In this episode, we discuss artificial intelligence large language models (LLMs) and how these will impact the future of the practice of pharmacy.
Key Concepts
- Generative AI with large language models (LLMs) have already changed how healthcare is delivered to patients. In the future, these changes will be more substantial and require pharmacists and other healthcare professionals to understand the benefits and downsides of this technology.
- Commercial LLMs, such as ChatGPT, are not HIPAA compliant and should not be used with protected health information. Companies currently offer software products that are HIPAA compliant and can integrate directly into electronic health records in a HIPAA-compliant manner.
- Currently, most commercial use cases of LLMs for healthcare providers focus on expediting or simplifying the documentation process (e.g. generating a first draft of a progress note or summarizing a patient encounter from an audio recording).
- In the future, LLMs will be used to perform a variety of clinical tasks, including drug interaction checking, renal dose adjustments, duplication of therapy, and even the appropriateness of a patient’s drug regimen for a given medical condition. These clinical tasks will almost certainly be done as a “first pass” to highlight or flag specific aspects of a patient’s chart and will then be reviewed by a licensed (human) healthcare provider as a final check prior to clinical decisions being made.
References
- Large Language Models (LLMs) referenced in the episode: https://chat.openai.com, https://coral.cohere.com, https://claude.ai, https://gemini.google.com.
- Prompt Engineering Guide (https://www.promptingguide.ai/techniques)
- OpenAI - Prompt engineering (https://platform.openai.com/docs/guides/prompt-engineering/six-strategies-for-getting-better-results)