PERSIMMON
Predictive Elective Surgery Management
Persimmon is a suite of digital products designed to help patients book elective surgeries and healthcare teams optimize limited resources, particularly during a constantly changing public health crisis, through leveraging machine learning and predictive modelling.
Problem
In response to ongoing waves of the pandemic waves and rapidly shifting conditions, medical institutions and their clients find it a challenge to schedule elective surgeries.
Patients struggle to know when they can schedule surgeries.
Hospitals are too occupied to effectively predict and coordinate surgical priority, allocate bed capacity and optimize resources efficiently.
Solution
Persimmon.health: Is a service that allows patients to plan and book elective surgeries based on shifting hospital capacities. Patients receive predictions on timing and availability and receive advice based on the optimal time to book a specific surgery.
Persimmon.insights: Is a dashboard tool that helps hospital executives and decision makers organize schedules to perform elective surgeries based on outstanding demand and capacity. Using the dashboard enables a hospital or medical centre to optimize their resources with data-driven predictive models.
Role
UX/UI Designer
Design Tools
Figma, Balsamiq
Type
Hackathon- Mobile & Desktop Web App
Date
March 2021
Introduction
In March 2021, I participated in Clinhacks, a hackathon focused on improving healthcare through innovation and technology. As part of a four-person team, we took on the Symphony Health challenge where we were challenged to develop insights and a feasible solution from data sets on Covid-19. As the UX/UI Designer, I developed two proof-of-concept products that constituted our solution.
The Setup
Our team was composed of two computer engineers, a health informatics consultant, and one designer (me).
We had 28 hours to dig into the data provided, generate insights, and put together some type of product that responded to the data.
The Challenge
Teams were given three datasets containing real data about COVID from the USA:
Elective surgeries for some key surgery types by week by state by age group.
State level restriction score to provide a relative score of strength/severity of community restrictions.
State level weekly totals of newly diagnosed COVID patients.
Teams were asked to focus on two areas:
What insights can you uncover?
Can you use this data and other publicly available data to design a decision support tool?
Our team was intrigued by patterns and correlations that can help policy makers, healthcare executives and patients during a global pandemic. Our team decided to organize data in a way to generate correlations, creating useful visualizations for insightful interpretation and building predictive models to provide a clear indication of future trends, useful for scheduling and planning.
The Approach
Understand + Define
Kickoff meeting
Data analysis and insights generation
Problem definition and voting on how to move forward
Ideate + Design
User personas and stories
Brainstorm and sketching
Task flows and wireframes
Build + Present
Creating the clickable prototype
Building the MVP data model
Producing the pitch deck and presenting our solution
Understand + Define
Working as a Team
Our team was composed of different backgrounds and skill sets, spread across three time zones. We used our first meeting to establish guidelines and a plan of action:
The engineers cleaned the data and took time to see what they could derive from it to uncover insights and actionable data correlations.
The team then brainstormed based on the results of the initial analysis to define specific problems and potential solutions to resolve.
We then built out an MVP that showed our data modelling and built a user-centered product that used the datasets and addressed our defined problem.
Analyzing the data
The engineers cleaned up the datasets using the Python Pandas library to help the entire team understand the potential of the data.
The data showed that we could develop a correlation between COVID-19 case numbers and the amount of elective surgeries completed at the same time. This information included surgery type and location.
This correlation helped us to decide to move forward in identifying the optimal time to complete an elective surgery considering rapidly changing COVID-19 crisis capacity.
Defining the Problem
Secondary research from the New England Journal of Medicine highlighted the already active discussion taking place around prioritizing elective surgeries during the COVID-19 pandemic. The research helped identify the decision-making aspects hospitals were using in deciding to proceed or not with elective surgeries.
Based on this problem set, we moved ahead to define a specific problem, user base, and potential application for the data. As some of the data included information of patient insurance claims, we brainstormed around problems insurance companies face in handling elective surgeries during the pandemic and how our data correlation could resolve these issues. However, as we were unfamiliar with the structure of the American healthcare insurance system, we were unsure how the data could provide a value add to insurance companies based on how patients file claims or book surgeries.
Pivoting our Thinking
Instead, our team decided to pivot our focus. I encouraged the team to take a step back and to focus on specific people as users and to consider how the insights from the data would impact them. The conversation shifted to a user-focused solution, rather than a corporate one. We saw a gap in the market:
Patients seeking to book elective surgeries are directly impacted because of the unknowns of the pandemic.
Decision-makers trying to optimize their hospital resources during a pandemic- both for improved patient outcomes and increased bottom line, had a hard time predicting surgery demand.
We distilled our thoughts into a question to take us forward:
How might we use predictive modelling to help both patients and hospital decision makers optimize their elective surgeries during the shifting conditions of a global pandemic?
This resulted in the development of two proof-of-concept solutions: one to help patients book and another to help hospital decision makers plan.
Ideate + Design
Through merging the datasets and creating visualizations using the Python Seaborn library, the team continued to develop the data modelling. I teamed up with our health informatics consultant to focus on the product design and what our solution could look like for both user types.
User Personas
In order to better focus our solution, we developed a proto-persona for each target user group.
With the limited time we had, we relied on our teamās industry knowledge to paint a picture of our target usersā needs, goals, and pain points.
User Stories
With a better understanding of who we were designing for, our team had to figure out which features to develop given our limited time frame.
I formulated key objectives that our users wanted to achieve through user stories. Using these our team was able identify the key features to include in the product.
As a patient, I want to know when the best time is to book my elective surgery so I can secure a spot and plan my life around it.
ā After going through a short questionnaire, patients will be shown all the options for booking a surgery based on availability and their preferences
As a patient, I want to not worry about continuing to check with my doctor when there might be a space so that I never miss out on an opportunity and Iām kept in the loop.
ā During the questionnaire process patients also sign up for an account so their preferences can be saved. They can have a specific surgery tracked with alerts and updates sent to them.
As a healthcare executive I want to have better information about the capacity of my hospital and insights to where I can be optimizing resources better so that I can improve my bottom line.
ā Through a personalized dashboard, hospital executives can review availability trends based on their region and the hospitals under their care.
As an executive, I want to be able to visually see data to help me understand future pandemic risk levels so that I can plan elective surgeries ahead and improve patient outcomes and surgery rates.
ā On logging into their dashboards, decision makers will get to see an actionable insight that is predicted and customized for them based on our model. They will also get a summary of actionable insights daily or weekly to their email.
Task Flows and Sketching
To build the wireframes we created a single task flow for each user and their products. This helped to identify the necessary steps towards reaching at least of their goals. I sketched out quick wireframes according to our hypotheses.
The team regrouped to reflect on our achievements to date. I used this session to solicit feedback on the wireframes from the whole team. With the feedback I jumped into fleshing out the entire prototype.
Build + Present
High-Fidelity Prototype
While half of the team built the visualizations from the data with scatter plots and correlations matrices, I worked with the healthcare consultant to prototype for each of the two products.
Using mood boards to decide on a branding direction, I put together the UI of each product while the health informatics consultant helped fill in the content.
This resulted in a proof-of-concept prototype for patients called: Persimmon.health
And one for executives and decision makers called: Persimmon.insights
Pitching our idea
As time came to a closer, the team worked on the data modelling MVP and the consumer facing presentations. We created a pitch deck and a video presentation to sell our ideas to the judges.
We presented to judges from the Clinician Engineering Hub as well as directly to Patrick Brundage from Symphony Health (our challenge sponsor).
We received extremely positive feedback from the judges, with them mentioning how they loved our product vision and the ability to put together āa very robust MVP that already has thought through the UI for both patients and clinicians. We were recognized for being bold and were to be commended.
At the end of the competition we were awarded second place for the solution we proposed.
Reflection
Key Learnings
Set up a framework and time goals at the start
While our group did come up with a high-level plan of our strategy to tackle this sprint, I think we would have all benefited from creating clearer milestones with deadlines from the beginning.
Because of the ambiguity of the data, we spent a lot of time brainstorming over the problem and potential solutions. If we had time-capped this activity we would have felt less rushed on both sides in creating our MVPs and would have had more time to create a stronger pitch presentation.
The power of cross-collaboration
Up until this point I had not worked directly across a mix of disciplines. Learning from my team-mates and their set of expertise was incredibly valuable. It not only motivated me to work harder but it helped to add value to a better and more robust solution. My work as a designer was strengthened by this cross-functional collaboration. A key takeaway is to engage with a broad set of expertise to build a successful project.
Keep it simple
Our team was ambitious and we took on a big vision. While this was really exciting, if we had simplified our scope we could have brought in more details to produce one strong product rather than produce multiple elements that split our efforts.
Final Thoughts
This was my first hackathon and it was incredibly rewarding to dive into creating solutions for the challenges in healthcare. It was even more rewarding to work with an exceptional team that pushed me to be a better designer and reminded me how much I love collaborating with passionate people. As much as I love the solo creative process, I thrive off collaboration, especially with those who have differing expertise to my own.
Even though there could only be one winner, the future of healthcare relies on multiple people fighting to improve it and it was incredibly valuable to be a part of that.