International Journal of Production Research

ISSN: 0020-7543 (Print) 1366-588X (Online) Journal homepage: http://www.tandfonline.com/loi/tprs20

A co-availability scheduling model for coordinating

multi-disciplinary care teams

Serpil Mutlu, James Benneyan, John Terrell, Victoria Jordan & Ayten Turkcan

To cite this article: Serpil Mutlu, James Benneyan, John Terrell, Victoria Jordan & Ayten
Turkcan (2015) A co-availability scheduling model for coordinating multi-disciplinary
care teams, International Journal of Production Research, 53:24, 7226-7237, DOI:
10.1080/00207543.2015.1018452

To link to this article: http://dx.doi.org/10.1080/00207543.2015.1018452

Published online: 20 Mar 2015.

Submit your article to this journal

Article views: 253

View related articles

View Crossmark data

Full Terms & Conditions of access and use can be found at

http://www.tandfonline.com/action/journalInformation?journalCode=tprs20

Download by: [Indian Institute of Technology - Delhi] Date: 15 July 2017, At: 14:59
International Journal of Production Research, 2015
Vol. 53, No. 24, 7226–7237, http://dx.doi.org/10.1080/00207543.2015.1018452

A co-availability scheduling model for coordinating multi-disciplinary care teams

Serpil Mutlua, James Benneyana*, John Terrellb, Victoria Jordanb and Ayten Turkcana
a
Healthcare Systems Engineering Institute, Colleges of Engineering and Health Sciences, Northeastern University, Boston MA, USA;
b
MD Anderson Cancer Center, Houston TX, USA
(Received 30 April 2014; accepted 23 January 2015)

We introduce a co-availability scheduling problem that arises in various healthcare settings in which personnel from dif-
ferent disciplines work together as care teams and for which synchronisation of their availability impacts scheduling flex-
ibility and procedure timeliness. Examples include breast cancer surgery involving oncologic and plastic surgeons,
primary and specialty care integrated visits, and vascular interventions involving cardiac surgeons, radiologists and radi-
ology technicians. We develop an integer programming model to help create optimal schedules that maximise the amount
of co-available time across the scheduling templates of the desired team members, while still satisfying each of their
clinic coverage, preference and extraneous responsibilities constraints. Application to breast surgery at a major cancer
centre increased team co-availability by 94%, with sensitivity analysis in other scenarios producing 64–152%, increases
in favourable team assignments, and without negatively affecting operating room neither utilisation nor surgery delays.
Keywords: healthcare; personnel scheduling; multi-disciplinary teams

1. Introduction
Coordinating multi-disciplinary healthcare teams is an increasingly common need in many medical contexts (Taylor
et al. 2010). Examples include multiple morbidities (Goldstein et al. 2004; Kuzma et al. 2008; Stocker 2003), chronic
care (Kasper et al. 2002; Levetan et al. 1995; Wagner 2000), cancer care (Forrest et al. 2005; Ruhstaller et al. 2006;
Sidhom and Poulsen 2006), mental health (Burns 2004; Liberman et al. 2001; Van den Berg et al. 2005) and surgical
care (Catchpole et al. 2008; ten Cate et al. 2004; Hu 2008). Forming the most effective teams is also important to
efficiency, care quality, safety, provider and patient satisfaction and disease management (Bunnell et al. 2010; Carter,
Garside, and Black 2003; Committee on Quality of Health Care in America, Institute of Medicine 2000; Davenport
et al. 2007; Gabel, Hilton, and Nathanson 1997; Makary et al. 2006; Mazzocco et al. 2009; Sexton et al. 2006).
Collectively synchronising schedules of a large number of staff in multiple departments, however, such that all
individuals of each desired team subsequently are co-available when needed, subject to their other weekly responsibili-
ties and scheduling constraints is a difficult problem, especially in complex settings involving multiple specialists (Kane,
O’Byrne, and Luz 2010; Kane et al. 2007; Wagner et al. 2001).
The resulting inefficient team scheduling has been shown to negatively impact workflow in radiology and pathology
(Kane et al. 2007), co-scheduling of dermatologic oncology specialists (Caudron et al. 2010), schedule matching in neo-
natal intensive care units (Brown et al. 2003), and care for diabetes (Ritholz et al. 2011), cancer (Wright et al. 2009),
palliative radiotherapy (Fairchild et al. 2009) and psychiatry (Fiddler et al. 2010) patients. In contrast, the value of effi-
cient teams has been well-studied in the literature, some via randomised studies, including the benefits on patient out-
comes (Goldstein et al. 2004), readmissions (Kasper et al. 2002), length of stay (Levetan et al. 1995), survival (Forrest
et al. 2005), quality of life (Van den Berg et al. 2005), patient safety (Catchpole et al. 2008), and timely care and patient
and provider satisfaction (Bunnell et al. 2010; Gabel, Hilton, and Nathanson 1997).
Rather than schedule day-to-day cases directly (e.g. surgeries), we introduce a schedule template optimisation model
that maximises the amount of time during which required or well-matched people are ‘co-available’, making subsequent
procedure scheduling itself easier and less constrained. The general idea is that by maximising mutually available
unscheduled times, such as by optimally scheduling other activities (e.g. teaching responsibilities, clinic coverage,
release time, etc.), healthcare procedures can be scheduled more efficiently, and with more desirable team compositions.
Whereas, many studies that primarily focus on scheduling rooms or procedures (e.g. Cardoen, Demeulemeester, and
Belien 2010; Cayirli and Veral 2003) assume provider schedules are given, our problem is more of a staff timetable

*Corresponding author. Email: j.benneyan@neu.edu

© 2015 Taylor & Francis

 International Journal of Production Research 7227

template problem (Ernst et al. 2004). To our knowledge, no studies have considered this problem of simultaneously
coordinating individual scheduling templates in order to ease subsequent scheduling difficulties and increase team
composition quality. Personal preferences, historical data on high functioning teams and technical or personality
incompatibility also can be taken into account in this approach.

2. Optimisation model
We define co-availability as the per cent of time slots in which all required or desired individuals are simultaneously
available, such that at some later time they could be scheduled to work together on patient procedures. We formulate
this problem as an integer programming optimisation model, with a group Gi defined as the set of all providers from the
same discipline i who will be scheduled into teams that each perform one of T distinctive types of tasks (see Figure 1),
and with the objective function seeking to maximise within-team co-availability subject to coverage, obligatory tasks,
personal requests and preference constraints. Table 1 summarises model inputs, notation and definitions.
The planning period is divided into S time slots across which each individual m ∈ Gi must spend a fixed number of
time slots Rmt performing each task (t = 1, … , T). Coverage requirements exist for each task in each time slot, in which
a fixed number of members Cit from skill group Gi must be assigned to perform task t. The set Oi is defined as all
obligatory tasks for members of skill group Gi such that individual m must perform task t during time slot s.

Figure 1. Relationship of personnel skill groups and multi-disciplinary teams.

Table 1. Input notation used in IP model.

I Number of groups that consist of people with the same skills, where index i ∈ {1, 2, … ,I}
R Number of teams that consist of people from different groups performing a task together, where index r ∈ {1, 2, … ,R}
m, n Subscripts denoting members of a group or a team
T Number of types of tasks, where index t ∈ {1, 2, … ,T}
S Number of time slots, where index s ∈ {1, 2, … ,S}
k Subscript denoting number of members in partial or whole teams
Gi Set of providers from the same discipline with the same skill set
Dimt Set of members with whom member m of group i works together for the common task t
Uimt Set of members with whom member m of group i cannot work together for the common task t
Oi Set of obligations for members of group i, Oi ={(m,t,s)}, member m should perform task t in time slot s
Qt Set of teams, members of which prefer to work together for task t
Qtr rth team in set Qt
Rmt Total number of time slots member m must allocate to task t
Cit Total number of members from group i that are required to cover each time slot for task t
wkt Increase in objective function by scheduling k members of a given team for task t in the same time slot
Ktr Number of team members of rth team in set Qt
Goalt Minimum number whole teams that must be assigned to task t in each time slot
7228 S. Mutlu et al.

Some providers additionally may have strong preferences or requirements regarding with whom they work, either
due to technical compatibility or personality dynamics. Preferred and not preferred individuals by member m of group i
for task t are represented by the sets Dimt and Uimt, respectively.
Defining Qt as the set of R teams Qtr , r = 1, … , R, whose membership is desirable to perform task t for any of the
above reasons, the co-availability problem then assigns providers to perform each task at each time slot in a manner that
maximises the number of co-available desirable unscheduled teamings, while satisfying all other coverage, obligations,
preference and compatibility constraints. Since partial team assignments may still create partial value, weight coefficients
0 ≤ wkt ≤ 1 denote the value of k members of a desired team being co-available for task t. This also causes the IP model
to assign as many team members together as possible, where pktrs denotes k members of team r are assigned to task t in
time slot s, the weight factors can be chosen to reflect the relative value of fewer whole or partial teams, and the param-
eter Goalt is defined as the minimum number of whole teams that must be assigned to task t.
The integer programming model then is formulated as follows:

X XXX
Max wkt  pktrs (1)
k
Ktr t r s

X
s.t. Amts
1 8i; s and 8m 2 Gi (2)
t

X
Amts ¼ Rmt 8i; t and 8m 2 Gi (3)
s

Amts ¼ 1 8ðm; t; sÞ 2 Oi (4)

X
Ants  Amts 8i; t; s and 8m 2 Gi ; Dimt 6¼ ; (5)
n2Dimt

Ants þ Amts
1 8i; t; s and 8m 2 Gi ; 8n 2 Uimt (6)

X
Amts  Cit 8i; t; s where m 2 Gi (7)
m

X X
Amts ¼ k  pktrs 8 t; r; s where m 2 Qtr (8)
m k
Ktr

X
pKtr trs  Goalt 8 t; s (9)
r

Amts 2 f0; 1g 8 m; t; s (10)

pktrs 2 f0; 1g 8 k; t; r; s (11)

where the decision variable Amts = 1, if member m is assigned to task t for time slot s (and 0 otherwise) and the decision
variable pktrs = 1 if k members of team r are assigned together to task t for time slot s (and 0 otherwise).
The objective function in Equation (1) maximises the total weighted number of unscheduled common time slots of
compatible or required team members over all tasks. Constraints (2)–(3) are assignment constraints that ensure each
group member is allocated to a time slot for at most one task, and that each assigned group member m spends Rmt time
units for task t, respectively. Constraint (4) ensures fulfilment of all obligatory tasks. Constraints (5)–(6) define
compatibility requirements and provider preferences. Specifically, constraint (5) ensures that all time slots of each
member m who is assigned to the common task t coincide with task t time slots of members that belong to his/her
Dt set, meaning that at least one member with whom it is desirable to work on task t is available for each provider.
 International Journal of Production Research 7229

Constraint (6) ensures that tasks of each member m are scheduled to different time slots than those of the members that
belong to his/her Ut undesirable set, so that none of the strictly-not-preferred members are scheduled to work together
on task t.
Constraint (7) is a coverage constraint to ensure that at least Cit members of group i are assigned to cover task t in
each time slot. Note that the coverage parameter Cit could be extended to include an additional subscript, Cits, if require-
ments change according to time slot. Constraint (8) determines if a whole (k = Ktr) or partial (k < Ktr) team is assigned
to a time slot for any defined team with any number of members. Constraint (9) ensures that at least Goalt whole teams
are assigned to task t in every time slot and avoids team accumulation in some time slots. Constraints (10)–(11) are
standard binary integer assignment constraints. To solve the below examples, this IP model was implemented in CPLEX
12.1 on an Intel Core 2 CPU with 4.00 GB RAM.

3. Case study: breast cancer surgery

We used the above model in a major cancer centre to schedule the co-availability of oncology and plastic surgeons per-
forming both mastectomy and reconstruction during the same case, where removal of breast tissue by an oncologic sur-
geon is followed by cosmetic reconstruction by a plastic surgeon under the same episode of anaesthesia (Figure 2)
(Scanlon 1991). Scheduling large number of such cases (roughly 110 per week across 8–10 operating rooms) requires
sufficient time slots in which both types of compatible specialists are available.
The cancer centre has 13 oncology and 18 plastic surgeons, respectively, performing tissue removal and reconstruc-
tion (denoted here as O1, O2, … , O13 and P1, P2, … P18) each working morning and afternoon sessions 5 days a
week in surgery or some other task (e.g. clinic appointments, research, administrative work and teaching). Since all
teams in this application are composed of two members (Ktr = 2 8r), the weight factor used was wkt = 1. Typically, most
surgeons work two days a week in surgery, two days in clinic, and one day on other tasks. Plastic surgeons perform
reconstruction of breast and other types of patients (e.g. throat, neck, ears, etc.), with on average 40% of their surgery
time spent on composite breast cases, 25% on other composite cases, and 35% on elective operations (where only a
plastic surgeon is involved), although these percentages are heterogeneous across surgeons. Of the current plastic sur-
geons in this cancer centre, six spend all their surgery time on composite breast cases, eight most of their surgery time,
and the remaining three only a little time.
Figure 3 summarises weekly available operating rooms for breast surgeries. Since some complex cases require a full
day, weekly schedules should include some days in which surgeons are assigned to surgery for both morning and after-
noon sessions, referred to here as whole day assignments. This is represented in our model by forcing some per cent of
team assignments to be consecutive.
On average, historically three breast surgeons are in clinic each day, so we assumed in the model that at least two
breast surgeons must be in clinic each day, with other non-clinic tasks deferrable to days, the surgeon is not performing
surgeries nor seeing patients in clinic. For preference relationships, three pairs of surgeons prefer not to work with each
other (Surgeon P2 and O11, P12 and O2, P12 and O12), with 10 pairs historically performing better than others ((O2,
P11), (O6, P7), (O6, P8), (O8, P11), (O8, P4), (O12, P12), (P3, O6), (P3, O4), (P3, O3) and (P3, O13)). In this particu-
lar application, no obligatory tasks were identified by the surgeons. Using these inputs, the resulting optimal weekly
schedule is shown in Figure 4, with the optimal solution resulting in a 94% increase in total desired co-availability
improving from 18 to 35 oncology-plastic surgeon pairings. This result represents a significant improvement in cancer
team schedules.

Figure 2. Breast cancer oncology and reconstruction composite surgery.

7230 S. Mutlu et al.

MON TUE WED THU FRI

AM PM AM PM AM PM AM PM AM PM
1
2

Operating Rooms (k)

3
4
5
6
7
8 O9
9 O8 O8 O1 O9 O9 O1
10 O9 O9 O1

Figure 3. Breast centre OR availability. Between 7 and 10 ORs are assigned to breast surgeries in each time slot. Black cells indicate
unavailable ORs and coloured cells indicate ORs assigned to specific oncology surgeons (e.g. OR 9 is assigned to oncology surgeon
O8 on Tuesdays).

Monday Tuesday Wednesday Thursday Friday

AM PM AM PM AM PM AM PM AM PM

O11 A C A A S S A A S S
O2 S S C C A A C C A A
O3 S S C C A A A A S S
O4 S S A C C A C C S S
O5 A C S A A S A C S S
O6 S S A C C A A A S S
O7 A A C C S S S A A S
O8 S S S S A C C C A C
O9 C A S S A C S S C A
O10 C A S S S A A A C S
O11 S S C A A C A A A A
O12 S A S S A S S S A C
O13 S S C A C A C C S S

P1 C S S A S S C A A A
P2 S S A C C A A A S S
P3 A A S C A C S S A S
P4 S S S S C C A A A A
P5 S S S S A A A A C C
P6 A S S S A A S C C A
P7 S S A A A A C C S S
P8 S A A A C A S C S S
P9 C S S A A C S S A A
P10 A A S S A A S S C C
P11 S S C S A A A S C C
P12 A A C C S A S A S S
P13 S S A C S C S A A A
P14 C S A A S C S A A S
P15 S A S S S A A C C A
P16 C C S S S A S S A C
P17 C C A A A A S S S S
P18 S S S S S S C C A A

Figure 4. Optimal co-available solution of breast cancer surgery case study (with set of task types A: administrative, C: clinic and S:
breast or other surgery).
 International Journal of Production Research 7231

(a) Mon Tue Wed Thu Fri

AM PM AM PM AM PM AM PM AM PM

O4

P2

O6

P8

Total 1 1 1 1 3 3 1 1 3 3 18

(b) Mon Tue Wed Thu Fri

AM PM AM PM AM PM AM PM AM PM

O4

P2

O6

P8

Total 9 8 2 3 0 0 1 0 6 6 35
Surgery Clinic Admin. Task

Figure 5. Example of improved co-availability scheduling template. (a) Current schedules of oncology (O) and plastic (P) surgeons
O4, P3, O6 and P8, showing only four co-available time slots (with 18 total co-availabilities for the whole schedule). (b) Optimised
schedules of oncology (O) and plastic (P) surgeons O4, P3, O6 and P8, now showing 11 co-available time slots (with 35 total
co-availabilities for the whole schedule).

While Figure 4 shows the full schedule, to provide some insight to the solution Figure 5 illustrates the improved
co-availability for certain pairs composed of four specialty surgeons (O4, P2, O6 and P8), with the ovals highlighting
co-availability. The bottom row of each part of Figure 5 shows the resulting total number of desirable team assignments
across all surgeons in the full model.

4. Performance
4.1 Breast cancer surgery case study
To evaluate how these optimised templates would perform in practice, we simulated the actual scheduling of breast can-
cer surgeries with random demand assumed to be Poisson estimated from 17 months of historical data (January 2011
through May 2012). To place these surgeons into scheduling templates some logic assumptions needed to be made.
Cases were scheduled for the first available time slot of the assigned surgeon with an available OR, with two wait time
scenarios assumed for composite cases, summarised in Figure 6. In the first scenario (referred to here as less waiting or
LW), composite patients are scheduled for the first availability of the assigned breast surgeon, after which an attempt is
made to co-schedule any compatible plastic surgeon; if none is available then any non-compatible plastic surgeon is
scheduled.
The second scenario (referred to as more matching or MM) assigns a composite case to the first time slot at which
both the assigned breast surgeon and any compatible plastic surgeon are available. In both scenarios, if there is no
option in the given week then the case is scheduled in a subsequent week to the first availability of the assigned breast
surgeon partnered with any available plastic surgeon.
While neither simulation logic exactly matches the actual process, they provide some sense for how the performance
of current vs. proposed schedules compare. In practice, OR requests for composite cases are made by breast surgeon
scheduling assistants after finding candidate dates with plastic surgeon scheduling assistants. Hence, the MM strategy is
closer to the actual process, although randomly choosing between options that fit the requirements for the surgery.
The simulation was coded in the C programming language and run for five replications of 52 weeks each to com-
pare the current and optimal schedules. Add-on, emergency and emergency return cases were ignored, since scheduled
7232 S. Mutlu et al.

Assign first available

time of breast
Breast surgeon and OR

(a)
(a)
Find first available
Assign a random
time when both an
plastic surgeon with
OR and breast
(a) determined OR and
surgeon are

Composite
Any matching N breast surgeon
Surgery request Breast, available
in that time slot?
(randomly determined type composite, or
Assign a random
and primary surgeon) other? Find first available
Found? first available plastic
time when both an
(b) surgeon and OR
OR and matched
with determined
surgeons are
breast surgeon
available Y
(b)
(b)

Assign matched
Other Assign first available surgeons and OR at
time of other found time
surgeon and OR

Figure 6. Simulation logic for breast cancer surgery case study: (a) LW surgery scheduling strategy, (b) MM surgery scheduling
strategy.

300 2000 600

Composite Other Breast
Number of surgeries
Number of surgeries
Number of surgeries

250 500
1500
200 400

150 1000 300

100 200
500
100
50

0 0
0
120
220
320
420
520
620
720
20
130
240
350
460
570
680
20
120
220
320
420
520
620
720
20

Surgery duration (min) Surgery duration (min)

Surgery duration (min)

Figure 7. Durations for different types of cancer surgeries.

cases account for over 95% of all volume. Historical frequencies and durations for breast, composite and other cases are
summarised in Figure 7 and Table 2. Since no standard probability distribution was a good fit to actual surgery duration
data, tabulated values were used to define an empirical distribution function.
Figure 8 summarises the increased number of surgeries assigned each week to desirable teams, with the LW and
MM strategies resulting in 152 and 64% improvements, respectively, and with no negative consequence on OR utilisa-
tion (Figure 9) nor patient waits (Figure 10). (Note that the remaining utilisation typically is absorbed by the omitted
emergency and add-on cases.) The difference in performance of the two simulation logics provide a general feel for the
improvement that would result in actual practice.

4.2 Other examples

To develop insight into potential improvements in other settings, we studied the effect of several factors (size of person-
nel groups, number of obligatory tasks and balancing goal value) in a similar hypothetical scenario, as summarised in
 International Journal of Production Research 7233

Table 2. Scheduled vs. actual case durations.

Scheduled duration Actual range Breast (%) Composite (%) Other (%)

60 20–60 6.49 0.03 18.63

90 65–90 6.04 0.16 6.56
120 95–120 6.52 0.23 6.03
150 125–150 3.43 0.12 3.32
180 155–180 3.53 1.29 3.54
210 185–210 0.60 0.53 0.93
240 215–240 1.30 2.99 5.12
300 245–300 0.63 1.55 3.58
360 315–360 0.94 2.19 4.11
480 375–480 0.48 1.41 5.39
600 500–780 0.09 0.34 1.88

10 10

Average # of desired matches

Average # of desired matches

9 current 9 current
8 proposed 8 proposed
7 7
6 6
5 5
4 4
3 3
2 2
1 1
0 0
1 5 9 13 17 21 25 29 33 37 41 45 49 1 5 9 13 17 21 25 29 33 37 41 45 49
Weeks Weeks
(a) LW scheduling strategy (b) MM scheduling strategy

Figure 8. Improvement in weekly number of desired team assignments for the (a) LW scheduling strategy and (b) MM scheduling
strategy.

16 15 16 15
14 14
14 14 13
Number of weeks

Number of weeks

12 current 10 12 current 10 10
10 proposed 9 9 9 10 proposed
8
8 7 8 7 7
6 6 6
6 5 5 6 5
4 4
4 4 3 3
2
2 1 2 1
0 0

OR Utilization OR Utilization
(a) LW scheduling strategy (b) MM scheduling strategy

Figure 9. Weekly OR utilisation under the (a) LW scheduling strategy and (b) MM scheduling strategy (no statistical significance,
paired t test, p > 0.90).

Table 3, here requiring the co-availability of a surgeon and an anaesthesiologist (Ktr = 2 8r). Table 4 compares the num-
ber of hours that each class of personnel (surgeons, anaesthesiologists) allocate to various tasks (Rmt), with each clini-
cian also classified by clinical or research orientation (Leitch and Walker 2000; Soh 1998; Winslow, Bowman, and
Klingensmith 2004). The listed values for the balancing goal constraint Goalt indicate the minimum number of available
whole teams in each time slot. Although greater balanced team availabilities may be desirable, this may not always be
7234 S. Mutlu et al.

60% 60%

Percentage of patients

Percentage of patients
50% 50%

40% 40%

30% current 30% current

proposed
20% proposed 20%

10% 10%

0% 0%
1-3 4-7 8-14 15-21 22-30 31-55 1-3 4-7 8-14 15-21 22-30 31-55
Waiting days Waiting days
(a) LW scheduling strategy (b) MM scheduling strategy

Figure 10. Patient waiting days for the (a) LW scheduling strategy and (b) MM scheduling strategy (no statistical significant differ-
ence, Kolmogorov–Smirnov test, p > 0.99).

Table 3. Factors varied in sensitivity analysis study also showing number of each type of clinicians in each size level.

Factor 1: Personnel Pool

Surgeon type Anaesthesiologist type
Factor 2: Balancing Factor 3:
Size 1 2 3 1 2 goal (Goalt) Obligations (Oi)

Small 3 3 2 4 3 0, 1, 2, 3 1%, 2%, 3%

Medium 5 4 3 5 5
Large 6 5 5 7 6
XL 12 9 9 13 12

Table 4. Number of hours surgeon and anaesthesiologists are allocated to tasks, by clinical vs. research orientation.

Surgeon with Surgeon in Surgeon with Anaesthesiologist with Anaesthesiologist with

clinical orientation between research orientation clinical orientation research orientation
Tasks Type 1 Type 2 Type 3 Type 1 Type 2

Administration 1 1 1 1 1
Research 1 2 3 1 2
Teaching 1 2 3 1 2
Clinical duties 4 4 3 3 3
Operating 5 3 2 6 4
Total 12 12 12 12 12

feasible; similarly the number of obligated time slots can limit the solution. Obligation levels indicate how flexible a set-
ting is, since they are used to determine how many of the assignments are fixed while generating the set of obligatory
tasks.
For the coverage requirements Cit, we assumed that at least one surgeon of each type is available throughout the
week for clinic appointments and another for performing surgeries. We considered three types of pair relationships
between surgeons and anaesthesiologists: Of all surgeons 40% must work with at least one of some individuals (type-1),
20% cannot work with some individuals (type-2) and 40% prefer to work with some individuals (type-3). These per-
centages were used to randomly choose members of the Dimt, Uimt and Qt sets, respectively.
Five problem instances were randomly generated and solved for each of the 48 factor combinations, with the optimi-
sation program terminated if an optimal solution was not found after 4 h. (Some factor–level combinations reflect set-
tings with extensive requirements that are not possible to be satisfied as a whole.) Figure 11 summarises the impact of
larger problem sizes, percentages obligatory tasks and goal levels on problem feasibility. Figure 12 summarises the (a)
average computation times for problems that were solved optimally and (b) optimality gap from the upper bound for
those that were not, both appearing reasonably good. As would be expected, larger percentages of obligatory tasks and
 International Journal of Production Research 7235

100% 100%

Percent generated problems

XL Near-optimal

Percent generated problems

large Goal=3
80% 80% Goal=2
medium
small 25% Goal=1
60% 60% Goal=0

20%
40% 25% 40%
15%
20% 15% 20%
5% 23%
5% 5%
0% 5%5% 0%
1% 2% 3% small medium large XL
(a) Percentage infeasible problems by obligatory task levels (b) Percentage of feasible problems by problem size

Figure 11. (a) Percentage infeasible problems by obligatory task levels and (b) percentage of feasible problems by problem size.

250 5
Goal=0
Goal=0 4.5

Average relative gap (%)

Average CPU time (min)

Goal=1
Goal=1 4
200 Goal=2
Goal=2
3.5 Goal=3
Goal=3
150 3
2.5
100 2
1.5
50 1
0.5
0 0
small medium large XL small medium large XL
(a) CPU time by problem size (b) Optimality gap by problem size

Figure 12. Model performance in terms of (a) CPU time by problem size and (b) optimality gap by problem size.

problem sizes result in more infeasible problems (11a), larger goal levels result in challenges to solve problems only to
near-optimality within the time limit (11b), larger problem sizes and goal levels result in longer average computation
time (12a) and smaller goal values result in smaller average gaps (12b). No XL problem could be solved to optimality
(12a), and similarly no small-size problem was solved only to near optimality (12b). An unpredicted result is the
decrease in the optimality gap when problem size changes from ‘large’ to ‘XL’, since one would expect the solutions to
be further from the upper bound as the problem size increases, although this apparent improvement may be a result of
using relative gap, where the difference is reported as a percentage. Additionally, further improvements from further
search in branch-and-bound tree should be relatively smaller when the objective value is large.

5. Discussion
Multi-disciplinary teams are increasingly important in many healthcare and non-healthcare contexts, with co-scheduling
often being a difficult problem due to asynchronous availability of necessary or desired team members. While many
studies focus on scheduling specific cases to rooms and personnel, limited co-availability often causes even these ‘opti-
mal’ schedules to be fairly poor or to require effort to reschedule conflicts. In such cases, focusing first on improving
the more macro scheduling templates can allow for the eventual day-to-day schedules to improve significantly, as well
as streamline scheduling efforts that often can be largely manual and time-intensive. As demonstrated across a range of
applications, optimising co-availability schedule templates can enable more efficient case scheduling, with significant
increases in desired team compositions and without adversely affecting OR utilisation nor patient waits.
Moreover, although implementation of many healthcare optimisation models can be technically and culturally chal-
lenging to implement on an ongoing basis, off-line solution of models of the type described here may be easier to
implement than those requiring integrating into daily practice. Finally, note that our simulation assumes patients are
scheduled via a first-come-first-scheduled rule, assigning them to the first available time, whereas some patients in
7236 S. Mutlu et al.

reality may request later dates based on to their schedules. Results nonetheless give a fairly good indication of the
improvement potential of this approach.

Funding
This research was supported in part by National Science Foundation (NSF) [grant number IIP-10341990]; Center for Medicare and
Medicaid Innovation (CMMI) [grant number 1C1CMS331050]. Any statements or opinions expressed herein are those of the authors
and do not necessarily reflect those of NSF nor CMMI.

References

Brown, M. S., J. Ohlinger, C. Rusk, P. Delmore, and P. Ittmann. 2003. “Implementing Potentially Better Practices for Multidisciplinary
Team Building: Creating a Neonatal Intensive Care Unit Culture of Collaboration.” Pediatrics. 111 (Suppl. E1): e482–e488.
Bunnell, C. A., S. N. Weingart, S. Swanson, H. J. Mamon, and L. N. Shulman. 2010. “Models of Multidisciplinary Cancer Care:
Physician and Patient Perceptions in a Comprehensive Cancer Center.” Journal of Oncology Practice 6 (6): 283–288.
Burns, T. 2004. “Community Mental Health Teams.” Psychiatry 3 (9): 11–14.
Cardoen, B., E. Demeulemeester, and J. Belien. 2010. “Operating Room Planning and Scheduling: A Literature Review.” European
Journal of Operational Research 201 (3): 921–932.
Carter, S., P. Garside, and A. Black. 2003. “Multidisciplinary Team Working, Clinical Networks, and Chambers; Opportunities to
Work Differently in the NHS.” Quality and Safety in Health Care 12 (Suppl 1): i25–i28.
Catchpole, K., A. Mishra, A. Handa, and P. McCulloch. 2008. “Teamwork and Error in the Operating Room.” Annals of Surgery 247
(4): 699–706.
ten Cate, G., E. Fosse, P. K. Hol, E. Samset, R. W. Bock, J. F. McKinsey, B. J. Pearce, and M. Lothert. 2004. “Integrating Surgery
and Radiology in One Suite: A Multicenter Study.” Journal of Vascular Surgery 40 (3): 494–499.
Caudron, A., G. Chaby, A. Dadhan, C. Andrejak, F. Dhaille, M. Bagot, and C. Lok. 2010. “Multidisciplinary Team Meetings in
Oncology: First Analysis of Benefits and Evaluation of Activity in a Dermatology Unit in France.” European Journal of
Dermatology 20 (6): 778–784.
Cayirli, T., and E. Veral. 2003. “Outpatient Scheduling in Health Care: A Review of Literature.” Production and Operations
Management 12 (4): 519–549.
Committee on Quality of Health Care in America, Institute of Medicine. 2000. “To Err is Human: Building a Safer Health System.”
edited by L. T. Kohn, J. M. Corrigan, and M. S. Donaldson. Washington, DC: The National Academies Press.
Davenport, D. L., W. G. Henderson, C. L. Mosca, S. F. Khuri, and R. M. Mentzer Jr. 2007. “Risk-adjusted Morbidity in Teaching
Hospitals Correlates with Reported Levels of Communication and Collaboration on Surgical Teams but Not with Scale Mea-
sures of Teamwork Climate, Safety Climate, or Working Conditions.” Journal of the American College of Surgeons 205 (6):
778–784.
Ernst, A. T., H. Jiang, M. Krishnamoorthy, and D. Sier. 2004. “Staff Scheduling and Rostering: A Review of Applications, Methods
and Models.” European Journal of Operational Research 153 (1): 3–27.
Fairchild, A., E. Pituskin, B. Rose, S. Ghosh, J. Dutka, A. Driga, P. Tachynski, et al. 2009. “The Rapid Access Palliative Radiother-
apy Program: Blueprint for Initiation of a One-stop Multidisciplinary Bone Metastases Clinic.” Supportive Care in Cancer 17
(2): 163–170.
Fiddler, M., G. Borglin, A. Galloway, C. Jackson, L. McGowan, and K. Lovell. 2010. “Once-a-week Psychiatric Ward round or Daily
Inpatient Team Meeting? A Multidisciplinary Mental Health Team’s Experience of New Ways of Working.” International Jour-
nal of Mental Health Nursing 19 (2): 119–127.
Forrest, L. M., D. C. McMillan, C. S. McArdle, and D. J. Dunlop. 2005. “An Evaluation of the Impact of a Multidisciplinary Team,
in a Single Centre, on Treatment and Survival in Patients with Inoperable Non-small-cell Lung Cancer.” British Journal of
Cancer 93 (9): 977–978.
Gabel, M., N. E. Hilton, and S. D. Nathanson. 1997. “Multidisciplinary Breast Cancer Clinics.” Cancer 79 (12): 2380–2384.
Goldstein, M., T. Yassa, N. Dacouris, and P. McFarlane. 2004. “Multidisciplinary Predialysis Care and Morbidity and Mortality of
Patients on Dialysis.” American Journal of Kidney Diseases 44 (4): 706–714.
Hu, S.-S. 2008. “One-stop Hybrid Approach for Cardiovascular Disease: From Conception to Practice.” Annals of Thoracic and
Cardiovascular Surgery 14 (6): 345–346.
Kane, B., S. Luz, D. S. O’Briain, and R. McDermott. 2007. “Multidisciplinary Team Meetings and Their Impact on Workflow in
Radiology and Pathology Departments.” BMC Medicine 5 (1): 15.
Kane, B., K. O’Byrne, and S. Luz. 2010. “Assessing Support Requirements for Multidisciplinary Team Meetings.” In 2010 IEEE
23rd International Symposium on Computer-based Medical Systems, Perth, Australia.
Kasper, E. K., G. Gerstenblith, G. Hefter, E. Van Anden, J. A. Brinker, D. R. Thiemann, M. Terrin, S. Forman, and S. H. Gottlieb.
2002. “A Randomized Trial of the Efficacy of Multidisciplinary Care in Heart Failure Outpatients at High Risk of Hospital
Readmission.” Journal of the American College of Cardiology 39 (3): 471–480.
 International Journal of Production Research 7237

Kuzma, A. M., Y. Meli, C. Meldrum, P. Jellen, M. Butler-Lebair, D. Koczen-Doyle, P. Rising, K. Stavrolakes, and F. Brogan. 2008.
“Multidisciplinary Care of the Patient with Chronic Obstructive Pulmonary Disease.” Proceedings of the American Thoracic
Society 5 (4): 567–571.
Leitch, K. K., and P. M. Walker. 2000. “Surgeon Compensation and Motivation.” Archives of Surgery 135 (6): 708–712.
Levetan, C. S., J. R. Salas, I. F. Wilets, and B. Zumoff. 1995. “Impact of Endocrine and Diabetes Team Consultation on Hospital
Length of Stay for Patients with Diabetes.” The American Journal of Medicine 99 (1): 22–28.
Liberman, R. P., D. M. Hilty, R. E. Drake, and H. W. H. Tsang. 2001. “Requirements for Multidisciplinary Teamwork in Psychiatric
Rehabilitation.” Psychiatric Services 52 (10): 1331–1342.
Makary, M. A., J. B. Sexton, J. A. Freischlag, C. G. Holzmueller, E. A. Millman, L. Rowen, and P. J. Pronovost. 2006. “Operating
Room Teamwork among Physicians and Nurses: Teamwork in the Eye of the Beholder.” Journal of the American College of
Surgeons 202 (5): 746–752.
Mazzocco, K., D. B. Petitti, K. T. Fong, D. Bonacum, J. Brookey, S. Graham, R. E. Lasky, J. B. Sexton, and E. J. Thomas. 2009.
“Surgical Team Behaviors and Patient Outcomes.” The American Journal of Surgery 197 (5): 678–685.
Ritholz, M. D., E. A. Beverly, M. J. Abrahamson, K. M. Brooks, B. A. Hultgren, and K. Weinger. 2011. “Physicians’ Perceptions of
the Type 2 Diabetes Multi-disciplinary Treatment Team: A Qualitative Study.” The Diabetes Educator 37 (6): 794–800.
Ruhstaller, T., H. Roe, B. Thurlimann, and J. J. Nicoll. 2006. “The Multidisciplinary Meeting: An Indispensable Aid to Communica-
tion between Different Specialities.” European Journal of Cancer 42 (15): 2459–2462.
Scanlon, E. F. 1991. “The Role of Reconstruction in Breast Cancer.” Cancer Journal 68: 1144–1147.
Sexton, J. B., M. A. Makary, A. R. Tersigni, D. Pryor, A. Hendrich, E. J. Thomas, C. G. Holzmueller, A. P. Knight, Y. Wu, and P. J.
Pronovost. 2006. “Teamwork in the Operating Room.” Anesthesiology 105 (5): 877–884.
Sidhom, M. A., and M. G. Poulsen. 2006. “Multidisciplinary Care in Oncology: Medicolegal Implications of Group Decisions.” The
Lancet Oncology 7 (11): 951–954.
Soh, K. B. K. 1998. “Job Analysis, Appraisal and Performance Assessments of a Surgeon: A Multifaceted Approach.” Singapore
Medical Journal 39 (4): 180–185.
Stocker, D. J. 2003. “Management of the Bariatric Surgery Patient.” Endocrinology and Metabolism Clinics of North America 32 (2):
437–457.
Taylor, C., A. J. Munro, R. Glynne-Jones, C. Griffith, P. Trevatt, M. Richards, and A. J. Ramirez. 2010. “Multidisciplinary Team
Working in Cancer: What is the Evidence?” BMJ 340: c951.
Van den Berg, J. P., S. Kalmijn, E. Lindeman, J. H. Veldink, M. de Visser, M. M. Van der Graaff, J. H. J. Wokke, and L. H. Van den
Berg. 2005. “Multidisciplinary ALS Care Improves Quality of Life in Patients with ALS.” Neurology 65 (8): 1264–1267.
Wagner, E. H. 2000. “The Role of Patient Care Teams in Chronic Disease Management.” BMJ 320 (7234): 569–572.
Wagner, E. H., B. T. Austin, C. Davis, M. Hindmarsh, J. Schaefer, and A. Bonomi. 2001. “Improving Chronic Illness Care:
Translating Evidence into Action.” Health Affairs 20 (6): 64–78.
Winslow, E. R., M. C. Bowman, and M. E. Klingensmith. 2004. “Surgeon Workhours in the Era of Limited Resident Workhours.”
Journal of the American College of Surgeons 198 (1): 111–117.
Wright, F. C., N. Lookhong, D. Urbach, D. Davis, R. S. McLeod, and A. R. Gagliardi. 2009. “Multidisciplinary Cancer Conferences:
Identifying Opportunities to Promote Implementation.” Annals of Surgical Oncology 16 (10): 2731–2737.