README.md

metadata

language:
  - en
license: mit
task_categories:
  - tabular-classification
  - text-classification
  - feature-extraction
tags:
  - medical
  - pharmacy
  - healthcare
  - drug-recommendation
  - symptom-checking
  - clinical-intelligence
pretty_name: AYUVANT Clinical Intelligence and Pharmacy Dataset
size_categories:
  - n<1K
configs:
  - config_name: default
    data_files:
      - ayuvant_cleaned_dataset.json

🏥 AYUVANT Clinical Intelligence and Pharmacy Dataset

Curated, structured, and continuously maintained via Adaptive Data — an intelligent dataset management platform that enables versioning, schema validation, and real-time collaboration across clinical data pipelines.

---

📋 Dataset Description

The AYUVANT Clinical Intelligence and Pharmacy Dataset is a structured, multi-config dataset designed to power clinical decision-support systems, drug recommendation engines, and pharmacy intelligence tools. It covers the full clinical data lifecycle — from symptom detection through disease mapping to medicine dispensation and transaction logging.

This dataset was built and is actively maintained on Adaptive Data, which provides:

• 🔄 Adaptive versioning — every schema change and data update is tracked with full lineage
• ✅ Automated validation — JSON configs are validated against clinical ontologies on every push
• 🤝 Collaborative curation — multi-contributor workflows with conflict resolution built in
• 📊 Usage analytics — real-time monitoring of how each config is consumed downstream

---

🗂️ Dataset Configs

This dataset is organized into 7 focused configs, each representing a distinct clinical domain:

Config	File	Description
diseases	diseases.json	Structured disease ontology with ICD-adjacent categorization
symptoms	symptoms.json	Granular symptom taxonomy linked to disease clusters
medicines	medicines.json	Drug catalog with dosage, category, and indication metadata
salt_compositions	salt_compositions.json	Active pharmaceutical ingredient (API) profiles
side_effects	side_effects.json	Adverse effect registry mapped to medicines and compositions
suspicious_health_issues	suspicious_health_issues.json	Flagged symptom patterns for triage and clinical escalation
transaction_records	transaction_records.json	Anonymized pharmacy dispensation and prescription records

---

⚡ Quickstart

Load a Single Config

from datasets import load_dataset

# Load the disease ontology
diseases = load_dataset("Anshulpj12/Adaptive_Dataset_med", name="diseases")
print(diseases["train"][0])

# Load the medicine catalog
medicines = load_dataset("Anshulpj12/Adaptive_Dataset_med", name="medicines")
print(medicines["train"][0])

Load All Configs

from datasets import load_dataset

configs = [
    "diseases",
    "symptoms",
    "medicines",
    "salt_compositions",
    "side_effects",
    "suspicious_health_issues",
    "transaction_records",
]

ayuvant = {cfg: load_dataset("Anshulpj12/Adaptive_Dataset_med", name=cfg) for cfg in configs}

# Example: cross-reference symptoms with diseases
for record in ayuvant["symptoms"]["train"]:
    print(record)

Symptom-to-Medicine Pipeline Example

from datasets import load_dataset

symptoms_ds   = load_dataset("Anshulpj12/Adaptive_Dataset_med", name="symptoms")["train"]
diseases_ds   = load_dataset("Anshulpj12/Adaptive_Dataset_med", name="diseases")["train"]
medicines_ds  = load_dataset("Anshulpj12/Adaptive_Dataset_med", name="medicines")["train"]
side_fx_ds    = load_dataset("Anshulpj12/Adaptive_Dataset_med", name="side_effects")["train"]

# Build lookup maps
disease_map  = {d["id"]: d for d in diseases_ds}
medicine_map = {m["id"]: m for m in medicines_ds}

def clinical_lookup(symptom_query: str):
    """Minimal clinical decision-support lookup."""
    matched_symptoms = [
        s for s in symptoms_ds
        if symptom_query.lower() in s.get("name", "").lower()
    ]
    results = []
    for sym in matched_symptoms:
        for disease_id in sym.get("linked_diseases", []):
            disease = disease_map.get(disease_id, {})
            meds    = [medicine_map[m] for m in disease.get("recommended_medicines", [])
                       if m in medicine_map]
            results.append({
                "symptom":  sym["name"],
                "disease":  disease.get("name"),
                "medicines": [m["name"] for m in meds],
            })
    return results

print(clinical_lookup("fever"))

---

🔬 Intended Use Cases

• Drug Recommendation Systems — map symptoms → diseases → medicines with side-effect filtering
• Clinical Triage Tools — flag suspicious symptom clusters for escalation
• Pharmacy Intelligence — analyse dispensation patterns from transaction records
• Medical NLP Training — structured labels for symptom extraction and entity recognition
• Healthcare Tabular ML — tabular classification benchmarks on clinical data

---

🏗️ Adaptive Data Platform Integration

This dataset is actively managed through Adaptive Data.

What "Active Usage" Means Here

Platform Feature	How It's Used in This Dataset
Schema Registry	Each config's JSON structure is registered and enforced — malformed entries are rejected at ingest
Version Lineage	Every medicine, disease, or symptom addition is logged with contributor ID and timestamp
Diff Reviews	Pull-request-style reviews gate any changes to medicines.json or suspicious_health_issues.json
Automated QA	On every push, Adaptive Data runs null-check, duplicate-ID, and cross-reference integrity tests
Dataset Snapshots	Tagged stable releases (e.g., v1.0, v1.1) are pinned so downstream models can lock to a version

Syncing from Adaptive Data

# If you have Adaptive Data CLI configured
# adaptive pull Adaptive_Dataset_med --config medicines --format json

# Or via the Adaptive Data Python SDK
from adaptive_data import AdaptiveClient

client = AdaptiveClient(api_key="YOUR_API_KEY")
dataset = client.dataset("Adaptive_Dataset_med")

medicines_snapshot = dataset.config("medicines").pull(version="latest")
medicines_snapshot.to_json("medicines.json")

---

📊 Dataset Statistics

Config	Approximate Records	Format	Updated
diseases	—	JSON array	Tracked via Adaptive Data
symptoms	—	JSON array	Tracked via Adaptive Data
medicines	—	JSON array	Tracked via Adaptive Data
salt_compositions	—	JSON array	Tracked via Adaptive Data
side_effects	—	JSON array	Tracked via Adaptive Data
suspicious_health_issues	—	JSON array	Tracked via Adaptive Data
transaction_records	—	JSON array	Tracked via Adaptive Data

Record counts update continuously. See the Adaptive Data dashboard for live statistics.

---

⚠️ Limitations and Bias

• This dataset is intended for research and development purposes only and is not a substitute for professional medical advice.
• Drug recommendations and clinical associations reflect the curation logic at time of release and may not capture the latest clinical guidelines.
• Transaction records are fully anonymized — no personally identifiable information (PII) is present.
• Coverage is weighted toward common conditions; rare disease representation may be limited.

---

📜 License

This dataset is released under the MIT License. See LICENSE for full terms.

---

🙏 Credits and Acknowledgements

Dataset curation and maintenance platform:

Adaptive Data — Intelligent dataset lifecycle management for structured and clinical data. Schema validation, version control, collaborative curation, and automated QA pipelines powering the AYUVANT dataset from ingestion to Hugging Face publication.

Project: AYUVANT Clinical Intelligence System Maintainer: (your name / organisation) Contact: (your contact / repo link)

---

📎 Citation

If you use this dataset in your research, please cite:

@dataset{ayuvant_clinical_2025,
  title     = {AYUVANT Clinical Intelligence and Pharmacy Dataset},
  author    = {Anshul Prajapati},
  year      = {2025},
  publisher = {Hugging Face},
  url       = {https://huggingface.co/datasets/Anshulpj12/Adaptive_Dataset_med},
  note      = {Curated and maintained via Adaptive Data (https://adaptivedata.io)}
}