Hello! I’m Lillian Jensen. In this video I will be presenting some of the work products that I developed this summer during my practicum experience at Kent County. I am super excited about what we were able to accomplish. First, a little about me. I have a bachelor's degree in Microbiology from MSU and I have professional certificates in advanced data analytics and data science. My background is in biomedical research. After I received my bachelor’s degree, I spent several years researching the healthy human gut microbiome and its protective effect against enteric diseases. In this picture, you can see me working in a body-temperature anaerobic chamber - I was either inoculating bioreactors with healthy human fecal samples, or I was plating those samples. This work was often very hot and sweaty. In addition to this wet-lab research, I also have experience building bioinformatics solutions to simplify the genomic sequencing workflow in the lab. I’ve always had a special interest in communicable disease epidemiology. After a career break to care for my late grandmother, I decided to return to MSU for my MPH, after being inspired by public health during the pandemic.
A little about my internship: I completed my internship with the Communicable Disease Division, with Julie Payne as my mentor. This was my MPH practicum experience. My project goal was to improve the process for generating the county’s monthly Notifiable Disease Report, which you can see on the left. The old protocol involved many redundant steps, that took many hours, to produce a monthly chart of disease case counts. The task has been a serious time-sink for Julie and has been preventing her ability to do actual epidemiological analyses. After examining this report and its associated workflow, we noted two opportunities for improvement. To improve the useability of the report, we shifted the format to an interactive data dashboard. To reduce the time and resources needed to build the report, an automated data management tool was developed.
My first task was to create the Power BI data dashboard. What you see here is a final product that’s been through multiple iterations of feedback and improvement. What’s really cool about this dashboard is that it’s interactive. So, what even is a data dashboard? A dashboard is a type of graphical user interface that provides at-a-glance views of key performance indicators that are relevant to a particular objective. Dashboards are effective communication tools which can save time and improve decision making among surveillance data stakeholders. Creating an intuitive dashboard design was a high priority - I wanted it to be understandable within 5 seconds of looking at it. This is where color comes in. For instance, cumulative case counts for COVID are much lower this year compared to the 5-year average, so, the number is green. When a diseases YTD is higher than average, like how Lyme Disease is right now, the number would be red. The line graph also includes color. The solid blue line is the Lyme Disease case count by month, and the purple dashed line is the county’s average case count for that month. This use of color provides contextual information to help the viewer gain a quick understanding of the data. This helps the overall accessibility of the report for stakeholders outside of the public health sector. Developing the dashboard was a gratifying challenge for me. It was my first-time using Power BI and Power BI’s data languages, Power Query and DAX. I am proud of the final design, especially its interactivity which makes it so useful for varying stakeholders.
Next, we’ll look at my automated data pipeline software, which I named Epi_Helper. If you're interested in seeing how it works, the QR code will direct you towards the software documentation, GitHub repository, and video demos. Before this tool, the process for making the monthly Notifiable Disease Report was repetitive and redundant. When querying MDSS for surveillance data, the user must specify how to count cases based on case status and investigation status. For example, for diseases like Campylobacter and the Flu we count both confirmed and probable cases, but we are only counting confirmed cases of Lyme disease and West Nile. There are many possible combinations of these case statuses and investigation statuses, and the reportable diseases in the county’s report are divided into 5 unique groups based on these settings. Each of the five disease groups required three separate MDSS data exports, for a total of at least 15 exports. The user would then manually examine each of the 15 exported csv files to find the specific diseases of interest, copy them line-by-line, then paste them into a master excel sheet to calculate measures and create the monthly report. That seemed like way too much work to me, so I used python to develop a tool called Epi_Helper to optimize the workflow. Basically, the tool does 2 things: The Query Assistant remembers which diseases have which case and investigation status settings, and guides the user through MDSS data exporting step-by-step The Automated Pipeline intakes those data exports, extracts the relevant data from each, then processes and formats the data before outputting one master dataset file that’s optimized for use in Power BI. With the way it’s set up now, the last thing the pipeline does is pipe data from its output file directly into the dashboard in the previous slide.
I’m proud of this tool. It lowers the necessary MDSS queries from 15 down to just 5. It also lowers the time needed to update the Notifiable Disease Report by 94%.
If you’re thinking “wow! This would make a great quality improvement project!” - you’re absolutely right! The combination of the dashboard and the software will lead to a savings about 100 hours per year. Assuming an average epidemiologist salary, that’s over $4,600 saved in personnel time alone. There are other, less quantifiable benefits, too, like increasing employee satisfaction. It can also decrease errors in the disease report. The dashboard can be updated more quickly, which can lead to faster responses to public health problems, and better health outcomes It also makes the disease data more accessible to the public, and other stakeholders, which is helpful for public health decision-making.
I had an excellent experience at Kent County. I successfully completed the objectives for my practicum; learned new technical skills; made amazing professional connections; and I had a lot of fun doing it. Special thanks to Julie Payne for conceiving of this project, and for supporting my project over the summer. She’s been an amazing resource, and I hope to collaborate more with her and with Kent County moving forward.