 Hi everyone, I am Karan Vanood. Today I am going to talk about Mortality Minder, a shiny based open source tool for visualizing social determinants of mortality, developed by my team at Rensselaer Idea. Mortality rates have been increasing across the United States for the past several years. Figure 1 on the right describes the midlife mortality rates for deaths of despair. Midlife refers to individuals within the age of 25 to 64 and death of despair includes deaths due to suicide, substance abuse, overdose, etc. Basically, self harm. The x-axis is the range of years from 2000 to 2017, while y-axis is the midlife deaths per 100,000 people. We observe that the mortality rate has increased by a whopping 90.4% from 2000 to 2017. This is very alarming where the county and state rates vary by communities, socioeconomic factors and regions. There is a need to do something about this and thus we decided to create Mortality Minder. Mortality Minder is an open source, web based visualization tool that utilizes our shiny and full-page JavaScript frameworks for interactivity. It explores social, economic and geographical factors associated with premature mortality among midlife adults across the United States. Mortality Minder was submitted to AHRQ's visualization of social determinants challenge, which required developing a visualization tool to enhance research and analysis of community level health services. We not only cleared phase one, but we also secured the third position among all the 12 semifinals. Mortality Minder provides a geographic view of trends as well. Figure 2 shows one such case for the mortality rates of deaths of despair from 2015 to 2017, where the darker color represents a higher mortality rate. The goal of Mortality Minder is to enable health care researchers, providers, payers and policy makers to gain actionable insights into how, where and why midlife mortality rates are rising in the United States by investigating cancer deaths, cardiovascular deaths and deaths of despair. It will help health care payers and policy makers as the national, state, county and the community levels to identify and address unmet health care needs. Further, it aims to support analysis by healthcare organizations for development of programs, policies and procedures to improve longevity. The data presented in Mortality Minder is obtained from several data sources. Mortality rates from 2000 to 2017 are obtained from CDC wonder. Eight specific mortality rates were calculated in three-year chunks for each cause of death. For completeness, the suppressed county data was imputed with mortality rates of the state. Social factors for the data 2015 to 2017 was obtained through county health rankings, an aggregate of county-level data curated by the Robert Wood Johnson Foundation. We gathered factors addressing health behaviors, clinical care, education, employment, social supports, community safety and physical environment domains and filtered them down to 70 factors that were relevant to at least one cause of death at the national level. The counties are clustered into risk groups using the K-means algorithm, which also smooths out inherent noise in estimated mortality rates. The risk groups represent groups of counties with similar mortality trends by the mean. After extensive analysis, we decided to cluster the counties into three risk groups, low, medium and high for each state. If the state has six or less counties, we considered it an exception. Clustering is not done and mortality trends for each county are shown separately. In figure three, we can see the deaths of despair risk groups from low to high for Ohio. The same geographic plot can also be obtained for all other states. For United States, we grouped the counties into six different risk groups. Yellow means less risky, while red means more risky. Figure four shows the risk groups for deaths of despair across the United States. The red regions in the west indicate that the counties are high risk in comparison to the counties in the east. As mortality rate is centered around identifying factors that cause mortality across counties, we use the Kendall-Tau test to identify the correlation between factors and risk groups. The Kendall-Tau test measures the correlation between two measured quantities. The correlation values ranges from minus one, indicating the two quantities are dissimilar to plus one, indicating that the two quantities are similar. The correlation is used to identify relevant factors for state and national level. In mortality-minder, positive correlation implies potential destructive determinants of health, where a greater value implies greater association. These factors potentially hinder better health in the given region. Figure five shows the potentially destructive factors for California. These range in decreasing order of destructivity as food insecurity, rural area, non-Hispanics, unhealthy days, and so on. We refer to such a plot as the lollipop plot. Similarly, negative correlation implies potential protective determinants of health, where a lower value implies greater association in the negative way. These factors potentially support better health in any given region. Figure six shows that the potentially protective factors for California include Hispanic, not being fluent in English, and so on. The application is designed in R Shiny and was accompanied by a number of very useful R packages and other frameworks that made the tool interactive, appealing, and informative. We combined R Shiny with styling and scripts from full page, which enabled us to create full-screen web applications with scrolling pages. Each page is connected by a common state and cause of death and includes touch and click horizontal scrolling. We use leaflet and ggplot R packages to create stunning visualization in the form of geographic plots, trend lines, factor lollipop plots, and more. The state and county shape files for each state were obtained from the Tigris R package. FortalityMinder is developed as a four-page full-screen application. The four pages are nationwide view, state view, factor view, and about. Figure seven shows the home page for MortalityMinder. The national view highlights mid-life mortality rates across the United States. It depicts the mortality trends as geographic plots of all counties. Further, it compares a given state's mortality trend to the nation's over the years 2000 to 2017. As we can see in the figure number eight, the mortality rate of the United States is compared with California. California has performed better than the United States on average in the years 2000 to 2017. The mortality rate in California rose by 44.3%, which is less than half the national rise of 90.4%. Next, the state view categorizes the selected state into risk groups as we discussed earlier and plots in a ten plot. We can select a given county in the state and it will compare its rate to the state's average. Figure nine shows the risk group's mortality trend for California and compares them with the national average depicted by the blue line and the selected Los Angeles County. We observe that Los Angeles County did better than all risk groups on average. The state view also visualizes a lollipop plot showing significant destructive and protective socioeconomic factors like we saw before. Finally, the factor view takes a deeper dive into those socioeconomic factors. We can explore individual factor here with their definition, identify quantitative protective or destructive relationships between factors and counties and data source for each of those factors. Figure 10 shows the food insecure factor distribution across various counties in California. The factor view also uses box and scatter plot to show the distribution of factors across counties and risk groups. The last page in the application is the about page, which details all the information about mortality finder, including innovation, insights, implementation, useful links and more. Further, as we can see in figure number 11, there are options to download the data set that we actually used for the mortality rates and the factors. Our results in the form of clusters and correlations can also be downloaded for future exploration. And now let's dive into the application and see it in action. This particular nationwide view allows you to view the transition of mortality rates with time. So as you can see for the year 2000 to 2002, the mortality rates were very low all across the United States, depicted by light colors. But as we progress in time over the years, we see that the by the year 2015 and 17, almost all the regions are darker, creating pockets of high mortality rates. Similar information is also depicted in the form of this trend plot. As you can see the blue line shows that over the years 2000 to 2017, there has been a more than 90 percent rise in the number of mortality, the mortality rate in United States. We, as we selected California, can compare this to the California rate, which was only about 44.3. Now mortality minor also allows you to deep dive into each state. So let's go and check the state view at the very top you can interact with this plot zoom in and look at the various colors. So the darker the color indicates higher the mortality rate, where lighter color indicates lesser mortality rate. We can choose various time periods here as well. And this plot shows the three risk groups that we discussed before low, medium and high. As we can see some areas are high risk, while other regions are low risk. And this can be seen as a transition between years in the form of this trend plot. The yellow line depicts the lower risk group, the orange line depicts the medium risk group, and the red line depicts the high risk group. The blue line here shows the national average. We can go ahead and select a county of choice. For example, we select this county and we see that a new line shows up on the trend plot. This means that this county is performing worse than the average high risk group, as well as the national average. But if we scroll around and say select Los Angeles, we'll know that the line has shifted to the very bottom indicating it's performing significantly better, even better than the national average. This plot on the very right gives us the destructive and the protective factors associated with this particular state. As you can see, food insecurity, rural, non-Hispanic, wise, mentally unhealthy days are some destructive factors, while constructive or protective factors include food environment index being in Asian and so on. Mortality monitor enables you to go even one step further, that is, you can deep dive into a given factor. Let's say we want to explore food insecure. If we can directly click on here and reach to this factor page. You can select any factor that you like. Here we have selected food insecure. We get a definition of the factor. We get an explanation. We identify its text source. We identify the data source where it came from, as well as whether it's protective or destructive. We can change the county as we like. And the given factor is correlated to counties in the form of shades of different colors of blue shown by this particular plot. We can zoom in and look at the each county name as well. The food insecure and mortality relationship for California is shown by this scatter plot. This diamond shape shows Los Angeles County where in general the trend can be seen as going upward in this particular direction. The risk groups are also depicted in the form of box plots to highlight the width and the distribution of each of these risk groups. Similar to the previous factors lollipop plot, this is exactly the same plot allowing us to keep looking at the factors that we might want to explore. This brings us to the last and final page, which is the mortality minor about page. It tells you what mortality minor is, what are the innovations that we did, risk groups and the various views we created, the insights that we drew as well as how we implemented everything. By developing mortality minor, we learned a lot about application development, healthcare and mortality trends and we were ready to take it to the next come COVID minor. A pandemic application that studies and reveals regional disparities of COVID-19 across the United States and New York. The information is split into outcomes which are the direct effects of COVID-19, determinants such as diabetes that impacted these outcomes and mediations which are resources to combat the pandemic. Figure 12 shows the homepage for COVID minor, which provides a state report card, a one stop solution for extensive COVID innovation for that particular state. Thank you for joining in today. If you have any questions for us, please reach out to me through email.