well good afternoon everyone and welcome to today's ifam webinar i'd like to begin some introductions while the room is filling up i'm ron ackerman i direct ifam and i'm senior associate dean for public health here at feinberg school of medicine i'm very delighted today to welcome you to our presentation which is co-sponsored by our program in public health in commemoration of the program's 25th anniversary in the coming year today's presentation is titled integrating clinical decision support into everyday care and it's given by ethan molich howe who is a graduate of the program in public health in 2008 and now is assistant professor of medicine in the section of hospital medicine at university of chicago as a reminder during all of our webinars we encourage the audience to pose questions of our presenter if you have any questions during at any point during the presentation simply type them using the q a function at the bottom of your zoom screen please don't use the chat function as we won't be monitoring it routinely for your questions ethan molecho is an academic hospitalist who is board certified in internal medicine as i said he's a graduate of our program in public health as well as the northwestern school of medicine dr mohlich hoe is interested in how electronic medical records and health information technology can impact direct patient care patient physician interaction and quality improvement in hospital care he's currently ufc's section of hospital medicine's lead hospitalist for the care of patients with covet 19 which includes responsibilities in areas of administration operations and clinical care health information technology education quality improvement and research with this introduction i now would like to turn it over and it's my pleasure to introduce to you dr molag hode dr moloko thank you so much for the uh kind introduction and afternoon to everyone uh in the audience here i'm excited to be uh virtually back at northwestern if not in person at least uh virtually kind of uh coming back to my roots in medicine here and so i'm excited to be here and talking with you all and so just a little bit of the um overview of what i'll be talking about first i'll kind of start just with some of our basic challenges in clinical decision making um then provide uh an introduction to what clinical decision support systems are and and how they integrate and then discuss some of the current practical applications that i've been involved with as well as some of my colleagues and some of the ongoing efforts here at the university of chicago in our section of hospital medicine so i want to start with the case um so mr ree i know we've all probably had our mr reed cases uh in the past but mr reid uh he's a 66 year old gentleman with a history of hypertension diabetes and a smoking history who came to the emergency room with the chief complaint of difficulty and breathing and so those of us that are clinically trained in medic in medicine you're probably already starting to think of your differential uh diagnosis for this patient what's bringing them in with uh shortness of breath is it a new onset heart failure from this long-standing hypertension is it new copd from a smoking history or a new lung cancer and given the climate of our times could this be a new covid diagnosis you're already starting to think of the next questions that you're going to be asking um you know what brings on your shortness of breath are you having chest pain with it are you having any lower extremity swelling are you coughing are you having fevers and you know are you having any long car rides recently you're probably already thinking about what tests you're going to be ordering you know are you going to be ordering a chest x-ray or an ekg or an echocardiogram all these things are going through your head just based off of your years of reading about this chief complaint of shortness of breath and maybe you're reading about hypertension and smoking history and all of that is kind of informing your uh question asking and what uh what clinical decisions you're going to be making and so uh you've probably also read kind of the most recent literature and are up to date on the guidelines that help drive the patient care and so when we think about how we go from our large differential to the beginning down to our diagnosis we're taking in that chief complaint we're asking our medical history we're getting our physical exam and getting diagnostic testing and our differentials narrowing down to an organ disease specific disease state or getting our diagnostic impression then get coming up with our diagnosis but along this process there's so many challenges that come into places to when we are making these clinical decisions and coming up with our diagnoses medicine involves so much uncertainty i don't think you'll ever run into an internist that says i i am 100 accurate about uh most things usually we provide a lot of caveats every time we're talking about a disease process and what might happen and what steps we might take there's so many variables that go into every clinical decision um some of which are patient-based go beyond kind of our human capabilities of uh processing all the variables and then do costs and risks of testings we can't test for everything so someone that comes in with shortness of breath we're not gonna get tests for everything that's on that broad differential that we started off with um we're going to narrow our differentials based off of the most likely uh diagnoses and try and narrow our differentials towards coming in we have these biases that uh influence us when we're thinking about our patients not just the biases of gender and race that often come into play but biases and heuristics that come in as to what we're thinking about and i'll go through them we have so many pressures that limit us when we're practicing and we don't have all the information available to us this is just a schematic about you know our human cognitive capacity and the sheer number of variables that are out there for every decision that we're making and as medicine has advanced and improved and we're able to test for so many more things including you know genetics and proteomics and gene expression that can influence how people respond to medications or their likelihood of disease all of these factors that we can test for um just logarithmically expand beyond what our human capacity can really comprehend and goes beyond that these normal decisions that we're making by these clinical phenotypes we have these biases and decision making and i believe i first learned about these during my mph training uh back at northwestern and so just to go through a few of the the biases and heuristics that help that hurt us and making our decisions and maybe testing excessively so there is the availability bias where a recent memory or recent event can make us think about something more prevalent than it actually is for example i recently had a i went to a case conference about a rare uh acquired factory deficiency a few days ago and i had to admit a patient a few days after that who had bleeding uh trouble and this patient had a wide variety of reasons for bleeding but because that fact rate deficiency acquired factor eight deficiency was on my mind i started to think do i need to get a mixing study on this guy do i have to look into the acquired factory deficiency when it was something that is far more rare than it actually is we have things like representative bias where um due to how a patient looks we might raise up the probability of uh of a rare diagnosis because a tall man with a heart murmur walk through i start thinking about marfan syndrome and mitral valve prolapse before thinking of more common diseases that could be going on with this patient there's anchoring so in hostile medicine i often get sign out from the emergency room from some of my colleagues and oftentimes when i get that sign out you know they give me a diagnosis to move forward with and sometimes that anchors my thinking for example i just had a patient with um wilson's disease causing hepatic failure and the patient was encephalopathic wasn't thinking clearly but two weeks prior to my meeting she was interacting with her family was um eating on her own but when i met her she was not saying any words coherently and wasn't feeding herself and it took me a couple days of kind of getting more of this history and more of the story and talking with family to realize that maybe we were missing something going on um we ended up testing for a catheter associated uti uh urinary tract infection and saw that that was there so started antibiotics removed the catheter and she got back to the point where she was interacting with her family and feeding herself again but because of that anchoring component did delay me for a couple days to finding that infection there's value induced by us or something i like to call the webmd it's cancer effect so every time that you google your headache uh and one of the things that comes up is that you know brain tumors on that differential and so for the person that's seeing this could be cancer this could be something that helps raise this rarer possibility um to the forefront when it really doesn't need to be as high but it does influence us in medicine you know oftentimes we'll be getting unnecessary stress tests or um uh pe protocol pulmonary embolism protocol ct scans uh to rule out uh these high risk elements and sometimes those are appropriate but other times they they're tests that can be avoided um but because of that value-induced bias sometimes with limit cells and so this is a accurate uh sketching i believe of myself on the wards a couple weeks ago where i had my iv espresso bag going in as i was taking care of patients but when we're in these kind of high stress environments with lots of patients admitting multiple patients from the er it really does affect you we have our limited short-term memory we're often late and hurried admitting one patient but having three more we have to admit end up being high stressed and high fatigued we have this limited ability to multitask some of us are better than others but uh we definitely have a limited ability and then as we walk around we carry around one of these pagers that interrupts us every time we uh start trying to think or talk and uh get interrupted for another thought and so having those interruptions definitely affects our train of thought and so maybe as we're thinking through our differential or thinking through tests that we want to get those distractions then uh limit us from caring for what our plan of care is going to be we have incomplete data for the patients that are in front of us an old study by uh the marco foundation found that um in a clinic 81 of cases were missing data uh whether it was a medication that was a patient was on a diagnosis of past surgery that a patient had they were missing up to an average of four items of critical data uh about a patient and 18 of medical errors are due to inadequate availability of patient information you know we're lucky in chicago a bit um in that a lot of the major institutions here at university of chicago at loyola northwestern we all use similar health records and so they can communicate and we can see what's happening at the other hospitals we have all these little community hospitals near us where we often get patients from these community hospitals and we don't know uh exactly everything that's happened there um and sometimes reliance on patients who are on whatever that facility has sent us about the the clinical care that was done so now going back to mystery uh i think think about that first diagnosis that popped into your head about what you thought was going on were you being influenced by bias were you distracted by something maybe you didn't hear me say the whole first sentence and so you missed kind of uh uh something that was going on and then that might have uh influenced your decision-making and things and so even though this is a case that's relatively straightforward that we see many times a day the uh all these biases and influences can affect how we're processing this clinical care and so now that we're already kind of having some struggles to come up with a diagnosis now we have to figure out how to treat that diagnosis and so we have this you know medical knowledge resource hierarchy that's there we have our original studies that our basic scientists and our clinical researchers produce the kind of guide a lot of our clinical care those are then combined into larger studies meta-analyses and systemic reviews that are available in pubmed and then on cochrane those are then synthesized down to um i think probably what's used most commonly on the words up to date or dynamed or the guidelines from the uspstf task force to guide our clinical care but that time that it takes to go from those original studies all the way to clinical practice going from bench to bedside on average can take up to 17 years to truly become routine clinical practice some of that is the kind of the research and development component of things some of that though is on clinical practitioners like myself who get locked into things that we're doing and maybe we don't adjust here despite guidelines and despite the most recent evidence until it becomes more routine practice and so you know guidelines are kind of what uh drive a lot of us and these you know great public health and population based resources that help provide kind of best levels of care for common clinical scenarios but some of the limitations that come through so they're not always able to account for the individual that's in front of me so while i have these population level data um based off of um you know well-controlled studies oftentimes older and more complex patients aren't always included in that underlying data that's there we have these when we're following just a guideline directly we're not always able to pursue new ideologies diagnoses and treatments aren't always integrated into a guideline of something strained from the path that they're on these guidelines can be incomplete and not frequently updated difficult to integrate into the electronic health record and there's not always clear evidence for strength they're actually very good about this but um when it gets in front of the provider we don't always know exactly what the strength level is of what's in front of us and whether we need to aggressively follow it and then we can be in a time like today uh where we have this active pandemic and new information is coming out at a rapid pace where we're getting information from social media from uh the news from uh relatives we're getting uh press releases about papers before we actually get to see the paper themselves the papers get put up online before peer review uh and so we can see the paper but they haven't gone through the uh rigorous processes and then at a time like this when paper after paper after paper just is coming out and bombarding us uh it can be overwhelming for the clinician that's on the ground uh to know what is the latest uh element to attack covet and and treat it uh with all this new information that's out and that happens not just in an intense environment like pandemic it happens kind of with every routine patient care so again going back to mr e um you know we've already thought about you know some challenges and coming up with our diagnosis but you know how do i know that i'm giving the latest and greatest care for this patient is this acs do i still need to load them uh with uh clevita coral or to assure that they're getting good care for their acs um are the guidelines in europe and the guidelines in the united states differing on these things and uh how do i know which one to follow and so oftentimes the uh guidelines and the uh events of the day can limit us as to how we're approaching giving that best care so the hope is that clinical decision support systems um so integrating things like up to date and this latest evidence into patient care every day can can help us navigate some of those barriers that i mentioned towards coming up with diagnosis and treatment i think a lot of times when people think about clinical decision support they're thinking about watson and dr watson coming up with diagnoses and coming up with uh finding that rare form of leukemia when doctors misdiagnosed it but we're not talking about physician replacement systems we're not talking about computers taking over we're talking about um support systems so there is always a user on the other end of these things and and we want to make sure that these systems are foolproof they're able to take any use or any position on those ends and be able to guide them towards care so they don't necessarily end up like uh andy dwyer here typing in and thinking that there's just network code activity problems and that's the true diagnosis so just a few definitions about you know what clinical support systems are so one definition is an active knowledge system that takes two or more items of patient data to give case specific advice clinical decision support systems or systems that provide information to the clinician in ways that help them make better clinical decisions computer systems decides to impact clinician decision making about an individual patient at the point of time that these decisions are made so why do we think these are important so helping us bridge that gap between translating uh that science that's done from that discovery point and helping disseminate that information to hopefully influence patient care and change behavior of the physicians on the ground i want to keep the initial promise of the electronic health record where you know everything that was touted patient safety and healthcare quality played a role they become a key requirement for meaningful use defined by the centers for medicare and medicaid services and then ways to help us absorb all that information both on the sheer number of variables that go into a clinical decision as well as all that overload of medical knowledge that's out there so a few different types of clinical decision support systems so one of the first branching points we think about are diagnostic versus interventional so diagnostic you know what is true that's where you're trying to figure out what that diagnosis is in front of you versus interventional about what to do component of which tests to order and what interventions to take tdss symptoms can be passive versus active so passive being the information's there it's in front of you but sometimes you have to go out of your way to click on it to read through it or you can take or leave what's there to see it active is probably what most people are familiar with you know those alerts that kind of come up saying this these two drugs interact or um the uh element in an order that kind of guides you along the way to um help you come up with a decision there's this element of consulting versus critiquing um so if i have a patient in front of me that has a new blood clot and i want to start them on a pixaband for instance you can think about when i'm ordering that apixaban uh it kind of gives me some questions as to what am i treating my treating atrial fibrillation or am i treating a new blood clot um does the patient have any factors that might influence what dose i want to use is are they elderly are they do they have renal insufficiency uh or do they have a low body weight should i use a lower dose of apixaban when i'm giving it and so it can be integrated into that order so it asks those questions as you're going through and then guides you on what dose to use for it versus critiquing so i order my five milligrams twice a day of apixaban and uh after i click my order i get an alert back at me saying uh no uh this patient is 80 years old and has a poor creatinine clearance please consider using uh 2.5 milligrams do you want to change your order um or this is a new clot you should use your higher dosing and so that's that consulting versus critiquing element so knowledge based clinical decision support systems are the majority of clinical decision support so they take this underlying medical knowledge that's programmed into the system so thing if then rules if my patient is greater than 80 and has renal insufficiency a lower dose should be used it takes in probabilistic associations and signs and symptoms so um while i uh did work i still remember my days uh during mphl working with bing chang in um my small group uh trying to determine whether a d dimer uh influenced my pre-test probability and then whether i needed to get a ct scan but all that uh at that time i didn't say let's all program that into a computer as opposed to doing it ourselves but now that can be programmed into the computer so that you can help limit any error that a physician can make you can enter things like the known drug drug and drug allergy or drug food interactions that are there all this medical knowledge base gets programmed in and then it gets combined with patient data and goes through an inference engine it uses formulas that combine these rules and associations in the knowledge base with the patient that's in front of you um and then it can give you the case specific advice uh right uh as you're caring for that patient uh giving an alert at the time of uh order entry so we're hoping that you know the cdss can uh be it'll help us improve clinical care so a person with uh technology is better than a person without support alone so these external aids link medical knowledge and patient specific data bridge the gap between the knowledge and practice and identify different options for the user without these tools providers may make errors overlook available knowledge and don't always account for the uniqueness of patients and we still even with these supports can still make these errors but we're hoping that the the role of cds can can improve that so just thinking of some of the early systems that were developed for computer assisted diagnosis one of the first ones was called internist eye or now called qmr it's an expert system to help with diagnosis of patients and narrowing things down to based off of symptoms that are entered and data that's entered it was actually designed to capture the expertise of just one man no no it wasn't dr house it was uh jack myers the the chairman of internal medicine at the university of pittsburgh who uh his knowledge base was uh attempted to be captured into a computer program as part of a educational experiments uh that was going on these actually were studied some of the diagnoses uh related um devices and uh the performance wasn't great back in 1994 you know our average uh success with these um about 50 to 70 of the time we actually got the diagnosis using these things so one of the early clinical decision support systems at vanderbilt so i apologize i know that the clinicians in the audience probably are already trying to click on the screen here to close this pop-up window but this is one of the first pop-up windows that existed so for icu level care in uh at vanderbilt um they wanted to assure that they had a target for uh the rast scores to help monitor sedation of their icu patients and making sure that they weren't getting too much or too little um sedation to go along with them so this is one way that the vanderbilt system uh helped trigger you to order um a target for their ask score so if you clicked on this you then got an order that you could enter to enter your level of grass that you wanted the patient to get this is one of the early systems at brigham women's and partners in health was a longitudinal medical record that took into all of the ambulatory accounts they really did focus on medications in helping give clinical decision support for them and helping you reduce adverse drug events by assessing dosing and assessing um drug drug interactions as i had previously mentioned they also did give some anticipatory medication to decision support so when we're thinking about our ideal clinical decision support our goal is to increase adherence to guideline-based care decrease medication error provide cost effective recommendations improve the ability to process information and reduce redundant testing um and be population based then narrow down to patient specifics for the patients that were in front of us but as we all know computers are far from perfect and those of those that have worked uh clinically know that you know the what can be produced is only as good as what goes into it and so some limitations so think about the definitions that get brought into um uh that are entered into the electronic health record so how do we define weight in our system on which data point are you being pulled are you pulling the patient weight the ideal body weight the weight that was measured at home in the hospital before hydration after hydration which weight should the computer use to help you make decisions i'm sure somebody diagnosed someone with hypertension today and documented it and how was it documented than the electronic health record did you write hypertension htn high blood pressure high bp did you let the blood pressure speak for itself i'm sure you build so you had to write an icd-10 code that tried to capture the information uh to capture all of that maybe you misspelled hypertension so maybe it wasn't encountered as much so how well these systems work depends on how they're structured and whether the vocabulary can match the terms that clinicians use there are temporal considerations that haven't come to place you know time related to symptoms are difficult to express in a controlled vocabulary so physicians use a lot of different elements to talk about time you know days prior to admission or has been going on for some time can be very non-specific and but these temporal modifiers can be pretty difficult to build into the system and even small changes um as you think about you know kind of goals of when to get someone into the cath lab or to get tpa initiation these small changes in timing can make a big difference into the conclusion that a system ends up reaching so i just want to talk about some ways you can see clinical decision support in the electronic health record there are things like info buttons that you can click on uh to give you reference material um about either a device a medication a lab value think about calculators and calculations that are done like your um fractional expression of sodium or urea or other calculators that can be used order sets are kind of the lifeblood of the hospitalist and so it's how we admit our patients and enter orders uh to care for our patients and alerts i think these are the ones that everyone's aware of but medication alerts that kind of give the drug drug interactions uh drug diagnoses and drug allergy alerts you can have systemic alerts that can tell you hey you've already ordered an echo a week ago are you sure you want to get a second one um or there's a better option you can just look at the ejection fraction so these systemic alerts can help limit duplicate procedures and testing and then the more ideal ones are these conditional alerts so um they can give you best practice advisories so you have a patient that um is getting prescribed and opiate undischarged and um clinical clinical best practice advisor um launch to tell you you should give a narcan prescription as well so the conditional alerts are all from there and then there are suggestions for um possible diagnoses that can come through so just to show you a few of these this is a info button that you can click on to get information about hematocrit and what the testing is and what the clinical significance might be this is an example of a calculator looking at the framingham risk score that you can enter in appropriate information and get their 10-year cardiovascular risk this is an order set and these are getting smarter and smarter as we uh as the systems advance and so now we can use these order sets to uh if somebody has a penicillin allergy we can then hide things that might cause an anaphylactic reaction if that's the reaction the patient has to help prevent use of things that might not be accurate or we can if we have a diagnosis of community acquired pneumonia we can then um show those antibiotics and the preferred antibiotics based off of the antibiogram in our hospital and so these are becoming smarter and smarter for use and then there's our alerts that again i'm sorry for the clinicians in the audience that probably are already trying to close this alert but this is um one that popped up as uh to notify somebody that there's an acute kidney injury so based off of a uh crap mean a few days ago of two and then now being uh four do you want to order an order set to look into this acute kidney injury or do you want to add it to a proper problem list so when you're making these changes and providing clinical decision support it's good to know kind of what works best what makes physicians uh use things as opposed to just become you know annoyed by things and so a few system reviews and that analyses found some key factors so it's best to make it part of the workflow so i don't have to click elsewhere into a different portion of the ehr or somewhere else to get that clinical decision support it's better when it's automated rather than making a user activated and occurs when the decision is being made so it doesn't launch right when you open the chart but maybe it launches when you're at the order entry site so that when you're truly making those decisions then it can be activated it's usually better when it's a recommendation um and not an assessment it's better when it promotes action over inaction although at times in action but promoting actions more likely to be taken on by positions um did not require any additional data entry so um as a user i would not want to ever add any additional information when i am entering orders and it's key to have uh local people on the ground that are actually doing this day to day be a part of any intervention that's uh created um just because uh it'll be more likely to be accepted and you'll probably get a better intervention there so i'll talk about a little bit of ways that we put this into practice and so um we had the issues with uh insulin pens and pen needles at emory we were giving uh insulin pen prescriptions that weren't covered by their insurance and then when we were giving the uh insulin pens um people were forgetting to give a corresponding pen needle order so we were getting extra calls from the pharmacist when it was our pharmacy uh they would call us and ask for the additional pending order or to change the prescription but when they when it was other pharmacies then oftentimes patients would get delayed prescriptions and maybe we don't hear about the issue with the insurance for a day or two or we uh forget a pen needle and we actually had a patient who tried to inject themselves with their insulin pen but didn't have the needles and so i ended up going into dka and so we thought this was a real problem you know part of that was the education about um the pen needles but this was uh something we thought we could fix pharmacy wasn't you know it's not allowed to automatically substitute your ad without a phone call and so initially a discharge order set was created to include pen needles and other supplies the order said instead of having atlantis solostar so brand name insulin pen in the discharge order we made them generic uh and substitutable so even if i sent a prescription for um the sole start pen it could then be swapped out for a different pen based off of the insurance coverage that was there at the time a lot of the insurances were switching over to a different type of insulin and we were getting a lot of calls and a lot of issues and this allowed it to um have that substitution without issue this still didn't totally solve our pen needle problem we're still getting folks sent out without it um we tried to have it just automatically had uh so anytime a prescription pen was added it would just automatically put it in but unfortunately at that time um the cpoe didn't account for wasn't able to account for that so we ended up doing a best practice alert that was not terribly invasive but we did end up doing that anytime a pen was prescribed without a pen needle so even for this simple intervention we had to have a lot of people involved swede hospitalists we had our outpatient retail pharmacists our inpatient clinical pharmacists endocrinologists diabetes educators we had patients that serve patient advocates um i.t specialists and quality data analysts and so we couldn't capture how the number of phone calls decreasing or the number of delays for uh insurance issues we were able to capture the pen needle orders um and so these were the four main folks prescribing pens and pen needles and uh we were able to push folks to prescribe that eventually you know there's still the hope that it can just be automated so we can get to 100 without any uh brain power or clinical thinking uh going into it for something that's that simple but this is one way that we help trying to improve uh our outcomes for patient care we have an issue with our correctional insulin dosing there as well and so this was the order that was used for corrective insulin precision scale insulin at the hospital so it took um the blood glucose that was drawn on the patient then subtracting 100 and divide by 40 to get the number of units that had to be given for blood glucose greater than 140. so already we're looking and you can see there's a lot of math involved for the nurses to to carry out it doesn't take into account you know how sensitive a patient is uh to insulin and so overall this was a poorly uh individual poorly written individual order and people don't like doing math especially when they're married and have to run around so our nurses would have to carry around a calculator with them um or use the calculator that was built into the ehr to get the appropriate dosing and they didn't know which way to round uh whether to round out round down and so there was one hospital in the emory system when i was working here that used uh table-based orders so they made all these calculations and likely that's what it's used in northwestern um but they compared table-based orders to this formula based order and so we found that our formula based order was doing pretty poorly the nurses were not administering what was truly ordered luckily they were under dosing more often than they were overdosing but only 68 of the time were they getting what was truly ordered by the system and comparing that to the table based order 92 of the time they were getting what was supposed to be ordered so we went through and changed the pre-written table rip changed them to pre-written tables and scheduled them to be standard sensitive or resistant so that people had some guidance as to how to choose and then we mentioned kidney function and patient weight and the total daily insulin needs to help people pick should i choose standard sensitive or resistant and we also gave guidance as to when to move beyond sliding scale alone given the guidelines of what is recommended by the american diabetes association and elsewhere and so we put that guidance into the order and included basil and bolus insulin um in there as well and so we were relatively successful in terms of improving the correctional dosing so the two hospitals that were performing kind of poorly not having having the nurses often using their calculators and delaying getting insulin folks uh ended up improving their um accuracy here again a lot of people had to be involved just because even for this small change because we were changing a high risk medication like insulin at the hospital so we had to go through our pnt committees our nutrition committees and our quality boards to get this change done in the it system there were some now here with the university of chicago where there's a slightly smaller intervention that we're working on that uh dr martinez uh here is leading and so she investigated the appropriate use of physical therapy consultation so she using the activity measure post-acute care in patient mobility short form this is a form of six questions that asks um can you um put on and take off your clothes on your upper body your lower body do you help with bathing do you help with toileting can you brush your teeth and eating meals so it takes those six questions and you get a score and this score can help influence whether physical therapy is beneficial and whether rehab might be needed this is actually being done already every shift by our nursing staff and so they looked at it and they found that 38 of physical therapy consults were over utilized so uh after just our initials of the patient when we first met them we might have been ordering unnecessarily but there were also significant delays for folks that had poor mobility or when decreases in mobility happen later on in the hospitalization for somebody that was in the hospital for some time maybe it wasn't captured uh because of that that clinical change and so those low mobility delayed placement uh because we didn't know right away that uh patients needed that and our physicians were totally unaware that this assessment was being done i personally was unaware that this assessment was being done uh every day three global three times a day by our nursing staff and so dr martinez and dr sarahzil um ended up integrating it into the notes so we have templated notes in hostile medicine to kind of guide us on some of the our key elements and so they put the ampex score right in front of us and then provides us with that clinical guidance right next to it as to whether we should order ptot and whether it's indicated and then you know we could always move beyond it this updates with every progress note right now it's only documentation but the hope is that it then gets directly connected directly to order entry to help lessen the extra step that's viewed there luckily this was an internal one so it didn't have to involve entire hospitals to change so something that was just done internally to um help us with our physical therapy utilization and now broadening out to our residents and our general medicine teams so one of the more complicated interventions that's been done so again some colleagues of ours looked at a calculation to help us predict cardiac arrest in the hospital they took a innumerable number of variables here uh to see do they influence whether a patient is likely to have a cardiac arrest or not and so looking at all these variables and how they're weighted you know that almost impossible for a unique clinician to look at this and look at each one of these individual values and say yeah i think my patient might have a cardiac arrest in the next eight hours they were able to bear you know and validate this uh this data um and compared it to the modified early warning system but given these this complicated uh number of variables it was hard to integrate into clinical practice luckily we were able to use a program called agilemd to take all of these variables and time to give us our is a patient likely to decompensate in the next eight hours so it takes in all these variables and if you look at my patient here on the 21st of september at 8 34 pm he had a 1 in 74 chance of deterioration in the next eight hours which put him in this red category and someone to pay higher uh higher higher amounts of attention to it then highlights you know which of the variables uh are influencing uh the patient the most and for this patient it was his uh oxygen saturation and so you know this actually did uh pushed me a little bit to um i think how i could help him in this situation and so i ended up giving him extra dose of diuresis that day and it gets integrated right into our patient list so uh this is a list of patients on not necessarily my patients distorted by on that ecart risk and i can see who's most sick and who's not um when i'm looking at this it's not perfect you know sometimes we have um some patients that we know are more stable than others and have a higher risk of deterioration that this isn't perfect for but it does give us some guidance and some objective objective data towards pushing us to think the other thing it does is it helps trigger our rapid response team uh to get involved and so there's another set of eyes that gets seen by the any patient that goes into the red a group of nurses and the rapid response team that looks through and um they can also see as where those hairy physicians getting our espresso ivs on the floor if we miss something and can help guide us one thing we're hoping to do as you can see that next column over we're not great about having our code status order entering uh there which is um is something that you know we're pretty passionate about improving um so we're hoping we can use these ecard scores to help influence that order so last i wanted to talk about um that we're using pretty extensively is something called clinical pathways um so these are structured multidisciplinary plans of care that translate um our evidence space into the structures that are there they provide details about that algorithm that goes in and criteria that you can go through so every step um that can go into a clinical decision uh and what what to do about it we're hoping to use this to aim uh to standardize care for a specific clinical problem in a specific population um but that's a lot of words so a diagram might be more helpful so this is one of our pathways for um inpatient febrile neutropenia i've scrolled down past kind of the initial workout portion of it on to some of the branching logic for choosing antibiotics and so it has this uh info button there to say you know when you can click on that to get a lot more details as to when you should initiate therapy and then i can order directly from it um i can see whether there is a severe beta-lactam allergy in this patient just by hovering over that portion i can then uh order cephepem directly from this um where i can order my estran and vancomycin orders if somebody has a betalactin allergy and then i can go to the area where should i order vancomycin in addition and it gives me some criteria for that so these pathways you know help us ensure that the latest clinical research and guidelines um uh inform our care teams we use these were how we uh kept a source of truth for uh kovit during all the clinical changes that were happening and so we uh have these road maps that um patients could follow as to when to give certain interventions and additional interventions that people could try especially as things were changing so quickly and people could stay on top of um this one source of truth we had our coveted leadership uh involved conditions quality improvement performance clinical informatics and so this was our just another example of one of these pathways that has the disclaimers on the side it has when things were most recently updated you can place orders and you can page directly from it as well as getting the additional information for about them so this is how we combated this this this information overload that was being occurring and help keep people updated as to everything that was happening so that our clinicians could focus on their safety on the patient safety and providing the best care and human care that we possibly could thank you all so much i'm happy to take questions about anything i presented here thank you ethan i think i can speak for the audio exciting presentation and lots of really good information that has led to a few questions already in the q a section which i'll help facilitate um if others have questions please uh feel free to put them in the q a box there at the bottom of the of the screen um i i have i want to start with one question that just uh you you talked a bit about uh the types of strategies that are helpful um for um successfully getting physicians and teams to use uh clinical decision support tools like putting it in the workflow and involving stakeholders in the design and and uh you know put positioning it at the time decisions are made um there's a lot of strategies that administration or improvement teams might use in order to accomplish those goals um a lot of different you know approaches like uh you know standing meetings or certain numbers of trainings or ways to provide um you know to sort of audit uh implementation and provide feedback to providers and reinforce uh their behaviors things like that i'm wondering what sort of um strategies your team is using and um you know broadly we've gotten increasingly interested in implementation science or the science of sort of understanding these strategies so that we can make these um uh these types of attempts to implement uh things like decision support more sustainably and more routinely and others can replicate it so i'm just curious if you're are you using any specific approaches that you found particularly helpful for those goals yeah and i think some of the additional you know support that we end up using every time we roll out one of these interventions does kind of depend on the complexity of the intervention themselves um and so for um big changes um to the system you know oftentimes we do have an education plan that we do plan out before um unveiling any big rollouts and so um having clinical teams uh and um everything that's there making sure that we have the plan to um go to the users themselves whether it's myself talking to the residents and to the hospitalists um or some and if there's a surgical team that's involved making sure that um we get in front of the teams that are um most used but we do use those in-person uh education uh elements um to uh uh ensure that we're getting in front of folks for major changes to things um we do sometimes utilize the continuing education component where a module is made those are not quite as successful just because people end up trying to click through and can't always ask the questions that are needed for bigger changes but for smaller changes oftentimes um using one of those modules or even just a standard email a system-wide email um going out for what the um uh what the change ends up being you know when with a lot of the changes um when they're smaller we're trying to make them so that they're so intuitive that you don't necessarily have to do a big portion of extra training um so for uh specifically for like rendezvous use in in the hospital um we have tried to build in uh the indications into the order itself um so that you know uh as i'm ordering um is my patient on oxygen now or not um is there symptomatology within you know seven days of um of onset and when the best uses and so trying to make it as simple as possible where somebody might if they missed the education event if they missed weren't able to do the training and still can be absorbed that's been working the best out of all of that is having it kind of right there when the order's being placed there there are two questions that i think relate in some level uh it has to do with i think the broader theme is that as you pointed out earlier there's such an explosion of information and such an uh evolution of of uh knowledge to on uh diagnostic and therapeutic um you know sort of options for the for the physician and you know as as these areas are being developed um at the physician level uh you know has the potential that if we're creating decision support tools for all of those things it's just a fireworks display of different pop-ups and how do you manage this and orchestrate that so that it doesn't become overly burdensome and then the second is just from a system perspective and building those tools it also creates a cue for the technology people that program all of that and how does that get prioritized and how do you sort of not um you know get such a backlog that you can't even keep up with the need for the development of these tools uh you know from a technological perspective yeah uh so you know obviously alert fatigue is something that uh everyone's pretty concerned about um in every intervention that we make we're trying to say uh are we helping the provider or annoying the provider as opposed to truly having somebody absorb what's there so there are the information technology teams that help kind of guide this for any time a an alert is kind of recommended it is uh saying well we only have we're only allowed a certain number of alerts uh for this patient so for this patient that's discharging um we have the um best practice alert for narcan prescribing we have this um best practice alert for can should we preside i use this uh insulin order set uh at discharge and so um anytime a new alert is uh presented the question is always asked should we remove something else and so to try and limit and limit that overwhelming uh uh component of uh alerts and and trying to prevent alert fatigue so that's something that's high on the minds of um uh the information technologists and absolutely right that uh things have to get prioritized uh based off of thing and obviously when coveted hit everything code related got prioritized but when there are smaller interventions like this sometimes you know you can do try and think of something simple that uh doesn't get prioritized right away because you do have that limited supply one way that we've found to work around that is we use epic at our hospital and epic has a physician builder program and so a lot of the physicians in the in both our department as well as um other departments have gone to these three-day trainings now they're virtual otherwise you get to make a nice trip up to verona uh wisconsin uh which is uh not doesn't have a ton going on there but it's pretty um but you can get this extra training that's there and then um as a physician coming back to the wards and saying i am constantly annoyed by this factor in the ehr i wish there was an order set for this um you can then propose it and they'll okay it for the uh you as a physician to then uh build it out so um while uh they're kind of implementing and putting power into the hands of the providers on the on the ground uh to help um prioritize things and things like alerts and best practice uh alerts that that can um really uh affect um how patients how providers provide that care those go through a much more rigorous process as to whether to implement or not but for more simple things like our templated notes or our uh order sets for certain things those are much simpler process than them can be put into the hands of providers so that they don't have to be they can be implemented more quickly sorry another rookie mistake on mute again um ethan we have one more question um and maybe i'll let you just uh with the two minutes left uh answer it quickly while we sort of close you can add any closing remarks um before i ask the question i want to thank you again for uh presenting to our ifam webinar today we've really enjoyed having you and and thank you for uh giving us this time uh which we know is precious and um you know our final question as you close out just relates to the challenges of fundamentally teaching clinicians to do to behave differently and and to practice really uh precision medicine or or a very specific patient-centered care that's you know the best evidence so that the training aspect and then and what are the biggest challenges of of actually getting physicians to use this and and to behave differently in the practice yeah so it's it's interesting so with with copic um those agile md athletes that i mentioned uh where we were um trying to show uh we were kind of laying out all of the ways to um apply care they were utilized um aggressively during the first few months of the pandemic and then every time there's been a wave they pick up um in terms of their use for best patient as people that maybe haven't seen it for a while are reminding themselves um but when people are seeing it every day um the uh updates um when we are updating something based off of new evidence um when people get kind of in their routine and kind of know um this is what i do for every patient that walks in with this they don't um jump in and kind of look at the new evidence or any updates that are there and so anytime there is a reminder there have there does often have to be that other intervention like a email or education session uh to kind of say hey there are some big things that have changed uh and these are the updates and um that's probably one of the biggest challenges uh that we still have is making sure the updates are getting in front of folks at the right time well thank you again ethan we really appreciate you joining us today and on behalf of the membership of ifam um i have a happy thursday and uh and thank you for giving us your time thank you so much everybody