Inspiration
We were inspired by the challenges clinicians face when patient information is scattered across different medical records and providers. In emergency situations, having an incomplete picture of a patient's medical history can make treatment decisions more difficult. We also recognized the problem of alert fatigue, where doctors receive so many generic medication warnings that important alerts can get overlooked. We wanted to build a solution that makes patient information more accessible and medication safety checks more personalized.
What it does
Patient360 is a clinical decision-support platform that consolidates a patient's medical history into one centralized profile. It gives clinicians an organized overview of medications, allergies, existing conditions, and previous treatments, with critical information highlighted first.
It also includes a medication safety checker that evaluates proposed prescriptions against a patient's existing medications, allergies, and conditions. Instead of overwhelming clinicians with generic warnings, it prioritizes significant risks, explains why they matter, provides supporting evidence, and suggests potential alternatives for clinician review.
How we built it
We built Patient360 using React and Tailwind CSS for the frontend and Python with FastAPI for the backend. We use SQLite to store patient profiles, medications, allergies, conditions, and interaction results.
Our system processes synthetic patient records in the HL7 FHIR R4 format using Synthea data. We use RxNorm to normalize medication names, RxClass for drug-class relationships, and DDInter 2.0 as our primary source for drug interaction information.
We implemented deterministic, rule-based checks to identify medication interactions and allergies. We also incorporated an AI explanation layer to translate verified findings into understandable language. AI does not make clinical safety decisions; it only explains results identified by our underlying logic and data sources.
Challenges we ran into
One of our biggest challenges was integrating different healthcare data sources. In particular, DDInter identifies medications by name, while patient records use RxNorm identifiers. Mapping these different formats accurately was essential for identifying relevant drug interactions.
We also had to think carefully about alert prioritization. Simply displaying every possible interaction would recreate the alert fatigue we wanted to solve. We focused on highlighting the most critical findings while keeping lower-priority alerts accessible.
Working within a 48-hour hackathon also challenged us to prioritize core functionality, coordinate development across the team, and build a reliable demo without depending entirely on live APIs.
Accomplishments that we're proud of
We are proud of building two connected capabilities around a single patient record: a consolidated medical summary and a personalized medication safety checker.
We also developed an approach to reducing alert fatigue by prioritizing critical findings rather than displaying every warning equally. Each alert includes an explanation and supporting evidence, helping clinicians understand the reasoning behind it.
Most importantly, we created a prototype that demonstrates how healthcare data, deterministic safety checks, and AI-powered explanations can work together without relying on AI to make clinical decisions.
What we learned
We learned how complex healthcare data integration can be, especially when different sources use different formats and medication identifiers. We also gained a better understanding of the importance of explainability and reliability in healthcare technology.
Another important lesson was that adding AI does not automatically make a solution better. In a high-stakes environment like healthcare, it is essential to establish reliable logic first and use AI in a supporting role.
Finally, we learned how to divide responsibilities, collaborate across different parts of a full-stack application, and make technical decisions under tight time constraints.
What's next for Patient360
We want to expand Patient360 to support more comprehensive patient profiles and additional types of medication safety checks. Future improvements include incorporating factors such as kidney function, pregnancy, and supplements, which are currently outside our prototype's scope.
We also hope to improve medication alternative recommendations, integrate with real electronic health record systems, and expand our interaction database.
Ultimately, our vision is to make Patient360 a more comprehensive clinical decision-support tool that helps clinicians access relevant patient information, identify potential risks, and make safer, more informed treatment decisions.
