Heath Halo

Problem

• Hospitals are data-dense environments where life and death decisions are made every day
• Data is spread across a maze of siloed systems and interfaces (EHR, imaging, paper notes), increasing the cognitive burden for health providers and increasing the risk of medical error
• Hospital errors are responsible for 10% of all deaths in the USA. 250K, just behind heart disease and cancer. These deaths could be prevented by better technology

Solution

• Provide a unified interface for care teams to understand the entire patient context
• Harness the medical reasoning capabilities of LLMs to constantly monitor the patient data and detect errors and critical clinical events
• The autonomous medical record monitor automatically notifies the care team when appropriate

How we built it

• LLM is given a description of data types available and is then fed to query data sources to build up its own context to solve a task
• With the appropriate patients' context and medical history, LLM is able to answer questions from care teams
• Running autonomously, it can monitor the changes to the patient’s medical records and appropriately evaluate in the context of the entirety of the patient’s history and then alert the care team

What's next for Heath Halo

• As LLM medical reasoning improves, AI will play an increasing role in care delivery, proposing diagnosis and treatment plans
• LLMs can also help keep patients and families informed and educated given their ability to understand the entirety of a patient’s medical record

Built With

• claude2
• nextjs
• python