Skip to content
Andreas Vesalius

Data

What the data contain

A longitudinal, de-identified record of a defined population in Argentina, the first source in a Latin American network, assembled from a private payer and a high-complexity hospital that belong to the same group. The platform that ingests, pseudonymises and models the data is called Fabrica.

Layers

Four layers, one person key

Every layer is linked through the same pseudonymous person identifier, so a question can move from a reimbursed claim to the laboratory result behind it and to the cost of the admission that followed.

  1. 01

    Financial

    The payer’s view of care: what was requested, authorised, delivered and paid, with the real cost to the payer.

    Source
    Private payer
    Evidence type
    Claims-inferred
    Typical content
    Authorisations, reimbursed procedures and visits, dispensed drugs, copayments, plan periods, cost per event.
  2. 02

    Clinical

    The provider’s view: what was observed and done, as recorded by clinicians at the point of care.

    Source
    High-complexity hospital
    Evidence type
    Primary clinical
    Typical content
    Outpatient consultations, laboratory results with units and reference ranges, imaging reports, admissions, critical-care stays, deliveries, procedures, discharge diagnoses.
  3. 03

    Network

    How patients move through the system: who refers, who treats, and in which setting.

    Source
    Payer and provider
    Evidence type
    Derived
    Typical content
    Physician and specialty of record (pseudonymised), referral sequences, site of care, in-network versus out-of-network use.
  4. 04

    Behavioural

    Patterns over time: how often people use care, whether they return, and whether chronic treatment is sustained.

    Source
    Derived from layers I–III
    Evidence type
    Derived
    Typical content
    Visit frequency, gaps in care, refill-based adherence proxies, persistence on therapy, continuity with the same physician.

Model

OMOP CDM domains covered

Data are standardised to the OMOP Common Data Model with OHDSI standard vocabularies. The table lists the domains populated at launch. Domain-level counts and completeness are documented in the data dictionary.

DomainWhat it holdsPrimary sourceEvidence
person Demographics: year of birth, sex, reduced geography Both Primary
visit_occurrence Outpatient, emergency and inpatient visits with setting and dates Both Primary / claims
condition_occurrence Diagnoses coded to SNOMED CT (mapped from ICD-10) Both Primary / claims
drug_exposure Dispensed and administered drugs, coded to RxNorm with ATC classification Payer, provider Claims / primary
procedure_occurrence Procedures, surgeries and interventions Both Primary / claims
measurement Laboratory results coded to LOINC, with values and units; vital signs Provider Primary
observation Clinical observations and findings not covered by other domains Provider Primary
cost Real cost per event as borne by the payer; copayment where applicable Payer Claims
payer_plan_period Enrolment periods and plan type, enabling person-time denominators Payer Claims

Imaging studies are represented through procedure_occurrence and DICOM metadata (modality, body region, date). Pixel data are not part of the offering.

Granularity

Granularity and temporal coverage

Unit of record
Event level: one row per visit, claim, result, dispensing or procedure, linked to a pseudonymous person.
Temporal depth
Longitudinal history per person bounded by the enrolment period, with plan periods to compute person-time.
Date precision
Dates are shifted by a consistent per-person offset. Intervals between events are preserved exactly.
Refresh cadence
Periodic data cuts agreed in each project protocol.
Population
400,000 members of a private payer in Argentina and the patients of a high-complexity hospital in Buenos Aires.
Denominator
Enrolment-based. Incidence and prevalence can be computed on a closed population, not only on people who happened to appear.

Evidence types

Claims-inferred versus primary clinical

The two sources answer different questions, and the data dictionary labels every variable with its origin.

01

Claims-inferred

Derived from the payer’s authorisation and reimbursement records. Strong for utilisation, spend, persistence and the full breadth of a member’s contact with the system, including care delivered by other providers in the network. Diagnoses are those stated for billing.

02

Primary clinical

Captured by clinicians in the hospital’s electronic record. Strong for severity, laboratory values, imaging findings and in-hospital outcomes. Covers the subset of the population treated at the hospital.

Where both sources describe the same person, the record carries both views and the dictionary marks the overlap, so analyses can test claims-based definitions against the clinical record.

Depth

Specialties with particular depth

Volumes reflect the hospital’s clinical profile. They indicate where the primary clinical layer is densest.