00:00:01
we're here today with Wharton operation
00:00:03
information and decisions professor huh
00:00:04
me song to talk to her about some of her
00:00:06
latest research how many thanks for
00:00:08
being with us oh thanks for having me so
00:00:10
first can you give us a short summary of
00:00:11
this research so so my work is in
00:00:14
healthcare operations and what I'm
00:00:16
focusing on is thinking about how we can
00:00:18
design systems and processes that will
00:00:20
enable physicians to work more
00:00:22
productively this particular paper I'm
00:00:25
focusing on queue management systems
00:00:27
thinking about how can we better design
00:00:29
this queue management system to enable
00:00:31
physicians to be more productive so
00:00:33
first what do I mean by queue management
00:00:35
so this is thinking about the kinds of
00:00:39
you know think about a line that you
00:00:40
waited in this morning let's say we
00:00:43
encounter these things all over the
00:00:44
place and we're trying to think about
00:00:45
different ways of structuring them so in
00:00:48
this particular case it's a queue
00:00:49
structure thinking about what kinds of
00:00:51
implications that has for productivity
00:00:53
outcomes that were interested in in this
00:00:55
kind of a healthcare setting we're
00:00:57
thinking about the length of stay of the
00:00:59
patients the waiting times of the
00:01:00
patients and as we'll get to talk about
00:01:02
this particular paper we're thinking
00:01:04
about in an emergency department setting
00:01:06
so these are patients in the waiting
00:01:08
room of an emergency department thinking
00:01:10
about how you should design those cues
00:01:12
as they wait to be seen by a physician
00:01:14
so you can naturally think of two
00:01:17
different ways of structuring these
00:01:18
queues one is what we'll call a pooled Q
00:01:21
so this is a case where you have
00:01:23
essentially a single line where the
00:01:26
patients are waiting to be seen by one
00:01:28
of the many physicians who are working
00:01:29
there they're not pre assigned to
00:01:32
particular physicians Q the other kind
00:01:35
of cue structure that we consider is a
00:01:37
dedicated queue and this dedicated Q is
00:01:39
essentially as soon as they arrive and
00:01:41
get registered they're getting assigned
00:01:43
to a specific physicians dedicated queue
00:01:46
right so from the beginning you're
00:01:48
waiting in the line that belongs to a
00:01:50
particular physician so ultimately we're
00:01:53
asking given these type type two types
00:01:55
of Q structures which one of these might
00:01:58
lead to more efficient or productive
00:02:00
outcomes which is a good question for
00:02:02
any of us who've ever been ill I at the
00:02:03
ER you know it can be a very frustrating
00:02:05
experience so when you look at these two
00:02:06
types of queues what it what were your
00:02:08
key takeaways what did you find sure so
00:02:09
what we find is surprisingly because
00:02:12
this is actually counter to what
00:02:14
tradition
00:02:14
queuing theory would predict we find
00:02:17
that having the dedicated queues
00:02:18
actually led to faster throughput time
00:02:21
so in this case shorter lengths of stays
00:02:23
and shorter waiting times for the
00:02:25
patients we found this to be interesting
00:02:28
because like i said it's
00:02:30
counterintuitive in a way and what we're
00:02:33
finding is that it's it's not just that
00:02:35
you know with pooled cues you are able
00:02:37
to buffer against the variability that
00:02:39
you might see in these kinds of cues but
00:02:41
in this case with the dedicated cues
00:02:43
given that the physicians are people who
00:02:46
they're people first of all and they
00:02:48
have a lot of discretion in terms of how
00:02:50
they organize their work when they have
00:02:52
these dedicated cues they start feeling
00:02:55
a greater sense of ownership and
00:02:56
responsibility over those queues of
00:02:59
patients in this case and they're trying
00:03:01
to actively think of ways to get people
00:03:03
into the beds quicker which means that
00:03:06
they have to think about how can i get
00:03:08
the patients who are in the beds out
00:03:09
faster so they're doing things
00:03:11
differently like pulling for the test
00:03:13
results if they ordered tests they're
00:03:15
proactively you know calling up to see
00:03:17
what kinds you know what are the results
00:03:19
rather than just waiting for those
00:03:21
results to be pushed to them and we can
00:03:23
validate a lot of these hypotheses that
00:03:25
we generated from our interviews and
00:03:27
observations with the physicians through
00:03:29
the data that we collect from the
00:03:31
electronic medical records of the
00:03:33
particular emergency department that we
00:03:35
worked with for this project so looking
00:03:38
at some of the practical implications of
00:03:39
this research I guess with I'm an
00:03:42
emergency department how could I I mean
00:03:44
it seems like this is indicating that I
00:03:46
would want to implement this sort of Q
00:03:47
and I guess first of all is that in fact
00:03:49
the case and if it is I guess how would
00:03:51
I go about doing that I mean are there
00:03:53
things you would say how do hospitals
00:03:55
take this and really apply it sure so
00:03:57
that's a great question and since this
00:04:00
paper we've actually spoken with
00:04:02
different emergency departments that
00:04:03
have started implementing this because
00:04:05
let me just backtrack just a little bit
00:04:07
and say a little more about our findings
00:04:09
and why it was so significant we find
00:04:13
that you know this change that I told
00:04:14
you about the reduction in the length of
00:04:16
stay in the reduction in the waiting
00:04:18
times they're actually quite significant
00:04:19
so we're finding a seventeen percent
00:04:22
decrease in the length of stay of the
00:04:24
patients and a nine percent decrease in
00:04:26
the waiting times when you translate
00:04:28
that into
00:04:28
numbers minutes for an average patient
00:04:31
who's coming to this particular ed
00:04:32
that's about 40 45 minute reduction in
00:04:36
that total time so that's huge and if
00:04:38
you think about given the fact that
00:04:40
they're about 200 patients who are
00:04:42
showing up to this emergency department
00:04:44
per day that's a large amount of time if
00:04:47
you add that up for all 200 such that
00:04:49
they're able to see about 30 more
00:04:51
patients per day with the same amount of
00:04:53
resources that's exactly why practically
00:04:56
speaking these other emergency
00:04:58
departments that tend to have pooled
00:04:59
queuing systems most EDS or emergency
00:05:03
departments in the US still have pooled
00:05:05
queuing systems that's why they're so
00:05:07
interested in thinking about oh how can
00:05:09
we make this transition to a dedicated
00:05:11
queuing system given that that might
00:05:13
mean that we're able to see much more in
00:05:15
terms of kind of a patient load per day
00:05:18
without adding more resources so the
00:05:20
kinds of things that they're doing
00:05:21
differently so a lot of times so this
00:05:24
will defer by emergency department so
00:05:26
each place we speak with is slightly
00:05:28
different but in this case you can think
00:05:31
of you know if there's a computer
00:05:32
assignment system in terms of how you
00:05:34
assign patients to physicians you can
00:05:37
make that assignment happen just
00:05:39
essentially at the beginning of when
00:05:41
they get registered rather than when
00:05:43
they get into a bed typically what is
00:05:46
done with these pooled queuing systems
00:05:47
is that they wait until the patient has
00:05:49
an available bed to get into to assign
00:05:52
the physician so in the again in the
00:05:54
waiting room they're really waiting in
00:05:55
this cold queue and when we think even
00:05:58
beyond emergency departments thinking
00:06:00
about how might we apply these kinds of
00:06:02
things it's very much a case-by-case
00:06:03
basis but what I would say is the key
00:06:05
thing to think about is you know are
00:06:08
these workers in this case it was
00:06:10
physicians but in other cases it could
00:06:11
be all sorts of other service sector
00:06:14
workers are they people who have lots of
00:06:16
discretion over how they work if they
00:06:18
don't I i would say pooled queues are
00:06:20
still probably more efficient but if
00:06:22
they have that capacity to really change
00:06:25
the way they structure their workflow
00:06:27
and use their discretion to make those
00:06:29
real-time decisions you might start
00:06:31
thinking about these dedicated cues as a
00:06:33
way to try to improve productivity and
00:06:38
efficiency in ways you wouldn't have
00:06:39
thought of before because it sounds like
00:06:41
one of the keys here
00:06:42
is that not only are you switching the
00:06:44
way you do the queue but you're also the
00:06:45
doctors know is that they have their
00:06:48
given ownership over this list of people
00:06:50
and so instead of just waiting what
00:06:52
comes next they see what's next and how
00:06:54
do what they can think about what they
00:06:55
can do about it that's a really
00:06:56
interesting point you mentioned about
00:06:58
seeing it there's actually a separate
00:07:00
study not mine that's recently come out
00:07:03
and I've seen it in the press actually
00:07:05
even just yesterday in the New York
00:07:06
Times we're having visibility into the
00:07:10
length of your queue is really important
00:07:12
so if you design this queue where the
00:07:14
physicians in this case might not
00:07:16
actually be able to see how many
00:07:17
patients are waiting in their queue it
00:07:19
might not actually make much of a
00:07:20
difference because what's important is
00:07:22
you know you can actually see oh wow
00:07:24
I've got 12 patients waiting for me like
00:07:26
I should really get to them they belong
00:07:29
to me if you don't have that kind of
00:07:30
information maybe it won't change your
00:07:32
behavior now what's next for this
00:07:34
research or what are you going to what
00:07:35
are you studying next yeah sure so most
00:07:38
directly related to this project so with
00:07:40
any kind of operations management
00:07:41
question it's really important to think
00:07:43
about trade-offs and thinking about the
00:07:46
boundary condition so with this
00:07:47
particular paper we looked at things
00:07:49
like quality of care and look what it
00:07:52
means for utilization those kinds of
00:07:53
things but I want to take it a step
00:07:55
further and think about less so in an
00:07:58
empirical setting like this paper was
00:08:00
but more so in kind of a jet more
00:08:02
generalizable modeling setting so I'm
00:08:03
working with some other colleagues
00:08:05
writing a modeling paper trying to see
00:08:08
what are the conditions under which this
00:08:10
might be the case we're dedicated queues
00:08:12
outperformed pulled queues given
00:08:14
different types of work a version that
00:08:16
you know servers might have so we
00:08:18
consider things like whether they're
00:08:20
averse to having high levels of workload
00:08:22
or just have it or just being busy at
00:08:24
all and then they would prefer idle time
00:08:26
so we're trying to understand what is
00:08:28
the threshold or what are the conditions
00:08:30
under which this would be the case so
00:08:32
that's most directly related to this
00:08:34
particular queuing paper but more
00:08:37
broadly I'm very interested in thinking
00:08:38
about different levers that we might be
00:08:40
able to use to try to think about ways
00:08:44
to improve efficiency and productivity
00:08:45
in these healthcare settings not just
00:08:47
the emergency department but also
00:08:49
thinking about impatient settings or
00:08:51
outpatient settings more broadly so I've
00:08:53
been looking into things like different
00:08:56
ways of leveraging performance feedback
00:08:59
that you might provide to physicians
00:09:01
different ways we might thinking about
00:09:03
matching the supply and demand of
00:09:05
hospital beds because there are all
00:09:07
sorts of different patients who show up
00:09:09
in different types of beds that are
00:09:10
available at different times and
00:09:12
especially given the enormous amount of
00:09:15
data that are being collected these days
00:09:16
electronically in hospital settings how
00:09:20
can we better leverage that to make
00:09:21
predictions that we can act on in real
00:09:23
time so those are the kinds of things
00:09:25
I've been really interested in great
00:09:27
thank you so much for being here thank
00:09:28
you
00:09:45
you
