WEBVTT

00:00:09.490 --> 00:00:12.020
In this recitation, we'll
discuss operating room

00:00:12.020 --> 00:00:13.420
scheduling.

00:00:13.420 --> 00:00:17.510
That is, how hospitals can
be made to run smoothly.

00:00:17.510 --> 00:00:21.030
In particular, we'll discuss how
an operating room manager can

00:00:21.030 --> 00:00:23.850
use integer optimization to
make sure the hospital runs

00:00:23.850 --> 00:00:25.790
smoothly.

00:00:25.790 --> 00:00:28.730
So hospitals have a limited
number of operating rooms,

00:00:28.730 --> 00:00:30.290
or ORs.

00:00:30.290 --> 00:00:33.330
And operating room managers
must determine a weekly schedule

00:00:33.330 --> 00:00:37.040
assigning ORs to different
departments in the hospital.

00:00:37.040 --> 00:00:38.540
If you look on the
right, you'll see

00:00:38.540 --> 00:00:40.870
a picture of an operating room.

00:00:40.870 --> 00:00:44.360
You can see immediately how much
specialized and fancy equipment

00:00:44.360 --> 00:00:45.590
there is.

00:00:45.590 --> 00:00:49.080
And you have to remember that
for any surgery, staffing

00:00:49.080 --> 00:00:51.790
the OR doesn't just involve
the surgeon performing

00:00:51.790 --> 00:00:56.270
the surgery, but also other
doctors, nurses, residents,

00:00:56.270 --> 00:00:56.980
and fellows.

00:00:56.980 --> 00:01:00.490
There's usually an
entire surgery team.

00:01:00.490 --> 00:01:03.540
So due to the specialized
equipment in an OR,

00:01:03.540 --> 00:01:06.200
as well as the
specialized staff members,

00:01:06.200 --> 00:01:08.910
it's been estimated that
staffing an operating room

00:01:08.910 --> 00:01:12.140
costs over $100 a minute.

00:01:12.140 --> 00:01:14.600
Therefore, for a hospital
to run on a budget,

00:01:14.600 --> 00:01:17.050
it's very critical for
the operating room manager

00:01:17.050 --> 00:01:20.340
to create a good schedule.

00:01:20.340 --> 00:01:22.820
However, this isn't
without difficulties.

00:01:22.820 --> 00:01:25.860
Creating an acceptable schedule
is a highly political process

00:01:25.860 --> 00:01:27.890
within the hospital.

00:01:27.890 --> 00:01:31.220
Surgeons are frequently paid on
a fee-for-service basis, which

00:01:31.220 --> 00:01:34.380
means they get paid for
every surgery they perform.

00:01:34.380 --> 00:01:36.450
That means that when you
change their allocated

00:01:36.450 --> 00:01:39.210
number of operating
room hours, it directly

00:01:39.210 --> 00:01:41.300
affects their income.

00:01:41.300 --> 00:01:44.240
Therefore, the operating room
managers' proposed schedule

00:01:44.240 --> 00:01:46.740
must strike a delicate
balance between all

00:01:46.740 --> 00:01:49.770
the surgical departments
in the hospital.

00:01:49.770 --> 00:01:51.740
However, there are
many logistical issues

00:01:51.740 --> 00:01:53.690
for the operating room
manager to consider

00:01:53.690 --> 00:01:56.350
when designing the schedule.

00:01:56.350 --> 00:02:00.140
Operating rooms are staffed
in eight hour blocks.

00:02:00.140 --> 00:02:02.380
Each department sets
their own target number

00:02:02.380 --> 00:02:06.810
of allocation hours,
which may not be integer.

00:02:06.810 --> 00:02:08.340
In addition,
departments may have

00:02:08.340 --> 00:02:10.630
daily and weekly requirements.

00:02:10.630 --> 00:02:15.270
For example, gynecology might
need at least one OR per day.

00:02:17.970 --> 00:02:23.470
Ophthalmology might need
at least two ORs per week.

00:02:23.470 --> 00:02:25.230
And for example,
the oral surgeon

00:02:25.230 --> 00:02:30.180
might only be present on
Tuesdays and Thursdays.

00:02:30.180 --> 00:02:32.870
The operating room manager
needs to take into account

00:02:32.870 --> 00:02:34.550
all of these
logistical issues when

00:02:34.550 --> 00:02:35.870
designing a feasible schedule.

00:02:38.990 --> 00:02:41.750
In this recitation, we'll
consider a case study

00:02:41.750 --> 00:02:45.570
of Mount Sinai
Hospital in Toronto.

00:02:45.570 --> 00:02:49.140
There are 10 operating
rooms at Mount Sinai, which

00:02:49.140 --> 00:02:52.160
are staffed from
Monday through Friday.

00:02:52.160 --> 00:02:56.920
So 10 ORs times 5 days
times 8 hours per day,

00:02:56.920 --> 00:03:00.630
means that they have 400
hours to assign between five

00:03:00.630 --> 00:03:03.350
different surgical departments.

00:03:03.350 --> 00:03:05.160
The departments
they are assigning

00:03:05.160 --> 00:03:11.860
are ophthalmology, gynecology,
oral surgery, otolaryngology,

00:03:11.860 --> 00:03:14.300
and general surgery.

00:03:14.300 --> 00:03:16.100
Each of these
departments has come up

00:03:16.100 --> 00:03:19.740
with a weekly number of
target allocation hours.

00:03:19.740 --> 00:03:23.600
For example, ophthalmology
requests 39.4 hours

00:03:23.600 --> 00:03:28.180
of operating room time,
and otolaryngology requests

00:03:28.180 --> 00:03:33.850
26.3 hours of
operating room time.

00:03:33.850 --> 00:03:37.690
Now, it's very clear that just
by looking at these numbers

00:03:37.690 --> 00:03:39.950
they are not integer, and
they are certainly not

00:03:39.950 --> 00:03:42.400
multiples of eight hours.

00:03:42.400 --> 00:03:46.090
This means that it's impossible
for the operating room manager

00:03:46.090 --> 00:03:50.160
to give any department exactly
their weekly number of target

00:03:50.160 --> 00:03:52.230
allocation hours.

00:03:52.230 --> 00:03:54.340
However, the
operating room manager

00:03:54.340 --> 00:03:58.900
would like to try to give as
many hours to each department

00:03:58.900 --> 00:04:03.280
as possible up to their
target allocation number.

00:04:03.280 --> 00:04:07.800
We'll see how to solve this
with integer optimization.

00:04:07.800 --> 00:04:10.130
Let's consider the rest
of the problem data.

00:04:10.130 --> 00:04:13.360
For example, we need to consider
the number of surgery teams

00:04:13.360 --> 00:04:16.360
from each department that
is available each day.

00:04:16.360 --> 00:04:18.940
We also need to consider the
maximum number of operating

00:04:18.940 --> 00:04:22.210
rooms required by each
department each day.

00:04:22.210 --> 00:04:24.740
Frequently, these
numbers are the same.

00:04:24.740 --> 00:04:30.020
For example, ophthalmology
requires at most two operating

00:04:30.020 --> 00:04:33.170
rooms every day.

00:04:33.170 --> 00:04:35.030
And we see that they
have two surgery

00:04:35.030 --> 00:04:38.630
teams available every day.

00:04:38.630 --> 00:04:42.100
However, let's look at
the case of oral surgery.

00:04:42.100 --> 00:04:45.860
They require at most one
operating room every day.

00:04:45.860 --> 00:04:47.830
However, the oral
surgeon is only

00:04:47.830 --> 00:04:50.250
present on Tuesdays
and Thursdays.

00:04:54.640 --> 00:04:57.450
Additionally, each department
has weekly requirements

00:04:57.450 --> 00:05:00.160
on the number of
operating rooms they need.

00:05:00.160 --> 00:05:04.690
For example, gynecology needs
to have at least 12 operating

00:05:04.690 --> 00:05:07.470
rooms in a given
week and at most 18.

00:05:13.540 --> 00:05:15.320
The traditional
way of doing this

00:05:15.320 --> 00:05:18.580
was not by using
integer optimization.

00:05:18.580 --> 00:05:20.680
Before integer optimization
was implemented

00:05:20.680 --> 00:05:24.470
at Mount Sinai in 1999,
the operating room manager

00:05:24.470 --> 00:05:26.740
used graph paper
and a large eraser

00:05:26.740 --> 00:05:29.930
to try to assign the
operating room blocks.

00:05:29.930 --> 00:05:31.770
Any changes that
were necessary were

00:05:31.770 --> 00:05:35.050
incorporated by trial and error.

00:05:35.050 --> 00:05:37.170
The operating room
manager made a draft,

00:05:37.170 --> 00:05:41.310
and this schedule was circulated
to all of the surgical groups.

00:05:41.310 --> 00:05:44.409
However, incorporating
feedback from one department

00:05:44.409 --> 00:05:47.230
usually meant altering another
group's schedule leading

00:05:47.230 --> 00:05:48.820
to many iterations
of this process.

00:05:53.600 --> 00:05:57.400
In the next video, we'll
design an integer optimization

00:05:57.400 --> 00:05:59.950
formulation for this problem.