 Alright. Well, hi, everyone. My name is Rebecca Ramesh. I'm a first year path resident at the University of Pennsylvania, and I'm excited to talk to you today about our method here of blood culture contamination rate tracking and reporting. Okay, so we'll start with some background. First, we'll address these three questions. What are blood cultures and why do we collect them? What is blood culture contamination? And why is it a problem? Why should two sets of blood cultures be drawn? And then next, we'll discuss how we calculate the blood culture contamination rate based on our data. And lastly, I'll show you what our blood culture contamination report looks like. A blood culture is used to help determine if a patient has a bloodstream infection. These infections can be life threatening. Blood is naturally sterile, so bacteria should not be present in the bloodstream. So because blood is normally sterile, positive blood cultures with a known pathogen have an overall high positive predictive value for infection. However, blood culture contamination is a significant problem. In the era of modern blood culturing techniques, virtually all blood culture contamination occurs during collection. The source of contaminants is usually the patient's skin or the hub or cannula of an indwelling cavitar. Frequent causes include poor collection technique and insufficient skin disinfection. So for this reason, the bacteria that grow when a blood culture has been contaminated tend to be skin commensals, which just means the organisms that can actually be found on our skin. So the consequences of blood culture contamination are significant. They include unnecessary antibiotic exposure and prolonged hospitalization because the clinical team is not certain if the isolated organism is a result of contamination or represents a real bloodstream infection that needs to be treated. So here's a list of common skin commensals, just kind of so you're familiar with some of these names. One thing to note is that while these organisms can be found on the skin as part of normal skin flora, they can also cause bloodstream infections, which is what makes identifying contamination difficult. So one set of blood cultures consists of two bottles, one is an aerobic bottle and one's an anaerobic bottle. The best practice is to draw two sets of blood cultures within 24 hours of one another. And this is for two reasons. First, we want to increase the sensitivity for detecting a true pathogen by submitting a large volume of blood for testing. But second, having two sets helps us rule out potential contaminations. So if a skin commensal grows in only one of the two sets, it's more likely to be a contaminant, whereas if it grows in both that we might consider a real infection. So the blood culture contamination rate is calculated by taking the number of blood culture sets with growth of a skin commensal without the same organism growing in the other set, then dividing that by the number of blood culture sets collected during that same period. A few things to note include that if a skin commensal grows in both blood culture sets, then we consider it a pathogen and it's not a contaminant. And if a true pathogen grows in only one of the two sets, it's still considered a pathogen, not a contaminant, because we would never want to see that organism. And then just so you're aware the laboratory standards suggest a goal of less than 3% for the blood culture contamination rate. I just want to take a brief minute to talk about single set blood cultures. So if only one set of blood cultures is drawn, which is different than the recommended practice of drawing two sets, then there's no set for comparison and we cannot consider contaminated, even if it grows a skin commensal. So for this reason single set blood cultures are bad practice, they increase clinical uncertainty, and therefore they lead to increased antibiotic use and prolonged hospitalization, and they also falsely lower the blood culture contamination rate. Next, I'm going to give you a sense of what our data looks like and go through a few different scenarios just so we can review what we just talked about. So in this example, if a skin commensal like staff epidermis grows in one of two sets, that set is considered contaminated. In this next example, both sets grow something called candidate parapsilosis, which is a true pathogen, but one of the sets also grows stuff epidermis. So this set is considered contaminated even though it also identified a true pathogen. If a skin commensal grows in both sets, then we're going to consider it a true pathogen and it's not their set is contaminated. If a true pathogen grows in one of two sets, neither is contaminated. And if a skin commensal grows in a single set, and only that one set is drawn, it's not considered contaminated just because we don't have any set for comparison to rule out contamination. So just, so this is kind of what our data looks like. And in the past, we had one of our lab directors manually go through an Excel file with a list of all the blood culture sets for the quarter and decide if each set was contaminated. So as you can imagine, at a big hospital, all the data from a quarter is like thousands of rows, so he would do this manually and it would take a long time. And then he'd send this back to a colleague who would add everything up and create another Excel file with the aggregate contamination rates for each of the hospitals and the system. And because this process was so labor intensive, it would take multiple people coordinating and multiple months after quarter and to produce the blood culture contamination report. So this project that I'm speaking about involved automating this process so that we could save hours and manual work and provide more immediate feedback to the hospitals and departments other blood culture collection practices. And I'll just briefly talk through our methodology. So we started out by making a table containing all of the potential isolates and whether or not they should be considered potential contaminants. So here's just an example, but we had hundreds of rows in this table. Next we converted the data from wide to long format just so that each blood culture set slash isolate had combination would have its own row so rather than having multiple isolates on a row we just wanted one isolate per row. And then we joined these two tables as seen below. And then lastly we count contamination events by looking at the number of patient date combinations where the number of times an isolate appeared was only one. But the number of sets drawn for that patient on that date was greater than one and isolate that grew is considered a potential contaminant. And then we divide this number by the total number of unique blood culture sets for the relevant time period. So here's a screenshot of one of our quarterly reports that we also generate monthly and yearly reports as well. We include blood culture contamination rate aggregates and trends over time floor specific contamination rates, a table of common contaminants, the single set blood culture rate, which we said is important. And we also provide reports for various departments that list the individual contamination events. So here's an example with an aggregate plot. And this is an example of trends over time so for example for the quarterly report we show the individual months and for the yearly report we show the individual quarters. Here we show the most. Oh sorry. This is a table that just provides a floor level data so basically for each floor or department it's telling them what their contamination rate was. And then here, we show the most common contaminants across the hospital system. And then lastly we show the rate of single set blood culture draws just so each department can keep track and as you can see this is actually a pretty significant problem at least at our institution between 30 and 40% of blood cultures that are drawn or single sets. So there's no set for comparison so we're probably over using antibiotics and causing a lot of prolonged hospitalization with this. And so what are the outcomes of this project. So we've been able to identify floors departments and individuals that need additional training and sterile technique. And also identify floors departments and individuals that are consistently drawing single sets. We can track common contaminants across the hospital system which you know might change over time might, might be able to even detect some sort of, you know, outbreak or trends. We significantly reduce the time and effort required when compared to the previous manual process involving multiple people. We provide real time feedback to floors and departments, and then we fulfill the cap regulatory requirement for tracking our blood culture contamination rate, and we're able to ensure that we're compliant with laboratory standards of having a blood culture contamination rate of less than 3%. So here are the our packages we use we created the template in our markdown and then we knit to PDF to make the report. We use knitter cable extra de plier. We'll use reshape to to go between wide and long formats for my data and gg plot to and then we just use the latex table format. I just want to thank Kyle Rodino and Catherine McCulley for their help with this project and related to the last session I was listening to it sounds like we maybe should have started this out within our package, but we're going to try to turn this into an R package so other institutions can use this template that we created. Thank you so much for joining us. Thank you, Rebecca.