Applying Prompt Engineering in Healthcare: 10 Concrete Recommendations for Improving Human-AI Interaction

Prompt Engineering is a relatively new field of research covering the design, refinement, and implementation of prompts or instructions that guide large language model (LLM) output to be useful across a variety of tasks. This article summarizes current research on applying Prompt Engineering in healthcare, and offers practical recommendations to help healthcare professionals improve their interactions with AI.

1. The Emergence of Large Language Models and Prompt Engineering

In the world of artificial intelligence, AI language models such as ChatGPT are prominent applications growing at a remarkable pace. These models have revolutionized how we work and access data.

With the emergence of artificial intelligence language models, an intriguing concept called Prompt Engineering was born.

Numerous studies have shown how Prompt Engineering can contribute to medical tasks and processes in order to reduce the burden on healthcare professionals, improve efficiency, and cut costs.

1.1. The Emergence of Large Language Models

An LLM (Large Language Model) is a language model trained on massive datasets of text and code. Large language models typically convert tasks into prompt-based datasets, training the model through a method called prompt learning. This enables text generation, translation, creative content creation, and answering questions.

Healthcare institutions have begun investing in generative AI, medical companies are integrating LLMs into their businesses, medical associations have released usage guidelines, and medical curricula have begun covering this new technology. In this way, Prompt Engineering has emerged as an important new skill in healthcare.

1.2. What Is Prompt Engineering?

Prompt Engineering is a relatively new research field covering the design, refinement, and implementation of prompts that guide LLM output to be useful across various tasks. In essence, it is the practice of interacting effectively with AI systems to optimize their benefits. By deepening their understanding of Prompt Engineering, developers can craft better prompts and guide AI systems toward the desired results.

2. Applications of Prompt Engineering in Healthcare

Prompt Engineering is an advanced language technology built to facilitate effective interaction between humans and AI. It goes beyond mere mechanical exchange — as a digital language, it reflects the complex interplay of grammar, context, and semantics involved in learning a new language.

In the context of healthcare professionals and the healthcare industry, this may include the following areas.

• Decision support: Healthcare professionals can use Prompt Engineering to quickly optimize AI systems, supporting decision-making processes such as diagnosis, treatment selection, and risk assessment.
• Administrative support: Designing prompts for tasks such as patient scheduling, record management, and billing simplifies administrative work and improves efficiency.
• Patient engagement: Prompt Engineering can be used to improve communication between healthcare providers and patients. For example, AI systems can be designed to send prompts such as medication reminders, appointment scheduling, and lifestyle advice.
• Research and development: In research scenarios, prompts can be created to support tasks such as literature review, data analysis, and hypothesis generation.
• Training and education: Designing prompts to promote continuing education for healthcare professionals on the latest treatments and procedures.
• Public health: At scale, Prompt Engineering can support public health efforts by helping analyze population health data, predict disease trends, and support public education.

As such, Prompt Engineering holds the potential to improve efficiency, accuracy, and effectiveness in healthcare, and is becoming an increasingly important skill for healthcare professionals.

3. Risks of Using Prompt Engineering in Healthcare ( Limitations and Risks Associated with LLMs, Especially ChatGPT)

As trends and changes happen every minute around the world, trends in prompt engineering in healthcare, as well as its use, also give rise to several challenges and risks.

3.1. ChatGPT Is Not Updated in Real Time

ChatGPT is not updated in real time and only contains information up to 2021-11, which means it may miss important changes in medical research or clinical guidelines. This can also affect the quality and relevance of its responses.

Furthermore, since ChatGPT cannot access or process individual user data or context, its ability to give personalized medical advice is limited, raising the risk of data misinterpretation.

3.2. Dependence on Training Data Patterns and the Need for Verification

Model responses are generated from patterns in the training data, and are not necessarily accurate or safe. It is therefore critical that all ChatGPT responses be verified by qualified healthcare professionals.

3.3. Impact on Patient Experience and Legal Concerns

The inability to empathize with patients or handle confidential information can lead to a substandard patient experience.

Importantly, the potential leakage of confidential patient information could violate privacy laws such as the U.S. Health Insurance Portability and Accountability Act of 1996. Therefore, even when used merely as a supplementary tool, careful consideration is needed when applying it to healthcare, taking these constraints into account.

While these risks are significant, there are also potential benefits — which is why better prompt design is needed to improve the use of AI models like ChatGPT.

4. Tips for Improving Your Prompt Engineering Skills

As with any other important skill, improving Prompt Engineering requires understanding the underlying technology’s basic principles, hands-on experience with systems that use it, and continuous improvement based on feedback.

These are some of the general steps healthcare professionals can take to improve their Prompt Engineering skills.

• Understanding the basic principles of AI and machine learning models provides a foundation for Prompt Engineering skills, even without prior technical knowledge or coding skills.
• Understanding each system’s unique capabilities and limitations helps you craft more effective prompts.
• Practice makes perfect — there is real benefit in regularly interacting with LLMs and noting the most useful, accurate prompts.
• Since a prompt’s effectiveness is best judged in real applications, it’s also important to always test prompts in real-world scenarios.

5. 10 Concrete Recommendations for Improving Human-AI Interaction

5.1. Be as Specific as Possible

The more specific a prompt is, the more likely you are to get an accurate, focused response. Below are some example prompts.

Less specific: “Tell me about heart disease.”

More specific: “What are the most common risk factors for coronary artery disease?”

Describe the setting and provide context for your question

You should think of a discussion with ChatGPT as being like a discussion with someone you’ve just met, who may still be able to answer your questions and help address your challenges.

Here is an example prompt: “Could you give me some of these tips and tricks along with specific prompt examples?”

5.2. Try Different Types of Prompts

It’s important to try different types of prompts, since prompt style can significantly affect the response. Use ChatGPT to explore various ways of creating summary lists or topic overviews. Below are concrete examples.

Direct question: “What are the symptoms of COVID-19?”

List request: “List all the potential symptoms of COVID-19.”

Summary request: “Summarize the main symptoms and progression of COVID-19.”

Process: “Explain the diagnostic process for COVID-19 step by step.”

5.3. Clarify the Overall Purpose of the Prompt First

It’s important to clarify the overall purpose of the prompt from the start.

Explaining exactly what kind of output you need helps ChatGPT find the right answer. Whether you want a detailed explanation of a scientific topic or creative ideas for an article, providing example questions helps — for example, “I’d like a list of 5 ideas for a presentation at a science event.”

5.4. Ask ChatGPT to Play a Role

It’s also important to ask ChatGPT to play a role.

You can have ChatGPT streamline the process of getting information in a specific setting — as a basic explainer for a new topic, or a tutor helping you understand a detailed topic step by step. Here are some examples.

“Explain Prompt Engineering to a doctor.”

“As my nutritionist, give me tips on a balanced Mediterranean diet.”

5.5. Iterate and Improve

Even with advanced Prompt Engineering skills, LLMs are highly dynamic, so you will rarely get the best response on your first attempt. It takes practice to continuously iterate on prompts, and users are encouraged to ask the LLM to modify its output based on previous responses.

5.6. Use Threads

Clicking a specific thread on the left side of the ChatGPT dashboard returns you to a previous discussion, letting you build on earlier details and responses. This means you no longer need to re-explain the same situation to ChatGPT, saving significant time.

5.7. Ask Open-Ended Questions

Open-ended questions let you understand a user’s situation more comprehensively. For example, asking “How are you feeling?” enables a broader range of responses than “Are you in pain?”, potentially revealing more about a patient’s mental, emotional, and physical state. Open-ended questions also help generate large training datasets, effectively improving AI models, and let ChatGPT draw on its training across a wider range of topics — potentially producing unexpected, creative solutions healthcare professionals might not have thought of. Below is an example.

Closed question: “Is exercise important for osteoporosis patients?”

Open question: “What benefits does regular physical activity bring to osteoporosis patients?”

Asking for concrete cases clarifies the meaning of concepts and makes them easier to understand — especially for complex medical terminology or procedures, where examples provide practical context. Examples also help visualize abstract ideas: when ChatGPT gives concrete examples, it shows how a concept applies across scenarios, which is valuable for connecting theory to real-world healthcare.

5.8. Temporal Awareness

Temporal awareness refers to a model’s ability to understand time-related concepts and generate responses relevant to a given point in time. Explaining which timeline your prompt relates to helps LLMs give more useful answers. Below is an example.

Example without time reference: “Explain the healing process after knee surgery.”

Example with time reference: “What can patients typically expect during the first 6 weeks of healing after knee surgery?”

5.9. Set Realistic Expectations

Setting realistic expectations requires understanding the limitations of AI tools like ChatGPT. ChatGPT cannot access information from after 2021-11, does not provide individualized medical advice, and does not replace professional judgment. Below is an example.

Unrealistic prompt: “What is the latest research on Alzheimer’s disease published this month?”

Realistic prompt: “What were the major research advances in Alzheimer’s disease treatment up to 2021?”

5.10. Use One-Shot / Few-Shot Prompting

One-shot / few-shot prompting refers to ChatGPT’s ability to generate a response based on a single example or piece of context you provide. Below is an example.

“Generate 10 names for a new digital stethoscope device.

My favorite name is DigSteth.”

Using a few-shot strategy, ChatGPT can generate a response based on multiple examples or pieces of context you provide. Below is an example.

“Generate 10 names for a new digital stethoscope device.

Examples of names I like are as follows.

Digital

Steth

oscope”

One of the easiest ways to improve Prompt Engineering is having ChatGPT help design prompts for users. Below is an example. “What prompt could I use right now to get a better output for this thread/task?”

In addition to these general approaches, we have compiled specific recommendations and practical examples that healthcare professionals should consider to improve their Prompt Engineering skills. Please refer to our guide article.

How to Implement Prompt Engineering | Prompt Engineering Guide

6. Conclusion

Because Prompt Engineering skills are attracting significant interest worldwide, especially in healthcare, it is important to incorporate the practical methods in this article into medical curricula and graduate education. Outside healthcare, Prompt Engineering is already applied in marketing and customer service. Generative AI’s technical details will likely enter future curricula, but until then, medical students would benefit from learning practical tips for using AI models like ChatGPT.

Miichisoft is a Vietnamese technology company with numerous successful case studies in cross-platform development for Japan. We have built our track record through more than 150 successful projects. We are confident in our ability to lead IT projects and are proud to contribute to our clients’ business success.

Furthermore, Miichisoft itself has recently completed and released dxGAI Business | Enterprise, our own comprehensive, forward-looking AI solution built on ChatGPT. As a company ourselves, we understand the challenges of using AI solutions, and we can support you whether you want to start using Prompt Engineering technology in your business activities or want to use it effectively without worrying about safety.

If you are interested in AI technologies such as Prompt Engineering, please feel free to contact us — let’s explore business collaboration opportunities together.