Data Overview
Access 15M+ de-identified patient records across all major clinical domains, assembled into fit-for-purpose datasets tailored to your research needs.
15M+ De-identified Patient Records
Lighthouse maintains a repository of longitudinal, structured patient records spanning 10+ years of clinical history. All data has been de-identified using expert determination methods in accordance with HIPAA Privacy Rule standards.
Rather than providing access to the entire repository, we assemble configurable datasets tailored to your specific cohort definitions, time windows, and data element requirements. This ensures you receive only the data needed for your protocol or model—nothing more, nothing less.
Available Data Modules
All data elements are fully de-identified and structured for analysis
Demographics
Age, gender, race/ethnicity, geographic indicators, and insurance coverage
Encounters
Inpatient admissions, outpatient visits, ER encounters, and care setting details
Diagnoses
ICD-10 coded diagnoses with dates, position indicators, and diagnosis-related groupings
Procedures
CPT/HCPCS coded procedures, surgical interventions, and procedure dates
Medications
NDC-coded prescriptions, fills, dosing instructions, and therapy durations
Labs & Vitals
LOINC-coded lab results, vital signs, and clinical measurements with result values
Dataset Configuration
Every dataset is custom-assembled to match your protocol and analysis requirements
Cohort Definition
Specify inclusion/exclusion criteria based on diagnoses, procedures, medications, demographics, and time-based constraints
Time Windows
Define observation periods, lookback windows, and longitudinal coverage requirements for your study
Refresh Cadence
One-time delivery or ongoing refreshes to support longitudinal studies and updated analyses
Delivery Formats
CSV, Parquet, or custom formats delivered via S3, Snowflake, SFTP, or secure file transfer
Sample Schema Available
Review a sample data schema to understand our data structure, field definitions, and coding standards before requesting a custom dataset.
Request Sample Schema
