Enhanced Vancomycin Hemodialysis model now available in DoseMeRx!Learn more →

Vancomycin Hub

Do’s and Dont’s of Bayesian Dosing: Webinar Recording

In this article:

Curious what it’s like to successfully implement and optimize a Bayesian dosing program at a leading hospital system?

In this four part webinar recording, hear the firsthand experiences of Dr. Shivani Patel from Memorial Hermann Health System and Dr. Dustin Orvin of St. Joseph’s/Candler as they reveal their top tips and lessons learned.

What You’ll Gain

  • Key strategies and tools used by Memorial Hermann Health System while implementing Bayesian AUC-guided vancomycin monitoring within their electronic health record (EHR).
  • Insider knowledge of how the impact of vancomycin AUC-based Bayesian dosing was measured at St. Joseph’s/Candler, which achieved an overall reduction in AKI and improved therapeutic target attainment.

Watch the Recording


Dr. Kristi Kuper, PharmD, BCPS, FIDSA, CMWA, Director of Clinical Pharmacy at DoseMeRx, introduces you to our featured panelists and provides an overview of DoseMeRx.

Part Two

Implementing DoseMeRx Across a 13-Hospital System

Dr. Shivani Patel, PharmD, BCPS, Clinical Pharmacy Specialist, Infectious Diseases at Memorial Hermann Southwest Hospital reveals why her system opted for EHR integrated Bayesian dosing, how they amended their protocol and engaged key stakeholders throughout their implementation journey.

Part Three

The Clinical, Operational, and Financial Benefits of Bayesian Dosing

Dr. Dustin Orvin, PharmD, BCPS, Clinical Pharmacy Specialist at St. Joseph’s/Candler shares how they achieved an overall reduction in AKI and measured the impact of implementing Bayesian dosing software (DoseMeRx).

Part Four

Question & Answer

In the final part of this webinar our panelists respond to questions submitted by the audience.

Review Further Questions & Answers

Dr. Shivani Patel, Dr. Dustin Orvin, and Dr. Kristi Kuper answer more of our audience’s questions below.

Have other questions? Get in touch with us today.  

Protocol development

Do you have any recommendations for how to coordinate this type of project across a large group of hospitals and clinicians?

Shivani Patel, PharmD, BCPS
Clinical Pharmacy Specialist, Infectious Diseases
Memorial Hermann Southwest Hospital

We developed our project plan based on the following categories.  Within each category, we planned for the following items to work towards a smooth implementation:

Protocol development
– Review the clinical data
– Develop decision pathway for evaluating vancomycin orders
– Determine loading dose and maintenance dose recommendations
– Define how levels will be drawn
– Downtime procedures


– AUC background information to support rationale for conversion to Bayesian dosing
– DoseMeRx software education with screen shots
– Create staff education for any workflow changes associated with implementation
– Guidance on how pharmacists will handoff between shifts and days
– Create pharmacy competency and patient cases for assess understanding of educational sessions

– AUC background information to support rationale for conversion to Bayesian dosing
– Communicate changes in workflow (i.e., pharmacy automatically follows and adjusts all vancomycin patients)
– Create one page reference guide

– AUC background information to support rationale for conversion to Bayesian dosing
– Communicate changes in workflow driven by changes in lab draws
– Create one page reference guide

IT Work Efforts
– Revise ordering options for vancomycin
– Create vancomycin AUC level
– Update all physician order sets with new vancomycin orders and workflow changes

Pharmacy Workflow
– How do you keep track of vancomycin patients?
– How do you communicate day to day changes?

Pharmacy and Physician Go-Live
– Set a date/time period to launch program

Some of the keys to our success were to get a representative from each campus where DoseMeRx was being implemented and to get a service line representative across multiple hospitals to assure that a non-ID point of view was represented.  For each decision point outlined above, we had consensus from the core ID pharmacist work group, the campus representative, and the service line representative.  Additionally, we provided monthly updates to the pharmacy directors and system ID physician leadership.


How do you determine your initial doses? Do you have a hospital specific protocol that you use to guide dosing, or do you use DoseMeRx to calculate initial doses? Also, when do you use loading doses?

Shivani Patel, PharmD, BCPS
Clinical Pharmacy Specialist, Infectious Diseases
Memorial Hermann Southwest Hospital

We use a weight-based nomogram to determine loading doses.  The nomogram is based on targeting a dose of 25 mg/kg.  We then utilize DoseMeRx for all maintenance doses for population dosing and then individual (Bayesian fitted) dosing once levels are obtained.  We use loading doses for all our patients.  If the physician orders a one-time dose prior to the loading dose being given, a supplemental dose is not given, and we proceed with DoseMeRx recommendations from the population model.

Dustin Orvin, PharmD, BCPS
Clinical Pharmacy Specialist
St. Joseph’s/Candler

Our dosing protocol utilizes 25 mg/kg loading doses for all our patients. To determine maintenance doses, we use a combined approach of Bayesian and first order kinetics. Our floor-based clinical pharmacists utilize DoseMeRx to determine initial maintenance doses during day shift hours. After hours, our team uses an institution-developed calculator to determine initial doses, which are then reviewed and entered into DoseMeRx the next day.

Vancomycin level monitoring

What are your general practices around vancomycin level monitoring? When do you obtain 1 vs 2 levels? Do you also utilize random levels? How long do you wait?

Shivani Patel, PharmD, BCPS
Clinical Pharmacy Specialist, Infectious Diseases
Memorial Hermann Southwest Hospital

We determine ordering of levels based on the model selected for vancomycin dosing.  With the one-compartment model, all patients get a loading dose and appropriate maintenance dose.  We get a random level with AM (e.g., daily morning) labs following the administration at least one dose (either loading dose, maintenance dose, or both).  After the first level, we get subsequent levels if dose adjustments are required or every five days.  For the two-compartment model (patients with unstable renal function), we also give a loading dose and appropriate maintenance dose.  For this model, we get a level with AM labs following the administration of at least one dose and then a second timed level at 11AM to account for clearance between levels and improve model fit.  For these patients, we check subsequent levels every three days until renal function stabilizes or normalizes.  We only utilized timed levels (i.e., troughs) in patients using traditional dosing.

Dustin Orvin, PharmD, BCPS
Clinical Pharmacy Specialist
St. Joseph’s/Candler

We utilize troughs to monitor and drive AUC determinations. We do a single trough concentration for the majority of our patients. We only have the one-compartment model currently at our institution. If we see a clinical situation where volume of distribution is altered, or in severe infections, we might add a peak concentration to make sure we’re getting the most accurate AUC estimate, but it’s still trough-based monitoring. We monitor concentrations within the first 48 hours of therapy to optimize doses early then typically every three and five days thereafter. It’s up to the pharmacist’s discretion on how often that’s monitored, depending on a combination of the clinical situation, renal function trends, and dose recommendations from DoseMeRx.


Can you expand on your comment regarding differences in pharmacokinetic parameter estimates when providing education (slide 43)?

Dustin Orvin, PharmD, BCPS
Clinical Pharmacy Specialist
St. Joseph’s/Candler

Prior to implementing AUC our staff relied heavily on population-based first order kinetics estimates to dose vancomycin. Naturally our staff was very comfortable with those estimates and the various pitfalls of using first order equations. Many had developed practices to address certain pitfalls (i.e., capping creatinine clearance [CrCl] estimates initially or utilizing adjusted body weight) when performing these calculations. Upon implementing Bayesian dosing there were tendencies to compare first order PK estimates to estimates derived from the model in DoseMeRx before utilizing a model derived dose. For example, if a pharmacist estimated a lower CrCl than DoseMeRx, they may have given a lower dose than what the model recommended and thus underdosed the patient. We noticed this quickly in the post assessment and were able to give additional education to our staff regarding how model estimates were derived and differences from first order PK estimates. This situation highlighted the importance of ongoing monitoring of our dosing program after implementation.

Dosing special populations

How did you separate out patients on vasopressors from those with renal failure in your protocol?

Shivani Patel, PharmD, BCPS
Clinical Pharmacy Specialist, Infectious Diseases
Memorial Hermann Southwest Hospital

When we evaluate a new patient with a vancomycin order, the first thing we evaluate is the presence of vasopressor therapy.  If present, the patient is evaluated with a two-compartment model.  If no vasopressor therapy in present, we then evaluate for stable renal function.  If the serum creatinine is stable at <3 mg/dL, we utilize the one-compartment model.  If the answer is no, then we evaluate for dialysis using the HD model and then for acute kidney injury (AKI) or unstable renal function using the two-compartment model.


How are you dosing vancomycin for non-Staphylococcus aureus infections?  

Shivani Patel, PharmD, BCPS
Clinical Pharmacy Specialist, Infectious Diseases
Memorial Hermann Southwest Hospital

The rationale for using Bayesian dosing for non-Staphylococcus aureus infections are based on  trials that showed that in Enterococcal bacteremia, achieving an AUC of greater than 400mg.h/L improved 30 day all-cause mortality. This data is how we determined our AUC goal of 400 to 600mg.h/L for Enterococcus and other strep species. There is a significant lack of data for coagulase negative Staphylococcus aureus (CoNS) infections. We looked at the higher minimum inhibitory (MIC) breakpoints established by the Clinical Laboratory and Standards Institute (CLSI) and felt comfortable based on PK/PD data that an AUC of 400-600mg.h/L would achieve adequate levels in this patient population also.  In addition to AUC monitoring, we ensure that the patient is clinically improving by evaluating clearance of repeat blood cultures, clinical response, and improvement in infectious markers such as fever and white blood cells.  If the patient has persistent bacteremia, we will target a higher AUC or change to alternative agent. This decision is clinically based on patient response.

Dustin Orvin, PharmD, BCPS
Clinical Pharmacy Specialist
St. Joseph’s/Candler

Most of the data currently is specific to infections caused by Staphylococcus aureus, specifically methicillin resistant S. aureus (MRSA), but we felt confident in switching to AUC targeting a range of 400-600 mg.hr/L for a couple of reasons. First, by targeting AUC we felt we would be improving the safety of vancomycin in those patients. Second, we’ve historically targeted troughs, even outside of S. aureus infections, and troughs have always been used as a surrogate for AUC.


Are there any types of infections where you would target an AUC24 below 400 or above 600 mg.h/L?  

Shivani Patel, PharmD, BCPS
Clinical Pharmacy Specialist, Infectious Diseases
Memorial Hermann Southwest Hospital

We do not target any AUC24 values under 400, but we do set our AUC24 goal at 400 for skin and soft tissue infections. This typically gives us goal AUC values in the range of 350-450.  We are comfortable with this range if it is associated with clinical improvement.  We do not target any AUC24 values above 600mg.h/L.  I am not aware of any clinical data that supports improved clinical response or outcomes with a  higher AUC target.  We actively advocate against an AUC of >600mg.h/L in all models except hemodialysis (HD) due to this information combined with the increase in nephrotoxicity seen with higher AUC values.  For our HD patients, we will allow for higher AUC value because the kidneys have already been compromised.

Dustin Orvin, PharmD, BCPS
Clinical Pharmacy Specialist
St. Joseph’s/Candler

At my institution, all our patients dosed based on AUC are targeting a range of 400 to 600 mg.hr/L. There is potentially an argument to be made for lower AUC targets in some infections, but I recommend at least an AUC of 400 mg.hr/L for all our vancomycin patients. As far as targeting AUCs above 600mg.h/L, I would strongly advise against that practice as you begin increasing the nephrotoxicity risk to your patient without evidence of improved efficacy. Remember, even the higher trough targets of 15-20 were in place to achieve an AUC of at least 400 mg.hr/L.


Does either institution utilize DoseMeRx for pediatric patients?

Shivani Patel, PharmD, BCPS
Clinical Pharmacy Specialist, Infectious Diseases
Memorial Hermann Southwest Hospital

We focused our implementation on the adult population.  Our pediatric clinical specialists are currently working on the pediatric implementation across our system and at our children’s hospital.

Dustin Orvin, PharmD, BCPS
Clinical Pharmacy Specialist
St. Joseph’s/Candler

We utilize DoseMeRx in the adult population.

Evaluating incidence of AKI

Do you have any suggestions on how an institution might structure their medication use evaluation (MUE) to help identify AKI and associated cost savings?

Dustin Orvin, PharmD, BCPS
Clinical Pharmacy Specialist
St. Joseph’s/Candler

It’s important to note that MUEs do not always provide the most conclusive data to determine cost savings however it is often the most achievable way to get information at an institution. Despite this, the AKI trends identified through an MUE can be very useful to justify implementing AUC dosing and to estimate overall cost savings potential. I suggest that you define AKI in the same way that the financial studies did that you will be using to determine cost savings potential. As a general reference, most studies utilize a serum creatinine increase of ≥ 0.5 mg/dL or ≥ 50% decline in estimated creatinine clearance to define vancomycin associated nephrotoxicity.

Your data should align with the available financial evidence so that you can assess the potential return on investment as accurately as possible once you have measured your institution’s AKI rate and the impact of  AUC dosing.


When you report AKI rates, are you reporting overall at the institution or just those on vancomycin who develop AKI?

Shivani Patel, PharmD, BCPS
Clinical Pharmacy Specialist, Infectious Diseases
Memorial Hermann Southwest Hospital

Currently we are only reporting AKI rates in patients on vancomycin therapy.  We conducted an MUE prior to implementing the Bayesian dosing software that provided a baseline rate and supported the need for Bayesian dosing.

Dustin Orvin, PharmD, BCPS
Clinical Pharmacy Specialist
St. Joseph’s/Candler

We look at both, but the data presented from my institution is specific to patients on vancomycin therapy.

View The Transcript

Do’s and Don’ts of Bayesian Dosing Webinar Transcript

Kristi Kuper: My name is Kristi Kuper and I’m Director of Clinical Pharmacy for DoseMeRx. We’re excited to have you joining us today. On behalf of our entire team we just thank all of the health care workers that are on the phone today for all you’ve been doing for patient care.

Just a couple of quick announcements. This webinar will be recorded and a link will be e-mailed afterwards. This will be sent to the e-mail that you used to register for. The webinar will have all lines kept in listen only mode. We have actually 900 individuals that are registered for today’s webinar, so we will have to keep all the lines in listen only mode. You can see a handout of the slides available by clicking the document icon located on the right of your screen. And please, use the questions feature to submit questions, and these will be answered during our Q&A.

We’re going to go ahead and move into a quick poll, because we want to get a sense of who is on the line today. And the question should be populating on your screen here in just a second. So you’ll see on your screen here we just have a question for you. We want to know how you’re currently calculating vancomycin AUC. I know we have some folks on the phone today that are existing customers, but for those of you that are not currently using Bayesian dosing software, we just like to see how you’re calculating AUC, and I think it’ll be helpful for our presenters to know, as well. We’ll give it just maybe, a couple more seconds, and then we will look at the results.

It looks like we have a large percentage of people that are currently evaluating the options, about a third of you on the phone are already using a Bayesian dosing platform, and it looks like we have a couple of small percentage of you that are still using an online two point calculator or an Excel spreadsheet. I think we’ve got a good sense of who’s on the line today, so we can go ahead and close the poll. Great. Thanks so much.

If you’re not familiar with DoseMeRx, just a little bit about our company and our software. We actually have over 7000 clinician users and they have used DoseMeRx to calculate over 1.4 million medication doses. We currently have 10 different countries utilizing our software and we have 42 drug models, in five different therapeutic areas. Of course today we’ll be talking about vancomycin but we also have and we have a number of drug models that are available for neonates, pediatrics and adult, and those are all listed on our website. We take data security very seriously. We’re HITRUST certified as well as HIPAA, ISO, and FDA compliant. One of the things that we always like to promote is the fact that we have customer experience people on call 24/7, 365 days a year. If you message through the website or through the chat, we’ll respond to you within two minutes – and that’s a real, live human being that you’ll be speaking with. There’s a number of different ways that you can access DoseMeRx. We have a web based platform, and we also have a special partnership for small hospitals. With ASHP were integrated, and Cerner, and Epic. We also are integrated with three different pharmacy surveillance systems: Sentri7, VigiLanz, and ilumInsight.

OK, so now onto the exciting part of our presentation! We’re really excited to welcome two amazing clinical pharmacists today who will be presenting. Our first presenter is Dr. Shivani Patel, Clinical Pharmacy Specialist of Infectious Disease at Memorial Hermann Southwest Hospital.Dr. Patel is responsible for leading hospital wide quality improvement and cost savings initiatives and co-leads several antimicrobial stewardship projects across the Memorial Hermann Health System, located in Houston, Texas. She received her Doctor of Pharmacy degree from the University of North Carolina at Chapel Hill, and completed a pharmacy practice residency at Vanderbilt University Medical Center. Today she will be walking us through the journey that her health system took in implementing DoseMeRx in their electronic health record (EHR).

After her presentation, we will be joined by Dr. Dustin Orvin, a Clinical Pharmacy specialist in Internal Medicine at Candler Hospital, which is located in Savannah, GA. He has a number of clinical responsibilities in his institution, including serving as the Health Systems Formulary Liaison to the Pharmacy and Therapeutics Committee. Today he will be discussing his health systems journey in implementing DoseMeRx and has many good tips and lessons learned to share with us. He received his Doctorate of Pharmacy degree from the University of Georgia College of Pharmacy and completed a pharmacy practice residency at St. Joseph’s/Candler Health System. And with that, I’m going to go ahead and turn it over to Dr. Shivani Patel.


Dr. Shivani Patel: Thanks so much Kristi, for that lovely introduction. And thanks so much for joining us today. As Kristi said, my name is Shivani Patel. I’m an ID Clinical Specialist here in Houston and I’m really excited to share our DoseMeRx journey with you today. Thinking back to our DoseMeRx experience, we’ve had a long and rewarding path to where we are today and I hope that sharing our experience makes your journey just a little bit easier. Before I get started, though, I also want to acknowledge the work and efforts of the Memorial Hermann Infectious Disease Specialist Team that lead and supported our implementation. Everything that we’re going to discuss today was really a result of our entire group’s work and effort, and I want to thank them in advance, for everything that they do.

So, Memorial Hermann is a network of 13 acute care hospitals and the Greater Houston area. Although we are a system, each one of our hospitals operationalize and implement things just a little bit differently because of differences in staffing models and campus needs. I think this is really important because our experience really shows how we did this as a broad system level, but also potentially how smaller individual hospitals can use and adapt some of the things that we did to implement at their facility as well. We really needed a process that created standardization across the system, but also allowed for flexibility. Our system employs about 200 full-time pharmacists. We have another 200 part-time PRN pharmacists that support us. And then within our clinical pharmacy group, we have 13 dedicated ID physicians with seven of us being full-time, then six of us being split between other roles.

Prior going live with DoseMeRx, our hospitals were a mix of pharmacy dosing and pharmacy and physician dosing. So workflow was a really major factor for us. Because we needed to get everybody up to 100% dosing, as we thought about how to implement Bayesian dosing and an integrated platform, this was one of the things at the forefront for us. Also, in terms of volumes, we see about 100,000 vancomycin days across the system. Every patient for us is evaluated every day, so this is a really big work burden on pharmacy, and that was a really big consideration for us if you’re on similar hospital with size or shape.

So why we chose an integrated Bayesian dosing algorithm – we opted for a fully integrated version of DoseMeRx, and I want to review some of our big rationale for why. Number 1: scalability. We wanted to maximize the number of patients that were being dosed via Bayesian dosing, because number one, we think it’s the right thing to do. We think it’s the right thing for patient care and we were investing in the software and the resources to be able to train up our staff. We did not want to live in a model of two different worlds vancomycin dosing. The second thing is we wanted to reduce levels. We wanted to be able to offset some of the costs that we are incurring by reducing costs associated with levels. And then, the scalability really allowed us to get early levels and then reduce the number of overall levels.

In terms of guideline compliance we wanted to transition to AUC dosing, because it’s the best practice standard and some of the first order calculators that we looked at online as well as some of the online platforms required multiple levels complex calculations. We really wanted something that was a little bit more streamlined and easy for our staff to be able to do. In addition to that, we really looked at quality; for me, this is the beginning and the end of any sort of initiative, or any innovative practice that you’re going to implement at your hospital. We have to keep the patient at the forefront of everything. Bayesian dosing really allowed us to lower vancomycin doses and reduce AKI rates to improve safety. This was really our big selling point for the medical staff and C-suite. So all of these factors combined help us consolidate therapy and standardize care, and make implementation possible for us.

The benefits of integrating into our EHR were pretty simple for us. Number one, we had excellence and efficiency in workflow. The integrated solution allowed for easy access within the current EMR, I’m going to show you some screenshots of this, and allowed for our staffing models to accommodate the new workflow. We decided that was much more cost effective for us to invest in the integrated software than it was to hire additional pharmacy staff and get them trained up to be able to accommodate moving to 100% pharmacy dosing for us.

The direct data feed saves so much time, and I can’t say this enough, our pharmacists do not have to input labs, doses, levels. It was a really a game changer for us. It also provides an additional safety layer in terms of the fact that we limit data entry errors, because nobody’s actually having to manually input data.

Finally, I think a big selling point for our staff was the fully integrated notes that the integrated version allows us to do; prior to DoseMeRx pharmacists were hand building notes. They were manually inputting levels, labs, and dosing regimens, and all of that we were able to eliminate with the implementation of DoseMeRx, which really allows for additional volume and capacity for our pharmacist to continue to dose vancomycin.

So this is what our integrated model or integrated workflow looks like. The first thing that we’re going to do is we’re going to select a course where we are easily able to see the courses that are available to us and then pick the models that we want to choose. At our campus, at our system, we opted for the availability of four models. We have a one compartment or two compartment and obese and the hemodialysis models that our pharmacists have to choose from.

The next thing that we wanted to do was view our dosing recommendations. As you open up, as soon as you choose your model, it opens up your dosing recommendations that are very face front for you to be able to see. The pharmacist can decide whether they agree with the dosing recommendations or whether there’s something about the patient that requires customized dosing, and they can do this very quickly, which really improves efficiency for us.

The last piece of this is the integrated note and as you can see here, the DoseMeRx note really provides a lot of the data that our pharmacists were manually entering prior to implementation. Target levels, outcomes, peaks and troughs – all of this is rolled up for us.

Our journey has been a long one. We started with engaging stakeholders, evaluating dosing, building the protocol, and so I want to talk you guys through some of these steps, and how we got from the beginning to end. I think the first thing for us is, the fact that we got engaged. We started at the top hospital administration in the C-suite, we engaged system quality and safety, and then the pharmacy directors as well. And our selling point for all of these people, or that we’re able to streamline process.

We were able to improve productivity. There’s a value add now from the pharmacy and the fact that we’re able to do things like reduce AKI rates, reduce vancomycin levels. We were able to see a return on investment with Bayesian dosing. The next group that we had to get buy in for was the physicians, nurses, and the staff pharmacists. I feel like the medical staff was pretty easy for us. Bayesian dosing is complex – let us help you with improved dosing strategies, lowering your vancomycin dosing and reducing AKI in our patients and essentially helping you take care of them better when they require vancomycin therapy. Our goal was to get those last remnants of our medical staff who were still dosing Vancomycin by themselves onboard with Bayesian dosing and get us to 100% pharmacy.

For nurses, our big selling point was reduction in timed levels. We decided to move to random levels that we get the morning after vancomycin is started, and because of this, we’re able to do it with AM labs and so it’s not an extra trip to the room for the nurses anymore, and it’s not an extra blood draw, which I think has improved patient satisfaction as well. Then, finally, I think [with] staff pharmacists we really needed to introduce the concept to them. When we decided we were going to move to Bayesian dosing and really get them to buy in , get them not to revolt. Change is hard, and we really wanted to prepare them for how to do this. We wanted to make sure that they knew what was coming. And not only that this was coming, but how we were there to help and support them through a pretty significant change in their workflow.

Finally, we had to gain support. With IT, we needed to stay engaged with the implementation process and then [with] staff pharmacist again, education, support, follow up on their mental well-being, and making sure that they were OK once we went live.

You know, I had 6 to 9 months to get ready for the DoseMeRx conversion. I was readily vetted, and the planning stages, how we were going to do this, how we were building workflow. But our staff got some education, hands-on live training, patient cases, and then they had to be ready to go. They didn’t have nearly the amount of adjustment time that I did, and I think that’s something that we really had to be sensitive to as we move forward.

The next thing that we did was review the guidelines in the clinical data. We looked to [ASHP/IDSA/PIDS/SIDP] all of the big guidelines, to provide rationale for dosing. However, we did not provide a how-to for this one, we engaged our medical staff. Then we had to decide on an exclusion criteria. Who were we going to include? Who are we not going to dosage vancomycin? And what we ended up deciding was that we’re going to exclude patients with unstable renal function and we define this as a serum creatinine of greater than six, and that’s the cutoff that we felt comfortable with, but I think it’s very flexible in terms of what you want to do.

Central nervous system infections was another big point of conversation for us, and we actually decided to include these patients. Then, finally, CRRT and peritoneal dialysis patients. We have a separate algorithm to dose these patients. Our end goal here after we reviewed all the guidelines and clinical data is that we wanted to maximize the number of patients that we could dose with DoseMeRx, and use Bayesian dosing.

The next thing that we had to do was create the protocol itself. I think this was probably the most time-consuming thing for us. We had so many iterations of this protocol, and I think it really comes down to the people that you’re engaging to build a protocol for us. It was the ID physicians and the ID pharmacist. We needed to get buy in and education from nursing, lab, in our IT support services.

The next thing is how you’re going to actually build your protocol. Are you going to be renal, function based, disease, state based, or DoseMeRx model based? What I mean by this is that if you’re renal function based, the first thing that your pharmacist is going to assess is the patient’s renal function. That’s how you decide how you’re going to move through the pathway. Are you going to move through model selection, based off of whether it’s cellulitis or a CNS infection? Or are you going to start with a one compartment model always, and then go to two compartment model if you need to for different variations? What we learned was that the protocol development and the work efforts really should be done in parallel with your IT development to be able to keep on the same time. What happened to us is that we underestimated how long the protocol was going to take to build for us, and so IT was ready to go before our protocol was built and it delayed the clinical implementation.

Next is creating the protocol. What we did was we essentially dosed everybody that we could with a goal AUC24 of 400 to 600 [mg*h/L]with our default targets at 450[mg*h/L]. The reason that we chose 450 for the default is that when you choose 450, you’re actually targeting somewhere between 900, because the number is not going to be exact every single time. For trough based dosing we kept a trough of 10 to 28 micrograms per milliliter and that’s for CRRT and peritoneal dialysis patients. Then for anybody who was unstable, stable, or a serum creatinine that was greater than six, we just dose them to a trough of less than 20. And we dose based on levels.

So after considering all of this, this is the decision tree that that we came up with. This was really key for our pharmacists to be able to maneuver and make decisions on DoseMeRx in terms of what level do I choose, what my AUC goal is, and then what are their directions in terms of monitoring. The first thing that we look at for our patients is vasopressor therapy, if they’re on vasopressors they automatically are treated with a two compartment model. If they are stable renal function then they go to a one compartment model and then hemodialysis patients go to hemodialysis. As you can see as you move down the tree, we’re going to dose trough based dosing or one-time dosing by levels for those patients who aren’t incorporated. It was very confusing before we developed this for our pharmacists to be able to navigate. This was something that was really important for us to think through and build on the front end.

In terms of feedback and editing it was really important to remember the staff pharmacists, and their role in this. I think their questions and their engagement in how we used DoseMeRx was the key to our success. Us, as clinicians, were so deep into the process of building this. We just assume things that were not clear to everybody else. Their fresh eyes, their daily use of the software, and the questions that they were asking really led to great questions and significant change that improve the way that we’re taking care of our patients. I’m really glad that we were able to engage them in this process. And it was really the “little big” things for us. For example, understanding our serum creatinine and weight cutoffs and how that impacted what was happening to the pharmacist. In the very beginning, we had these very strict serum creatinine cutoffs and we did not realize how many patients we excluded from Bayesian dosing because of that, and we ended up adjusting ourselves and widening our net. Same thing with the implementation of the random level. This was new for us and we sold all the benefits, but we didn’t realize that nurses didn’t necessarily 100% understand and they were holding vancomycin doses because of these AM levels, which we expect it to be high. Also physicians were discontinuing Vancomycin therapy that was needed because they saw a high level, and they didn’t know what to do with it. And so it’s making sure that you’ve got that active surveillance for these patients, and how you’re going to keep track of all of your vancomycin patients. Are you going to use a work list, do you have a spreadsheet? How did your handoff look like? How do you hand off levels? I think all of those things were really important.

Our big lesson learned was that each staff member who used DoseMeRx, it took them only about a day, one shift, to feel comfortable with the software. It was very easy to navigate, it’s very easy to learn, but we realized that our staff needed 3 to 5 days of consecutive vancomycin dosing to really be able to navigate the protocol well and establish some sort of clinical expertise in how we thought in the thinking part of the protocol.

The last piece here is educating our staff. For our pharmacists, what we did was we did a didactic education, which was a prerecorded CE, so that they understood the rationale and the basis for AUC dosing. Because the questions don’t come to the clinicians 100% of the time. We wanted our staff to be very well versed in why we’re doing AUC dosing, what the benefits are, and to be able to answer basic questions for the medical staff and the nursing staff. We did software and protocol education, and there was a huge focus on our new workflow and logistics. How do you keep track of your patients? How are you ordering levels? What information needs to be included and handoff from one person to the other so that we’re optimizing workflow and efficacy? Then we created sample patient cases to make sure that our pharmacists could be able to reproduce some of the key principles of our protocol.

For physicians, we had to decide, whether we were going to give access to physicians, not the ID Physicians versus everyone else. This is something that they were very engaged in. We didn’t worry about the small details of the protocol. They really just needed to know that we’re targeting AUC now, and what the clinical benefit was to their patient. We really focused on quality safety and interpretation of levels in terms of their education. Then, finally, for the nurses, this was a really big education effort. As well as hospital wide, system wide education. We wanted to make sure that they had general talking points for rationale for AUC dosing, how to transition from tough based dosing to AUC dosing, when it was still appropriate to draw timed levels, and what those would look like. And then interpretation of levels, so that some of these questions weren’t coming to us, because we’re managing 100% of our patients instead of the physicians now.

This is what our volume looked like over time. The entire system did not go up at once. If you are at a big system, and you think that you know every single one of your campuses need to go up at one time, we’re proof that you can do this at a staggered fashion and end up working out really well, because campuses who went live early were able to share their experiences with campuses who went live late. We were able to build, you can see that our as our volume went up, our users went up and up. 80% of these patients that we’re dosing are using the one compartment standard Vancomycin model with maybe another 15% with a two compartment model for our unstable renal function patients and then maybe another 5% with our hemodialysis patients.

It was really nice to be able to celebrate some of our successes. You can see that almost 90% of our patients with the one compartment model – what this means to us is a loading dose maintenance dose as dictated by DoseMeRx. And then one level, we were able to have therapeutic AUC at 48 hours of 90%. This was really huge, because our traditional Vancomycin trough based dosing protocols prior to this at 72 hours, we were only able to achieve this 50% to 60% of the time. This was, this was really great. I think it says something to the intuitiveness of Bayesian dosing, and everything that it takes into account.

Finally, I wanted to show you guys our trough versus AUC plot for our system year to date. The blue box shows you the trough of 10 to 20 [mg/L] and the percentage of doses that fell within that trough range. The yellow box shows you an AUC24 target a 400 to 600 [mg*h/L] and what how many doses fell within that. I don’t know if you’ll notice but I actually have this yellow box stretched back to 350 because when you’re targeting an AUC of 400, you’ve got a little bit of leeway for severe skin  and soft tissue infections. For patients who were clinically improving, we didn’t necessarily push the dose just because they were at the lower end of our AUC range, and you can see, Dustin will talk about this as well, is really translated to an impact in terms of AKI and lower doses used.

I’d like to wrap up with some of our “Do’s and Don’ts” that we learned. First, we keep continuous lines of communication open with IT. Even after the launch, there were a lot of things that they were really helpful with, or things that we still had to navigate, and that relationship was really important.

Be ready to have to answer questions for staff. Then have the clinical specialist shadow the other pharmacist, to confirm dose elections [which we did] for the first month that we went live with DoseMeRx. My personal approach was to camp out next to the pharmacist who were doing Vancomycin. I would hang out there for an hour or two in the mornings, and an hour or two in the afternoons just because if you’re sitting there, they’re really more willing and ready to ask questions like what about this, what do you think about that? That either allowed me to address education deficits that they may not have picked up on when we did our didactic or our live education sessions, or if [something] was really unclear [to them]. We need to go back and adjust our protocol and our directions to the staff in terms of how to manage these types of situations.

There is going to be a learning curve and one of the things that was really helpful to us as we kept a living Frequently Asked Questions (FAQ) document, was that any question, whether it was an education based question or something that needed to be adjusted in the protocol, we added to this FAQ document. When we were off site, the pharmacists were able to scroll through these questions and up their understanding and be able to handle questions based on the FAQ document. Lastly, follow discontinue vancomycin orders for appropriate discontinuation, and then therapy not warranted or concerns over AM lab levels and being able to help the medical staff manage those.

Now, in terms of our “Don’ts”, we had to remember that the written protocol, don’t think that it’s not going to be defined. I feel like that’s a living document as well. It still is a year later, we’re still making adjustments and tweaks. I think that’s absolutely OK, because we’re constantly learning from our experience. Don’t assume that the staff will do everything exactly as you intended; we’ve already seen variation in practices. I think some of them are appropriate and some of them we’ve had to re-educate staff on. Those have been important lessons learned for us as well.

Don’t forget to prepare for EMR downtime situations. This is one of the caveats that we did not think about, since we’re a fully integrated DoseMeRx site, when the EMR goes down, DoseMeRx goes down as well. We had to prepare our staff or how to handle those situations, and then come up with initial dosing recommendations until the EMR was backup and we could run models in DoseMeRx and then adjust doses after that.

Thank you for letting me share our experience. I hope this was helpful to you, and I would like to hand it over to Dustin to talk about the clinical financial and operational benefits of Bayesian Dosing.


Dustin Orvin: All right, Perfect. Thank you Shivani for that excellent review of implementation at such a large and diverse health system, at Memorial Hermann. Now I’m going to try to share with you our experience, at St. Joseph’s/Candler, and through our implementation experience and we’ve been up and live with this for quite some time, now, about 18 months. I’m going to talk to some of the clinical, financial and operational benefits that we’ve observed with Bayesian dosing at our institution.

Alright, so to give you a little context, St. Joseph’s/Candler Health system, we’re a community health network, we have 714 beds divided between two anchor hospitals. Some quick stats, to give you some comparison: we have two hospitals, 57 inpatient pharmacists, many more outpatient. Pharmacy’s responsible for all of our Vancomycin dosing via an automatic consult. We have three dedicated IT specialists and numerous College of Pharmacy faculty.

Alright, so our journey so far. You know, our journey is very similar to what was just described to you at Memorial Hermann, we had to go through all the same exact steps to implement AUC at our institution. Now we’re in this monitoring phase, the post implementation phase, where we’re really assessing our program to make sure we’re getting the intended results, the results we set out for initially. I’m going to talk to you about how we monitor our program and the things that we’ve observed so far.

Anytime you’re defining and you’re going through a monitoring program and structuring what you have to figure out how you’re going to define success. Are you setting out to improve clinical outcomes? Are you looking to gain cost avoidance and cost savings? What metrics are you going to monitor, what metrics are available to you? This is something where we will discuss the benefits of Bayesian dosing in relation to first order kinetics throughout the rest of the slides are coming. There are specific metrics you can pull out for your patients to get benchmarks for your program. Then, of course, you have to know what your baseline metrics, what baseline metrics you have already. You might have to dust off some old MUE data, to see where you are, where you’re starting from, to really see where you’re going and the benefits you’re going to realize with AUC. You want to know where your baseline is, and what the impact of AUC has been at set intervals throughout.

So a little bit about our program structure. We actually went with a web-based version instead of an integrated version of DoseMeRx. We went live a little bit ahead of the guidelines in February of 2020, right before COVID-19 put everything into a gridlock tor most people. Pharmacy’s responsible again, like I said, for 100% of all Vancomycin dosing via an automatic console. And we brought Vancomycin AUC dosing to all of our patients, all of our adults receiving Vancomycin. We have target AUC range of 460, like the guidelines recommended, then we derive that from a single trough concentration in the majority of cases. With some cases we do recommend peak concentrations.

Some key metrics of our program looking at how many doses we’ve seen so far, who didn’t ramp up quite as quickly as, Memorial Hermann size. We have, at this point, we’ve just about 2500 unique patients in, about 2800 unique patient courses, so we had some repeat offenders in there. But that’ll pay off. That will come up later in slides when we’re trying to determine the financial offsets from our program implementation.

So why did we go with Bayesian? Well, first off, we have get to why we chose AUC. In about 2000, we did in a MUE and showed 14% AKI right across the board at our institution, and that was really concerning for us. There were, of course, confounders in there. We knew there were some things that we picked up, like concomitant nephrotoxins and IV contrast, whether these patients for high risk to begin with obesity, critical care, things of that nature. But we still felt like that was pretty high, and we had some room to improve. And, again, pharmacy’s doing all of the vancomycin consulted at this point. So this was really a pharmacy problem – we had to come up with a pharmacy driven solution. At that time, we started closing key changes to our dosing protocols, and AUC dosing was identified as a potential solution to improve our AKI rates. At that time, the literature was emerging more, and more and more piling up – this might be the move going forward. So we started evaluating it seriously to see if it was feasible at our institution. Know, historically, this hasn’t been feasible, it hasn’t been practical as while we’re using trough based dosing for so long. Well, now the invention of Bayesian dosing, and these commercial platforms, this has made it so much more feasible and practical to implement in our institution. We did evaluate first order kinetics to see if that was a good fit for us. Our pharmacists are familiar with that, with trough based dosing. We were used to the calculations, used to the estimations, but we really felt like it was still not practical, enough to implement, and we really had a goal of improving AKI across the board in all of our patients. We knew it was going to be difficult to roll AUC out with first order kinetics in that kind of volume patient volume.

This is where Bayesian really, really paid off for us. It allowed us to integrate AUC dosing almost seamlessly into our workflow, both for a pharmacist that were actually dosing vancomycin, but also our nurses, lab, staff, and clinicians interpreting Vancomycin values. It created reproducibility, and within our program, and improved the standardization. We’ll talk about some of that, and some of the results we’ve seen in some future slides and then clinician confidence was huge here. With Bayesian dosing, we knew we were going to get validated calculations. As Shivani alluded to earlier, first order kinetics has some really complex calculations associated with it at times, whether you get it from an online calculator or something in your EMR or even an Excel document that you’ve got home-grown within your institution. There’s still complex calculations involved and potential errors involved there as well. And this is something that Bayesian takes off the table. Once you get Bayesian, you get the information for your patients into the software. It does all the calculations for you. Your pharmacists have confidence, they are getting the validated clinical results and the clinicians interpreting this. These values also know that the results validated so it helps you implement in a very individualized patient care plan in your patients. And then, we didn’t have a lot of data at the time, but we really felt like this was going to be cost effective method for us. We knew first order kinetics was going to increase our lab monitoring costs by adding peak values. We’re going to get more concentrations. We didn’t really know how many times we’re going to have lab sampling errors or maybe those timing issues as well, where you had to repeat certain lab values. We knew that was going to increase cost further, so we felt like we could really offset some of those costs by implementing a Bayesian program there. So, all in all it made AUC dosing a lot more practical and feasible at our institution.

When we set out, to bring AUC dosing to our patients, we had a few goals in three distinct areas: the clinical, operational, and financial. Of course, we want to improve patient safety and maintain efficacy and vancomycin, I think that’s everybody’s motivation here. Operationally we wanted to do that in a way that didn’t disrupt our current workflow. We didn’t want to significantly change what our pharmacists were doing on a day-to-day basis. How much time they were spending on vancomycin? And we didn’t want to impact our nursing staff, our lab staff, or physicians as well.

So improving our standardization was a huge aspect that was something that really drew us to Bayesian financially, you know, the clinical patient safety and the efficacy markers there that we knew that was going to have financial benefit. And, of course, this was going to be an investment in patient safety. But, we wanted to implement AUC with minimal impact on our budget. Going forward that we felt like Bayesian was the best option, and we’ll talk about why in future slides.

Some of the clinical benefits we’ve observed since implementing, we’ve definitely improved our standardization and results have been consistent across the board from pharmacist to pharmacist, among different care settings. As we’ve rolled Bayesian dosing out, it has easily accessible dosing reports saved within the platform. I talked about some of our patients, how many patients have come back in and repeat courses – well, I’ve got all of those old dosing histories right there within the platform. If I have a new course, I can quickly review the whole course, see how they performed last time, see if there’s clearances, or the serum creatinine clearances are similar if they’re in a similar clinical scenario, if the dosing probably is going to be reproducible this time so we can get the dosing right on the front end.

It [Bayesian dosing] improved our pharmacist confidence. You’ve got validated, reliable, and accurate results, allowing us to focus on executing that patient care plan we wanted to set forth to optimize those outcomes. It also has increased our prescriber confidence. Physicians are placing more trust in pharmacy dosing plans, and this is directly in association with the reduction in AKI that we’ve seen, we’ll talk about some of the numbers we’ve found in our institution a little bit later. And then in the infamous rare toxic vancomycin levels, nephrologists are much more hesitant to blame acute kidney injury on vancomycin these days than they were, say two years ago.

But those are all anecdotal feel good stories, those are things we’ve observed. What are some hard metrics that we can report back? If we implement and spend and invest in implementing AUC, you have to come back and report some real hard data to your appropriate committees, whether it be the C-suite, administrators, or something like a pharmacy and therapeutics committee. Well, this is where Bayesian has a clear leg up on first order kinetics. You’ve got all this data within your platform and they can help produce hard benchmark metrics for your dosing program, at your institution. So, these are some of the options that you can pull out in DoseMeRx, for example, and we’ll talk about how you might use those to develop and refine your program along the way.

These are results in our program, actually, so we’re looking at serum creatinine increases, as the 3.3% as a big reduction from that 14% or so we saw, that MUE we presented in 2019, and you know this is some of the interpretations that DoseMeRx would generate for you, is that very few patients have those serum creatinine increases while on therapy. Very few doses are given of AUC above 600 within our program. This is something that is re-assuring to see.

Then some other metrics they can pull out related to efficacy. You can see how fast or how often your patients are getting to a therapeutic concentration, then at what time interval they’re getting to it. For the majority of our patients, our goal is to get our doses optimized at the 48 hour mark in the majority of our patients. Right now we’re currently meeting that at about 87% of our patients. Very similar to what Shivani was showing you earlier, Memorial Hermann has about 89% of the patients. But here’s the thing, let’s say we wanted to get to 90%, what we’re doing at our institution now is utilizing dosing to identify these patients that did not reach their therapeutic AUC in that timeframe. We can actually specify and get patient specific information for our problem patients that aren’t getting there, pull them right out of the platform, analyze what’s going wrong there on a solution and fix it. Directly, this is very unique and something we would not be able to do with first order kinetics.

Operationally when we implemented with DoseMeRx, it ensured that all of our pharmacists were getting their AUC estimates from a uniform source. We had a variety of trough based practices. Previously, some patient pharmacists were pulling out calculators and doing the first order kinetics by hand, others had Excel documents, or used an online calculator some sort. Now we’re getting all the results from a single place, all of our calculations from a single source. And it’s given us validated results. So, this was huge for us and improve standardization across the board.

Now, rolling out Bayesian dosing and AUC dosing with DoseMeRx allow for implementation with fewer vancomycin assays. Which greatly reduce the impact on our nursing and lab staff. This is what really allowed us to bring AUC to all of our patients. I don’t know that we could have brought AUC to everyone had we had to do first order kinetics due to the incremental costs associated with it and time constraints that would have been associated. So this is something huge selling point for AUC for our institution. Now, we still do trough based dosing to determine our AUCs, so prescribers are still seeing the customary trough. They’re not having questions about what’s going on, we’re not seeing random levels is off. And we’re not seeing peaks as often. So we’re not confusing our prescribers that are that are used to seeing troughs; we’re not changing too much on their behalf. And it’s really not changing anything for our nursing staff as well, because they’re used to drawing troughs right before they give a dose. By doing this and reducing our impact on our nursing and lab staff, it allows us to reduce our education needs, it wasn’t as intensive education needed for nursing, or lab staff or physicians. We could focus that education on our pharmacists, who actually managing vancomycin and our patients. That allowed us to rollout DoseMeRx with a little more focus on clinical concepts and utilization of software.

This is some useful data that you can pull out of some amazing platforms, specifically DoseMeRx. This is something we haven’t had before. We can get clinician specific data, performance. How often are you using the software? How many patients are you dosing there? You can identify your top performers, your people that are integrating AUC within your institution. Most often, you can figure out what issues, if any, they’re having. You can figure out how they worked around those and integrated it into the workflow, and then you can take those solutions and improvement at your institution and take this to somebody who’s maybe not performing as well. Not integrating AUC into the practices as easily and try to help them facilitate solutions to maybe raise up your average and standardization across the board. That’s something that’s pretty cool and unique to Bayesian dosing, that’s something we haven’t fully implemented at our institution, we’re still working on this, and how to use this data. This is a interesting, because we haven’t had this before.

After we implemented AUC, we decided to do a post launch an assessment here in order to evaluate our dosing practices, we compare trough versus AUC dosed patients, we try to do a case match here, and we’ll talk about how we did them. We had some unexpected findings early with that, the other was there was clinical uncertainty in the immediate transition period, which Shivani said there’s definitely some transition period you’re going to have to monitor and make sure everyone’s interpreting your protocol as you intended. As you see, we had to repeat concentrations in a protocol recommended in some cases. And then, AUC group ironically have more troughs in their goal range than our trough dose group.

From that, we dug a little deeper to understand where that clinical uncertainty was coming from. Because we notice that in the pharmacokinetics notes. It really dove down to an oversight on our part. When we rolled out education, we really focused on the platform, how to interpret that information and make sure they were comfortable with Bayesian there. But we didn’t focus so much on the PK parameter estimates. Our pharmacists are very used to and had very good clinical experience dosing patients on first order kinetics and population estimates. From that, they were calculating their own population estimates, comparing it to what DoseMeRx was generating. And saying, well, I don’t know if I want to trust DoseMeRx in this case, because I’ve got a different creatinine clearance estimate. I wouldn’t calculate that creatinine clearance in this patient. Then we saw them use different doses or alternative doses is from what DoseMeRx was recommending. And we found that there were more likely to have sub therapeutic AUCs. So it’s something we had to re-rollout, and really educate on that the differences in and where these PK parameter estimates are coming from within the platform versus population kinetics there.

This was a national survey we did as part of that post implementation assessment, simply published a letter to the editor, our team did earlier this year. We were really looking at transitioning from that guideline approval to practical implementation. We knew this is something that our facility, we went through, it was difficult, it was a long process – but definitely worthwhile. But we wanted to figure out where other people stood with that. And we were surprised to find that 18% of respondents stated they had plans to continue trough based dosing, they evaluate AUC to some degree and decided to stick with trough. We just wanted to know why, we didn’t fully capture that in our survey, but some people actually reached out to us and we found a trend that many were, you know, just very concerned about the ROI on this. They were looking at the financial aspects, they looked at their current AKI rate and decided, hey, we’re at a comfortable level, we like where we’re at, so we’re not really going to look to implement. You see at this time, more data’s out there, more data is available. So we’re going to talk about what financial data is out there in terms of return on investment and specifically in relation to AKI.

So how do you get that necessary Budgetary support, right? We’ve got the clinical benefit as well as well listed in the literature. Now, the operational benefits are there if you look at Bayesian and how it seamlessly it can be done, it can be implemented at your institution. It’s a very nice, very attractive option, but then financial approvals always seems to be the hiccup. That’s the question I get the most from other institutions that are going through this process, and how do you get that justified? So one of the things that I really, really worked out for us was you have to know where your current AKI rate is and your patient population. That you’re intending to bring AUC to all your patients. You want to know your AKI rate across the board. But if you’re looking at targeted populations, you need to know what your vancomycin associated AUC rate is in those targeted populations. And at that point, you can estimate of cost savings with a very conservative 50% reduction in a cap. So, looking at that, that’s pretty standard across the board. In the literature, you can see a nice 50% reduction. But some had a little bit more than that. So, think 50% is a conservative measure. And at that point, there’s a number of cost savings or associated cost data out there now that you can generate a cost savings estimate from.

Then you can decide from there is Bayesian or first order kinetics the right way to get there. Is it going to be the best way to implement to the fullest. And when you’re talking to your finance department, you’re trying to get budgetary support, you have to speak their language. You can tell them Bayesian is known as a fixed cost. You can plan for it, you can budget for a year after year, its irrespective of patient volume. So no matter if I dose 5 patients or if I dose 500 in a month, I’m going to pay the same fee. First order kinetics that’s not the case. If you want to roll this out to maximize the benefit of bringing AUC at your institution, there’s uncertain incremental costs associated for every single patient, right? You’re adding vancomycin assays or any unknown and impacts on nursing time lapse, pharmacist and the clinician interpretation issues as well. So this is something we knew we struggled with in terms of troughs when even, trough values are tough to get in appropriate time in clinical practice. Adding a peak to that, we thought that was just going to make things even more, difficult for us and potential for repeating. Redundant lapses is astronomical there.

Continuing on with that post implementation assessment that we did, we also included a financial analysis in that. It’s a little more specific, we had 100 patients in each group case managed based on demographics, Charleston Comorbidity index, as well as ICU 10 codes to identify them. We had a trough group, an AUC grouping – see the timeframes there and then our definition for acute kidney injury.

So, some of the interesting results. This was a student project that presented, a poster on mid year, this past year. We had some interesting findings come out this. The AUC group had a little bit longer length of therapy. On average there’s 100 patients, but number of dose changes tended to be fewer in those groups, troughs were actually in range, more often, than in the trough targeted group. We saw our average trough come down, which is re-assuring to see, is what you would expect to happen when you implement AUC. But the big thing here, this is a beautiful -reduction in an nephrotoxicity at our institution. We started off with 12% and the trough based groups are very similar to that 14% we saw in the MUE data. And we cut that down to 2% in this in this study. With the AUC dose group This is huge weight. We saw a big patient safety impact, and that’s exactly what we’re shooting for, why we implemented this. So seeing thes results was very re-assuring and big-time confidence building for our program. The other thing is the average cost per patient day is net neutral here. You’re looking at about $16.5 for trough, and $16.2 for AUC. This was awesome and re-assuring to see that we implemented with a net neutral impact on average daily vancomycin dosing.

So how did that translate to us financially at our institution? Well, first off, looking at the AKI rates we saw in the trough group comparing that with the AUC group at 12% versus 2% and that was statistically significant. That was six times higher in the nephrotoxicity in our trough group. And you can see the estimated cost reductions here. This is based on a study by Patel N, Huang D, Lodise T., published earlier this year. It found that a single dose of vancomycin  associated nephrotoxicity was an extra cost of about $8000 when you consider specialist consultation, length of stay. All the things associated with an AKI, especially if you get to the severe case where the analysis is concerned. No, sir, the impact on our institution is potentially massive right there. You’re looking at about almost 2 million dollars in net return, just from the safety offsets. This is a huge, huge number that we can consider. And I can assure you that definitely offsets our Bayesian dosing budget. But let’s say you don’t remember that, you don’t actually believe that number. Let’s say that AKI rate, you think, are the cost of $8000 seems a little high to you, let’s cut the number in half. What if it’s $4000, if that’s what every AKI cost at your institution. Well, great, that’s still a million dollars in cost savings, still very significant financial numbers. We can do that math and keep going down, over and over again. It still proves to be very, very beneficial. Financially offsets the cost of a basic platform.

To conclude, some of the these are the different things we want you to consider when implementing and learning, from my experience. Definitely consider how Bayesian dosing can meet your dosing and your monitoring need. It’s really cool some of the data you can pull out and monitor your program. Keep benchmarks, and make sure you’re hitting the goals you set out for.

Pick metrics to monitor your most important goals at your institution. You know, everybody’s rationale for implementation is unique, so make sure you’re picking the goals and metrics that most accurately represent those goals.

Communicate your success stories early and often. This improves provider and clinician confidence, those that are actually managing vancomycin patients and utilizing AUC dosing. Communicate those success stories early, showing that, hey, we are actually seeing those reductions in nephrotoxicity. You get increased clinician buy in.

Provide open lines of communication for stakeholder feedback, making sure everyone can communicate issues early and often and effectively if there are any that arise. You can come up with solutions in real-time, instead of waiting on things to progress and having to catch major issues that could have been prevented, then, of course, continuously utilize data to refine your process.

This is where Bayesian has a clear leg up on first order kinetics – you can really pull some unique metrics out of there and really determine where you are as a program and where you want to be, you can pull out patient specific issues to help you get there faster.

Some general “Don’ts.” First thing is don’t take monitoring for granted when you’re evaluating Bayesian dosing. It’s an amazing benefit of Bayesian. Don’t assume the AUC or Bayesian dosing is too expensive to implement without evaluating those potential cost savings to your patients and to your institution. Don’t underestimate the financial impact of AKI reductions and how much that can have at your institution; that was something we definitely didn’t expect to see that large of a no financial offset at our institution, but going through this is something it’s been a massive benefit to us.

And with that, I want to conclude and thank you all for the opportunity to present and turn it back over to Kristi for questions.


Kristi Kuper: Great. Well, thank you so much. I know Shivani’s going to come on here as well on camera and we’ll go into the Q&A. And if you want to share your screen Dustin, share your camera. Again, we’ll do this live. Thank you both for two really amazing presentations. You had a lot of great, practical tips in there that I think will be very useful for those who are not only evaluating to implement Bayesian dosing, but even those who have already implemented it. We have just a few minutes left for questions.

When I look at the questions we got, probably half of them were related to when you obtain levels, how many levels you obtain? When might you obtain one level versus two levels, and you know, how do you handle it? Do you do random level? So maybe each of you take a minute or two and just re-iterate your process around when you take levels and how frequently you take. Shivani, if you could start, that’d be great.

Shivani Patel: So we decided how we’re going to handle levels based on the model that we’re using because we have multiple models available to us. For our one compartment model that our patients get a loading dose, we start maintenance dose, and then we get a random level in the morning with AM labs. If based on the level, if there’s no dosing recommendations changes required, we don’t get another level for five days unless DoseMeRx, when we run the model every day, recommends something different. For our two compartment model, so those are the patients that are on vasopressor therapy for some reason have altered volume of distribution, or just have unstable renal function where we’re worried about clearance. What we do is we get to model, or two levels for those patients. We do a level with AM labs, just to make it easy for the nursing staff. And then we do a second level somewhere around 10 to 11 AM. So that both levels are back for day shift to be able to run the model and that allows us to look at actual vancomycin clearance for that patient with a two compartment model and make dosing recommendations based on that.

Kristi Kuper: Great. And Dustin?

Dustin Orvin: Awesome. So, for us looking at monitoring we utilized troughs and we stuck with troughs to monitor and drive AUC. We do a single trial for the majority of our patients and we only purchased a single one compartment model. So we don’t have the two compartment models as Shivani does at her institution. One of the things that, if we saw any situation that arose where volume distributions in question, maybe it’s fluctuant, and we wanted to see if the models capture it accurately, we might add a peak concentration at that standpoint. Or, in the most severe infections, we might add a peak to make sure we’re getting the most accurate AUC estimates within our institution, but it’s still trough based monitoring. We do that, typically, between every 3 and 5 days. And it’s up to the pharmacist discretion on how often that’s monitored, depends on clinical situation, and then, of course, patients stability with renal function and the dose recommendations now.

Kristi Kuper: OK, great, thanks, that was really helpful. This question is for Dustin. Do you have any suggestions, Dustin, on how an institution might structure their medication use evaluation, or MUE, to help identify AKI associated cost savings?

Dustin Orvin: Yeah, so, when you’re structuring those MUEs, it’s really tempting to go with a very sensitive marker or looking at some of the definitions set forth by the guidelines [inaudible] and things of that nature. And that’s something we did at our institution: We use the serum creatinine measure of 0.3 or greater, and increase to that extent, or a 1.5 times baseline. However, most of the financial data out there is actually going to go back to the more technical definition of vancomycin associated nephrotoxicity. So I think, the biggest thing for you to structure your MUE is to go find a financial paper that you really trust, really have confidence and see how they defined AKI and then model your MUE, definitions of AKI behind that paper that way, the results you find are directly translatable with the financial savings.

Kristi Kuper: That’s great, great suggestion. Another question that we got today and this one comes up a lot and I was wondering maybe Shivani, if you could comment and then Dustin. Are you applying AUC dosing to non staph aureus infections? And if so, what is kind of your general target range?

Shivani Patel: We are. So we are doing it for any vancomycin therapy across the board. The vast majority of them target staph aureus and specifically MRSA. But the reason that we felt comfortable using it for non staph aureus infections is there is a few small papers on Enterococcus where they showed that achieving an AUC of greater than 400 improved 30 day all cause mortality and secondary infections. So that was our surrogate endpoint for Enterococcus and other strep species. And then there’s really a lack of data for coagulase negative staph. But we just really looked at breakpoints and making sure that AUCs, the patient’s clinically improving, right? So, clearance of repeat blood cultures, and severe infections], if we’ve had persistent bacteremia, we either push the AUC target or we’ve changed to alternative therapy but it’s been clinically based, really, not AUC, based for those patients.

Dustin Orvin: Yeah, awesome. And like Shivani said, there’s not a lot of great data out there outside of Staphylococcus aureus, specifically MRSA, but we felt confident in switching. And we still utilize AUC dosing. We still target that 450 [mg*h/L] range in our vancomycin patients because we feel pretty confident about the efficacy line, as well as the safety markers they’re in that window. The issue we run into is, we’ve always those patients based off of trough and troughs are always a surrogate for AUC. So, we felt confident rolling out, AUC there based on the clinical data and the clinical experience we had in the past. But there’s definitely a need for more research in that area.

Kristi Kuper: OK, great, well, that takes us to the top of the hour. We have a number of other questions that we didn’t get a chance to answer today, and that that often happens, especially with this many people on the webinars. What we’ll do is, we’ll get those answered, and follow up with the participants, and also send out an FAQ after the webinar, when we share a copy of the slides. So we’ll definitely get a chance to answer those questions for you, and thanks to everyone who submitted those. So with that, we’re going to conclude. I would love to again thank Dr. Patel and Dr. Orvin for two amazing presentations. And thank you, everybody, for joining us today. And hope you have a great rest of the day! Bye!

Download The Handout

A copy of the slides is available for you to download and refer to at your convenience.


Explore Bayesian Dosing and DoseMeRx

HIPAA Privacy Statement | Privacy policy | DoseMe & the GDPR | Copyright © 2012 - 2024 DoseMe Pty Ltd. All Rights Reserved.
DoseMe® is a registered trademark of DoseMe Pty Ltd. DoseMeRx℠ and DoseMe℞℠ are service marks of DoseMe LLC.