This year at recent trade shows, spokespeople from Augmedix, HCA Healthcare, and Google Cloud gathered to discuss key elements needed to effectively implement generative AI (GenAI) solutions in healthcare. The panels addressed Augmedix Go’s success in implementing Ambient AI documentation in emergency departments (ED) and how GenAI is transforming clinical interactions in both acute and ambulatory care settings.
At HIMSS24, Augmedix CEO Manny Krakaris sat down with Vikesh Tahiliani, MD, VP of Care Transformation and Innovation at HCA Healthcare, and Aashima Gupta, Global Director of Healthcare Strategy and Solutions at Google Cloud. At Google Cloud Next, Ian Shakil, Founder, Director, and Chief Strategy Officer of Augmedix, was joined by John Doulis, MD, VP Data Services and Technology Innovation at HCA Healthcare, Ken Su, Senior Product Manager at Google Cloud, and Praney Mittal, Group Product Manager at Google Cloud.
Highlighted below are key takeaways from both panels, including the importance of effective change management processes, the complexity of GenAI, and the need for technology that is adaptive to clinicians’ needs.
The Need for Change Management
Effective change management is crucial for the successful implementation of GenAI solutions in healthcare. The panelists emphasized the following strategies:
- Ensure Champions at Every Level: From the enterprise level down to regional teams, having champions who understand the technology and its benefits is essential for smooth adoption.
- Set Clear Expectations: It is important to establish and communicate clear expectations for what innovative GenAI will achieve, ensuring that all stakeholders understand and align with the goals.
- Align on Metrics: Agreement on the metrics for success among all stakeholders helps in tracking progress and making necessary adjustments. This alignment fosters a shared vision and collective accountability.
The Complexity of GenAI
Creating a medical note with GenAI involves more than just feeding raw audio data into a large language model (LLM). The panelists highlighted several complexities:
- Utilization of Multiple Models: Implementing GenAI in healthcare requires the use of both large and small models, tailored and modular, by section and specialty. This ensures accuracy and relevance in the generated outputs.
- Constant Benchmarking: Models need to be continuously benchmarked and updated to adapt to new data and evolving medical standards.
- Achieving GenAI Optut Quality: Ensuring the quality of medical notes is challenging and requires a combination of human and automated systems for constant measurement and validation.
- Google MedLM Implementation: Despite these challenges, Augmedix’s use of MedLM on every note has been performing exceptionally well, showcasing the potential of specialized models in enhancing note quality.
The Need for Humans in the Loop
Even with “Pure AI” solutions like Augmedix Go, the role of clinicians as the human in the loop remains critical:
- Human Oversight: Clinicians provide essential oversight, ensuring that AI-generated notes are accurate and contextually appropriate before final submission into the EHR.
- Guardrailing AI Capabilities: As bidirectional capabilities in Augmedix Go become more advanced, it is important to implement guardrails to prevent the system from encroaching on areas that require clinical judgment and decision-making. This ensures that AI supports clinicians rather than replacing their critical decision-making roles.
Looking Ahead: The Future of GenAI in Healthcare
As we look to the future, the advancements discussed by Augmedix, HCA Healthcare, and Google Cloud at HIMSS24 and Google Cloud Next signal a transformative era for healthcare. The integration of GenAI promises to enhance the efficiency and accuracy of clinical documentation, freeing clinicians to focus more on patient care.