david-sprague/Medical-Health-QA-Articles-Dataset

Medical Health Q&A & Articles Dataset — iCliniq, HealthTap & WebMD

A multi-source medical Q&A and health articles dataset combining doctor-answered questions and medically reviewed content from iCliniq, HealthTap, and WebMD. Built for LLM fine-tuning, medical chatbot training, clinical NLP research, and healthcare AI development.

Dataset Overview

What Makes This Dataset Valuable

• —Verified doctor answers — every iCliniq Q&A pair is answered by a licensed medical professional, with author_speciality field covering 80+ specialities at 99.7% coverage
• —Multi-source unified schema — iCliniq, HealthTap, and WebMD all share identical field names and formats, no separate preprocessing needed per source
• —content_type field — cleanly separates Q&A pairs from articles, enabling task-specific filtering in one line of code
• —50,000+ iCliniq records — one of the largest verified medical dialogue datasets available outside academic institutions
• —author_speciality at 99.7% — enables speciality-specific fine-tuning for cardiology, dermatology, psychiatry, pediatrics, and 80+ other fields
• —100% body_text and answer_text coverage — no missing primary fields, production-ready out of the box

Data Fields

Available in Full Dataset at crawlfeeds.com

conditiontags, bodypart, contentsubtype, medicallyreviewed, reviewername, authorverified, helpfulcount, viewscount, publishedat, updatedat, wordcount, scrapedat

Use Cases

LLM Fine-Tuning & Pretraining

• —Medical domain adaptation — fine-tune LLaMA, Mistral, GPT, or any open-source LLM on verified clinical text
• —Instruction tuning — use title + bodytext + answertext as instruction-response pairs for medical assistant models
• —RAG knowledge base — index article content from WebMD for retrieval-augmented medical Q&A systems

Medical Chatbot & Virtual Assistant

• —Symptom checker training — natural symptom-driven questions from HealthTap mirror real user inputs
• —Doctor-patient dialogue models — iCliniq's verified Q&A pairs are ideal for clinical dialogue training
• —Triage assistant development — train models to route patient questions to appropriate specialities

Clinical NLP Research

• —Named entity recognition — extract medical entities from bodytext and answertext
• —Relation extraction — identify symptom-condition-treatment relationships
• —Text classification — use category and author_speciality as classification labels
• —Question answering benchmarks — evaluate LLM medical knowledge using Q&A pairs

Healthcare AI Products

• —Medical knowledge bases — build structured health information systems from article content
• —Clinical decision support — use verified doctor answers as evidence base for AI recommendations
• —Medical education tools — build study aids and exam preparation tools across 80+ specialities

Sample Data

{
  "uniq_id": "icliniq_00042891",
  "content_type": "qa",
  "source": "iCliniq",
  "source_domain": "icliniq.com",
  "title": "What causes persistent lower back pain after long sitting hours?",
  "body_text": "I am 34 years old and have been experiencing lower back pain for the past 3 months. The pain gets worse after sitting for long periods at my desk job. It radiates slightly to my left leg.",
  "answer_text": "Based on your description, this sounds consistent with lumbar disc irritation or early-stage sciatica. The radiation to the left leg suggests possible nerve involvement at L4-L5 or L5-S1 level. I recommend an MRI of the lumbar spine and physiotherapy.",
  "category": "Orthopedics",
  "author_speciality": "Orthopedic Surgeon",
  "language": "en"
}

Loading the Dataset

from datasets import load_dataset
import pandas as pd

dataset = load_dataset("crawlfeeds/Medical-Health-QA-Articles-Dataset")
df = dataset["train"].to_pandas()

# Filter Q&A pairs only
qa_data = df[df["content_type"] == "qa"]

# Filter articles only
articles = df[df["content_type"] == "article"]

# Filter by source
icliniq = df[df["source"] == "iCliniq"]

# Filter by medical speciality
cardiology = df[df["author_speciality"].str.contains("Cardiol", na=False)]

# Prepare instruction tuning pairs
training_pairs = qa_data[["title", "body_text", "answer_text"]].dropna()

Full Dataset & Custom Data

This is a 1,000 record sample. The full dataset contains:

• —50,000+ verified iCliniq Q&A pairs
• —HealthTap and WebMD content
• —Complete conditiontags, bodypart, and review fields
• —Custom subsets by speciality, source, or content type
• —Monthly refresh options available

Get the full dataset: crawlfeeds.com

Request custom medical data: crawlfeeds.com/contact_us

Related Datasets by Crawl Feeds

• —Booking.com Hotel Reviews Dataset
• —Trustpilot Reviews Dataset – 20K Sample
• —IKEA Home Decor & Furniture Dataset
• —Home Depot Smart Home Dataset
• —Walmart Reviews Dataset
• —Reddit Medical Discussions Dataset (coming soon)

License

Available under CC BY-NC 4.0 for research and non-commercial use. Commercial licensing available at crawlfeeds.com/contact.

Citation

@dataset{crawlfeeds_medical_qa_2025,
  author    = {Crawl Feeds},
  title     = {Medical Health Q&A and Articles Dataset},
  year      = {2025},
  publisher = {Hugging Face},
  url       = {https://huggingface.co/datasets/crawlfeeds/Medical-Health-QA-Articles-Dataset}
}

Data collected and maintained by [Crawl Feeds](https://crawlfeeds.com) — structured web data for AI, analytics, and business intelligence.