Online Surgery Rescheduling - A Data-driven Approach

for Real-time Decision Support

Norman Spangenberg

, Moritz Wilke

, Christoph Augenstein

and Bogdan Franczyk

1,2

University of Leipzig, Information Systems Institute, Grimmaische Straße 12, Leipzig, Germany

Wroclaw University of Economics, ul. Komandorska 118/120, Wroclaw, Poland

Keywords:

Online Surgery Scheduling, Decision Support System, Operating Room Management, Real-time Architecture.

Abstract:

The operating room area is still one of the most expensive sections in the hospital due to its high and cost-

intensive resource requirements. Further, several uncertainties like complications, cancellations and emergen-

cies as well as the need to monitor and control the interventions during execution distinguish the operational

planning tasks of surgery scheduling from more tactical and strategical planning activities. However there

are few solutions that support monitoring and decision-making in operating room management at this level

since they focus on creation of initial schedules or the efﬁcient resource allocation. In this paper we describe

a solution approach for supporting online surgery scheduling by a real-time decision support system. It allows

the rescheduling based on intra-surgical information about the current surgical phases and predictions about

remaining intervention times and further allows replanning due to emergent or canceled patients.

1 INTRODUCTION

Many business processes on operational level are

characterized by uncertainties and frequent changes

in the environmental setting. Hence, online opera-

tional planning methods address the monitoring and

control of the process during execution and encom-

pass to react to unforeseen events (Hans and Van-

berkel, 2012). As well in every-day hospitals oper-

ations and especially the operating room (OR) area

are locations where these traits and vaguenesses are

ever-present. The OR manager is responsible for op-

erational planning in the OR area, in particular for the

supervision of all surgery-related resources and the

guarantee of efﬁcient accomplishment of the initially

created surgery schedule according to diverse perfor-

mance indicators. Uncertainties like urgent or emer-

gent patients require the immediate integration in the

schedule and complications or cancellations lead to

time delays and shifting later procedures.

By this reasons, (May et al., 2011) describe the online

surgery scheduling (OSS) as an contemporaneous job

with a very short-term perspective that includes the

execution, monitoring and control of schedules that

were constructed the day before. At the beginning of

each day a surgery schedule exists but is often out-

dated within a few minutes and needs to be modi-

ﬁed on-the-ﬂy as the associated uncertainties and dy-

namics occur. Since complications, cancellations and

emergencies happen frequently it is uncommon that

a schedule stays all day through. For this reasons,

the OR manager needs latest information of the sit-

uations in the ORs as well as predictive information

about future states and the impact of possible deci-

sions. Despite, the well-known and often described

surgery scheduling problem, there are few systems so

far that tackle intra-day surgery scheduling and al-

low OR managers to get necessary information and

support the decisions based on this information. Ac-

cordingly, in this work we address the problem of

rescheduling of surgeries and describe a solution ap-

proach. The corresponding research question reads as

follows: How should a decision support solution be

designed for supporting the online surgery schedul-

ing problem?

In contrast to other approaches, we describe an inte-

grated solution that allows real-time rescheduling and

schedule modiﬁcation based on intra-surgical infor-

mation about the current surgical status and predicted

future developments.

The paper is organized as follows. In section 2 we

present a state of the art of approaches for operational

support for real-time scheduling as well as for online

surgery scheduling. Section 3 provides a description

of the underlying decision and optimization problem.

Subsequently in section 4 we introduce the solution

336

Spangenberg, N., Wilke, M., Augenstein, C. and Franczyk, B.

Online Surgery Rescheduling - A Data-driven Approach for Real-time Decision Support.

DOI: 10.5220/0006805103360343

In Proceedings of the 20th International Conference on Enterprise Information Systems (ICEIS 2018), pages 336-343

ISBN: 978-989-758-298-1

approach and the inherent components of the deci-

sion support system. Section 5 describe the efforts so

for to validate our solution and give insights into the

use case setting and its impact. Finally we conclude

with a discussion of our results and describe future

research directions.

2 RELATED WORK

Surgery scheduling in general is one of the highly

adapted problems of operations research and schedul-

ing research community. (Demeulemeester et al.,

2013) as well as (Erdogan et al., 2010) state that

operational support for real-time scheduling is not

researched well in contrast to other domains where

real-time approaches can be found. (Atkin et al.,

2008) developed an approach for operational support

for online scheduling of airport runways with a de-

terministic scheduling algorithm. (Ngai et al., 2012)

describe an approach to compose primitive context

information of location sensors to support real-time

accident handling in ﬂeet management use case.

The problem of monitoring and scheduling multiple

production plants is tackled by a information system

including a algorithmic pipeline is described by (Guo

et al., 2015) Since, the OSS problem differentiates

in aspects of uncertainties and unpredictablities to

the characteristics of these domains these approaches

cannot be replicated to the operating room area. E.g.

in manufacturing use cases the production process

can be paused and proceeded within the same state of

the item, which is not possible within a surgery (May

et al., 2011).

Nevertheless in surgery scheduling literature sev-

eral papers address the OSS problem and suggest

approaches for supporting decision makers. (De-

meulemeester et al., 2013; May et al., 2011;

Guerriero and Guido, 2011; Erdogan et al., 2010)

provide comprehensive reviews of existing literature

approaches tackling the various levels of the surgery

scheduling problem. Further, we focus on approaches

that face the OSS and are published after the men-

tioned reviews.

(Dios et al., 2015) provide a decision support sys-

tem for operating room managers to plan different

decision tasks like medium-term and short-term

schedules. Further, it is focused on handling elective

patients so it lacks in supporting very short-term

planning tasks like handling deviations in interven-

tion times or emergency patients.

(Erdogan et al., 2015) describe a stochastic integer

programming model for dynamic sequencing and

scheduling of appointments in hospitals with the

goal to minimize the weighted sum of direct waiting

time and waiting time until appointment for patients.

Though, they include different kinds of uncertainties

like process durations or number of customers,

the model isn’t directly portable to OSS since it

doesn’t involve important surgical characteristics like

urgency.

(Riise et al., 2016) propose an approach for a

generalized operational surgery scheduling problem

that is able to support decision making on different

planning levels and with different characteristics.

Hence, it helps planning elective patients as well

as rescheduling by integrating urgent and emergent

patients. Since, they argue that it is also applicable for

intra-day rescheduling, the evaluation only focuses

on scheduling on a weekly or daily level.

(Samudra et al., 2016) used a discrete event sim-

ulation model for the patient scheduling model

considering uncertainties like varying estimations

and arrivals of unplanned surgeries to avoid excessive

overtimes in the OR area. They handle rescheduling

of elective patients as well as including non-electives

in the current surgery schedule since it represents the

hospitals reality. They also use a estimated surgery

duration model based on mean values of similar

OR sessions but without feature-based machine

learning model. As well it doesn’t include real-time

remaining intervention time estimations based on

current phases.

(van Essen et al., 2012) developed a DSS that is

providing the three best adjusted OR schedules

according to variability in surgery duration and

emergencies. This system is based on a linear integer

programming model with the goal to accomplish

the preferences of all stakeholders and departments

as good as possible. Further, the objective function

includes penalties for canceling surgeries or overtime

minimization. It doesn’t include the reassignment of

surgeries to different ORs which leads to a reduced

ﬂexibility in scheduling and hence reduced efﬁciency.

The previously presented approaches treat the OSS

problem on an algorithmic level, but don’t take

into account that information collection and DSS

architecture considerations could also show improve-

ments. This research papers assume that necessary

information is already present in the scheduling

system and further exclude the aspect of real-time

information systems.

Online Surgery Rescheduling - A Data-driven Approach for Real-time Decision Support

337

Monitor

environment

Rescheduling

Uncertainty

event occur

elective

planned duration

elective 3

elective

planned durationplanned duration

elective 2

elective 3

elective

real duration real duration

real duration idle

elective 2

elective emergency

real duration real duration real duration overtime

elective

Figure 1: Model of the OSS problem for a single operating room including the challenges due to changing intervention

durations and introducing emergent cases (extension of (Hans and Vanberkel, 2012)).

3 PROBLEM FORMULATION

In this section we give a short formulation of the

OSS (rescheduling) problem for a particular day and

the corresponding mixed integer linear programming

(MILP) model that can be used to generate valid OR

schedules within a surgery day. Since, it is a dynamic

scheduling problem, it implies updating the schedule

deﬁned the previous day in reaction to external effects

like incoming emergencies or internal changes like

deviations (see ﬁgure 1). According to the reschedul-

ing framework of (Vieira et al., 2003) the OSS prob-

lem described in this work can be seen as a dynamic

scheduling problem with variable arrivals of patients.

The OSS problem consists of a number of character-

istics, assumptions, resources and constraints that are

introduced next. The corresponding MILP model to

solve the OSS problem is formulated by:

A set of indices with capacities and resource require-

ments:

• I: Set of interventions to be performed within the

day, with elements i ∈ I

• O: Set of operating rooms available for surgeries,

with elements o ∈ O

• S: Set of surgeons, with elements s ∈ S

• T: Set of available time slots within the day, with

elements t ∈ T =1,...,X

A set of parameters describing properties of resources

related to the OSS:

• l

o,t

: Available time of OR o in working hours

• c

s,t

: Time capacity surgeon s is available for per-

forming interventions

• d

: Estimated duration of Intervention i

• u

: Urgency status of intervention i

• m

: Modiﬁcation status of intervention i

Two planning variables are available to optimize the

schedule according to the given constraints, assump-

tions and resources:

• t

: Assigned starting time of intervention i

• o

: Assigned OR o of intervention i

The number of surgeries to be scheduled on the

tagged day is not known in advance, since it is likely

that emergencies occur. A surgical intervention i is

characterized by its surgeon s

, the estimated duration

before or during the intervention and its urgency

according to the scale elective, urgent and emer-

gent. Further, a modiﬁcation parameter is introduced

to block interventions in a speciﬁc OR at a speciﬁc

time manually or after start. For each surgeon in-

dexed s, I

denotes the set of jobs that are performed

by that surgeon. Several assumptions are made to re-

duce complexity and develop a sparse model of the

OSS problem:

Assumption 1: ORs are interchangeable, e.g.

there are no equipment constraints.

Assumption 2: Unexpected incoming patients

receive preferential treatment in case they have

higher priority than scheduled elective patients.

ICEIS 2018 - 20th International Conference on Enterprise Information Systems

338

Assumption 3: There are enough surgeons and

surgical teams to treat electives, as well as accom-

modate non-electives.

Assumption 4: Surgery durations are estimations

that change during interventions.

The above assumptions and the following constraints

represent some of properties of the rescheduling pro-

cess resulting from the situation in the OR area. Some

hard constraints are deﬁned, which, if they are vio-

lated, lead to an invalid surgery schedule.

Constraint 1: Only one surgery at the same time

in a operating room. A surgery cannot be assigned

to a OR that is occupied.

Constraint 2: A surgeon/surgical team can per-

form only one surgery at the same time.

Constraint 3: Surgeries tagged as not movable

must not be reassigned to other ORs or time slots

Further, four soft constraints are modeled:

Constraint 4: Don’t assign elective intervention

after operating room working hours.

Constraint 5: Do urgent and emergent interven-

tions as soon as possible.

Constraint 6: Avoid reassigning or canceling al-

ready assigned surgeries.

The problem is now to ﬁnd an assignment σ : I ×T 7→

O of interventions to available time slots of operating

rooms and surgeons according to the intervention du-

ration. Hence, the solver optimizes the rescheduling

result according to the following goals.

The most important optimization criteria for the OR

manager (besides treatment quality) is maximizing

OR utilization of each operating room ω

) (1).

Since, there are several methods to calculate OR uti-

lization we use the deﬁnition of (Hans and Vanberkel,

2012).

Max ω

) =

∑

i=1

(1)

The 2nd objective minimizes waiting time ω

(σ) and

should lead to fast integration of non-electives:

Min ω

(σ) =

∑

i=1

(2)

describes the cost efﬁcient for the waiting time

of an intervention, while u means the urgency fac-

tor (higher urgency, higher integer value). The solver

should minimize the penalty costs for waiting, so

more urgent interventions are assigned fast (2). Fur-

ther, all types of surgery are assigned as early as pos-

sible, thus a by-product is minimized overtime ω

(σ).

Min ω

(σ) =

∑

i=1

canc

(3)

Adding penalty costs β

for each canceled or reas-

signed intervention should lead to the effect that valid

schedules with fewer reassignments/cancellations are

preferred (3). Canceled interventions have a higher

penalty beta than the reassigned and urgent interven-

tions have higher β

then electives.

4 SOLUTION APPROACH

According to this problem, formulation a predictive-

reactive rescheduling strategy is utilized and sup-

ported with software tools to generate and partially

update the current schedule based on incoming events

with planning-relevant information. In this section

we propose an architectural approach with an online

surgery rescheduling engine. To realize this approach

several software components are needed to collect and

enhance the necessary information (see ﬁgure 2). The

segmentation of the solution approach into three parts

is because of separation of concerns. Nevertheless

they build on top of another each subsystem uses dif-

ferent type of data and information.

4.1 Situation Detection Subsystem

(SDS)

This component supports the information gathering

tasks of the OR manager and automatizes it to

ease and advance this process. Based on low-level

real-time data of e.g. cameras, surgical devices, OR

equipment or other connected devices information

about intra-surgical phases in running interventions

can be gathered. Besides processing the incoming

data streams, SDS realizes methods for the phase

detection. Lots of research exists for surgical phase

detection methods. Some are image- or video-based,

e.g. (Dergachyova et al., 2016). Others relying on

electronic signals of surgical devices are described

for example by (Padoy et al., 2012; Spangenberg

et al., 2017). All of these methods have their pros

and cons, e.g. some detect minimal invasive surgical

phases better then others and vice versa. We used

Complex Event Processing (CEP) for modeling

surgical phases based on surgical device data and

operating room equipment e.g. OR lights. According

to the taxonomy proposed in (Lalys and Jannin,

2014) this component classes into the data-to-model

analysis methods.

Online Surgery Rescheduling - A Data-driven Approach for Real-time Decision Support

339

Rescheduling

subsystem

Situation

detection subsystem

Event Processor

Raw data stream

Prediction

subsystem

Prediction model

computation

Online prediction

component

Model

database

Event

history

database

Phase event

model repository

Phase pattern

repository

Resource

database

Online rescheduling

component

Domain model

repository

Constraint

database

Surgery

schedule

Figure 2: Components of the solution approach and their interactions.

4.2 Prediction Subsystem (PS)

PS utilizes the data of the SDS two-fold: First, a batch

layer that uses historic surgical phases and other fea-

tures to create a machine learning model that predicts

the remaining intervention time of running surgeries.

This model is built with a random forest algorithm

based upon ﬁve features: an identiﬁer for the surgery

type based on the ofﬁcial German classiﬁcation of

medical procedures (OPS), a time stamp representing

the time passed since start and the operating room.

Further, the current phase as well as an event history

based on previously detected phases is factored into

the model. Second, an online prediction layer (speed

layer) that loads the model and aligns detected phases

in running surgeries with the model to update the esti-

mated intervention time. Further, this starts triggering

the rescheduling process.

4.3 Rescheduling Subsystem (RS)

After collecting information of the PS, the RS starts

to adapt the current schedule to events and changes

in surgeries. The RS is responsible for the genera-

tion of valid surgery schedules based on the resources

and constraints described in section 3. Rescheduling

is triggered by several factors, for example changes

in remaining durations of running interventions based

on the machine learning model. Further, the adding of

emergent or urgent patients to the set of interventions

leads to the execution of the rescheduling procedure.

We use a metaheuristics approach for solving the op-

timization problem of the rescheduling task. Meta-

heuristics don’t guarantee ﬁnding an optimal solution

for the optimization problem, but ﬁnd an appropri-

ate solution in a given amount of time, which is nec-

essary for our goal to give real-time decision sup-

port. The search space is deﬁned by two vectors:

One for the OR assignments of each surgery and sec-

ond a vector for non-overlapping time-slots including

surgeons, surgical team and ORs. Hence, the plan-

ning variables are operating room and the combina-

tion of starting time slot and the intervention dura-

tion. The cost function incorporates all cost factors

of the constraints described in 3. Violations of the

hard constraints, e.g. two surgeries at the same time

in the same OR, are not allowed. The quality of a

valid schedule is determined by the minimization of

the soft constraints. Our metaheuristic consists of the

following computational steps, based on the princi-

ple of local search. The used algorithm is Simulated

Annealing, described in more detail by (Kirkpatrick

et al., 1983). Since, it has been successfully used in

dynamic scheduling domain before it is as well scal-

able and ﬁnds near optimal solutions (Ceschia and

Schaerf, 2016).

ICEIS 2018 - 20th International Conference on Enterprise Information Systems

340

4.3.1 Initial Solution

To get a satisfying, but non-optimal and mostly not

feasible solution, initial solution of a schedule that af-

terward could be optimized, we use the First Fit Ap-

proximation algorithm. The algorithm assigns the in-

terventions to a available planning value (in our case

ORs and available time slots) and further takes the

already initialized interventions into account. Since,

First Fit doesn’t change an planning entity after as-

signing, it terminates after initializing all interven-

tions.

4.3.2 Move Selection

Moves are chosen indiscriminately as it is common

for Simulated Annealing algorithm. A move is se-

lected if it is equal or greater than the best move.

Furthermore, non-improving moves are also picked

with a certain probability according to its score and

the time gradient. In the early phase of the calcula-

tion process the probability of selecting sub-optimal

moves is higher than in later phases.

4.3.3 Cooling Schedule

Since, an ideal cooling method cannot be determined

in advance, a cooling calculation for temperature is

used. Depending on a time gradient decreases from

time to time by a constant quantity.

4.3.4 Acceptance and Stop Criterion

Moves are accepted in every case if they improve the

solution. Moves leading to a worse solution the ac-

ceptance probability is determined by e

− f

temp

, where f

describes the cost function and temp the current tem-

perature. The whole procedure stops when the cal-

culation gains a ﬁnal temperature or exceeds a given

amount of time due to the near real-time requirement

of the system.

5 EVALUATION

The evaluation of the solution approach and its im-

plementation in a case study in a real-world setting

is planned for advanced research. So far we used a

simulated environment of an operating room area rep-

resenting 10 operating rooms each with 10 hours of

operation/day and 4 starting time slots/hour. We used

a data set of 15 surgeries with real-world data that

produce a low-level events stream to simulate a surgi-

cal day and feed the SDS. The detected intra-surgical

phases trigger the calculation of remaining interven-

tion times and use this information afterward to start

rescheduling. In this stage the interventions can have

ﬁve different states:

• Planned: Are introduced to the system, but OR or

time slot are not assigned yet.

• Scheduled: OR or time slots are assigned, but in-

tervention didn’t start already.

• In progress: Intervention is running and changes

in running intervention time are likely but OR

isn’t moveable.

• Reassigned: Scheduled intervention is reassigned

to other OR or time slot.

• Canceled: Are delayed with higher priority for

next day.

Observations showed that each running intervention

updates its predicted remaining time on an average

of 20 times so the rescheduling is triggered the same

number. Further, the observations indicate that the

metaheuristic provides good solutions according to

tardiness and schedule stability. Few reassignments

or cancellations are done by the algorithm and non-

elective interventions, that are fed into system as well,

are assigned fast (see ﬁgure 3).

First implications of the proposed system for produc-

tive operations are that the notable number of updates

of the surgery schedule will increase. This is be-

cause that domain knowledge of the OR manager ,

e.g. for remaining intervention times, is now mod-

eled and leads to a higher degree of transparency,

since less information and thus decisions are based

on experience-based knowledge and human estima-

tions. For the main user of the system (the OR man-

ager) two major improvements can be noted. First,

the whole process for information collection in the

operating room area is simpliﬁed. Second, the cogni-

tive efforts for combining current states, estimations,

available resources and potential emergencies, which

is done without software support so far, is reduced

signiﬁcantly. These performance aspects will later be

investigated in more detail by comparing it to deci-

sions made by the OR manager.

Compared to other approaches tackling the OSS our

work provides some beneﬁts. (Li et al., 2016; Riise

et al., 2016; Dios et al., 2015) also address short-

term scheduling, but focus on optimization and man-

ual adjustments on the day before. Hence, intra-day

rescheduling is still unsupported and conducted by the

OR manager. (Bruni et al., 2015) and (Heydari and

Soudi, 2015) describe a similar problem of handling

emergencies and uncertainties in surgery reschedul-

ing and formulate new solution strategies from a algo-

Online Surgery Rescheduling - A Data-driven Approach for Real-time Decision Support

341

(0) Set of interventions I

(1) Initial schedule at time t

(2) Updated schedule at t

time with emergent case E

and predicted intervention duration changes d

and d

Figure 3: Components of the solution approach and their interactions.

rithmic point-of-view but lack with an integrated ar-

chitectural approach and implementation details that

would lead to simpliﬁcations for decision maker.

6 CONCLUSION

In this paper we presented a new solution approach

for supporting the OSS problem by a real-time deci-

sion support system for rescheduling. Based on intra-

surgical information about the current surgical phases

and predictions about remaining intervention times it

allows updating the surgery schedule and replanning

due to emergent or canceled patients. The proposed

approach denotes an innovative solution since most

of the current approaches operate on the tactical and

strategical planning and scheduling with longer time

horizons. We focused in this work on modeling OR-

related resources and constraints and for now omit

other related entities like intensive care (ICU) unit

or the like. But the approach can easily be extended

in this directions, e.g. by modeling other personnel

resources (nurses, porters, anesthetists) or facilities

(equipment, devices, ICU capacities). It was shown

that the beneﬁts of our approach focus on the sup-

port of the OR manager and improve his daily tasks

twofold. First, the process for information collection

in the operating room area is simpliﬁed since it re-

duces communicative efforts, e.g. for monitoring cur-

rent system status the status of running interventions

in particular. Second, the cognitive efforts for com-

bining current states, estimations, available resources

and potential emergencies, is reduced signiﬁcantly.

The prediction and the rescheduling subsystem pro-

vide an automatized solution for tasks which so far

are dispatched without software support.

In future work, we will focus on methodologies for

the appropriate delivery of information to the OR

manager. For instance a situation-aware user inter-

face would beneﬁt our approach concerning for bet-

ter representation and prevention of information over-

load. Further, modeling more resources and con-

straints would lead to a more realistic Finally the eval-

uation of the integrated system in a real-world setting

in a operating room area will be done to compare the

performance of the system against human decision

makers.

ACKNOWLEDGMENTS

This paper was funded by the German Federal Min-

istry of Education and Research under the project

Competence Center for Scalable Data Services and

Solutions Dresden/Leipzig (BMBF 01IS14014B) and

by the German Federal Ministry of Economic Af-

fairs and Energy under the project InnOPlan (BMWI

01MD15002E).

REFERENCES

Atkin, J. A. D., Burke, E. K., Greenwood, J. S., and Ree-

son, D. (2008). On-line decision support for take-off

runway scheduling with uncertain taxi times at london

heathrow airport. Journal of Scheduling, 11(5):323–

346.

ICEIS 2018 - 20th International Conference on Enterprise Information Systems

342

Bruni, M. E., Beraldi, P., and Conforti, D. (2015). A

stochastic programming approach for operating the-

atre scheduling under uncertainty. IMA Journal of

Management Mathematics, 26(1):99–119.

Ceschia, S. and Schaerf, A. (2016). Dynamic patient admis-

sion scheduling with operating room constraints, ﬂex-

ible horizons, and patient delays. Journal of Schedul-

ing, 19(4):377–389.

Demeulemeester, E., Beli

en, J., Cardoen, B., and Samudra,

M. (2013). Operating room planning & scheduling.

Handbook of Healthcare Operations Management.

Dergachyova, O., Bouget, D., Huaulme, A., Morandi, X.,

and Jannin, P. (2016). Automatic data-driven real-

time segmentation and recognition of surgical work-

ﬂow. International Journal of Computer Assisted Ra-

diology and Surgery, 11(6):1081–1089.

Dios, M., Molina-Pariente, J. M., Fernandez-Viagas, V.,

Andrade-Pineda, J. L., and Framinan, J. M. (2015). A

decision support system for operating room schedul-

ing. Computers & Industrial Engineering, 88:430–

443.

Erdogan, S. A., Denton, B. T., Cochran, J. J., Cox, L. A.,

Keskinocak, P., Kharoufeh, J. P., and Smith, J. C.

(2010). Surgery planning and scheduling. In Wiley En-

cyclopedia of Operations Research and Management

Science. Wiley Online Library.

Erdogan, S. A., Gose, A., and Denton, B. T. (2015). Online

appointment sequencing and scheduling. IIE Transac-

tions, 47(11):1267–1286.

Guerriero, F. and Guido, R. (2011). Operational research

in the management of the operating theatre: a survey.

Health Care Management Science, 14(1):89–114.

Guo, Z. X., Ngai, E., Yang, C., and Liang, X. (2015). An

rﬁd-based intelligent decision support system archi-

tecture for production monitoring and scheduling in a

distributed manufacturing environment. International

Journal of Production Economics, 159:16–28.

Hans, E. W. and Vanberkel, P. T. (2012). Operating theatre

planning and scheduling. In Handbook of healthcare

system Scheduling, pages 105–130. Springer.

Heydari, M. and Soudi, A. (2015). Predictive / reactive

planning and scheduling of a surgical suite with emer-

gency patient arrival. Journal of Medical Systems,

40(1):30.

Kirkpatrick, S., Gelatt, C. D., and Vecchi, M. P. (1983).

Optimization by simulated annealing. science,

220(4598):671–680.

Lalys, F. and Jannin, P. (2014). Surgical process modelling:

a review. International Journal of Computer Assisted

Radiology and Surgery, 9(3):495–511.

Li, F., Gupta, D., and Potthoff, S. (2016). Improving oper-

ating room schedules. Health Care Management Sci-

ence, 19(3):261–278.

May, J. H., Spangler, W. E., Strum, D. P., and Vargas, L. G.

(2011). The surgical scheduling problem: Current re-

search and future opportunities. Production & Opera-

tions Management, 20(3):392–405.

Ngai, E., Leung, T., Wong, Y. H., Lee, M., Chai, P., and

Choi, Y. S. (2012). Design and development of a

context-aware decision support system for real-time

accident handling in logistics. Decision support sys-

tems, 52(4):816–827.

Padoy, N., Blum, T., Ahmadi, S.-A., Feussner, H., Berger,

M.-O., and Navab, N. (2012). Statistical modeling

and recognition of surgical workﬂow. Medical image

analysis, 16(3):632–641.

Riise, A., Mannino, C., and Burke, E. K. (2016). Modelling

and solving generalised operational surgery schedul-

ing problems. Computers & Operations Research,

66:1–11.

Samudra, M., Demeulemeester, E., Cardoen, B.,

Vansteenkiste, N., and Rademakers, F. E. (2016).

Due time driven surgery scheduling. Health Care

Management Science, pages 1–27.

Spangenberg, N., Augenstein, C., Franczyk, B., Wagner,

M., Apitz, M., and Kenngott, H. (2017). Method for

intra-surgical phase detection by using real-time med-

ical device data. 30th IEEE International Symposium

on Computer-Based Medical Systems 2017, 30:1–8.

van Essen, J. T., Hurink, J. L., Hartholt, W., and van den

Akker, Bernd J. (2012). Decision support system

for the operating room rescheduling problem. Health

Care Management Science, 15(4):355–372.

Vieira, G., Herrmann, J., and Lin, E. (2003). Reschedul-

ing manufacturing systems: A framework of strate-

gies, policies, and methods. Journal of Scheduling,

6:39–62.

Online Surgery Rescheduling - A Data-driven Approach for Real-time Decision Support

343