Problem Statement

To solve the disruption in the health care system for elective surgeries during a time when the hospitals are pressed for resources.

Inspiration

Our team attempted to undertake the Symphony data challenge as we were intrigued by what patterns and correlations we can make that can help policy makers, healthcare executives and patients during a global pandemic. In this age of Big Data there is no lack of an abundance of data points being available to us and the challenge lies in being able to make sense of it by organizing the data in a way that correlations can be made, creating useful visualizations for insightful interpretation and building predictive models to get a glimpse of future trends. It is in this conjecture that our team is excited to provide our solution: Persimmon.

What it does

The vision behind “Persimmon” is to facilitate visibility into the different elective surgeries dependent on COVID rates, location and hospital resource availability. The model behind “Persimmon” has been strategically designed to provide patients with schedule prediction for elective surgeries with respect to COVID rates and to help hospital personnel responsible for managing operations to preemptively know when the best times to book an elective surgery via an insights dashboard with state-wise data. Our aim is to reduce the apprehension around accessibility to an elective surgery during a time when resources are limited, so that all stakeholders involved can feel confident in their decision to pursue the surgery.

The goal of the Persimmon product suite is to show the potential of how a robust data analysis model can be used to empower patients, clinicians and decision makers in challenging times. Persimmon.health provides a welcoming and user friendly experience for patients and clinicians in finding the best time book your elective surgery. Persimmon.insights provides executives and decision makers a robust yet simple dashboard where they can analyze information and receive recommendations from our predictive machine learning model for how best to allocate hospital resources during a shifting public health situation.

How we built it

We initially started by cleaning up the datasets available to us using the Python pandas library ensuring that the fields that we wanted to focus on were binned correctly. We chose to focus on the age of the patients, the American state the data is being collected for, the week the data is collected over and the number of COVID claims, elective surgery claims, patients and physicians. Once we merged the datasets, we discussed what visualizations we wanted to focus on and used the Python Seaborn library to create time space graphs. For the patients and the clinicians we mapped out a visualization that showed the number of claims, patients and physicians for a particular elective surgery over time, a scatterplot time space graph to indicate physician availability with relation to COVID rates over time and a kernel density correlation matrix that shows the normal distribution of all variables involved.

Challenges we ran into

Couple constraints that we faced throughout the hackathon were a lack of real time data or historic hospital resource specific data. This would have allowed up to create a more accurate prediction and story of what happened with hospital resources once the pandemic hit and how that impacted elected surgeries which then could have been used to better understand bed allocation for elective surgeries. Having data specific to the average length of stay for each elective surgery is also a key metric that could have been used to better plan and book elective surgeries within hospitals. Knowing which surgeries require a longer post-op monitoring period will give a better understanding for priority bed allocation, especially during a pandemic. Due to this there were certain assumptions that needed to be made when coming up with out POC. Additionally, another challenge faced during this time was understanding how the American healthcare system operated in terms of process flow and knowing how insurance claims work with in-state and out-state elective surgery procedures. Due to time- constraints we did conduct research to the best of our abilities to understand the American payer system. This also means understanding how insurance companies work with the hospitals in America and understanding what KPI metrics would matter for users such as Hospital decision makers versus insurance.

Accomplishments that we're proud of

We are extremely proud of the fact with the way our team has worked together. Despite all our team members being located in completely different time zones(three to be exact: EST, MST, GST ) we were able to work together cohesively and each bring in our own expertise to the table. From when we first met at the start of the hackathon first we all knew we had a passion for the integration of health, innovation, people, processes and technology! We also knew the importance of creating an end result that will ultimately be a benefit and optimize healthcare outcomes to the end users. The graphs have been able to show time periods where the most amount of physicians are available with relatively lower COVID claims and when frequency of elective surgery are higher. From starting of with a relatively large dataset to parsing it and being able to make meaningful connections we are amazed by the power of data analysis and what it can do for the healthcare system.

What we learned

Together we learned from each other's expertise and gained insightful knowledge on different tools used to create our final products. We also were able to go through the various processes for prototyping starting from discovery to ideation. Additionally, we attended the story telling session for insights and was able to gain more knowledge and understanding on how to derive more meaningful and actionable insights. Some of our team members learned how to use python to clean data while others were able to learn how story board and go through the process of ideation. We realized that coming up with an end product is an iterative process that can be complex when people from different backgrounds are collaborating, but through communication and an ability to tackle the problems from different angles we were able to converge to a solution.

What's next for Persimmon : Predictive Elective Surgery Management:

With more data we would be able to further iterate the Persimmon health project and even potentially have an integration scheduling software which could be embedded with EMR systems in the hospitals with the use of the Persimmon Health tool. This could later be used to notify patients when their pre-op, and surgery is scheduled or even notify them if there is a potential chance for a cancellation due to the hospital resources being constrained due to epidemics, natural disasters, or any emergency that will impact hospital resources.

Additionally, with more selective hospital resource data we can accurately create more predictive models which can also benefit future on site scheduling. This would include information about the number of Operation theatres, Average Length of Stay for surgeries, Payer information - insurance. This would be embedded in the dashboard insights tool. From there we would also want to partner with insurance companies so that we can create further specific insights to the Insurance companies for them in terms of how their claims are being impacted by region with COVID-19 or future pandemics. With this insight they can also predict and financial costs which will potentially impact their operations.

Built With

Share this project:

Updates