Actions models
2 The operating rooms
2.2 Mathematical problem
The problem we consider is the scheduling of surgical procedures over a short time period (typically one week). The project management goal is to define and schedule the different surgeries and to assign them the available resources (human or material) under costs and time constraints (Roland et al. 2007).
The project has a length of D opening days. We consider that there are T periods in
We consider J non preemptive activities, the elective surgeries. Each surgical intervention j has a deterministic processing time . To be performed, the surgery requires a certain amount of renewable resources. The renewable resources we consider are the personnel as the anesthetists, the nurses and the scrub nurses. The availability of renewable resources is constant for each period over the day. The availability of the resource can though vary over the period to take into account the variation over the week (there are fewer nurses during the weekend). The availability of nonrenewable resources is defined for the whole day. Non renewable resources are for instance specific materials as prosthesis, or blood.
Each operation j has to be carried out between an earliest and a latest starting day, and respectively. An operation cannot start before the admission day of the patient and has to be carried out within an acceptable amount of time to maximize the patient’s satisfaction (Chaabane et al. 2007).
Each surgical treatment is assigned to a particular surgeon, c, who has his personal availability a day. However, it could be possible to affect a surgery to several surgeons to model the fact that in some hospitals surgeries are led by a pool of surgeons.
Each activity can be performed in one of the S operating rooms. If the surgeries can be performed in any one of the ORs, we model an open scheduling management policy; if the interventions can only take place in some specific OR at some determined time, we model a block scheduling management policy. Each of these rooms has a regular availability for day d, representing the normal opening hours of the room defined by a number of periods. Since we allow an elective operation to occur after the normal working hours, in overtime, a room has also a maximal availability for day d which depicts the total amount of periods available this day for this room (opening hours and overtime).
Each time an operating room is opened the hospital spends €, and pays a fixed amount for each time unit an OR is opened in overtime, € per time unit. Hence the objective is to minimize the costs associated with room openings and with overtime pay.
2.2.1 Assumptions and notations
The model has the following characteristics:
• Indices
‐ j: number of surgical interventions to schedule; j =1,...,J.
‐ s: number of operating rooms ;s =1,...,S.
‐ d: number of opening days; d=1,…,D.
‐ t: number of periods in a day; t=1,…, T.
‐ c: number of surgeons; c =1,...,C.
• Parameters
‐ : operating duration of surgical intervention j
‐ : earliest starting day for surgical intervention j
‐ : latest starting day for surgical intervention j
‐ : amount of renewable resource Κ requires for intervention j
‐ : availability of renewable resource k for day
‐ : OR s where the surgical intervention j can take place
‐ : availability of the surgeon c for day d
‐ : amount of non‐renewable resource Κ requires for intervention j.
‐ : availability of non‐renewable resource k for day
‐ : availability of room s for day d
‐ : the total amount of periods available for room s the day d (normal and overtime hours)
‐ : opening cost of a room
‐ : cost of opening a room in overtime, per period
• Decisions variables
‐ : binary variable ; take value 1 when the surgical operation j starts in room s during day d and at time period t
‐ : binary variable; take value 1 when room s is opened in day d
‐ : overtime work in room s for day d
2.2.2 Mathematical programming formulation
The model formulation for our problem takes the following expression:
min, , 0 . 2.1
where the closing hour of a room s in day d is given by:
max, . 2.2
subject to:
, . 2.3
1, . 2.4
, , , . 2.5
, , . 2.6
, , , . 2.7
, , , . 2.8
1, , , . 2.9
, , . 2.10
, 0,1 , , , , . 2.11
where the binary variables (2.11) take value 1 when the surgical operation j starts in room s during day d and at time period t while take value 1 when room s is opened in day d.
A number of constraints are defined to ensure the feasibility of the OT planning:
• A surgical intervention has to take place between an earliest and latest starting day (2.3).
• A patient undergoes the surgery only once (2.4).
• There are a limited number of renewable resources (nurses, anesthesiologists) available to operate (2.5); the amount of renewable resources consumed at one given time by all operations undergoing cannot exceed the amount of resources available.
• The OT can use a certain amount of non renewable resources (surgical materials, medicines, etc.) each day: the total consumption of all surgeries of a day cannot go beyond that amount (2.6). The OR are also renewable resources. Since OR are managed according a block scheduling policy, an operation can only take place in the OR dedicated to the appropriate surgical specialty that day.
• Each surgery is performed by a surgeon and has to take place when he is available (2.7). In these latter, we sum over the operations attributed to each surgeon c and defined by the set O(c).
• A surgical operation only occurs during the opening hours. An operation cannot finish after the closing hours (2.8).
• Surgical interventions cannot overlap in an OR (2.9).
• A surgical intervention can only take place in the operating room dedicated to the appropriate specialty (2.10)
Expressed like this, the programming is nonlinear since the objective function contains the nonlinear term , 0 . This difficulty can be overcome by introducing a new variable 0 and by adding two new constraints. The objective function (2.1) becomes:
min, . 2.12
and the additional constraints are:
, , , , . 2.13 0, , . 2.14
The variables represent the overtime work in room s for day d, if indeed there is overtime work, and are null otherwise. This new formulation makes the model a linear mixed‐integer programming. The amount of overtime in each room is computed through the constraints (2.13); the overtimes are always positive numbers (2.14).
The MIP program is then solved with optimization software using a branch‐and‐
bound algorithm. This way of resolution ensures to find the optimal planning, exploring all the possible solutions. However, while the number of resources and the number of surgical interventions are increasing, the number of feasible solutions explodes and it becomes hard to solve. Hence there exists no efficient method to solve it to optimality, only on small scale problems.