Table 2.
T1 (DATA REORGANIZATION) | |||
---|---|---|---|
NETWORK | RECODING | QUALITY: DATA COMPLETENESS | QUALITY: DOCUMENTATION |
EU-ADR | Does not require mapping to external standard: original coding and/or free text is maintained | Demanded to local partners, no formal procedure | No formal documentation |
Mini-Sentinel | Source data are homogeneous in coding systems | Local report on specific issues + feedback from standard programs checking for completeness and consistency | Data model, data elements and guiding principles approved by partners. ETL formal document, ad hoc per DB |
OMOP | Source data standardized to common vocabulary by domain: Drug (RxNorm), Condition (SNOMED), Labs (LOINC) | Formal procedures: OSCAR and GROUCH tools | ETL formal document, ad hoc per DB |
MATRICE | Source data are homogeneous in coding systems | Formal procedures checking data completeness | Local configuration of the TheMatrix software (text file) |
D2 (GLOBAL SCHEMA) | ||||
---|---|---|---|---|
NETWORK | NAMES OF TABLES CHOSEN ACCORDING TO | NAMES OF ATTRIBUTES CHOSEN ACCORDING TO | EVERY TABLE OF THE CDM HAS A VIEW IN EVERY DB | ATTRIBUTES ARE CODED UNIFORMLY ACROSS DBS |
EU-ADR | Reason/setting of data recording | Clinical contents | N | N |
Mini-Sentinel | Clinical content and data source (diagnosis, procedures, encounters, lab results) or reason/setting (outpatient pharmacy, death, enrollment) | Reason/setting of data recording for diagnosis and similar, clinical contents for pharmacy and death | N | Y |
OMOP | Clinical content | Reason/setting of data recording | Y | Y |
MATRICE | Reason/setting of data recording | Clinical contents | Y | Y |
T2 (DATA DERIVATION) | ||||
---|---|---|---|---|
NETWORK | LOGIC | SINGLE DEFINITION PER DERIVED DATA | QUALITY: PROCESS CONTROL | QUALITY: VALIDATION |
EU-ADR | DB-specific algorithms, harmonized through a formal negotiation process | Y | No common procedures were implemented., although logic of local procedures was shared | Internal incidence rates comparison, comparison with literature, some validation with external gold standard (PPV) |
Mini-Sentinel | The same algorithm was used across all DBs | Y | Shared SAS script | Systematic review of previously published validation studies, expert clinical, data, and epidemiologic guidance, medical chart review for PPV and assessment of difference in dates |
OMOP | Multiple alternative algorithms were adopted to derive the same data, some were DB-specific | N | Shared parameterized SQL queries stored in common procedure (RICO) | Internal prevalence rates comparison, no external validation performed |
MATRICE | Multiple algorithms were explored, decision was taken by means of a validation study | Y | Shared script in a scripting language developed ad hoc (TheMatrix) | Validation of algorithms with external gold standard: sensitivity, specificity, PPV, NPV |