>> Desney Tan: It's my pleasure today to introduce our speaker, Dr. Justin Gatewood. Justin graduated from the University of Washington School of Medicine some number of years ago. I won't say how many I guess. So he's -and he's got family around the region, so he's plenty familiar with the beautiful weather we have this time of year. Justin did his residency at the University of Chicago Hospitals where he was also education chief resident and is now an attending physician at the emergency department at Washington Hospital Center out in DC. I met Justin about a year ago when I was out visiting the Microsoft Medical MediaLab out there. In fact, Liz is out here from there. And since then we've been collaborating on several projects, mostly to do with patient facing information displays. It's been a blast working with Justin, I must say, both because of his breadth and depth of expertise and interest in the medical domain but also in technology, which is sort of a difficult combination to come by. Justin's here for the day. I've got a couple of open slots later in the day if anyone would like a slot catch me after this meeting. And you know, hopefully these exchanges start to foster more collaboration both with Justin, but also with Washington Hospital Center and the MedStar unit as a whole as well as the various health solution group labs out there. Without further ado, I'm going to hand the mic over to Justin. >> Justin Gatewood: Great. Thank you. Thanks, Desney. And you know, just to kind of get sentimental a little bit here, Microsoft has been great. Washington Hospital Center and Microsoft have kind of an established relationship I think starting back from the development of Azyxxi, which clearly we know now is an Amalga by Dr. Craig Freied and Dr. Mark Smith. And so that's just sort of continued to roll, and things have been great and we're kind of forging ahead and I'm asking some interesting questions. So thanks for having me here. I appreciate it. So again, Justin Gatewood. And I'm an attending physician in emergency medicine. And this is sort of your chance to pick the brain of an ER doc to get under the hood of a working emergency department. And I want to kind of take a systems view of the of the emergency department. It can be sort of apparently chaotic place on the outside, but it actually makes sense. There's a method to the madness and I want to kind of elucidate some of that. So the objectives for today -- and we've got the room until 12, I want to spend roughly half of that kind of talking -- I don't want it to be too didactic, and then I want the other half to just be a lot of sort of question and answer discussion, brainstorming of potential collaborations. Please feel free to stop me at any time to ask questions. A couple disclaimers here before we start. I don't have any formal training in systems theory or in computer science or mathematics. Sometimes I may flirt with terms that -- and feel free to say, hey, that's not -- you know, we're purists here, that's not the correct usage of the term. Another disclaimer, I'm a physician. I have a very physician centric view of workflow, so that bias is inherent as well. With any talk about workflow, there's going to be lots of flow diagrams. This is not up to UML convention. I didn't use Visio. I should have. But that being said, hopefully you'll be able to grasp some of the concepts. They're quite -- they're quite straightforward. So we'll deconstruct the flow in the emergency department, looking at two systems, one the flow of patients and also the flow of information. Sometime they can be one and the same. And a very high volume active emergency department. And then I want to examine some of the concepts that underline medical decision making. Because I hope that our discussion can move past just examining workflow but also get towards really understanding how a physician's mind works and how decisions are made. And then I want to take a second to highlight some existing MSR and Washington Hospital Center collaborations and then get into areas that we can work together. So a bit about my hospital. MedStar Health is a large health organization. We have several hospitals in the Washington, Baltimore area of which had Georgetown is also one. Washington Hospital Center is the largest, it's the largest in the District of Columbia. It's got 926 beds. It's a very large hospital. It's also a major teaching hospital. We accept patients from all over the area, so it's a tertiary care referral center. It's also the level 1 trauma center and the only burn center in the area. And a plug for the hospital, consistently ranked by US News and World Report in several areas. About the department itself, we provide care, emergency medicine and trauma care for adult patients. There's a children's hospital directly across the parking lot. So we don't see -- we don't see kids. Sometimes they'll wander in. We see about 85,000 patients a year. That breaks into, you know, as much as 300 patients a day on a very busy day. We have a very -- a population that reflects the District of Columbia. It's largely African-American although we have a surprisingly diverse international population. There's several universities near by. More than a third of the patients that we serve are illiterate, and from that I'll deduce have health illiteracy as well. The 37 percent is something that was put out by the DC government to describe actually all of DC functionally illiterate all of DC. And as far as payors go, the majority have public assistance Medicaid. As far as a degree of the acute of patients, we see -- we roughly admit to the hospital -and this is a -- this is sort of a proxy for how sick our patients are, we roughly admit about a quarter of our patients. 10 to 15 percent of those admissions are about two to three percent of the entire people that we serve end up in the critical care unit. So just to kind of give you a benchmark. This is a -- we see really, really sick folks. The department itself is -- and I'll go into the layout. It's 42 beds. At any time there can be up to six attending level emergency physicians. An attending physician is just a physician who has finished their residency training. Anywhere from 10 to 20 emergency trained nurses. We also are a emergency medicine residency training site. And we have scads of great Georgetown medical students as well. So first I just want to talk about the principles of information flow. As you sort of -- as we go through the flow of patients and information you'll see a couple things emerge here. First of all in really the classic sense of the term, they're both the deterministic and a nondeterministic system. You know, some things are measured output for every input, and other things you put a patient in the system and you really don't know what you're going to get. And this is really important to realize. You put in a therapy, you administer a medication and you really don't know what you're going to get. And that's something inherent in emergency medicine and in medicine itself. Another concepts, there is asynchronous flow, but you have these parallel sort of pathways. Patients go if the system at different times they converge at different nodes, whether it be in a radiology suite, labs. They're all different bottlenecks, and they all sort of manage their way out of the system. And we'll go through that as well. Another thing is this is not just a linear pathway. You will see that patients will come back to different nodes. We are constantly reevaluating the patient. The machine sort of will go, start and stop and will do different iterations of therapy. So it's not just we're not going to straight through here. And another thing is it can be very chaotic. And I mean this in, you know, the lay sense of the word. You walk in and it can be very chaotic. And sort of more of the system sense of the word, it's chaotic in that there are identifiable patterns within what appears to be chaos. And hopefully you can make some connections there as well as we go through. How many people have ever been to the emergency department? I think most folks have been to the emergency department. And how many of those people -how many kind of understood what was going on? Okay. Great. That's fantastic. That's fantastic. And kudos to that emergency department that they were able to have it make sense. I wish we were always able to do that. A lot of people I think see the emergency department as a black box. I mean patient in, patient out. The same could be said for a hospital as well. And I'll use the words of my mentor Marc Smith and also Dr. Craig Freied who wrote a manuscript back in '99 looking at the emergency department as a complex system. In a sense, it really looks much like a hospital. People go in, they're evaluated, they're treated, tests are run, decisions are made, hopefully cures occur and then the patient comes out on the other side. So the emergency department is kind of a small temporally compressed version of that. In fact, though, the emergency department looks much more like this. And we'll fill in each of these boxes. So if you can kind of keep this framework at the front of your mind as we're walking through flow, I think it will be kind of easy to understand where things fall into place. So you know, the first kind of space in flow is a prehospital space. It's not even in the department itself. Second is when people come, they register and they're triaged. We'll of course walk through this. Evaluation and management is by far the most complicated, in a sense the most chaotic and it's a place where we'll be able to explore some of the decision making. And then lastly, the disposition; that is, where does the patient go once we're done with them? The diagram itself looks like this, and we'll walk through each part. So keep that in mind. Starting from the beginning. So the prehospital setting is anywhere that the patient comes from. So patients arrive by ambulance, they arrive by helicopters. We run MedStar Transport, we have several helicopters and ground units that fly and pick patients up. They go to scenes, get people off of the freeway, they get them from other hospitals. So they arrive by ambulance, helicopter. DC Fire Department brings them. Sometimes they're transferred via private ambulance. All those people are funneled in. Clearly they walk in as well and are dropped off by friends. And at times we will get people from the hospital. These are patients who are in the cafeteria visiting a family and who become acutely ill and are brought to the emergency department. So once a patient is -- enters into the physical space of the emergency department, the first thing that happens is that they're registered. We have to be able to generate a chart for them, a computer chart or an electronic chart. We generate a medical record number and a visit number. And that happens when a patient presents to a front desk and states their chief complaint. The chief complaint I wish it would be something like well I have Chess pain of cardiac etiology. But often it's I don't feel well. So -- and we'll see how that can be kind of confusing in terms of codifying why people are actually in the emergency department. Their demographics, how they're going to pay of course, we have to get paid. And then that's when the paper chart is generated. I won't get into all disparate electronic medical records but in our department currently we have paper charts that are filled out and then scanned into an electronic medical record. So after the chart is generated, then they'll be triaged. So a triage, if anybody speaks French is French to sort. And what triage does is when a person enters, it's a way of determining essentially how sick you are. You'll hear an emergency medicine of sick or not sick. A lot of what we do especially is determining -certainly we want to find a diagnosis, but we want to know are you sick; IE, are you likely to die in X amount of time or are you not? And by assigning a triage priority that's a way initially of sort of deciding how quickly a person should be seen and what resources that they're likely to need. This is done by a nurse. And it's based on symptoms, age, vital signs that are taken initially. That's just heart rate, blood pressure, temperature, oxygen level. And just getting a quick past medical history. There are two ways of doing this. There is an ESI algorithm, couldn't even tell you what it stands for. But this is an algorithm that basically you can go through that will then assign you a triage number. I will tell you seasoned nurses can look at a person and say you look like you're about a two. You look like you're a four. One is the most acute and then four are, you know, sprained ankles, hang nails, that sort of thing. Okay? And then patients are sequentially assigned to the treatment teams. We have four treatment teams. We have the blue team, the red team, the green team. The color has no significance at all. They are just geographic areas. And then we have the ambulatory care area, which is like an urgent care or fast track. Most, if not all, larger emergency departments have these areas. From there, the patients will often go through a triage team. This provides a lot of confusion for patients and for people who aren't familiar with the flow of the department. And I'll try to make this as clear as possible. The triage team is sort of an optional team that's in place through which all patients are funneled on the way to the teams to which they're actually assigned. The purpose of the triage team is to do just sort of a rapid evaluation to sort of get the ball rolling and to anticipate any diagnostic procedures, any diagnostic tests that the team physician on the other end would likely order. A quick example would be someone who comes in and has Chess pain. Instead of -- after their triaged, instead of putting the patient in the waiting room you can say well, Dr. Gateway is going to see you on the green team, the wait may be a while, but in the time being we'll go ahead and draw some blood and sent to you chest x-ray so that by the time all of these things are back Dr. Gateway will be able to make a more informed decision. The triage team at our hospital emergency department is only operates between 10 a.m. and midnight. So actually the majority of the -- you know, around the clock we do not have the triage team. But we do during the busiest times. And we've actually seen that that does sort of increase throughput, depending on how you measure it. So patients wait in the waiting area until they're seen by the respective team after they see the triage doctor, okay. So you go, triage doctor says hi and then you still sit out in the waiting room. It can be several hours on busy day until you're actually able to see the physician to which you're assigned. Medical decision making is incredibly complex. It takes a period of observation. It takes some thought. And it takes getting a very full history of the illness and to go a good physical exam in a period of observation. So the evaluation by this doctor in triage is not a substitute for a full examination on the other respective teams. That's something that patients often have a hard time with, they say well, a doctor already saw me, why can't I go home? And it's sometimes difficult to explain that it's not always that simple. So, after they've waited, patients are placed in rooms on the respective geographic teams as they become available. They're first placed in order of acute and then duration of weight. This causes a degree of duress because patients don't necessarily understand that well, this guy is -- if I have an 85 year old with chest pain who has had medical heart history stints placed and has a history of a heart attack, I understand that you've been here for five hours with a sprained ankle and that this patient just came in. She's much more acute. She has a higher triage number or a lower number, higher acuity, so she needs to come back first. And then after acuity of course it's then placed by duration of wait. As I mentioned, the team colors have no relation to acuity. And the least acute patients during the hours at least at our hospital of 10 a.m. and midnight go to the ambulatory care area, okay? And so they're sort of fast tracked through. So during the evaluation, this is where the patient physician relationship really begins. This is when the doctor takes a full medical history, performs a physical exam and begins the workup or thinking about the diagnostic therapy, the diagnostics and the therapy. That's when they're initiated. And this is really when medical decision making begins. Some of it can begin back around where the triage doctor is. But all that is is to anticipate the medical decision making that will take place here. Any questions thus far? I think that's kind of a good place to -- okay. Yeah? >>: Yeah. You might get this -- is it your -- so you're describing how it kind of works at the abstract. At least in some places it -- it actually opportunity always work that way in the sense that the nurse -- that people may be assigned an emergency [inaudible] and next number, it's not actually -- it doesn't actually correspond to the one that's in the book. >> Justin Gatewood: Absolutely. >>: Right. >> Justin Gatewood: And this is part of the -- that's part of the chaos that can happen. As you mention, the ESI algorithm is not 100 percent, clinical gestalt by even the most seasoned of nurse is not. Patients will often declare their true acuity. And remember acuity can change because the natural course of disease can change. And so bad outcomes can happen in the waiting room someone who comes in with chest pain then begins to have abnormal vital signs and may they're themselves as being more acute. So the idea is that we can get people back to see doctors as soon as possible so that someone doesn't start off a four and then a three and then a two, all this happening within the waiting room. And that's why we've sort have put the triage team out front, to be able to kind of capture some of that. But that's a very good point. This is best case scenario how it's happened. And I'll actually go through a couple of scenarios as well. So after the -- we'll talk just briefly about physician documentation, how it happens at our hospital in our department. Physician documentation happens around the patient-physician encounter. So after taking a good history and a physical exam, the physician will either handwrite the note at the bedside, will go out depending on an individual sort of microworkflow, write it outside of the room. Sometimes they'll busy -- they'll just keep it all in their brain. They'll go do other tasks. An hour later they'll come back and hand write the note or type the note. Currently notes are not being typed bedside. Workflow in the environment just doesn't facilitate that. Handwritten notes and also the typed notes are sometimes scanned, and then into the electronic medical record. Now we have Paladin which is sort of a module on to Amalga, which will allow you to electronically sign. And so we've got different physicians refer to handled documentation differently. And that's only in our department. So after the evaluation, which is when the workup begins. And a workup basically is just, you know, the diagnostic tests. So patients and their appropriate specimens flow to the appropriate laboratory or imaging suite. So what happens is a physician comes out and says this person could have this, I want to order this. They go, they write down the order, they give it to a clerk, the clerk inputs it into the system and then the sort of ancillary diagnostic suites are activated so the CAT scan will say we see an order here for CAT scan of the brain, then we will -- when time allows, send patient transport out to get that patient. The patient will go over, get the CAT scan an then go back. Urine is sent, blood is sent, spinal fluids, et cetera, are sent to the labs. What happens is the clerk -- and this is getting pretty granular, but the clerk will print out a sticker to go on the tube with the patient's name, the type of test ordered. That will be sent down to the lab. So the physician orders the says and takes this -- the nurse takes the specimen which is sent either to the hospital laboratory or also the emergency department itself has its own laboratory that does sort of quick, you know, time critical tests. Sort of ultra time critical. Everything is time critical in the emergency department. Those results, if they are critical, will often be called directly to the physician for really high values or really low values depending on the test. Once those tests are run, they'll also put into the local database that's sort of specific to the lab, and then the electronic medical record, Amalga is able to locate those from disparate databases and then pull them into one field, which the physician can view. And then the physician views the electronic medical record. Radiology happens similarly. The physician orders the tests, either performs -sometimes them self will perform some bedside imaging techniques in the case of ultrasound more often they will go to the respective imaging suite. That information is then put into a local database, which is then viewed by the radiologist. So emergency physicians are credentialed to read certain types of imaging studies but we can't read them all. The radiologist will then call back the results to us or dictate them and have a written report and an electronic medical record. Often in the case of a simple x-ray, the emergency physician will just look at the image in Amalga and then make a decision. Yeah? >>: Is the quality of the in and out of the department tests -- assumed to be roughly the same, or is there also a discrepancy, the trading off time for quality? >> Justin Gatewood: No. There are certain standards that are established by the department of laboratory medicine, and so although the tests are run in different ways, you are -- you can hang your medical decision making hat on the tests that you get. It's -- we would love to do everything in department but sheer volume prevents us from doing that. So therapy. Once you've decided a workup, you start to administer therapy, the appropriate IVs, medication, oxygen, sometimes procedures are performed and then often consultants or specialists are involved, depending on the case. Now, all of this happens at sometimes simultaneously. Some processes are dependent on other things. In terms of therapy, often you are administering therapy and then you are going back to reevaluate the response to your therapy based on that decision sometimes you'll order another x-ray, another CAT scan, order more plod or urine tests. So these things are interdependent, often they're also happening in parallel and there are multiple feedback loops happening. This is for one patient. So, you know, superimpose this same flow -- you know, as an attending emergency physician it's not uncommon that on a very busy Monday afternoon I will have 25 people at a time assigned to my team. So imagine that big nasty diagram 25 folks deep and, you know, you've only got one lab. You've only got one -- you know, two or three CAT scan machines. You've only got one orthopedic surgeon in house available for consults. So things can get -- that's really where the bottlenecks occur. And then lastly we'll talk about disposition. Clearly certain medical conditions and responses to therapy dictate where the patient goes. If they are being discharged, it implies usually that they're going home. Sometimes they're being transferred to another institution. Often they will leave against medical advice, which is unfortunate, but sometimes what we deem as being sort of a priority does not always -- and if you've got five dogs at home and no one's going to feed the dogs, then maybe you just don't want to stay to get that other test. And so sometimes, you know, that's important for the patient. Others will just sort of disappear and leave AWOL, and then unfortunately some patients do die as well. If the patient is not discharged, they're then admitted. So we admit patients based on how ill they are based on the need for extra more tests, more interventions, things that cannot be accomplished while they're in the emergency department. And there are several areas in the emergency department to which they're admitted. And that really depends on how ill they are and what resources they need. So when someone is just admitted to the floor, this is just sort of a general medical or surgical bed. These patients are sick enough to stay in the hospital, but, you know, they don't -- they're not critically ill. Critically ill patients stay in the ICU. We have different ICUs depending on specialty. Sometimes patient meet criteria to be in the intermediate care, which is sort of in between. And I won't get into that because it just goes into a lot of medicine. And then some folks go to the operating room. So here again is that big nasty flow chart. And one thing I want to point out is you'll hear -- there's several metrics that we track and are easy to calculate. The door to doctor time is probably the one that you'll hear the most. And that's basically how long it takes you to see a doctor. Door to decision time is the time that they enter the door until the time that a decision to either discharge them or admit them is made. After they're admitted, they may reside in the emergency department proper for some amount of time. That's the boarding time, and it's completely -- it is motion dependent upon what else is happening in the hospital. And then of course the length of stay is from the time they enter the door to the time they physically leave the department. So we talked a little bit about, you know, kind of individual flow of patients. And I wanted to talk about sort of stepping back and viewing the department and all the chaos. And here's sort of a face based diagram of the emergency department kind of the state of emergency department. And you could probably -- there are probably several different types of states. But the one that I wanted to talk about is sort of the progression from normal operations to disaster. It's great you get there at seven o'clock in the morning. There may be some intoxicated folks sleeping it off from last night. Someone slips on the ice in the morning. The department's slow. Folks are not that acute. You're down here, hey, great. Normal ops we're getting people through. And on this axis here is the number of patients. This is basically just the measure of acuity. And so when there's a time when you have a lot of patients and really none of them are that acute but there's still so many, that can be classified really down kind of in the disaster zone. In fact, when you talk about emergency preparedness and disaster medicine, the majority of people who present even after a major bus crash or a dirty bomb detonation are people who are not very acute. It's usually the walking wounded, people who are just freaked out and they just want to get checked. People who have -- you know, rolled their ankle and aren't very sick. But occasionally, let's say an example where like we all heard unfortunately in Lakewood yesterday where the four police were shot, it was early in the morning. They may have been taken to a local emergency department where there were a few patients, but four shooting victims are all very acute, could be classified as a disaster. The funny thing is, and just sort of think about this, is at some point we say this is a disaster and we flip a switch and we make a call to the head of the hospital and we go into disaster mode. And when we do that, we open these pelican boxes, we put on -- we put on these vests, we have walkie-talkies and we sort of get into disaster mode and we follow, you know, sort of a different flow of patient care. And we change some of our resources and we change the way that we look at patients a little bit. >>: [inaudible]. >> Justin Gatewood: Once every couple of years probably. At our hospital it happened I think during the anthrax scare, it happened during 9/11. >>: Those cases where the large [inaudible] Y disaster tons of people coming in because they were scared? >> Justin Gatewood: I wasn't there. But I will tell you that when I hear them talking about them, especially the anthrax, you're here. Now, this can actually -if you take this and you look at it over time as well, you can have several weeks that will sort of stay around here. For example, Monday afternoon you have a lot of people who have flu like symptoms. None of them are very sick, but they all want to know dock, check me out. Tell me if I've done Swine flu. So we sort of hover around here. And so it's just sort of a kind of abstract question. When do we pull the trigger? And ideally we wouldn't pull a trigger. Ideally our operations would be scalable enough that we could just adapt to, you know, a time when you're up in the red zone. I wanted to do this in 3D to show you that you can easily put another parameter or another axis, and that is that the rate at which patients present. A classic example would be you are -- it's eight o'clock in the morning, nothing's happening, and all of a sudden there's a bus accident full of special needs adults. And they just have to get checked out. Nothing's wrong with any of them. And they drive them in. And all of a sudden the bus drops off. You know, 40 people. You know, then you're up around here. So that the just another sort of parameter. But a very interesting question, and one that hopefully we can explore later as well in discussion. When do you -- what do you call disaster and why do you change your operations and what happens to workflow? So, let's talk about some of the types of information. It's hard to actually show a separate flow of information. But I wanted to talk about the types of information and where they -where they're generated and the pattern in which they're generated. So demographics just to sort of review occurs in registration, and it's a one-time thing. People's demographics don't change during the visit. The medical database really rear to consists of the chief complaint, the history of present illness, basically the story behind their chief complaint, their past medical history, the vital signs and then the physical examination. So the chief complaint as you can see happens up front sort of in lay terms. It's recorded by a clerk up front. It's often reiterated in triage. And then really the doctor figures out -- their job is to figure out really why are you here? You say you're sick, you say you're here because you just don't feel well, but it turns out that you've got, you know, the worst chest pain of your life and you're wife just made you come in because you're a pretty stoic guy. And often this is not explicit, the chief complaint. And it's our job to make sure that we find it. The history of present illness is expanded upon in the physician encounter. And it actually is kind of ongoing as well. Patients will add things. Hey, doc, I forgot to tell you this. So it's not a one-time thing. Past medical history is usually a one-time thing because it's available from the medical record. Vital signs are taken by the nurse, and they are taken serially, depending on how ill the patient is. The physical exam is ongoing. It's often goes through a period of reevaluation. If you are classic example is someone with abdominal pain. You're not too sure if they need a CAT scan, but they've got enough risk factors that it could be something bad so you did you have them some pain medicine, do some lab work, go back fill their belly again go back before you're able to make a decision. So -- and when something like that happens, the documentation of that may not end up in the medical record until the patient leaves the department. Until I'm as a physician able to say okay, this is what I want to put into the medical records to describe this patient's exam. Laboratory results come from the laboratory either after being put into the medical record or often if it's a critical result by phone that's ongoing the same as we discussed with radiographic images and the dictations. And then consultants come down and help us decide how to best take care of patients. Sometimes they'll come down and follow a patient through the course of the stake and be ongoing, or sometimes they'll come and say, no, I don't think they have an appendicitis and they can go home. And then also, this is really big, is kind of the nonverbal and subjective things that are -- can be inferred from communication with other staff members can be inferred from talking to the patient. And sometimes ways that other consultants or physicians will write a note in the medical record you can sort of read between the lines and the subtext may tell you something that's not explicitly stated. And that's very important as far as medical decision making as well. So let's talk about principles of decision making. The first is at its core decision making is sort of hypothetico-deductive, and that is you present with some chief complaint, I will then initiate a battery of tests that will aid me in sort of proving my hypothesis wrong or proving myself right. They can also be algorithmic. And this really depends on how ill the patient is or how serious the chief complaint is. If someone comes in and says, you know, my belly just hurts and I've had a fever, there really isn't any strict algorithm for abdominal pain and favor. If you are 65 and you twist your ankle and it hurts here, there is, however, a clear algorithm for how to take care of those patients. In reality it's often a mix of this hypothetico-deductive decision making and an algorithm. So this is sort of getting under the hood and into the brain of an emergency physician's mind. So crossed with this algorithmic and hypothetico-deductive sort of framework and approach is time dependence. The emergency department is intensely time -- and that's why I love it. What you do matters. Everything that you do matters. You have to anticipate the next step because we don't have all day to take care of patients. We have other patients waiting. Some illnesses and some disease processes will reveal themselves as I mentioned earlier over, you know, varying degree of time. And so we're doing, number one, you know on an X axis and so things need to be -- things are very time dependent as well. And time is we're always trying to sort of get ahead of the curve of the natural progression of this disease process. And I'll detail this in an example here is the idea that diagnoses can either be presumptive or confirmed. If it quacks like a duck, it often is a duck. However, if we have the CAT scan that says it's a duck, well then it's a duck. And often we'll treat those two things like ducks. As far as treatment is concerned, we treat things empirically or we use sort of more evidence-based approach. And sometimes they're not mutually exclusive. A good example is a patient who comes in, has an x-ray that has a confirmed pneumonia, but we know based on the patient's age and other things that this is likely a community acquired pneumonia, that there are ive to seven different bacteria that will cause that pneumonia. We will initiate empiric treatment of a particular antibiotic that we know will take care of all of those bacteria. And then as I mentioned before, we're constantly reevaluating the patient, making sure that we're measuring the response to our therapies and ordering any additional tests. Taken in the context of several patients, sometimes up to 25, we may need to reshuffle our priorities. Sick patient take precedence. It's very frustrating, but also understandable when you come out of a room and the patient who two hours ago you just said I'm sorry, I know you don't feel well, I'll be right back, they're calling me, you run in and you spend the next, you know, hour or two hours running a code, resuscitating a dying patient and then you go back in and you say I'm so sorry, where were we? It's born out of necessity. It can be very frustrating. And it's hard sometimes to let other patients know that, hey, I was in there with a dying patient for the last two hours, you know, give me a break, I'm sorry that that happened. That's certainly a part of the job. And then any intervention or any interpretation needs to be taken in its appropriate context, both in terms of the individual or the group. A group sort of context would be in a disaster situation where you say, hey, I got to let the walking wounded kind of just term and often the disaster is we will turn them around and send them home. For the individual, a patient who has a blood level of X who also we know had a blood level of X for the last 10 years and has a reason for it, you don't really get too excited about blood level X, although it's below normal. However, a healthy male or female who comes in, who we assume had a blood level of Y that was normal two weeks ago and now has a blood level of X, we know that that's abnormal. So deciding how we interpret labs and imaging results completely needs to be taken in context, and that provides -- it can be sometimes tough as well. I'm going to quickly go through two examples, and then I think that's about it, and we can open it up. So to display some of these concepts. The first is a 43 year old guy. He calls the ambulance after twisting it while stepping off of a curb. By 7:22, he's in the department. He provides his demographic and insurance information. After hearing his chief complaint and you know looking him over, they say, hey, you're level 4, go to ambulatory care area. He's taken by a wheelchair. The patient waits there for, you know, a little more than half hour to be seen by a doctor. The doctor looks at him and says, great, want to make sure you don't have a fracture. Let's get an x-ray and here's a percocet for your pain. He gets the percocet, he goes over to x-ray, the doctor reviews the x-ray and says you don't have a broken ankle, asks the nurse to splint it. The nurse gives the patient some -- a wrap, some crutches and the patient's discharged. Easy-peasy right? So that was problem a very deterministic way of approaching that system. Depending on how many people were there, there was that converging note of radiology. Maybe the nurse was busy with some other patients. And not particularly chaotic, and there was no reevaluation. That was as I mentioned very algorithmic decision making process. Time was still of the essence. And we had to anticipate get the patient out in a timely manner so that we could see other patients. We confirm that diagnosis by looking at an x-ray, and we use some evidence based practices that say ankles heal better if we wrap them. And there is really a broken ankle is a broken ankle. Case 2 is a little more difficult. And this is kind of more what we live for. It's a 25 year old woman. She didn't call the ambulance like the other guy. She walked into the emergency department. She's 25. She says I've got a little chest pain. She provides her demographic information. She was a three on the emergency sever it index. She had normal vital science, kind of a reassuring story. But, you know, chest pains don't usually go the fast track because more can go wrong in the chest. But in any case, she was assigned a 3, and she was sent over to the blue team. Before going to the blue team, she was seen by a triage doctor who said well, you got chest pain, you look pretty well, but you know, chances are that the doc on the blue team is going to want an x-ray and EKG and I'll order some blood tests that they probably will find pertinent. The lab is drawn by the nurse and sent to the lab. After seeing the triage doctor, the patient goes out into the waiting room for another two plus hours and then waits. At 5:47, finally a room opens up on the blue team and the patient's taken back. And the physician starts the patient-physician encounter, takes a good history, and performs a physical exam. You know, after that, the patient notes that, you know, doc, I'm actually having a bit more pain and the doc says, okay, well, we'll order some pain medicine and the nurse gives it to her. And then she's tape off to x-ray to get this -- to get the chest x-ray. So at 6, though, the patient returns from the x-ray. The nurse takes another set of vital signs and says looks like the heart rate is a little elevated, and it looks like her oxygen level had dropped a little bit. The doctor is at the work station and reviews the x-ray and says that it's normal. But then also reviews the -- reviews the lab work that was done that was actually sent by the triage doctor and says well, I see here that you have an elevated D-dimer. D-dimer is a blood test that checks for degradation products of blood clots. So it's not very specific. A lot of things can give you blood clots. But the utility of this test, and I won't belabor it, is if it is normal, you essentially have excluded the possibility of any blood clots. In this case, it was sent to rule out a blood clot in the lungs. He notes that it's high though. He or she notes that it's high, orders a CAT scan of the chest to rule out a blood clot in the lung for this patient. That's at 6:07. And then at 6:10 the nurse goes to the doctor and says, you know, these vital signs actually look a little worse. So the doctor seeing that and then seeing that the D-dimer is evaluated says well, let's presumptively treat this patient for a blood clot. So the doctor orders an injection and a blood thinner Lovenox, just to prevent the clot from further forming. And then the nurse administers this. Well, at 6:21, the patient not great vital signs but still stable, goes over to the CAT scan where the CAT scan of a chest is performed and then comes back to the emergency department. When the patient comes back, and I hope your heart is just fluttering right now like mine is, in anticipation of what's going to happen to this patient, the blood pressure actually is dangerously low when the patient comes back. The doctor thinks this is probably a worsening blood clot, gets the cardiologist to come and perform an ultrasound of the heart to see if this blood pressure is caused by that. At the bedside the cardiologist performs this echocardiogram and notices that there's strain on the heart, presumptively from the blood clot. And then the doctor then orders TPA, which is a clot butting sort of clot chewing medication. This is administered by the nurse. The doctor then calls -- requests a bed up in the cardiac care unit, critical care unit. Yeah? >>: So you're waiting on results of the CT scan, you're treating presumptively for embolism that the CT scan can pretty much confirm or deny? >> Justin Gatewood: The CT scan as soon as it's performed and is in the local database and then is in the electronic medical record and that takes a while to happen ->>: How much pressure do you have to -- for a case like this with the CT, you're already treating presumptively for some things is pretty critical for how much pressure can you exert on the radiologist or on whatever other [inaudible]. >> Justin Gatewood: As much as the system will allow. The radiologists often are very overwhelmed. May be you know, four gunshot wounds just came in and, you know, they're booked up and they say well, we got a bunch of trauma cases ->>: They have their own internal sort of triage system that's not necessarily first come first serve, they also have a severity system roughly for other parts of the hospital so ->> Justin Gatewood: Imaging from some areas of the hospital take precedence. So CAT scan -- you know, we have dedicated critical care radiologist who only read our scans and scans from the trauma unit. But in a smaller hospital, all the routine daily x-rays and CAT scans often are read by the same radiologist. So often we need to make time sensitive decisions without confirmed -- yeah. And that's absolutely the point of that. And then as you can see, then at 7:07, after we've already administered this TPA, the clot busting mechanics, the doc calls and says -- or the radiologist calls and says hey, I just want to let you know that your patient has a blood clot. And we say yeah, thanks. We kind of, you know, we needed to act on the information that we had, and the patient's now tee'd up for a CCU bed. And then at 9:12, after more than two hours boarding and a length of stay of more than six hours the patient is then transported up to the critical care unit. That is much more typical of kind of what we're dealing with and patient by patient, day to day. That does represent a very sick patient. Remember what I said. About three percent of the patients that we see are admitted to a critical care unit. So. And then we talked about clearly this is deterministic and non deterministic flow. You know, the asynchronous and parallel flow. We talked about the feedback loops in terms of maybe administering pain medicine and going back and seeing what effect it has. And clearly there's chaos involved in that, especially if you sort of zoom out and you see nurses running and phone calls being made. But it all is embedded within a relative -- you know, relatively -I don't want to say easy, but defined system. And then more or less most of these decision making concepts are utilized here. So quickly we'll talk about currently projects. And I just kind of want to open it up, and we can chat. One thing that we're doing and this is -- I've been working with Greg and Desney and Dan -- yeah? >>: Just before we transition, you said that the patient was randomly assigned to the blue. Do you literally mean randomly, or it just so happened that based on availability ->> Justin Gatewood: I shouldn't have said randomly. Honestly it's very low tech. How it happens is there is a rack for blue team, for red team, for green team. The triage nurse after triaging the patient takes the chart. There is a clip. Dan was there, saw it. Des was there, saw it too. There is a clip on the rack, whatever clip is on the rack you put the chart in that rack, you move the clip to the next rack. So not randomly, it's just sort of sequentially. >>: [inaudible]. >> Justin Gatewood: But the point is that there's no relation to acuity. >>: Sure. I ->> Justin Gatewood: Yeah. >>: Just randomly sounded like you were doing some sort of trial to figure out which team was the best and the patient might wait [inaudible] just so that you can do that. >> Justin Gatewood: Embedded in that flow chart I have all of the colors going from all of the teams to all of the other ones. And that is because, come on, if -you know, if the green team doc is having a bad day and is very slow and they're piling up, it makes sense to do a little reshuffling. And although we set up the system so that shouldn't happen, often that happens. >>: The teams are not fixed to people, the physicians are randomly assigned to teams, also, so it's not really like you would evaluate one thing at a time in that sense? >>: Yeah, it just seemed a little odd for it to be random. >> Justin Gatewood: Yeah. >>: [inaudible]. >> Justin Gatewood: Any other quick questions just about the flow before we talk about quickly the projects that we're working on so far? And they'll be plenty of time for questions afterwards as well. So one thing that we wanted to do was in working with Dan and Greg and Desney on this is how do we effectively model throughput, you know? There are quite a few papers out there on neural networks. There's some work out of classic machine learning stuff. But people have been doing kind of neural networks, artificial neural networks of simplified and small emergency departments for quite a long time, for the last, you know, 10, 15 years or so. And this particular project we're using various machine learning techniques, none of which I'm at all qualified to talk about, although they sound pretty cool. And the training data that we're using is from Microsoft Amalga-generated timestamps. So most of what happens in the department requires a click by somebody that generates timestamps. They all reside on a SQL server database and we can utilize those and feed them through fancy SQL server algorithms. And so we're able to predict with and increasingly so all the time with reasonable accuracy a lot of different milestones in the visit. The one that we're utilizing most in trying to display is trying to predict wait times based on these models and display the wait times in a way to patients waiting out in the waiting room. This is important because patients often wait -- I mean, you can wait six hours pretty easily on a busy Monday afternoon. And the idea behind this is that we are given these kind of receding flood bars to monitor progress. The problem that we're finding -- and this I think is actually one of the more interesting problems is, you know, on one side the patient is given a ton -- is given no information, kept in the dark. And the other side if we give the patient too much information and our model is incorrect or is not accurate enough, we're setting up unachievable expectations for the providers where the patient will come up and say hey, you said you were going to see me at 2 hours, 37 minute, at 2 hours 38 minutes what's going on. So this is just a screen shot of what is -- embedded in this is a dynamic model that will, you know, display these receding flood bars. And there really is no time anchor for this. And this is one of the things that we're trying to work on is how do we display this information so that it's -- it provides the appropriate amount of information. Yeah? >>: How often do you see patients lie about [inaudible] in order to get prioritized [inaudible]. >> Justin Gatewood: It is not uncommon that patients will -- especially those who know the system will manipulate the system for secondary gain, whether that's pain medication, whether that's to be seen more quickly. It's very common. >>: And I assume -- is it safe to assume that the more information you give them about weight time such the more likely they are to do that? >> Justin Gatewood: Yes. To an extent. Often our triage nurses are great and we can say, well, I understand that you say you're having the worse pain in your life. We see here that your vital signs are totally normal, and we know that people who exhibit -- who are in true pain usually have an elevated heart rate. And there are kind of ways of -- but it's dangerous. A lot of patients know very keywords they can say that will get them straight back. But generally patients are -- patients wait a long time and often don't say anything. Yeah? >>: So perhaps a slightly different direction of that is [inaudible] the system in the other way. So you talked about like Monday afternoon I'm feeling a little sniffly, I'm going to, you know, appear at the ER or whatever. The flip of that is so before we start it sounded like a bunch of people this weekend probably you know put up holiday declarations and it's Thanksgiving weekend and I fell off my roof and whatever. I'm wondering if like we know that there are rush hours for car traffic and like we know that even the DMV says don't show up to renew your driver's license on a Wednesday because we have higher than normal volume on all Wednesday. Do you -- is there something interesting that can be abstracted so like for example I didn't know that there was on Monday afternoon -- didn't occur to me there was a Monday afternoon phenomenon in the ER. When should I put my holiday lights up? I could -- if I could do it Wednesday morning [laughter] and like then I fall off my roof [inaudible]. >> Justin Gatewood: So, you know, unless you -- that's a great ->>: [inaudible] as much as I can schedule the things that cause me to get damaged, can people? Is there anything interesting there? Or is that ->> Justin Gatewood: So essentially. >>: [inaudible]. >> Justin Gatewood: You know, using as a parallel the traffic on a map program, is there any way to adjust to your commute or seeking care? Yes. That could be done. The problem is I think what makes that a little difficult is unless you are using the emergency department a lot. >>: Right. As your primary care provider. >> Justin Gatewood: As your primary care provider. Or even if you're chronically ill but no hey I'm going to go to Atlantic City this weekend and I'll probably eat a lot of salty food and come back and my congestive heart failure will be acting up, you know. But I think that's very useful. Another thing is we do not want to dissuade for two reasons anybody to come in the emergency department. Number one, the more patients we see, the more people we can take care of, and the more money we can make. Right? Number two is it gets very tricky when you're putting anything out there that would ever dissuade a person from seeking care. And that's one thing that we -- a risk that we run with that and we'll monitor that is I understand that you have chest pain, you've been made a triage level two, but the wait for a two, believe it or not, is still two hours because we're so busy. Assume that the other twos are either sicker twos or got there before you did. I understand that two hours is a long time, but at least you're in our sights and we can see you. We've got some vital signs. We would hate for anybody to come and say oh, man, my wait's this, well, I'm going to go somewhere else, or I'm just going to come back tomorrow. >>: Yeah, you might keep them in there. It's great to give them data when you've got them on the facility but if they're at home like the wait is three hours, I'm going to just hang out here, oh, my God, I'm dead. >> Justin Gatewood: Absolutely. And that's a huge fear. >>: A pulmonary embolism, your wait is four hours. I'll just hang out here. >> Justin Gatewood: And it's much better for the natural course of that disease process to really manifest while you are in a waiting room or monitored bed where things can still be done than to happen at home when you're by yourself. >>: [inaudible] my weird outlier I fell off my roof because I was doing this you know largely self-inflicted thing. >> Justin Gatewood: Yeah. >>: Almost no value to exposing that data outside the hospital. >> Justin Gatewood: Well, plenty of healthcare systems have displayed average wait times on bill boards like along the highway or on websites. And what they found is there's obviously sort of a redistribution. But we don't want to redistribute patients. We want to see everybody and anybody that comes to our emergency department. But we want to do it quickly and safely so that we can make room for the next person who's coming in behind you. >>: I see ->>: Tell him to put up the Christmas tree ornaments before Thanksgiving so you might beat the rush. [laughter]. >>: I assume relative wait times would be relatively harmless to the patient and potentially harmless to the bottom line. If you are at a hospital system where the same hospital system -- where the money all went to the same place ->> Justin Gatewood: And the experiment that I mentioned were in those settings. >>: Well, I think on the flip side, your idea capturing the wait times and time of year and weather is how do you stack. You know you're going to have a -- >>: Yeah, yeah, yeah, that's actually great. >>: And then you know stack for those things. And then if we collect the right data, since Amalga can take data from anything, data is data, then it would be very interesting to see is there actual causation or just correlation in an assortment of data. >> Justin Gatewood: And just so that you know, when I gave those ranges of staffing, we staff for averages and the only variations that we -- we have a couple less docs during the nighttime. And we ramp up during the day. But -- and sometimes some of the actual like the triage team shifts we don't have on weekends because weekends actually tend to be pretty slow. Besides that, there really are kind of just staffer averages. But the idea is that the system is scalable that we can, you know, take -- that we can work quickly in a surge situation. Unfortunately that doesn't always happen because, you know, union laws govern some of these things. Just plain old ethical laws govern some of these things. You cannot have a nurse, even though it just takes me -- it may take me 15, 20 minutes to go see a patient, write a documentation, and then order some stuff, well a nurse has to do that and he or she may have another critically ill patient that is boarding in the emergency department for two hours because the hospital's so full that there's not a room upstairs, and so you can't assign although there's no law against assigning, you know, a physician to N number of patients. The nurses and the nurse management, and I think rightfully so, once they start getting four, five patients, they're like it's not safe. >>: But that is just data too. And presumably some, you know, future or maybe current version of Amalga can consume that same type of information, thou shalt have this number of minimum level of nurse care. >>: I think that you can also add patient acuity to that. You have a nurse with five patients and they're all the worst acuity, that would be a lot harder to manage than five patients that have sprained ankles. And so we could do some sort of algorithm and then add in staffing with historical on Christmas Eve you know somebody's going to call in sick. And so then staff, those sort of staff fluctuations you might stand a better chance of really evening things out. >> Justin Gatewood: One thing I hope that you'll find too the more we can work together is -- and Dan and Desney especially having subjected to this, you know, there's a certain culture in the emergency department. There's a certain hierarchy within nurses, within physicians, within the technologists within the clerks. And between those as well. And so certain cultural potentially cultural attitudes often make some of these things difficult. As we all know culture sort of lags behind in medicine often what makes sense. I mean, it's only until recently that residency programs, training programs have put caps on work hours. Before that, you know, working 120, 140 hours as much as you needed to in a week was the norm, and the idea was, well, you know, tough, it's sort of a right of passage. So the first thing I think would be to prove that there are bad outcomes because staffing levels are low and then that would give you the leverage to be able to use that data to fix things. But it's incredibly interesting. And I -- you know, I invite you all to come out and hang out, and you're all welcome any time to get an idea of how subtle of these cultural nuances shape implementation and also valuation of the data. >>: Are there approximate simulators that have been built to try to figure out what happens, how long it takes for these pipelines of -- how long it takes to do a CT scan and how long it takes to do each thing where you'd say okay, if we threw another CT scan in here, how would that change is workflow? Do you know if this has been done. >> Justin Gatewood: I don't -- I'm sure that there have been -- I glanced over a couple papers that related to using neural networks to show if you perturb the system in this way how well your outcomes for example heart attacks or -- but in -- as far as resource utilization, I have not. I mean, I think there have been ->>: There have been [inaudible] if you could add another machine or if you could add more staffing on that. There's a lot of literature, but they tend to be ED specific sometimes, so they can ->>: [inaudible]. >>: So I looked at some of the simulation picture and [inaudible] and one of the problems in the literature is that the models that [inaudible] doesn't take into consideration the culture there [inaudible]. Do you have any idea about how you could incorporate that [inaudible]? >> Justin Gatewood: As far as -- well, I think there are two ways. There is -- you try to modify -- you essentially try to change culture or at least modify what you can to streamline it. As far as trying to codify that and having it be another parameter in your model, to account for it, one thing that you may be able to do is if you were -- right now a doctor is a doctor, right? A nurse is a nurse. You could look at differences in throughput or care given by different providers and then sort of, you know, weigh -- well, Dr. X is a little slower, Dr. X is a little faster, this nursing team -- even these nurses when together are more efficient and use that as sort of a proxy. I would love to do that. I would tell you that you have to be very sensitive with that data. No one wants to -- you end up putting people in competition. >>: This is a small end but I've looked at a hospital system in Atlanta and huge cultural difference between teaching hospital and non-teaching hospital. So if you were starting down that path, that might be one ->> Justin Gatewood: Yup. >>: And it leads to first of all staffing differences and then other cultural differences as well. So. >> Justin Gatewood: And you may be able to actually use that, though, if it is a teaching hospital it's a lot easier to evaluate residents and medical students and nursing trainees than it is to say to the 30 year emergency attending veteran we're going to be watching you, you know. >>: Do you think number of years in the field could be a determiner or do you think that it's based -- once you're an attending? Because I think there's a clear line between residency and attending that once you're an attending that years of experiences ->> Justin Gatewood: Some of our residents see more than our senior attendings. Some of our senior attendings are so fast it's -- you know, are faster than some of the young bucks. I mean, it's really ->>: It's all over the map? >> Justin Gatewood: I think it's all over the map. And I think things would ultimately would kind of center towards the middle. Which then you beg the question, well, is that even -- is that even a pertinent factor. I wanted to just get your brief idea. I'll show this as well, something we're working on with Lauren Wilcox. She's out at Columbia working on her PhD. And she was here as an intern. And we wanted to find out -- she wanted to find out is there any benefit to having an in-room display that may potentially be manipulated by the patient or that they would have access to at the bedside that could show, you know, what's the care team information, what are some of the results of my diagnostic tests, you know, things like, well, what's next. And then also what patients and what the physicians and providers deem would be relevant or appropriate information to put on one of those displays. And so we were able to complete a complex paper prototype, a big poster that was actually assembled by Desney and Lauren and Dan in realtime using information from the physician and the electronic medical record. I'm sorry, I wish I could make that a little bit bigger. But we short of just finished kind of a informative study, and we've got a lot of other questions. But I think a great step is just received really well by hospital administration. Very sort of high profile thing that was implemented relatively easily. I mean, you guys put a lot of work into it and has received -- and most importantly the patients absolutely love it and felt quite empowered by it. So as far as potential projects, things that I've been thinking about and that have been mentioned as well by other physicians is interest -- and Dan and I are -and I read your -- is there a way to sort of intelligently automate these patient displays; that is, do we have physician defined rules as to what is okay to display, taking into account that the interpretation of labs and images is contextual based on that individual person if that person has never been there before, been there once, you have relatively sort of shallow pool of data to choose from to be able to make those decisions. But I think that's a ton of fun to think about. And then also, you know, right now our documentation kind of impedes workflow. I mean, you're sitting there. And if you're like me, you're writing madly in the room, but how much eye contact am I making with the patient? Well, you could also talk to the patient and then go outside and sit down and then type your note and stay after your shift or type your note and then nurses and other docs are coming up to you. I mean really we just don't have a way to seamlessly sort of capture that encounter. And is there a way sort of passively of doing it. You know, perhaps using voice recognition to sort of populate, sort of predetermine templates that would kind of listen in and then populate sort of like an e-Scribe. Right now there are some places that have scribes. We're usually medical students or pre-med students who listen and take notes and then the physician will review those before they enter the medical record. But that -- that would be incredibly useful. And the question would be, you know, how does that improve efficiency and also what impact does it have on the patient-physician relationship and ultimately patient satisfaction? >>: [inaudible] just stay on that idea just a minute. And when you were talking I visualized you standing by the bedside with a voice-activated device like a recorder. Based on your prompts ->> Justin Gatewood: This would be ->>: [inaudible]. >> Justin Gatewood: Yeah, and again, I don't know what technology exists. This would not be just simply a dictation machine that's voice activated that will turn on, but you know, if you had -- if you had some device that had preloaded templates so I understand you're here for chest pain. Opens up chest pain. And then so how long -- you know, what duration of your chest pain. Listens to duration and may populate that portion. Because when we document, we document for three reasons. To relay medical information between providers. We document for medical legal purposes, to make sure that we're doing the right thing if we get sued. And then also we document for billing. So a lot of the templates that exist in these documentation systems that are already in the market, they exist to capture sufficient information for documentation. And I think using Protex someone in medical MediaLab has already used some natural -- Pete was telling me using natural language processing to sort of parse out relevant things for billing. But if we could do that using some sort of voice system, I think, you know, that may be a pipe dream but it seems like ->>: No, no, I think it's very doable. If you had a template that you could hit a tab on the keyboard and then it record your conversation and then pull out through the natural language processing pull out the relevant information, once it then comes to you, you do a little cleanup and sign it so that you then captured all the voice interactions between you and the patient, which we can leave as voice tags in the actual document. >> Justin Gatewood: You know, that would be something ideal. But ->>: I don't think that's beyond ->>: I wonder if there's even a -- the systems that are regularly used in particular [inaudible] where the medical professional is someone who may or may not actually have specific medical training and they're walked through a very form laic kind of flow chart thing, you could imagine a similar kind of tool that, you know, the doc is working from and so that because they're walking through ostensibly a logical template, the system then knows I'm here, I'm there, and you can enter the tab on the little thing on the screen that says I'm going down this branch and so then you have some contextual awareness presumably about the kinds of data that the system will be expecting. Would you describe the chest pain as mild or really not at all mild? Okay. And then that becomes data rather than nearly a string that somebody needs to text ties as well as keeping your -you know, the voice stuff around and somebody says but this doesn't make any sense and then they turn back. >>: Reporting [inaudible] voice that says what is the duration of your chest pain? >> Justin Gatewood: I just pulled out the [inaudible] [laughter]. >>: Or like with the tablet that you have, you know, you could stand at the bedside and click that, record everything as being said, click the next button, record it. >> Justin Gatewood: Now, there are T systems that exist, they're called T systems that exist for documentation that allow I to sort of electronic systems where you just sort of check boxes. And that could be. But it still requires kind of deflecting your attention. >>: Right. >> Justin Gatewood: And it also doesn't leave room for -- there's nuance verbiages that -- verbiage that often needs to make its way into electronic medical record. So you need to have a space to be able to edit that afterwards. And I by no means am an expert on all of those products that exist. But we can keep talking about that. And then also using sort of fuzzy text classification schemes to be able to make better use of the text inputs that we all right have now. So program, the chief complaint field is often misspelled, may have some loose association with what it really -- what the chief complaint truly is, and that is a ton of data that right now we're just not able to use. And if we were able to incorporate that into our model, clearly chest pains look much more -- look like kind of going down the pipe than does a chest pain and an abdominal pain even though they both may be a two, may come in at the same time and be the same age. So we think that there's probably some wait in that as well. So those are things that we've brain stormed and hopefully we'll continue. I know we're at 12 now, and I want to be cognizant of time, but is there anything that anybody -- I hope this was, you know, somewhat useful to kind of give an idea as to kind of what goes through our brains. Great. >>: Very interesting. Thank you. >> Justin Gatewood: Well, thank you so much for your time. [applause].