Health Vision

There aren't enough hours in the day for clinicians, or any frontline worker, to review everything in a patient's record and take the next best action. Each patient generates thousands of data points: notes, labs, scans, prescriptions, and histories. Most of it goes unused, adding complexity and burden instead of clarity.

Health Vision, built by the Health Data Analytics Institute, is an AI-powered, clinician-focused platform that uses real-time predictive analytics and generative AI to turn a patient's full medical history into a clear summary of their health, highlighting and ranking the most important clinical risks.

I led product design from the earliest concepts through the Series B and C funding rounds, working with product, engineering and clinical stakeholders, and established the information architecture, interaction patterns and design system the platform grew on.

The data isn't clean or uniform. It comes from structured EHR encounters, unstructured notes, Medicare records, billing and payer data, census and ADT feeds, and more. Useful in theory, overwhelming in practice. The real value isn't just in what happened, but in predicting what comes next: machine learning and generative AI build and validate hundreds of predictive models. At this scale, product design becomes mission-critical.

The patient roster is often where a shift begins, and where every minute counts. I redesigned it from a static list into a real-time, prioritized view of risk. As new notes, diagnoses, or encounters come in, the roster updates automatically, surfacing the patients who need attention first, whether they're scheduled, admitted, in surgery, or recently discharged.

Clinicians needed a way to cut across hundreds of data points, such as past encounters, diagnoses, utilization, adverse events, costs, and forward-looking risks, without digging through endless menus. I kept the most common filters visible on the surface, moved advanced options under a "More" menu to keep the workspace clean, and made applied filters sticky and visible, so clinicians always know what view they're working with.

When clinicians need more precision, they can choose whether patients match every filter or any of them, refine a broad group step by step, filter by risk level, baseline or percentile, and look back in time, for example at patients with an event in the last month. This made it possible to spot patterns and follow up on recent changes without digging through records manually.

During testing, clinicians told us that even a refined list could still feel overwhelming. To help them act faster, I added a Focus feature that highlights only the most critical patients within a filtered group.

For example, after selecting admitted patients with multiple conditions, a clinician can instantly see just the five with the highest mortality risk. What was once a long dataset becomes a clear, prioritized shortlist for action.

Many clinicians were rebuilding the same filters again and again. To save time, I introduced Smart Cohorts: dynamic groups that update automatically as patient data changes. They can be reused anywhere in the platform, they remove repetitive setup, and they give teams a reliable way to track the patient groups that matter most.

The Patient Spotlight gives clinicians a quick, scannable snapshot of a patient's health. It surfaces the most important risks and insights in real time, so clinicians can preview the essentials before deciding whether to dive deeper. Every insight links back to its source note, making it easy to trace details and review the full history when needed.

The Spotlight is built around customizable widgets. Clinicians can configure their layout to highlight what matters most, such as cardiovascular risk, recent hospitalizations or medication changes, so the snapshot adapts to different roles and workflows.

Health Vision is in use at some of the top healthcare systems in the U.S., including Houston Methodist and Cleveland Clinic. Here's how they apply it: