PrecisePK
Real-time Monitoring and Data-driven Analytics for Acute Kidney Injury Management
The Basics
Company
PrecisePK
Team
7 people
Time
4 months
My Role
Product Designer
The Problem
Acute Kidney Injury (AKI) is an abrupt decrease in kidney function, resulting in the renal failure to filter waste products from blood. AKI occurs frequently in patients who have already been hospitalized, particularly in critically ill people and patients with concurrent use of nephrotoxic agents, like Vancomycin. AKI can be fatal and requires intensive treatment. It develops rapidly, usually in less than a few days. A large increase in Serum creatinine (Scr) in a short period of time indicates AKI. On the one hand, detecting incidences of AKI responsively to patients’ Scr changes would be crucial in preventing further worsening in patients and helping pharmacists adjust their dosing regimen to provide the most appropriate care. On the other hand, retrospective analyses on the institutions’ treatment strategy in dosing and treating patients are also essential for clinicians to evaluate and improve their practice. Therefore, our team was inspired to design a tool that would provide our users with real-time monitoring for AKI assessment and a dashboard with data-driven insights for retrospective quality evaluation.
Usere & Audience
The target users of PrecisePK are pharmacists working in hospitals who treat patients with infection diseases. Pharmacists usually work in a team with rotating shifts, and need to attend tens and hundreds of patients. In their day-to-day work, it’s a great challenge for them to identify patients most in need, prioritize care accordingly and record all necessary lab results and biomarkers across multiple shifts. Manual documentation just doesn’t work swimmingly.
Team & Role
Working as the only Product Designer, I led a team with 6 other developers, with occasional support from out in-house pharmacists and Sales & Operational team.
​
I coordinated and led all facets of design including: design research, information architecture, user task flows, interactions, visual, product and prototyping. I also defined KPIs for this feature and broke them down to specific metrics and users actions to track in our program. I was also responsible to provide internal training to other team members and worked closely with Sales team in marketing campaigns.
Design Process
In order to identify what and how this product needed accomplish, I started out by assembling literatures, pharmacological guidelines and dosing protocols from different institutions to fully understand the possible causes and symptoms of AKI, different widely adopted classification criteria, and what users needed to know to take further actions.
After sifting through all of the resources, I was able to identify the key pain points: pharmacists wanted a real-time assessment to identify AKI incidents as soon as possible, and get the notification seamlessly in their workflow. As for retrospective analysis, they wanted insightful statistics that inform them the efficacy and quality of their dosing practices, and the ability to look into specific patient cases for further inspection.
With the ultimate goal to Improve Patient Outcome, minimizing the risk of negligence of AKI events, I nailed down to 3 Design Principles that we should follow throughout the design and development process.
-
Information transparency – provide comprehensive information relevant to AKI assessments and analytics to support clinical judgement.
-
Intuitive and insightful – provide meaningful statistics that are easily digestible to guide users to take further actions.
-
Customizability – give user options to choose the criteria & other relevant data that best fit to their protocols.
We identified the key specs that had to be included in this project, and prioritized these key specs in terms of how efficient they were in meeting users’ needs and how technically feasible to develop.
Based on the key specs, I constructed the information architecture applying the Model-Controller-View framework so that our developers would know which aspects of the application they should focus on and how they would be reflected onto user user interfaces. Although there could have been more customizability added to this feature, like allowing users to manually add an AKI event to the patient case based on their own judgement, but we decided to focus on the core jobs-to-be-done, which was to provide real-time, automatic AKI assessment to support clinical workflow. Besides, adding another layer of complexity would also significantly put off the deliver time of this project.
The Challenge
As we started to explore the interface and interaction design for the analytical dashboard, we encountered the biggest challenge in the process: After assessed the patients’ AKI status with patient biomarkers, how might we translate these unsorted data into meaningful statistics and visualize them into insightful and easily understandable infographics, so that when users looked at our dashboard, they would know clearly what they were looking at and what kind of actions they could take to improve their care.
Our initial idea was to compile data into only 2 charts -- one line chart showing data changing over time, and a pie chart showing numerical proportions of different segments. But as one single patient could experience multiple AKI events at one or different stages throughout the time frame of interest, we were struggling in defining the most meaningful way of group these data from backend and visualize them in these 2 types of charts to the front end.
Because we were working as team, I brought everyone together to brainstorm what the best statistics and infographics we should deliver to our users. At the beginning we had lots of different thoughts throwing out —“we can only count patients by their latest AKI event”, or “let’s only count the most serious events”, and later "no we should still show the number of patients..".
But we got stuck really fast because all the ideas either didn’t provide much value to users or they actually contradicted to each other.
The Solution
While listening to them closely, I started to realize where the problem was: everyone was thinking of solutions main based on technical feasibility — what kind of data we could pull up and what graph was most code-able. The team was more like solving problems for ourselves instead of for our users.
​
That was when I stepped out again. As the only product designer on the team, I acted as the voice of customers. So I started with writing out user needs/stories into simple clear statements, and then translating these statements into programmable logics.
For example, one statement could be like:
“In a given month, as a pharmacists, I want to know the number of patients at AKI stage 1, 2, and 3, and the total number of AKI patients within that month.”
​
And to translate into statistic logic:
“if the patient has multiple AKI events at different stages, then we count this patient each time in each stage, but only once in the total number of patient.”
User Interface Design
Color coded tags indicating AKI status at different stage. Prominently showing these tags on Home page and allow advanced search to help pharmacists prioritize and coordinate care accordingly.
Automatically assess patient’s AKI status based on their lab results (SCr values) and individual patients characteristics, adhering to the widely adopted AKI classification criteria.
This assessment happens in real time and is seamlessly integrated to user's current workflow.
A customizable Filter that makes it easy to quickly generate analytics on data of interests.
Interactive and data-driven infographics that help you visualize patient data over time and over the overall population
The infographics compile patient data into meaningful visualizations and are truly interactive. The stacked line chart shows users an overview of how AKI Events at different stages change over time at their institutions, and the pie chart shows them AKI Events at each stage as a percentage of the Total AKI Events detected in the program.User can hover on the graph to read more detailed statistics.
To account for the fact that single patient may have experienced multiple AKI events during their treatment, we created a bar chart that compares the counts of AKI Patients at different stages and their ratios to the Total AKI Patients and to the Total Patient Population as recorded in the database.
Within the Patient Case List, one CTA button is the "Quick View", which opens an overlay that gives user a snapshot of the patient's SCr history trend and how their AKI status change over time. Users can also hover on the graph and then read the corresponding SCr level and time of event. If user wants to inspect this patient case more, they can also Load Case directly from the table.
The Outcome
Taking this user-centered approach, everything we designed and developed made great sense. Different elements on the dashboard could be complementary to each other and provide helpful insights to our users. We launched this feature in Mid October 2021, and the new release has led a great increase in traction to our platform, our daily site visiting increased 48%.
Next Steps
I believe defining the key metrics for success of a product is crucial for its growth. As the owner and the designer of this features, I defined its KPIs and the specific user actions we need to track to measure its success.
