Table 3.

Comparison with T3 and D4

T3 (APPLICATION OF STUDY DESIGN)
NETWORK	LOCAL PARTNERS EXECUTE SHARED PROCEDURE	COMMON AMONG DBS	SCORES ESTIMATION	SPECIFIC SOFTWARE	PROGRAMMING LANGUAGE
EU-ADR	Y	Y	N	Jerboa	Java & Jerboa scripting language
Mini-Sentinel	Y	Y	Y	Modular programs and macros; PopMedNet	SQL, SAS, Java, R,
OMOP	Y	Y	Y	—	SQL, SAS, R, C, Java
MATRICE	Y	Y	N	TheMatrix	Java & TheMatrix scripting language

D4 (DATASETS FOR ANALYSIS)
NETWORK	TYPE	FORMAT	QUALITY: STUDY RESULTS VALIDATION
EU-ADR	Intermediate files that can be shared among partners, analysis will follow	csv	Drug safety methodology: comparison of observed drug-event associations with previously classified true and false causal associations; impact on this of different definitions of the derived data
Mini-Sentinel	Level of granularity of dataset depends on study needs; always transfer minimum necessary. Some analyses transfer aggregate data, some use highly-summarized patient-level data Intermediate files saved locally by data partners	csv, SAS datafiles, HTML	To test code known associations are used. Rapid Response queries include data characterization and are reviewed manually by a data expert and an epidemiologist. Results are also reviewed by data partners. Protocol-based assessments might include chart reviews.
OMOP	Final estimates, intermediate files are discarded	csv, SAS datafiles, SQL tables	Drug safety methodology: comparison of observed drug-event associations with previously classified true and false causal associations; impact on this of different definitions of the derived data; estimate of residual bias per event by means of known non causal associations.
MATRICE	Intermediate files to be used for analysis or report generation	csv	Results are reviewed by data partners for comparison with similar analysis performed independently