Maccabi Library Discover

GenAI Prompt Tips in RAG

Whenever using GenAI for consultation or research, keep the Human-in-the-loop.

Remember that EBM must include 3 components: Research, Clinical Experience, Patient Values:

Find the evidence, evaluate the evidence in the context of their knowledge and the needs of the individual patient before making a clinical decision

Learn more about responsible AI

הצעות לבניית פרומפטים אפקטיביים

מספר דוגמאות לתבניות פרומפטים בשימוש ב GenAI לספרות מחקרית וקלינית, ולשימוש כללי

  • The T-C-G-R-E-I AI Prompt Framework - Task, Context, Goal, References, Evaluate, Iterate
  • PICO - Population (persona), Intervention (task), Comparison (contrast), Outcome (output, format)
  • Structured - task, context, example, persona, format, tone
  • SMART - Seeker, Mission, AI role, Register, Targeted question
  • The T-C-G-R-E-I AI Prompt Framework: How Top AI Users Get the BEST Results

    Full guide - PDF

    The video 

    Source: https://www.youtube.com/@AIandTechforEducation

  • PICO Reinterpretation for AI 

    Element Equivalent in prompt engineering User purpose AI interpretation
    P (Population) Persona: to whom the response is directed Determine to whom the response should be written (eg, a student, patient, clinician, etc) Adjusts the vocabulary, technicism, and discursive language
    I (Intervention) Task Define the action that is expected from the AI (eg, summarize, analyze, compare, etc) Chooses the type of semantic processing and the output structure
    C (Comparison) Contrast element (optional) This option may be included if you want the AI to test differences between 2 or more conditions, treatments, or mechanisms Structures responses in a comparative in differential format
    O (Outcome) Format + tone + output Specifies the type of output (list, narrative, table), the style (academic, informative, normative), and the end use (teaching, clinical decision, research, etc) Generates the output type closest to the instructed pattern

    source: Jaime Andrés Pereañez, Johan Granados, Pedro Amariles, PICO-like prompts for clinical pharmacy in generative artificial intelligence, American Journal of Health-System Pharmacy, 2025;, zxaf357, https://doi.org/10.1093/ajhp/zxaf357

  • A structured GenAI prompt for medical and clinical literature research and evidence.

    Check it out in Open Evidence

    Task Gather information on the impact of sleep disorders on mental health in adolescents.
    Context Specific to sleep disorders such as insomnia and sleep apnea and their psychological effects like anxiety and depression.
    Example Focus on studies that explore the correlation between sleep disorders and mental health outcomes in adolescent populations.
    Persona For a medical faculty member preparing an evidence-based lecture for medical students.
    Format A detailed summary with key findings from recent research studies, organized by type of sleep disorder.
    Tone Academic and informative.

    Prompt: "Identify and summarize the latest research on the impact of sleep disorders, such as insomnia and sleep apnea, on mental health outcomes in adolescents, specifically focusing on anxiety and depression. Present the findings in a detailed summary, organized by type of 
    sleep disorder. This information is intended for a medical faculty member preparing an evidence-based lecture for medical students, so the tone should be academic and informative."

    source: Robinson KL, Bontekoe KR, Muellenbach JN. Integrating PICO principles into generative artificial intelligence prompt engineering to enhance information retrieval for medical librarians. J Med Libr Assoc.  2025 Apr 18;113(2):184-8.  doi:10.5195/jmla.2025.2022  

  • The SMART Prompt Format

    Part in prompt Clinical scenarios Theoretical questions
    Seeker
    It refers to the identity and perspective of the user inquiring. The prompt must communicate the seeker's role and the clinical context to ensure that the AI understands the expertise level and specific needs of the user. I'm a head and neck surgeon with over 15 y of experience, working in a tertiary-level hospital, and I specialize in (type of surgery relevant to the scenario). I am a fifth-year resident specializing in head and neck surgery.
    Mission
    It defines the purpose of the inquiry. This involves clearly stating the objective or problem that the AI is expected to address. The mission statement ensures that the AI focuses on the most relevant information. I need your advice on how to manage a patient I'm treating. I need your help to prepare for my final residency exam.
    AI Role
    It describes the expected role of the AI in the interaction, which may range from providing information to offering clinical advice. Clearly defining the AI's role helps set expectations for the type and depth of the response. You are the world's leading expert in (type of surgery relevant to the scenario). You are a full professor of head and neck surgery at the most prestigious university in the world.
    Register
    It involves the tone and style of language to be used in the AI's response. The prompt should guide the AI to use language that is appropriate for the target audience, ensuring clarity and avoiding ambiguity. Correct and specialized scientific language. The information must be based on the most recent and solid scientific evidence. I would like you to also provide me with the bibliographical references from which you draw your information. Correct and specialized scientific language. The information must be based on the most recent and solid scientific evidence. I would like you to also provide me with the bibliographical references from which you draw your information.
    Targeted Question
    It refers to the specific question or query posed to the AI. This question should be direct and focused, designed to elicit a detailed and accurate response. (Noncontextualized question) (Noncontextualized question)

    Source: Vaira L.A., Lechien J.R., Abbate V., Gabriele G., Frosolini A., De Vito A., et al. Enhancing AI Chatbot Responses in Health Care: The SMART Prompt Structure in Head and Neck Surgery. OTO Open. 2025;  9 (1): n/a.
    DOI: 10.1002/oto2.70075 

RAG is the common type of tools we use to look for updated information. 

LLM = Large Language Model -> Thinks

RAG = Retrieval Augmented Generation -> Reads

AI Agent -> Acts

MCP = Model Context Protocol -> Connects

Source: LinkedIn

איך נבחר את הכלי המתאים?

Purpose Databases (Full text, Clinical) /
Search engines (Bibliographic)

GenAI tools

Clinical Evidence, EBM
 specific case, right to the point, concise, direct answer, decision support
UpToDate, Cochrane, DynaMed Open Evidence
Clinical Guidelines
 
general, update
UpToDate, Cochrane, DynaMed, Clinical Key /
PubMed
Consensus, Open evidence
Research
 learn more about a subject, stay current
Clinical Key /
PubMed, Google Scholar, TDNet discover,
Perplexity, Consensus,
TDNet AI
Research
 create my library, collaborate
  Perplexity, Consensus,
TDNet AI

Use EBM PICO query model to build your GenAI prompt for better focus and precision in the results. Example:

For the query: what evidence supports carbohydrate diet for adults with type 2 diabetes?

Use the prompt: for adults with type 2 diabetes (P), does a low carbohydrate diet (I) compared to a standard diet (C) improve blood sugar control (O)

Learn more about PICO

Prompt Engineering: The CLEAR Framework

Concise, Logical, Explicit, Adaptive, and Reflective (CLEAR).

ערוצי יוטיוב מומלצים עם הצגות, הדרכות וביקורות של כלי AI במחקר

Dr Andy Stapleton

      עוד AI אחד ודי: טל קמינסקי רוזנברג, דיאטנית ומידענית רפואית מספריית הטכניון

טוען...