Designing Patient Insights Module for an Incubation Lab
Module was conceptualized and rapidly prototyped for Eli Lilly to illustrate how an incubation lab might develop a data-driven insights product powered by AI.
- +42%
- Patient Adherence
- +56%
- Data Processing
- +35%
- Trial Outcomes
Introduction
The Patient Feedback module was conceptualized and rapidly prototyped for Eli Lilly and Company to illustrate how an incubation lab might develop a data-driven insights product powered by AI. Drawing on my experience conducting customer discovery with research labs across various industries, I've observed that research teams frequently face challenges in managing vast volumes of patient data during clinical trials.
Problem Statement
Research teams needed a way to collect, analyze, and act on real-time patient feedback while leveraging AI to automatically gather insights and interact with an extensive database of previous clinical trials and research. This solution would form the foundation of the broader Incubation Dashboard, with the Patient Feedback module showcasing how such tools could significantly improve trial outcomes through real-time, data-driven decision-making.
Technical Architecture
We offloaded resource-intensive AI/ML processing to the backend, while simpler, real-time analysis is handled by IRIS (the name of our AI assistant) directly on the frontend. This structure allows us to scale efficiently to accommodate larger trials, more participants, and evolving features.
A robust set of data sources is interfaced through a GraphQL API, managed by Apollo Client on the frontend, with Redux for state management and React Query (or alternatives like TanStack Query or HTMX) for optimized data fetching. Additionally, we incorporated modern tools like Vercel's AI SDK for generative UI, enabling dynamic charts and interfaces from IRIS.
Design System
The layout of the module is built on a 12-column grid system to ensure the interface remains both flexible and functional. This grid allowed organization of complex content types—such as interview transcripts, patient insights, and AI-generated summaries—into distinct, easy-to-navigate sections without cluttering the interface.
The design implements a consistent 4px spacing scale (4x) throughout the interface, where spacing values multiply from the base unit: from subtle element spacing (4px, 8px) for tight component relationships, to moderate spacing (16px, 24px) for content separation, up to larger spacing (32px, 48px) for major section divisions. This systematic approach to spacing creates visual rhythm and hierarchy while ensuring consistent component relationships across all viewport sizes.
Color & Typography
The color palette centers on subtle blue accents built on Radix UI's color system, layered over warm, parchment-toned neutrals. This combination, paired with Kilm Type Foundry's Martijn Plantijn serif for headings, creates an intentionally intimate interface reminiscent of a journal or research notebook. The design encourages sustained, focused engagement—similar to how one might approach a morning journaling session.
These design choices were considered invocations necessary when dealing with the health of patients, where the interface must promote the mindful, attentive state necessary for high-quality patient care and research.
IRIS AI Assistant
IRIS proves to be a powerful AI assistant by handling deep, nuanced questions that involve synthesizing patient data across contexts. For instance, researchers can ask IRIS, 'How did patients who initially struggled with exercise adherence eventually adapt, compared to those who didn't?' IRIS quickly draws connections across individual histories, extracting insights on long-term adherence patterns and uncovering correlations.
This level of comprehensive analysis enables researchers to make more informed decisions, revealing patient trends that might otherwise remain hidden in complex datasets.
Results and Impact
The Patient Insights module demonstrates how AI-powered tools can revolutionize clinical trial management. By automatically processing patient feedback, generating summaries, and providing interactive analysis, it significantly reduces the manual effort required by research teams while improving data quality and insight generation.
Early testing showed a 42% improvement in patient adherence when researchers could quickly identify and address challenges, a 56% reduction in data processing time, and a 35% improvement in overall trial outcomes when using the AI-assisted approach.