Helpful prompts to transform the way healthcare professionals use conversational generative AI.
Welcome to the Prompts-For-Health repository! This is a collection of AI prompts, which are tasks and instructions that can be given to generative AI that use Large Language Models (LLMs), such as GPT-4 and Llama 2, via services such as ChatGPT and BastionGPT. It is dedicated to providing healthcare professionals with an array of safe, innovative, and transformative conversation prompts. These prompts are designed to facilitate more efficient, empathetic, and constructive interactions with patients, helping to bridge communication gaps, unearth crucial health insights, and empower healthcare professionals to spend less time documenting and more time providing quality patient care. Each prompt page contains a description, the prompt itself, and an example showing how it works in practice.
Brainstorm Differential Diagnoses
Determine a Patient's Eligibility for a Clinical Trial
Draft Replies to Patient Messages
Explaining Health Concepts Across Cultures Age Education Levels
Extract Information from Forms
Generate a Framework of Questions for a Patient Visit
Generate Prior Authorization Letter
Generate SOAP Notes from Clinical Notes
Generate SOAP Notes from a Transcript
For Healthcare Professionals
Browse through the prompts and find ones that resonate with your needs. Each prompt will need to be customized to meet your needs, often by appending relevant clinical information at the end of the prompt.
Warning
Only use prompts that incorporate PHI or PII in AI services that are compliant with applicable healthcare regulations such as HIPAA and GDPR. As of this writing, OpenAI ChatGPT and ChatGPT+ are not compliant with most healthcare data security and privacy regulations. BastionGPT, Microsoft O365 Copilot, and Azure OpenAI, can meet applicable regulations in most circumstances.
Contributions
We encourage healthcare professionals, AI enthusiasts, communication experts, and even patients with insights into effective communication to contribute to this repository. Before contributing a prompt, please check to see if a similar prompt already exists. If it does, we recommend enhancing the existing one with comments or revisions. You can contribute directly to the project via the instructions below, or you can reach out to FortaTech or another contributor.
To contribute directly, follow these steps:
- Fork the repository.
- Create a new branch in your forked repository.
- Add your prompt name following the PROMPT_TEMPLATE.md format.
- Submit a pull request against the main branch of the original Prompts-For-Health repository.
- All submissions are reviewed by moderators before they're accepted and merged into the repository. This review includes a check for safety, appropriateness, clarity, and healthcare basis.
As pioneers in healthcare technology, we recognize the crucial role we play in incorporating artificial intelligence into healthcare workflows. We firmly support and align to leading AI ethical and safety principles.
We have established a complimentary, and more focused set of principles, specific to healthcare generative AI, that act as our north star to guide our innovative journey to elevate patient care and reduce healthcare workforce burnout, focusing unwaveringly on the safety, privacy, and well-being of patients. These principles aim to foster trust and transparency with our diverse community and reflect our steadfast commitment to maintaining the highest standards of integrity, accuracy, and reliability in our services. We encourage other organizations to embrace these guidelines, helping to realize a safe and transformative integration of technology and healthcare, with medical professionals at the forefront, ensuring a human-centric and ethical approach to patient care.
Principle 5: Guard Patient Interaction
“Generative AI must not directly provide medical advice to patients unless monitored and strictly validated.”
Generative AI is advancing rapidly, creating boundless possibilities. However, when it comes to providing medical advice, these solutions should not directly communicate with patients unless the outputs are strictly controlled and thoroughly vetted by both industry experts and the relevant regulatory bodies. The distinct risk of AI hallucinations and biases in healthcare topics necessitates medical professionals, who are trained and experienced, to supervise interactions. Medical professionals are essential in ensuring the accuracy and reliability of the advice given, recognizing the limitations of AI, and preventing potential harm to patients.
Principle 4: Uphold Privacy and Security
“All personally identifiable information must be kept private and secure.”
The influx of insecure AI services, often cheaper and below standard, poses significant risks to the security and privacy of sensitive personal information. The integrity and confidentiality of patient information must never be compromised. All AI healthcare services must align with healthcare standards and comply with regulatory requirements to avoid breaches that could jeopardize patient wellbeing. Establishing robust privacy controls and security measures are critical in maintaining trust and safeguarding patient interests.
Principle 3: Prioritize Evidence-Based Medicine
“Source data for generative AI must prioritize evidence-based medicine and reputable sources.”
Misinformation and biases are prevalent, even in reputable sources across the internet. Generative AI in healthcare must be anchored in evidence-based medicine, relying on trusted, scientific sources to inform its knowledge base. A stringent, proactive approach is essential to mitigate the risk of circulating harmful and inaccurate data points, improving the accuracy, reliability, and safety of the AI-generated outputs.
Principle 2: Promote Transparency and Caution
“Generative AI must transparently communicate its propensity for errors even when used by medical professionals.”
The incredible capabilities of generative AI can lead to a false sense of security and overshadow its inherent limitations and potential for mistakes. It is crucial to maintain transparency about AI’s limitations and encourage both users to approach the AI outputs with caution and critical evaluation. The emphasis should be on ensuring the responsible use of AI, with medical professionals maintaining vigilance and scrutiny to detect any inaccuracies, thereby preventing potential harm to patients.
Principle 1: Maintain Human Oversight
“A medical professional must mediate the provision of medical advice or information from AI to patients.”
The interface between AI and patients, especially in delivering medical advice or information, must be sufficiently supervised by medical professionals unless appropriate safeguards are robustly implemented. These safeguards can include strict governance around AI outputs, ongoing monitoring by healthcare professionals, and rigorous validation of medical accuracy. We must keep healthcare human-centric, with clinicians retaining control to ensure the safety, accuracy, and appropriateness of the advice provided.
This project is licensed under the terms of the MIT license.
The prompts within this repository are illustrative use cases that have been screened to reduce the risk of harm and are intended for use by healthcare professionals. However, users must be aware of the risks that come with the use of Large Language Models (LLMs). LLMs may occasionally provide inaccurate or inappropriate responses. All users must review the responses from LLMs for accuracy and appropriateness prior to use in any healthcare setting, such as entry into an electronic medical record or communication with a patient.
Warning
While these prompts provide value in various healthcare applications, users should be aware of limitations and risks inherent to the AI
- Hallucination: Generative AI, despite its sophisticated learning algorithms, can produce "hallucinated" responses—information that appears plausible but is incorrect or not based in reality. This can occur even when the input data is both accurate and unambiguous. In the healthcare environment, this could lead to misleading or inaccurate health information, which could pose serious risks if not properly understood and managed.
- Biases: Despite extensive training and regular updates, Generative AI can sometimes exhibit unintended biases. These biases are reflections of the biases present in the data it was trained on and not an intentional feature of the model. While we strive to minimize these, they may still occur and could potentially influence the model's responses.
- Outdated Information: As of the AI model's information cut-off, most AI does not have the capability to update its knowledge base with real-time information or developments. Consequently, any advancements, changes in guidelines, or novel treatments emerging in the healthcare industry post this date may not be reflected in the responses of AI.
If you're experiencing any issues, have questions, or simply wish to connect with the core team, please feel free to open an issue in this repository or reach out to us at hello@forta.tech.