Baptist Health South Florida is testing a generative AI system to reduce clinical documentation time and improve patient care. By integrating AI with transcription technology, they aim to produce instant clinical notes, saving time for providers and enhancing patient experiences. The system utilizes GPT-4 for medical diagnoses and integrates with Snowflake for data analysis. It offers time savings, increased clinical productivity, and cost-efficiency while emphasizing the importance of human verification and patient data security.
Baptist Health South Florida, a healthcare system in South Florida comprising 11 hospitals, is piloting a generative AI system with the aim of reducing clinical documentation time and improving the overall patient experience. The challenge faced by providers at Baptist Health was managing the vast amount of information generated during patient-provider interactions.
One significant issue was the time cardiologists had to dedicate to documenting patient visits, a process that was both time-consuming and contributed to provider burnout. However, commercial solutions were expensive. To address this challenge, Baptist Health integrated generative AI into an AI-assisted documentation app that combined medical transcription technology with advanced AI, particularly large language models. This unique combination allowed AI to rapidly generate clinical notes from transcribed patient conversations, providing immediate access to comprehensive clinical notes after each patient visit and reducing the time-consuming manual documentation process.
The proposal included using generative AI, AWS HealthScribe for medical audio transcription, and other tools like Azure OpenAI, Snowflake, and DataRobot. Patient-clinician interactions were recorded with patient consent, transcribed into text, and then processed by a large language model to generate clinical summaries in SOAP format. This automation was expected to reduce documentation time to just two to five minutes post-visit and enhance personalization through Snowflake integration.
GPT-4, known for its proficiency in suggesting medical diagnoses, played a crucial role in this initiative. The goal was to improve patient experience, clinical productivity, operational efficiency, and cost savings. Data generated by the application was planned to be stored in the Snowflake data lake for further analysis and improvements.
The technology journey began with the recording of patient-clinician interactions, transcription by AWS HealthScribe, and AI-generated summaries verified by clinicians. AWS Lambda and AWS S3 were pivotal in building the AI service, and integration with Snowflake was essential for extracting patient appointment data. The finalized summaries would be integrated with the electronic health record system to ensure accurate and standardized clinical documentation.
The expected results included time savings for physicians, enhanced clinical productivity with quicker turnaround times, and cost efficiency compared to commercial alternatives. It was emphasized that AI should augment, not replace, human input in clinical documentation. A human-in-the-loop verification process was recommended to maintain the integrity of patient records. Organizations contemplating AI implementation should weigh the costs of vendors against in-house solutions while prioritizing patient consent and data security to maintain trust and adhere to healthcare regulations.
