Datasetpublic

adachankawai/healthcare-dataset

Vietnamese Healthcare Synthetic Patient Records This dataset contains synthetic Vietnamese healthcare identity records from multiple source systems, plus a canonical synthetic patient table used by the generator. All records are synthetic and are intended for entity resolution, record linkage, and Vietnamese identity-field preprocessing experiments. Included Files Only the following CSV files are included in this upload: File Rows Description… See the full description on the dataset page: https://huggingface.co/datasets/adachankawai/healthcare-dataset.

sourceHugging Faceupdated 4mo agoView on Hugging Face

0likes85downloads

Dataset Card

Vietnamese Healthcare Synthetic Patient Records

This dataset contains synthetic Vietnamese healthcare identity records from multiple source systems, plus a canonical synthetic patient table used by the generator.

All records are synthetic and are intended for entity resolution, record linkage, and Vietnamese identity-field preprocessing experiments.

Included Files

Only the following CSV files are included in this upload:

File	Rows	Description
`clinic_patients.csv`	6,053	Source table for clinic patient records
`hospital_patients.csv`	6,143	Source table for hospital patient records
`insurance_members.csv`	6,014	Source table for insurance member records
`lab_test_patients.csv`	6,105	Source table for lab test patient records
`pharmacy_prescriptions.csv`	6,068	Source table for pharmacy prescription records
`vaccination_records.csv`	6,049	Source table for vaccination records
`true_patients.csv`	10,000	Synthetic canonical patient table used by the generator

The six source tables contain noisy observed identity records. true_patients.csv contains the canonical synthetic patient attributes before source-specific noise is applied.

Source Table Columns

The six source-system files share the same schema:

Column	Description
`unique_id`	Unique record identifier
`source_system`	Source system name
`source_record_id`	Record identifier within the source system
`full_name`	Observed full name
`date_of_birth_raw`	Observed date of birth
`gender_raw`	Observed gender
`phone_raw`	Observed phone number
`email_raw`	Observed email address
`cccd_raw`	Observed CCCD/CMND-style national identifier
`national_id_type_raw`	Observed national ID type
`health_insurance_id_raw`	Observed health insurance identifier
`address_text`	Observed address string
`province_raw`	Observed province
`district_raw`	Observed district
`ward_raw`	Observed ward
`admin_address_version`	Current or legacy administrative address version
`facility_name`	Facility name associated with the record
`event_date`	Event or record date

These files intentionally use raw observed fields. Normalized fields such as phone_norm, identity_text, and name_address_text should be created during preprocessing if needed.

Canonical Patient Columns

true_patients.csv contains the canonical synthetic patient records:

Column	Description
`entity_id_truth`	Synthetic true patient/entity identifier
`canonical_full_name`	Canonical full name
`gender`	Canonical gender
`date_of_birth`	Canonical date of birth
`phone`	Canonical phone number
`email`	Canonical email address
`cccd`	Canonical CCCD identifier
`legacy_cmnd`	Canonical legacy CMND identifier
`health_insurance_id`	Canonical health insurance identifier
`house_number`	Canonical house number
`street`	Canonical street
`legacy_province`	Province in legacy administrative address
`legacy_district`	District in legacy administrative address
`legacy_ward`	Ward in legacy administrative address
`legacy_full_address`	Full legacy address
`current_province`	Province in current administrative address
`current_district`	District in current administrative address
`current_ward`	Ward in current administrative address
`current_full_address`	Full current address
`_hard_negative_partner`	Generator helper field for hard-negative construction
`_hard_negative_pair_type`	Generator helper field for hard-negative construction

Loading Examples

With pandas:

python

import pandas as pd

clinic = pd.read_csv("clinic_patients.csv")
hospital = pd.read_csv("hospital_patients.csv")
true_patients = pd.read_csv("true_patients.csv")

After uploading to Hugging Face:

python

from datasets import load_dataset

repo_id = "YOUR_USERNAME/YOUR_DATASET_REPO"
clinic = load_dataset(repo_id, data_files="clinic_patients.csv")["train"]
true_patients = load_dataset(repo_id, data_files="true_patients.csv")["train"]

Load all six source tables as named splits:

python

from datasets import load_dataset

repo_id = "YOUR_USERNAME/YOUR_DATASET_REPO"
data_files = {
    "clinic": "clinic_patients.csv",
    "hospital": "hospital_patients.csv",
    "insurance": "insurance_members.csv",
    "lab_test": "lab_test_patients.csv",
    "pharmacy": "pharmacy_prescriptions.csv",
    "vaccination": "vaccination_records.csv",
}
sources = load_dataset(repo_id, data_files=data_files)

Intended Use

This dataset is intended for:

—record linkage across Vietnamese healthcare source systems;
—identity-field preprocessing and normalization experiments;
—synthetic data quality checks;
—development of entity resolution pipelines.

Limitations

This is a synthetic dataset and should not be used as real healthcare evidence or as a demographic sample. The data is designed for controlled technical experiments only.