[Paper Review] Prompt Engineering for Healthcare: Methodologies and Applications

Jiaqi Wang, Enze Shi+22 more|arXiv (Cornell University)|Apr 28, 2023

Topic ModelingComputer Science116 citations

TL;DR

A comprehensive review of prompt engineering in medical NLP, covering prompt types, design methods (manual and automated), LLMs, and healthcare applications, with guidance for researchers and future directions.

ABSTRACT

Prompt engineering is a critical technique in the field of natural language processing that involves designing and optimizing the prompts used to input information into models, aiming to enhance their performance on specific tasks. With the recent advancements in large language models, prompt engineering has shown significant superiority across various domains and has become increasingly important in the healthcare domain. However, there is a lack of comprehensive reviews specifically focusing on prompt engineering in the medical field. This review will introduce the latest advances in prompt engineering in the field of natural language processing for the medical field. First, we will provide the development of prompt engineering and emphasize its significant contributions to healthcare natural language processing applications such as question-answering systems, text summarization, and machine translation. With the continuous improvement of general large language models, the importance of prompt engineering in the healthcare domain is becoming increasingly prominent. The aim of this article is to provide useful resources and bridges for healthcare natural language processing researchers to better explore the application of prompt engineering in this field. We hope that this review can provide new ideas and inspire for research and application in medical natural language processing.

Motivation & Objective

• Summarize the development and role of prompt engineering in healthcare NLP.
• Survey the types of prompts (manual and automated) and their design methods.
• Analyze healthcare applications of prompts across classification, generation, detection, augmentation, QA, and inference.
• Highlight challenges, future directions, and resource gaps for medical prompt engineering.

Proposed method

• Define the scope and conduct a literature review of prompt engineering in medical NLP from 2019 to 2023.
• Characterize large language models and their relevance to medical prompts (e.g., BERT, T5, GPT variants).
• Catalog prompt formats (cloze vs prefix; manual vs automated; discrete vs continuous) and their construction.
• Compare manual and automated prompting approaches, including zero-shot and few-shot strategies.
• Summarize applications of prompts to medical tasks such as classification, generation, detection, augmentation, QA, and inference.

Experimental results

Research questions

• RQ1 What are the main prompt engineering techniques currently used in healthcare NLP?
• RQ2 How do manual and automated prompts compare in terms of effectiveness and practicality for medical tasks?
• RQ3 What healthcare NLP tasks have benefited most from prompt-based methods, and what are representative applications?
• RQ4 What are the key challenges and future directions for prompting in medical domains?

Key findings

• Prompt engineering leverages LLMs to perform few- or zero-shot learning for medical NLP tasks.
• Manual prompts (zero-shot and few-shot) show strong performance, with domain-specific prompts yielding improvements in clinical text tasks.
• Automated prompts (discrete and continuous) offer scalable design and task adaptation, with methods like prompt mining, paraphrasing, generation, and scoring enhancing efficiency.
• Continuous prompting enables task-specific optimization in embedding space and supports multimodal and low-resource scenarios.
• Applications span classification, generation, detection, augmentation, question answering, and inference within clinical and biomedical contexts.
• The review identifies current challenges and suggests directions for future research to advance medical prompt engineering.