Mastering MeSH, Zotero/EndNote, CASP, PRISMA—and AI: A Step-by-Step Research Framework for Young Researchers

If this research framework advances your work, consider supporting my independent medical-climate and ministry publications. Every coffee fuels the next deep dive!

☕ Buy Me a Coffee ❤️ Support on Ko-fi

Why this matters (especially in the AI era)

Many early-career researchers start by using everyday search terms—or asking AI to “write a literature review.” That approach may feel efficient, but it is also where common failures arise: incomplete searches, weak appraisal, and citations that cannot be verified.

For Ethical use of AI with resources. 

The fix is not to abandon AI; it is to anchor AI inside proven research frameworks:

- MeSH makes your searching precise and reproducible.
- Zotero/EndNote makes your citations consistent, traceable, and auditable.
- CASP helps you separate robust evidence from poor-quality studies.
- PRISMA ensures your systematic review is transparent, complete, and defensible. 

Think of it this way: AI is the engine. Research frameworks are the steering wheel and brakes.

Step 1 — MeSH: Speak the language of medical literature

What MeSH actually does

Medical Subject Headings (MeSH) are the National Library of Medicine’s controlled vocabulary used to index PubMed/MEDLINE records. MeSH standardises concepts so that different author wordings map to the same underlying topic.

If one paper says “heart attack” and another says “acute myocardial infarction,” MeSH helps you retrieve both through the standard heading (Myocardial Infarction).

A practical MeSH workflow (beginner-friendly)

1. Open the official MeSH Database (NLM).
2. List your main concepts (usually 2–4): condition, intervention/exposure, comparison, outcome.
3. Search each concept individually and select the most accurate MeSH heading.
4. Check the hierarchy (“tree”) to decide whether to broaden or narrow your scope.
5. Add qualifiers when appropriate (e.g., therapy, diagnosis, adverse effects).
6. Build your Boolean strategy using AND/OR/NOT and export/save it.
7. Use this exact strategy in your AI tool (instead of conversational prompts). This reduces irrelevant retrieval and makes your process reproducible.

Clinical reality check: MeSH is powerful, but very new articles may not yet be fully indexed—so you still often pair MeSH with well-chosen keyword synonyms.

Step 2 — Zotero or EndNote: Build your “external academic brain”

Why reference managers are non-negotiable

A serious research project quickly becomes hundreds (or thousands) of PDFs, notes, and citations. If you manage that manually, errors and inefficiency are inevitable.

Zotero is excellent for most students and is free.
EndNote is feature-rich and commonly supported by universities.

A clean workflow that prevents citation chaos

1. Install the full setup
   - Desktop app
   - Browser connector
   - Word/Docs plugin
2. Create collections that match your writing structure
   - Introduction
   - Methods
   - Results
   - Discussion
   - Subfolders by theme (e.g., “Immunotherapy trials,” “Adverse effects,” “Meta-analyses”)
3. Import citations the right way
   Use the browser connector from PubMed/journal sites so you capture correct metadata early.
4. Use the manager to audit AI-generated references
   If AI suggests references, import them via RIS/BibTeX, then verify:
   - Does the DOI resolve correctly?
   - Do journal name, year, volume, and authors match?
   - Can you find the article in a real database?
     If not, do not cite it.
5. Cite-as-you-write
   Insert citations directly in Word/Docs and generate the bibliography automatically. This eliminates formatting errors and makes journal submission much easier.

Step 3 — CASP: Appraise evidence before you trust it

The problem CASP solves

“Published” does not mean “high quality,” and AI summaries can unintentionally obscure bias, confounding, or poor methods. CASP gives you structured checklists to assess whether a study is valid, what it actually found, and whether it applies to your setting.

A simple CASP routine for every paper you might cite heavily

1. Choose the correct CASP checklist (RCT, cohort, systematic review, etc.).
2. Answer the initial screening questions first
   If the research question is unclear or the design is inappropriate, stop early.
3. Evaluate internal validity
   Focus on the few items that most often break studies:
   - randomisation / allocation concealment (if RCT)
   - blinding (where feasible)
   - confounding control (especially observational studies)
   - completeness of follow-up
4. Interpret results properly
   Prioritise effect sizes and confidence intervals over “p-value excitement.”
5. Assess applicability
   Decide whether the population and context match yours.

How AI fits in: AI can help highlight where the paper discusses randomisation, blinding, outcomes, and limitations—but you must verify against the PDF text.

Step 4 — PRISMA: The gold standard for systematic reviews

PRISMA 2020 is a widely accepted reporting guideline built to ensure systematic reviews are transparent, complete, and reproducible. It includes a checklist and a flow diagram that documents how studies moved from identification to inclusion. [1]

A PRISMA-first workflow (do this from day one)

1. Download the PRISMA 2020 checklist and flow diagram. [1]
2. Pre-plan what you will report
   PRISMA works best when it guides your methods early—not when you try to “fix reporting” at the end. [1]
3. Document every search
   Record the database, date, full strategy (including MeSH/keywords), and any filters used. [1]
4. Track your screening in real time
   Maintain accurate counts for:
   - records identified
   - duplicates removed
   - titles/abstracts screened
   - full-text assessed
   - excluded with reasons
   - included in qualitative/quantitative synthesis [1]
5. Disclose AI use transparently
   If AI helped with screening, translation, extraction, or summarisation, report what was used and how it was used so readers can interpret the work appropriately. PRISMA is fundamentally about transparency. [1]

The unified workflow (the version you should actually follow)

MeSH → build validated, reproducible search strings
Databases + AI (controlled) → accelerate discovery and summarisation with verification
Zotero/EndNote → organise, deduplicate, and ensure citations are real
CASP → filter for quality and applicability
PRISMA → document the entire process transparently for publication 

Final thoughts

AI can dramatically improve efficiency, but credibility still depends on method. If you master MeSH, Zotero/EndNote, CASP, and PRISMA, you will not only write faster—you will write work that is defensible, reproducible, and publishable. PRISMA 2020 remains a cornerstone for systematic review reporting, and using it early strengthens both your methods and your manuscript. 

Author: Dr. William Lamptey (Dr.med.Dr.Dr.)

Fuel the Mission of Medicine & Ministry
If this deep-dive research framework saved you time or blessed your ministry work, consider supporting my independent writing! Your sponsorship directly funds free, accessible resources for young scholars worldwide. 👇