Hello, and welcome to today's webcast, Making the Data Shine, Using Data Visualization to Enhance Research Publications and Presentations. My name is Paul Jung, and I'm a professor of pharmacy practice at St. Louis College of Pharmacy in St. Louis, Missouri, and I will be monitoring today's webcast. The recording of this webcast will be made available within 5-7 business days. Log on to mysccm.org and navigate to the My Learning tab to access the recording. A couple of housekeeping items before we get started. There will be a Q&A session at the end of the presentation. To submit questions throughout the presentation, type into the question box located on the control panel, probably to the right of your screen. Please note the disclaimer stating that the content to follow is for educational purposes only. And now, I'd like to introduce you to your speakers today. First, we have Dr. Todd Brudder, who is a clinical assistant professor at University of Rhode Island in Kingston, Rhode Island, and Dr. Mohamed Mamou, who is assistant professor at West Virginia University School of Pharmacy in Morgansville, West Virginia. And now, I'll turn it over to our first presenter. Good afternoon, everyone. Before we begin, I first would like to take this opportunity to say thank you to SCCM for the opportunity to speak with all of you today regarding the importance of making your data shine to better engage your audience in the science of data visualization. So, before we begin, let's start with our learning objectives that we'd like to achieve by the end of our conversation today. We want to define and examine the importance of data visualization in critical care research. We also want to define and differentiate among various classifications of analytical data. And lastly, to compare and contrast the strengths and limitations of individual data visualization platforms. I believe when your audience is better informed of the important data metrics shared in a clinical setting, it will lead to improved decision-making, which translates into better patient outcomes. As we now live in a digital age, transforming healthcare using digital health has improved the quality delivered by reducing waste of profoundly limited resources, minimize professional burnout, and regain balance in healthcare resource distribution with the goal to improve outcomes by ensuring compassionate, equitable care. There are many ways in which we currently use healthcare analytics to improve our decision-making. Beginning with the top left spot depicted in blue, we are able to decrease the cost to develop medication. For example, convolutional neural networks have been able to predict the binding of small molecules to proteins by analyzing little hints from millions of experimental measurements and thousands of protein structures. In fact, in 2015, during the West African Ebola virus outbreak, AtomWise partnered with IBM and the University of Toronto to screen the top compounds capable of binding to a glycoprotein that prevented Ebola virus from penetrating into cells in vivo. This AI analysis occurred in less than a day, a process that would normally take months to years, if ever, which enabled the development of a treatment for the Ebola virus. Further, AI algorithms are able to identify new drug applications, tracing their toxic potential, as well as their mechanisms of action, which is quite an appealing strategy for big pharma companies, since it is less expensive to repurpose and reposition existing drugs than to create new ones from scratch. Also, analytics can forecast kidney disease and early prediction of AKI. In 2019, the Department of Veteran Affairs teamed up with DeepMind Health and created a machine learning tool that can predict AKI up to 48 hours in advance. This AI tool was able to identify more than 90% of AKI cases 48 hours earlier than traditional care methods. So therefore, most recently, medication and medical imaging analysis has been considerably improved by detecting relevant diseases and presenting them to radiologists in a user-friendly view, which enables the design of more customized targeted reporting for better diagnostic decision process making. Lastly, the ability to analyze unstructured data from electronic medical records helps identify patterns that would otherwise go unknown. All of this wonderful technology will support health equity by removing biases, by promoting data transparency, and diversity. In 2013, Sahari Collins stated, data science is the study of generalizable extraction of knowledge from data. Therefore, data science has been defined as a multidisciplinary field that combines statistical modeling, specialized programming using data technology, and advanced analytics, such as AI or artificial intelligence, and machine learning. Data is then presented in a graphical and visualized manner to uncover actionable insights hidden within big data. Ultimately, in the health sector, we want it to empower clinicians such as yourself with actionable insights armed to make more informed decisions, ultimately leading to higher quality, more efficient, and safer care. Newer learners to the world of data science can likely benefit from reviewing a bit of nomenclature to better understand the components of data science. It can be thought as an umbrella for all disciplines to make sense out of large volumes of commonly termed big data. Machine learning is a set of methods, tools, and computer algorithms used to train machines to analyze, understand, and find hidden patterns within the data and make predictions. More simply put, simply learn from the data. In a supervised learning model, machines are trained to find solutions to a given problem with assistance from humans who collect and label the data to feed the system. The machine is then told which data characteristics to look at so it can identify patterns, putting objects into corresponding classes, and evaluate whether the prediction is right or wrong. Conversely, in unsupervised learning, machines learn to recognize patterns and trends in unlabeled training data without being supervised by learners or users. Data mining, on the other hand, is comprised of different techniques and use of tools to extract new, insightful information from a large set of data of previously unknown modern data technologies. Therefore, this concept is able to find hidden patterns and oddities that reflect the multifaceted relationship within raw data. Data visualization, when it's done well, provides a quick, easy way to convey concepts and summarize and present large data in easy-to-understand and straightforward displays, which enables readers to be insightful and gain worthwhile information. Good visualization tools' benefits include communicating your results or findings to tell a better story, support the monitoring of the model's performance at the evaluation stage, identify trends and patterns, correlations between data set features, data cleaning, such as outlier detection, and validating model assumptions. Here is a great example of a patient experience. This is a value-based care dashboard supported by data science. Value in healthcare is traditionally defined as health outcomes, or what I call quality care, with the achieved per dollar spent, or the cost of care. Healthcare dashboards depicting patient experiences, such as this one, which is an example of an ER wait time, for example, can lead to positive change and other implications that may otherwise go unnoticed. Dashboards seem to be the current trend, especially within acute care settings, particularly for medical administrators, to graphically depict, for example, in this graphic, the ICU bed capacity in Europe during the COVID-19 pandemic. The ability to visualize this vital information supports stakeholders to evaluate the healthcare landscape on a minute-to-minute, real-time basis. The last graphic here is what we call a heat map. This is from the CDC, and it will help identify U.S. obesity trends and patterns that can help raise awareness about modifying the lifestyles of patients, as well as to the general public. In the right-hand graphic, depicting opioid-related hospitalizations in the U.S., you can see that the darker colors refer to the higher incidence by state in which opioid-related hospitalizations occurred. The suggestion here would be states that have a higher prevalence perhaps should consider reviewing their dispensing policies, their regulations, compared to the lighter-colored states, maybe to help provide insight into the reasons why their hospitalization rates are considerably higher. Healthcare providers can use machine learning algorithms to identify patterns, correlations, and large data sets. To demonstrate how machine learning can be used for data visualization, let's consider a scenario where a healthcare provider has a large data set containing patient records, and they want to evaluate or predict heart disease risk, for example. This data set will contain both static and dynamic information, such as demographics, medical history, lifestyle factors, and laboratory results. The multidimensional patient data gets analyzed through a multifaceted neural network and identifies that patients, for example, over 50 years old, with high blood pressure, use of tobacco, have a much higher risk of developing heart disease. To better visualize this information, the health system EHR provides an interactive dashboard for the clinicians that displays the crucial risk factors for heart disease. The dashboard could contain charts and graphs that show the distribution of risk factors among the patient population. The dashboard might also have interactive filters that allow the user to dive deeper into a specific subset of data or patient phenotype grouping. In addition, the dashboard may, most importantly, include a predictive model that uses machine learning to estimate a patient's risk of developing heart disease based on the specific factors. Thereby, by visualizing this data, healthcare providers can quickly identify patterns for who's at risk for heart disease and develop targeted interventions to reduce their risk. They could also track changes in the patient population over time from an epidemiologic standpoint to further assess the effectiveness of the intervention in lowering the incidence of heart disease. The question I pose to all of you now is, how should you go from obtaining the data to the actual visualization process? Depicted here on your screen is a process that involves five key steps. First, the process should begin with the data cleaning process. Detection of outliers is vital to ensuring the accuracy of the results. For example, scatter plots can be used to detect outliers, and use of heat maps, like I showed you previously, can also check for multicollinearity. Next is the data exploration phase. Before you build any model, exploratory data analysis is required to identify key data set characteristics. Histograms can be used for the continuous variables to examine normality within the data. Scatter plots can be used to examine correlation between two features. Next, the evaluation of modeling output can be used with the confusion matrix and the learning curve to measure the performance of the model during its training. Next, identifying trends is essential and can be examined using time and seasonal plots in a time series analysis to identify unique trends over time. And lastly, and maybe arguably most importantly, presenting your findings to others that are less familiar with the data should be explained in relatable terms using informative plots that summarize your findings. I will now turn the presentation over to my colleague, Dr. Al-Momoon. Thank you, Dr. Brothers, for the great introduction and gluing data science with these data visualization techniques. And I also would like to thank SCCM for this opportunity to present on this topic. And I would say this is a timely topic. So why data visualization is important. So data visualization is important, not only in machine learning project, not only the projects related with AI. It is extremely important that we visualize our data from the very beginning, from the data cleaning stage. So what it does, it helps people to see, interact with the data that they have. And also it gives us a better understanding for exploring the data and analyzing the data and presenting it to the audience. So for data visualization is useful for data cleaning, exploring data structure, detecting outliers and unusual groups, identifying trends and clusters in the data set, spotting some local patterns in different variables, and also evaluating modeling output. If we do any statistical model to plot that, we need the data visualization technique and definitely presenting the results. And what types of data we have? We have two types of data. One is quantitative data. Another one is qualitative data. So we all may know quantitative data is a tabular data where we have numerical variables and we have categorical variables and we have different variables which tells us the measurements of the clinical practice or measurements of the other fields. But in qualitative data, we call it as text data, free text data. And in our practice, we call clinical notes as open-ended data. So where we don't have actually a tabular format of the data. So the purpose of the data visualization is to do exploratory data analysis. And if we want to create a data-driven tool. So the data-driven tool would be based on what data is showing us, we make a modeling tool and then plug into the traditional EHR systems. And the conceptual tool, the conceptual way of data visualization would be presenting the statistical models and how the data is fitting well with the, sorry, how the model is fitting with the data, what we have. And all we do just to make sure that we present the clarity in our hypothesis. So whatever the research hypothesis we have. And then increasing the precision of those measurements and also efficiently use those techniques, those visualization graphs to support our hypothesis. And, you know, at the end, we want to minimize the ink because if we make a table, probably a graph would be better than a table. Just to provide the visual patterns to the audience. So what do we detect in the data? So the purpose of the data visualization is to detect the pattern, trends, and correlation. And in other words, we want to show the association that we want to make in our research hypothesis. So in my next slide, this slide is very important because, you know, in next few slides, what I'll try to do, I'll try to connect the basic examples of data visualization to our scientific research. So before we even we think about our data visualization tool, we need to make sure that we understand that we need to select the appropriate data for our aim. So unless we select an appropriate data, we will not be able to answer what we actually ask originally to our data. So and also we need to think about who is our audience. So here audience is a term which can actually define that, OK, are we presenting these data or these research hypothesis in front of the clinicians, in front of the stakeholders, in front of the pharmaceutical companies. So based on that, your visualization will change. Also, if you think about that, OK, I'm presenting these as a part of a poster presentation. And we should be different than presenting as a podium presentation. And also it will change probably with the visualization will change when you present it as a manuscript or scientific paper in a scientific journal. So and the next thing we need to consider in our in our mind that what would be the structure of the visualization? So what do you mean? What do I mean by structure? So what kind of visualization would be appropriate? Is it a pie chart? Is it a histogram? Is it a density plot? What will fit my data? So then I can show that exact correlation, exact association between the variables. So the next thing is the demonstrating the relationships. So which variables we should choose to show the relationships. And, you know, believe it or not, in our clinical data, we have tons of variables that we can extract from the EHR data. And if we test each one of them in a statistical model, it would be impossible to feed in a statistical model. So it is very important to to look these data critically to see which relationships that we want to present. And the next thing is the scale of the visualization. So suppose if you have multiple plot in multiple plot in one figure, we need to see what are the scales they have so that we can merge them. Or if it is not mergeable, so that how we can present those, you know, scales, scales for different plots in one figure. So and and and relation relationship with that in relation with that, we need to define that, OK, whether our plots are univariate plot or bivariate plot or multivariate plot. So this is very important in terms of data visualization. So once we go through the data during the data cleaning phase, we need to we need to determine all of these. We need to think about all of these prior even in a plot of data. So what are the basic elements we have? Again, I'm going through very basic, basic structure of the data visualization. The purpose here is to connect that how we do data visualization efficiently and effectively. So we all know about this, you know, that this plot, this is our Y axis, this is X axis. And and there are different data elements we plot here. And we have limited and limited aesthetics for for data for plotting a data. We have definitely we have this position. We can have different vertical lines on it and we have different shapes. We have different sizes. We have different colors, line widths and line type. So, you know, it doesn't matter that which data visual visualization tool we we use to create this visualization. That tool will have all of these aesthetics built on it. So moving forward, I want to show you a real example. So I try to make a very, very naive example here using the enhanced data. So the enhanced data is National Health and Nutrition. Nutrition examination survey is done by HCA project from HRQ. And it it actually collects it actually does run a survey to collect the health and nutritional status of the adults and children in the United States. So I limited my data sets into a few variables so that it does not confuse you. And here we have ID, we have age, we have age group, we have race, gender and we have systolic and diastolic blood pressure. So I in next few slides, I will I will show you some of the basic flaws that we can make with that. And how do they make sense in terms of presenting these data points effectively and efficiently. So moving to my next slide. So first, probably I want to, you know, I want to check the relationship of the age group and the gender. So basically, I want to know the percentage of people stratified by gender for different age groups. So you can see in my Y axis, I have counts and my on my X axis, I have male and female. And that is stratified by age group. So you can you can use your question chat box. And then and then can you tell can anyone tell me what is the problem with this plot? Like, are we are we getting the information that we supposed to get from this plot from using this bar chart or is there anything missing? So you can see in in this Y axis, we have just count. So these are just raw counts. So it would be it would be really effective if we have put percentages here. So even it is a simple graph, but you can see a simple bar chart. But if we have percentages of these age group, that will actually give us more more clear picture that, OK, this is how these distribution of female male and female among these age groups are there. So basically, if our Y axis would have been a percentage of these people, sorry, percentage of these groups, that would have been better to present these and the next relationships I want to see. I want to see a scatterplot. So now we are looking at the now we are looking at the the the basically the systolic blood pressure by age stratified by gender. So what what a scatterplot does scatterplot actually give us whether there is a pattern or not. So and you can see in the Y axis, we have systolic blood pressure and in X axis we have age and it is stratified by male and female. The purpose here, I want to see the variation in systolic blood pressure among this population stratified by gender and for four different ages. So can anyone tell me what is the problem in this chart? Do we see any pattern? So very busy slide, maybe younger females with lower systolic blood pressure. All right. So, yeah, you're right. It's very busy. And these dots are actually overlapping each other. It's very difficult to understand what I'm trying to achieve from here. Another response for me, the mixing of male and female makes it difficult to see the patterns of the data might be better with the trend line. Yes, I really agree with you, all of you. So let's let's plot it in a better way. So let's move forward. And now what I did, I make an age group and now I'm plotting them in a scatterplot. So now we can see the distribution, the pattern a little bit better. I would say 90 to 40 age groups, they have their diastolic blood pressure has a little bit shrinkage pattern, like, you know, the pattern a little bit shrink. And you can see there's a variation in 65 to 89. Are you happy now with this plot or do like or is it more confusing? Do you see any problem in this plot? Is it telling us what is the distribution or what is the pattern among these age groups? Again, you can use your chat box. Is it better now? OK, yeah, I do agree that previous plot is showing us the higher values with the increase in age and but that is lost here. Yeah, I do really agree with that because I made the age group and it is better, but still busy, needs a trend line better than last one, but can can be made better. OK, so someone suggested using a box plot. Yeah, you actually are right. If we make a box plot, that will actually give us a better idea of what's going on with this age group, what's going on in terms of systolic blood pressure and diastolic blood pressure for these age groups. And also we can stratify it by gender. I cannot see who actually suggested that, but that a great, great suggestion. So see the process of process of, you know, depicting this data into visualization. It requires our medical knowledge. It requires our practical knowledge that, you know, whichever professional practice we are in. Actually, it requires a lot of things to think about how effectively we can visualize that. So what what does the box plot do? So basically box plot displays for five five number summary and you can see its minimum. So this is the minimum and and we have first quartile and then we have median. This straight line is median and then we have third quartile and then we have maximum. So in box plot, we draw a box box from the first quartile to third quartile and the vertical line goes to the box with the median. So basically box plot tells us how how how my data is spread out and and how they look like and what do they mean and do do we have outliers. So this is a better presentation after trying at least two or three different graphs. So and we can we can make it a little bit better. And if we move to a plot called violin plot. So what does a violin plot tells us? A violin plot actually gives us the peaks in the data. So basically it is used to visualize the distribution of the numerical data. So if we take box plot, we box plot shows the summary of the statistics, but violin plot actually give both. So if you see in this violin plot, these are smooth. So basically these data points were, you know, were made smoother so that we can see how the data is distributed. So basically the wider these violins are the yeah. So another I just I just see in the chat box that another suggestion was violin plot. Yes. So exactly. You know, you read my mind and you can see that the more wider these plots, it is showing that more data points are in there. So in that way, you know, again, this is very basic. I did not have any hypothesis to make these flat plots. But what I wanted to show that for different clinical for different vitals for different lab tests, we can actually plot this kind of this kind of plot to see how they vary by different groups. So you can see that how violin plot is helping us to show the density. So if even we want to make it more, much more better, we can plot we can actually move to a density plot. So what density plot does, density plot can be seen as an extension of histogram. And now in this plot, what I'm plotting, I'm plotting the distribution of their diastolic blood pressure by the race so that I can see what are the distribution of the diastolic blood pressure in different races, different race group, and then we can do it for systolic blood pressure. So what a density plot does, it visualizes the distribution of the data over a given period of time and and peak shows that where the values are concentrated. So basically, you can see if there is a variation in the distribution of of this systolic and diastolic blood pressure in in different race groups. So so next, I will show you. So we can actually complicate our visualization by increasing the number of variables. So basically, that's what I initially I say that we can you can plot univariate. You can plot bivariate that's been using two variables, you can actually do multivariate. So now in the next plot, what I'm going to show you what I what I did here, I actually divided into two groups. So I define hypertension. And then I said, OK, now I want to see their age distribution by whether they have hypertension or not, basically hypertension and normal. And then I want to see how that changes in different race. So again, based on the hypothesis, I think this would be very useful based on the disease, based on the condition, especially I know if a person have, you know, AKI, AKI, a person have septicemia and the person is in critical care, these these values are important. And again, these x axis and y axis, we can put our own variables to see how they're changing. So the purpose of this, this is a multivariate plot. The purpose of these to show that how we can show effectively a lot of information using a simple and effective plot. So so far, what we have done, we have we have seen some plot. And at the end of this plotting, what do we actually decide? We actually decide how we can compare two groups. So here is a summary. Again, this is I did not have any hypothesis, just I wanted to show you that how this data summary works for both of these groups. And, you know, that's how we present in the scientific literature in the poster. So moving forward now, I will walk you through that, how actually we understand how to better plot. So can anyone tell me what is the problem in this plot? So again, just to give you a little bit background. So here we are looking at the total dispensing rate of formaldehyde and gabapentin, gabapentin in a state prescription drug monitoring program where I plotted them yearly. Can anyone tell me what is the problem in this plot? This is I tell my undergrad student, grad student and my colleagues, whenever you are plotting, when you are plotting. You need to put your X axis label, Y axis label so that it makes sense to the people. Yes, the data is not normalized. And also you can see these are just raw rates. These are these are not normalized based on the population. So I really appreciate your appreciate your your input here. So actually, this plot, you know, this plot would be much more valuable if you present in a poster or in a scientific paper to add those elements in there, then we call it as an effective plot, effective plot. The definition of effective plot is not only that you use a lot of complex data elements in there, even a simple visualization plot, you can make it effective. So moving forward, I want to show another another another, you know, another plot, which is called heat map. So this is from one of my published paper, one of our published paper that Dr. Brothers and I worked together. And you can see. So what is this MRCI? So this is medication regimen complexity. This is the index. And we wanted to see how the value by low and higher MRCI and, you know, stratified by different different agents. So and, you know, at the end of the day, a chart should tell a standalone story. It should not depend on any other elements of the data. But whenever you are whenever you are making a chart, you should think about, OK, am I telling a good story with this particular plot? You don't need to say a lot of story with one plot. But I think it is better that your plot talks by itself. Like you don't need to depend. You don't need to support like support that plot using other tables. So basically, a plot should be independent. The reader should understand what is in that plot using, you know, using the elements of that plot. So moving forward. Moving forward, that what makes a good plot. So it is the preference. It is the preference of the presenter. And always think about that. Is it better to use a table rather than a figure? Because sometimes if it is multivariate plot, a figure can be difficult to interpret. And as I told you previously, that matching scales and alignment of the graphics are important. If you are putting multiple plots in one figure, it is important to match those alignment. And then you can use log scale. And if the data is not normalized and inclusion of the reference line labels, notes and keys, these are very important aspects and important tips for the, you know, effective plot. And again, color is another element that everyone, you know, everyone uses different colors in their plot. And you need to make sure that whenever you are making plot, that you need to make sure that it matches with your background color. If you are presenting in a podium presentation or in a poster. And also, you know, do not change color too much, intensity of the color too much in your scientific literature. Because, you know, if you're presenting systolic blood pressure in one color in one figure, in another figure, you change the color or change the intensity, actually it will confuse your readers. So basically, this is what I said, that, you know, you have a lot of different options for different colors, but that does not mean that you will choose them. So there's a standard guideline for using this color in the scientific literature. So you can follow those guidelines. And moving forward, this is my opinion, do not present any 3D visualization in your scientific presentation. What happened? 3D visualization is really cool, but they actually skew a lot of values and it can lead to very difficult interpretation. Because most of the time, this 3D visualization is presented in a slide when the slide is 2D. So we don't have glasses, you know, glasses on, 3D glasses on, and we cannot see, even we cannot move those slides to see their 3D visualization. So this is what my suggestion is that do not put any 3D visualization in your scientific literature. So moving forward, what sort of software you can use to plot? So I actually plot all of these, all of this plot that I showed you, I use ggplot. So ggplot is a standalone package where you can use it in R, you can use it in Python, you can use it in a lot of other softwares. Plotly is similar. You can use Rshiny, Rshiny is based on R, where you can actually make interactive plots. I will show you an example. You can also use another package called matplotlib. So these are the open source package where you will be able to, you'll be able to plot these, you know, your data, and you just need to install it again. You know, these are not difficult tasks to plot. You just need to get along with these softwares like R and Python. And there are some commercial softwares, everyone knows about Tableau, and Google started their Google Charts, and first two of them can be free. If you are in an academic institution and in a limited fashion, and we have Microsoft Power BI, we have other softwares like Domo and Infogram. So these are paid softwares, but a lot of companies, they use a lot of healthcare organization they use. So, but for me, I think I prefer more open source packages. So now I will show you CDS as an example, where we have, I will show you our Shiny platform, where they actually, you know, make a visualization tool for understanding antimicrobial use in inpatient, you know, inpatient admission. Yeah, it's a filler here, Dr. Amul, to help support there while we're waiting for it to load. One of the things that's really valuable to us as clinicians is we have so much EHR data that we really try to make sense of it and to find trends and be able to really evaluate how long is my patient as a group, for example, MSSA bacteremia, how many days of hospitalization is required, what antibiotics are they on, what's the dosing, you know, commonality. So here, Radar uses an R Shiny platform that really looks at granular EHR data and helps to graphically distribute and give us the data in a meaningful way. Yeah, and also you can see that we can see different antimicrobial use here. And you know, and also we can see the, you know, the daily dosage per 100 beds per month. So in that way, we are able to, we're able to understand the use of different antimicrobials in a hospital system. And also we can look at some diagnostics. So basically, we can see how frequently, you know, the diagnostics were performed, you know, among this population, and there are different distribution. And on the left panel, you can actually choose what actually you want. So you know, which can support your practice. So basically, if you're interested to a particular age group, if you're interested to particular antimicrobials, you can select your antimicrobials here, we have a list of antimicrobials here. So if you see here, we have, you know, the whole list of antimicrobials here. And also, you can look at different outcomes here. So basically, what are the length of stay, if a patient was prescribed these, if a group of patients was prescribed this antibiotic for particularly that long, you can see their length of stay. And also you can see their length of stay by gender. So this is one example. And my next example is from CDC. So CDC, this is again, CDC Atlas, and you know that this is one of the great project by CDC, and who actually, you know, make this data available to, you know, to all to the to all researchers that you can actually choose by different diseases, different infectious diseases here. And also you can see the distribution of them for, you know, in different states and how do they change, you know, in different state and region. And also you can do it by, you know, gender, you can do it by race. Another tool I wanted to show you, which is a web-based injury statistics and query reporting system. And they have different tools here, you can actually look at different tools here. And you can see if you click this one, it will take you to the fatal injury data. And they have these live interactive dashboard where you can actually go through and then see what's going on in there. I added this link here, feel free to explore this. And then the last thing I wanted to, I want to say thank you for this opportunity to speak about these data visualization tool. If you have any questions about this, feel free to, you know, reach me out. Thanks everyone. Thank you. Thank you for a great talk from Dr. Brothers and Dr. Almohan. Any questions that anybody will have, please feel free to enter within the question section and go to webinar. I guess one question I have for both of you, for someone novelist in terms of researcher, in terms of presenting data, is there anything you guys would suggest I could turn to to help them? This is a great question. And do you mean that any tool? Yeah, just any tool, any book, something like that. Okay, so it depends on what you are practicing. So the basic tool is Excel. So this is the one that we use day to day life. But if you're a little bit data nerds, and then, you know, you like tools like R, it's very easy to learn. So you can start with Excel plotting simple things in there. And there are also statistical software that you can use. But again, as I told you, in that slide, that particular slide, that actually depends what you want. If you want to present these data in front of your, you know, project manager in to do a quality improvement, actually, first thing you need to think about, think about, okay, do I have an appropriate data? And then you can actually accomplish more, accomplish these visualizations using Excel as well. So Excel is not, I would say Excel is much more friendlier than, you know, some of the tools, but it is not the fastest one or not the quicker one, I would say. And I don't know, Dr. Brothers, do you have anything to add? Yeah, I was gonna say, thank you for that question. One thing I think we're all struggling with out in the health informatics and, you know, this type of data visualization ability is the, just what it is, it's the ability to get open source data. So to be able to create prediction tools, it really requires substantial hundreds and hundreds of thousands of data points. I'm not sure, I'm familiar with some of the packages out there. They're extremely expensive to be able to get our hands on some data. So in the critical care world, for those folks that are unaware, there is a MIMIC, it's called MIMIC, M-I-C, MIMIC for data set, that is open source data that we commonly use in some of our research as a comparator group. So I really think the biggest challenge we have right now is just the access of this big open data. Thank you guys for that great answer to that question. Any other questions that we could ask our panelists today, speakers? I would say if I could maybe in the meantime, as folks are entering stuff into the chat, one thing I just want to encourage everybody listening is, you know, think of the end user. So when you're trying to construct your data set and the answer that you're really trying to obtain, folks in the room, you know, administrators in the room, we kind of know what they're most concerned with. So it seems kind of obvious, but I'd be remiss not stating it, being cognizant of what your data should be showing and what the, you know, the folks in the room are really interested in is such an important way to engage and kind of carry the ball forward. So that's where I really find that data, good data visualization tools, such as what we showed you today, hopefully will help you kind of convince the C-suite at times for some of your projects and to improve patient care. And just to add with these, actually it depends on the nature of the audience who are sitting in front of you. Suppose, I want to give you an example. So you can present, you know, a data, you know, a bivariate data using a bar chart. But if you're going to a regulatory meeting, I think using pie chart would be better because they're, they actually, you know, they are, so pie chart is much more easier to interpret compared to, you know, bar chart or histogram. So I would never prefer a histogram in front of the regulatory people because, you know, some of them may have been very difficult and not, and I would not prefer a density plot either because, you know, I would prefer which is easy to consume in front of the audience. So if it is regulatory or if it is, you know, who are naive of, who are not from the clinical world or who are not in that, you know, particular field that you are from. So basically deciding your plots based on the audience is another, you know, another important aspect. I don't see any other questions in the question box. Otherwise, if there's any other questions, feel free to enter now. Otherwise, that concludes our Q&A sessions. Thank you, Dr. Brothers and Dr. El-Mohamed for your excellent talk today. And thank you for the audience for attending today. Again, the webcast is being recorded. The recording will be made available to all registered attendees within five to seven days. Again, log out into the mysccm.org and navigate to the My Learning tab to access the recording. And that concludes our presentation for today. Thank you.

data visualization

research publications

communication

visualization tools

types of plots

open source software

clear labels

tailored audience