV (return visit) and NP
(new). 3) Overwriting the designated slots. 4) The physician
is needed once or twice to help with the patients scheduled
for the nurse practitioner. This occupied about 20 minutes
of the physician’s time from his own clinic session.
4.4. Implementation reasons and challenges
The underlying reasons for improvement for these three
clinics are: 1) The design of the scheduled time intervals
is not based on actual treatment time, which leads to in-
accurate estimated treatment time intervals that cause the
waiting. 2) Clinics’ insensitivity to the wait ratio between
patient and physician wait time results in long patient wait
time. For examples from Table 5, the original wait ratios
from each visit type of Case Study 1 and 2 ranges from
R = 25 to R = 2000 or not even assigned, which are rel-
atively high compared to the proposed R ∗ . This indicates
clinics are not sensitive to patients’ waiting and prefer new
patient over other patient types. 3) Clinics allow schedulers
to overbook and double-book on the designed schedule
template to keep physician busy at all times (see Table
6(a)). Case Study 3 is a typical example where wait ra-
tio is reasonable enough to create a good patient flow, but
the patient wait time is still large because of overbooking
and double-booking. 4) Clinics’ current scheduling systems
do not differentiate between the physician schedule and the
patient arrival schedule and are not flexible enough to take
into account pre-visit activities, which significantly com-
pounds the wait time. The challenges for implementation
are very minimal. As mentioned earlier, the systems can
109
be easily adjusted to the new templates. The only change
made is when patients are scheduled to come in. Medical
staff including physicians continues to perform their nor-
mal clinic functions. Hence, the impact is relatively minimal
when implementing the proposed solution.
4.5. Post-implementation performance
The results shown in Table 5 indicate that there is no statis-
tical difference on average patient wait time between the re-
sults of simulations and actual data collections according to
95% confidence intervals for each case. This validates that
our extended simulation models reflects clinical conditions
closely. The implementation results from a three-week data
collection indicates that the percent reduction on average
patient wait time are 53%, 50%, and 56% for Case Stud-
ies 1, 2, and 3, respectively; despite the estimated average
physician idle time varies less than one minute per patient.
Table 6(b) shows how the proposed physician and patient
arrival schedules are constructed. The proposed schedul-
ing sequences are based on distributing each patient type
as evenly as possible throughout a clinic session. The re-
sults also indicate that the clinics from Case Study 1 and
2 underestimated the provider’s treatment time for all pa-
tient types while the clinic from Case Study 3 overestimated
them given the clinical constraints. Besides the fact that the
proposed solution significantly reduces average patient wait
time, the physicians from each case study are still allowed
to see and schedule exactly same number of patients within
the same time frame as before without overbooking and
double-booking.
After the implementation, the participated medical staff
in the Orthopedic Surgery noted increased patient satis-
faction with the service, especially in respect to wait time,
as well as a significant drop in the stress the staff had ex-
perienced when behind schedule. The attending physician’s
assessment of the new schedule in the Plastic Surgery is pos-
itive. She said “I’m really shocked at what a difference this
has made. I don’t think I’ve run behind since your sched-
ule has been implemented. Plus, patients have the option
of morning or afternoon appointments, which they like.”
The participated physician and medical staff in the Vascu-
lar Surgery indicated that the stress level to the end of the
day has dropped significantly. All comments indicated the
success of the proposed approach.
4.6. Sustainability guidelines and recommendation
To sustain the proposed approach, some overwriting guide-
lines were developed for schedulers to adjust for the demand
fluctuation of different visit types. For example, in the first
case of the Orthopedic Surgery, when a slot is overwritten,
certain rules must be observed: 1. NP slot can be used to
schedule any other type of patient besides XR slot since
NP slot does not consider the x-ray time. 2. XR slot cannot
be used for any other visit type, since the XR treatment
 110
time is shorter and most of the other visit types do not
require the x-ray. If a clinic wishes to increase the num-
ber of time slots due to the increasing demand, the solution
needs to be revaluate rather than overbooking patient slots.
The approach was demonstrated through the imple-
mentation of three clinics. The results are positive and
it takes less than six month time for us to implement in
these three clinics. The implementation steps are data
collection (mainly focusing on the physician treatment
time), simulation model development for the process flow,
determination of the optimal treatment time intervals
for each visit type, development of physician schedule,
development of patient schedule considering ancillary
services, and system adjustment.
Downloaded by [University of Saskatchewan Library] at 09:45 11 January 2015
5. Discussion
Before an appropriate schedule template can be produced,
it is critical to generate the best scheduled time intervals
of each wait ratio. However, to do so first requires distin-
guishing between the dominant service provider and the
provider who will dominate the schedule. In most cases,
the physician is the provider, that is, the physician’s sched-
ule will be used to determine patient arrival time. However,
in some cases, for example, a physician in training, such
as a resident, may extend the encounter or visit time, and
may, therefore, become the controlling factor in scheduling.
This illustrates how the organizational purpose of the clinic
can influence choices about best scheduled time interval for
each wait ratio. This is further illustrated by the differences
we observed between a Private Clinic and a Teaching Clinic.
In the case of a Private Clinic, since the physician’s treat-
ment time is the most critical and costly portion of the
whole visit as well as the source of variation, then “Sched-
ule by Physician” is clearly preferable for generating the
best time intervals and should dominate the schedule tem-
plate. Case Study 1 is a perfect illustration of ‘Schedule by
Physician’ and implementation demonstrated that this ap-
proach could indeed successfully reduce patient wait time
by 53% while physician average idle time increases by about
half a minute (Table 5).
In the case of a Teaching Clinic, the fact that most physi-
cians are not paid by the number of patients seen intro-
duces flexibility and, so to speak, relaxes the demands on
the schedule. However, these clinics are required to provide
as much learning opportunity for students and physicians
in training as possible. In this instance, scheduling is best
done on a case-by-case basis since the decision will tend
to depend more on how the attending physicians and the
institution elect to run a particular clinic, rather than on
the preferences of a for profit owner/physician. In insti-
tutional clinics, the number of residents or medical stu-
dents involved in scheduling, i.e., Single or Multiple Res-
ident(s)/Student(s) will influence whether to ‘Schedule by
Physician’ or ‘Schedule by Resident’ or both.
Huang et al.
In the case of a Single Resident/Medical Student, if the
physician prefers to have the resident assess all of the pa-
tients beforehand, then “Schedule by Resident” is prefer-
able, as presented in Case Study 2. Implementation results
indicate a 50% reduction in patient wait time and the resi-
dent is 52% busier while physician idle time keeps consistent
(Table 5). On the other hand, if the physician prefers to keep
busy all the time, whether the patient is seen by the resident
or not, then “Schedule by Physician” is preferable.
In the case of Multiple Residents/Medical Students,
since there will always be enough residents or medical stu-
dents, “Schedule by Physician” is always the best option,
regardless of the physician’s preference, as illustrated by
Case Study 3. Implementation results indicate that “Sched-
ule by Physician” will result in a 56% reduction in patient
wait time while keeping physician idle time down to about
1 minute on average (Table 5).
6. Conclusions
This research demonstrates a step-by-step approach for pa-
tient scheduling developed to reduce patient wait time and
enhance patient flow, but without significantly increasing
physician idle time. The effectiveness of this approach is
proven by the implementation results from three case stud-
ies. The approach allows clinic management to quickly de-
termine the best scheduled time interval for different visit
types and then integrate clinical policies or constraints to
produce two scheduling templates, one for the physician
encounter and one for patient arrival. Separating the two
schedules makes it possible to create a template for patient
arrival that takes into account any patient processing tasks
or ancillary services that need to be conducted in conjunc-
tion with a given physician service. The implementation
results from three case studies indicate the reduction on
patient wait time as much as 56%.
The three participating clinics were able to implement
the method quite quickly without putting any additional
workload on medical staff. In those cases where staff com-
plained of the approach not working effectively in the initial
trials, the obstacle turned out to be an issue of organiza-
tional behavior and perception such as overbooking and
double-booking. Once it was demonstrated to the staff
how these old habits undermine their objective, the tri-
als proceeded with great success. Physicians and staff were
quickly convinced of the effectiveness of the schedule in
maintaining good patient flow without compromising pa-
tient or physician time.
The paper focuses on developing a solution for design-
ing a scheduling template without overbooking. However,
one of the limitations this paper has not considered is
how to deal with patient no-show when design the tem-
plate. Another limitation is the study of developing a cost-
effective scheduling sequence that accommodates the pa-
tients’ needs. The other challenge now is to further extend
 Improving outpatient experience
the scheduling approach to accommodate a larger range of
specialties and clinic management systems. The implemen-
tations to date have been invaluable in testing the strength
and adaptability of the approach, and implementation in a
wider range of clinic types can help refine these qualities.
Finally, extensions of this approach for scheduling ancillary
services should be investigated.
References
Carr R. F. (2006) Outpatient clinic. NIKA Technologies, Inc., Rockville,
MD.
Cayirli, T. and Veral, E. (2003) Outpatient scheduling in health care:
A review of literature. Production and Operations Management, 12,
Downloaded by [University of Saskatchewan Library] at 09:45 11 January 2015
519–549.
Cayirli, T., Veral, E., and Rosen, H. (2006) Designing appointment
scheduling systems for ambulatory care services. Health Care Man-
age Sci, 9, 47–58.
Denton, B. and Gupta, D. (2003) Sequential bounding approach for
optimal appointment scheduling, IIE Transactions, 35, 1003–1016.
Green, L. V. and Savin, S. (2008) Reducing delays for medical appoint-
ments: A queueing approach. Operations Research, 56, 1526–1538.
Guo, M., Wagner, M., and West, C. (2003) Outpatient clinic scheduling –
A simulation approach. Proceedings of the 2004 Winter Simulation
Conference, Washington, DC.
Harper, P. R. and Gamlin, H. M. (2003) Reduced outpatient waiting
times with improved appointment scheduling: a simulation model-
ing approach. OR Spectrum, 25, 207–222.
Ho, C. and Lau, H. (1992) Minimizing total cost in scheduling outpatient
appointments. Management Science, 38, 1750–1764.
Ho, C., Lau, H., and Li, J. (1995) Introducing variable-interval appoint-
ment scheduling rules in service systems. International Journal of
Operations & Production Management, 15, 59–68.
Huang, X. (1994) Patient attitude towards waiting in an outpatient clinic
and its applications. Health Services Management Research, 7, 2–8.
Kaandorp, G. and Koole, G. (2007) Optimal outpatient appointment
scheduling. Health Care Management Science, 10, 217–229.
Keller, T. F. and Laughhunn, D. J. (1973) An application of queuing
theory to a congestion problem in an outpatient clinic. Decision
Sciences, 4, 379–394.
111
Kim, S. and Giachetti, R. E. (2006) A stochastic mathematical ap-
pointment overbooking model for healthcare providers to improve
profits. IEEE Transactions On Systems, Man, and Cybernetics, 36,
1211–1219.
Klassen, K. J. and Rohleder, T. R. (1996) Scheduling outpatient ap-
pointments in a dynamic environment. Operations Management,
14, 83–101.
Klassen, K. J. and Rohleder, T. R. (2004) Outpatient appointment
scheduling with urgent clients in a dynamic, multi-period environ-
ment. International Journal of Service Industry Management, 15,
167–186.
LaGanga, L. R. and Lawrence, S. R. (2007) Clinic overbooking to im-
prove patient access and increase provider productivity. Decision
Sciences, 38, 251–276.
Meza, J. (1998) Patient waiting times in a physician’s office. The American
Journal of Managed Care, 4, 703–712.
Murray, M. and Tantau, C. (1999) Redefining open access to primary
care. Managed Care Quarterly, 7, 45–55.
Muthuraman, K. and Lawley, M. (2008) A stochastic overbooking model
for outpatient clinical scheduling with no-shows. IIE Transactions,
40, 820–837.
National Correct Coding Initiative Correct Coding Solutions, LLC.
(2009) National Correct Coding Initiative Policy Manual for Medi-
care Services. Version 16.3 Effective Date: October 1, 2010. Ameri-
can Medical Association, Washington, DC.
O’Hare, C. D. and Corlett, J. (2004) The outcomes of open access schedul-
ing. Family Practice Management, 11, 35–38.
Swisher, J. R., Jacobson, S. H., Jun, J. B., and Balci, O. (2001) Modeling
and analyzing a physician clinic environment using discrete-event
(visual) simulation. Computers & Operations Research, 28, 105–125.
Vanden Bosch, P. M. and Dietz, D. C. (2000) Minimizing expected
waiting in a medical appointment system. IIE Transactions, 32,
841–848.
Wijewickrama, A. and Takakuwa, S. (2005) Simulation analysis of
appointment scheduling in an outpatient department of internal
medicine. Proceedings of the 2005 Winter Simulation Conference,
Orlando, FL.
Wijewickrama, A.K.A. (2006) Simulation analysis for reducing queues in
mixed-patients’ outpatient department. International Journal Simu-
lation Model, 5, 56–68.
Yang, K. K., Lau, M. L. and Quek, S. A. (1998) A new appointment rule
for a single-server, multiple-customer service system. Naval Research
Logistics, 45, 313–326.