 My name is Robert Lovato. I'm an Associate Professor of Anesthesiology and Chief of the Division of Multispecialty Anesthesiology at the University of Nebraska Medical Center in the United States. Today I'd like to talk to you about the role of R and Shiny dashboards in facilitating quality improvement in anesthesiology and perioperative care. The goal of this talk is not to teach you how to do the things that I'm showing you, it's more about telling you why. What I want to do is demonstrate a use case for how R and Shiny dashboards can be used in a clinical service and identify what has been helpful or persuasive when addressing stakeholders. So to back up a little bit, I want to differentiate between quality improvement and clinical research. As most of us know, clinical research is meant to generate new knowledge and validate it in new populations. These are the things we're most familiar with when talking about medical sciences, clinical trials, observational trials, meta-analyses. These are really fundamental for establishing what the best practices are in evidence-based medicine. In contrast, quality improvement is a framework used to systematically examine ways that care is delivered to patients. So really translating our evidence-based practice into how we treat patients and take care of them on a day-to-day basis. There's an enormous delay in adoption, however. Going from basic science research to human clinical research to moving on into clinical practice, some studies have estimated that there is as big as a 17-year lag in implementing new evidence into general medical practice. Now, that's a tremendously long period of time, and my hope, at least as an academic physician, is that we shorten that to really as small as possible. That said, it's not uncommon for new publications, new evidence to take several years to make it into clinical practice in academic centers. And even into smaller community centers, it can take a decade, even more. So in designing these clinical dashboards, I think it's really important to understand who is your audience. As a physician, I understand how to talk to other physicians, but bringing change in a large institution, especially an academic medical center, requires buy-in from more than just physicians. Requires nursing staff, support staff, laboratory techs, scrub nurses, all the way up to the C-suite. Really, if there's not buy-in across the board, then really no change is going to be effective. And so I think one of the major things that are in shiny dashboards can do to help us is to accelerate this change. It can really help demonstrate why change is needed. Interactive dashboards like this can also give viewers a way to interact with the data, which I think is tremendously powerful. The last and most important thing is it allows stakeholders to explore subsets of data. And I think nothing is more convincing to someone who's investigating a problem than drawing their own conclusions and agreeing with what the path of change is being proposed. So let me move over and show you an example of an R&Shiny dashboard that we developed at our institution. This is a dashboard that's being used in actual clinical practice. It was written using the Flex dashboard tool and it demonstrates for viewers the total number of anesthetics that we perform over a 12-month period. And then a rolling count of the anesthetics we've performed over the last 30 days illustrates the number of serious events, postoperative events, and intraoperative mortality events that we really want to keep track of. Down at the bottom we have an interactive month-by-month summary of individual data points. And the hover data allows us to explore individual data points and highlight things like serious events or mortality events over time. The dashboard also includes tabs to investigate individual events in more detail. So the serious events tab would allow us to go through and look at individual surgical cases and explore exactly what happened that we would want to look into to improve the quality of our anesthetic and perioperative care. The dashboard allows users to click through and explore individual data items and filter in ways such as which clinical location we're looking at and subset the data that way. It also allows us to explore by date to look at any particularly high or particularly low trends. There's a similar tab for mortality events which in this case are meant to track death in the OR and death shortly after a surgery. And again we can subset those data filter by location or filter by time period. The last tab of the dashboard includes information from the MIPS and other quality measures that we use globally across the institution and across the United States to track perioperative care. So one that we're looking at in particular is prevention of postoperative nausea and vomiting in high-risk patients. This is MIPS number 430 and MIPS stands for Merit-Based Incentive Payment Program. So there is a financial reason why the C-suite would want us to be as good as possible on these metrics. Eventually this leads to us getting paid to do the job we do or at least to receive full reimbursement once we can show that our quality is high enough. The two sidebars here are meant to look at more recent guideline data. So the MIPS data here in the prevention of post-op nausea and vomiting is in high-risk patients. This was based on the 2013 guidelines and as you can see we've really done a pretty good job of integrating this care with success rates hovering right about 90 percent which is pretty darn good. The two pains to the right of that show the more current guidelines. These are taken from the 2020 publications for the prevention of post-op nausea and vomiting and when looking at this we can see that there really are some deficiencies in our care. There's a huge opportunity for improvement. We see that our care meets the guidelines as infrequently as 50 percent of the time for men and as infrequently as 30 or 40 percent of the time in women. We then go on to see that as a result of not meeting these guidelines our male and female patients require additional medications in the recovery area in order to treat nausea and vomiting. This subset male patients has to receive a rescue anti-imetic somewhere between five and ten percent of the time. Our female patients in contrast require rescue animetics 15 to almost 20 percent of the time so huge opportunity for improvement. Let me just touch on a few of the tools within R and Shiny that have made these possible. The first is the DB plier package. This is part of the tidy verse and has been absolutely central to how we do dashboard creation within our institution. So we use Epic as our medical record system and the database behind Epic is called Chronicles. There's an ETL of Chronicles that takes place every 24 hours into a database called Clarity which then filters down to another database called Kaboodle. The DB plier package allows us to query these databases on a daily basis and update our dashboard so that everything we're seeing is really no more than about 24 hours old. This is tremendously helpful for our leadership in being able to identify problems or trends very early on. Another package that I think is really important is the high charter package. This is a JavaScript library wrapper for the high chart tools and allows us to generate these interactive displays of graphs with hover data summary data highlighting. I think this is tremendously impactful. The goal of these dashboards is not necessarily for me to interact with them in front of people but to give people a URL and allow them to interact with the dashboards on their own. By being able to hover over data click receive subsets it's really a great way to convince people that there is a problem or is an opportunity that needs to be addressed. Again this is not really about how to code in R but the high charter package for instance is really simple. So this particular graph was made from just a few lines of code it has two y-axis and x-axis and three data series and really just came from these few little lines so very simple to use very powerful makes for an excellent presentation tool. The reactable database tool has been really very helpful it allows us to display a data table and when combined with crosstalk allows us to filter subsets very effectively. Again not a whole lot about code this is really just a demonstration of how little code it takes to build powerful tools. We start by creating a shared data object we insert filter widgets within crosstalk and then we plug that data into a reactable table which can subset for us along the way. So in conclusion R and Shiny dashboards are a great way to communicate opportunities for quality improvement and help accelerate change throughout an institution. I think it's very effective to present complex data to build a consensus around change. The interactive plots allow viewers to engage with the data allowing them to explore on their own and the dynamic tables allow people to drill down into details and come up with their own hypotheses about what could be improved or how changes can be made. Really I think the fundamental opportunity here is for these tools to transform your audience into analysts and to participate in the change process with you. I'd like to thank everybody in the R and Medicine Committee for organizing this fantastic conference. I'd like to also thank thousands of people behind the R project and the R Foundation especially the developers and maintainers of the high charter reactable crosstalk and tidyverse packages. They've been instrumental in generating these sort of tools. Thank you very much.