Design:
Incidence rates were obtained by applying a retrospective cohort design.

Study period:
In order to obtain stable rates and to look at trends over time, the study period ran from January 2000 up to January 2009

Population:
The population comprised all individuals who were registered within the databases during the study period and for whom an adequate start and end of follow-up could be defined. The start of follow-up was the date that the person was first registered in the database or start of data collection, whichever is latest. The end of follow-up was patient death, transferring out of the study population or end of data collection, whichever is earliest.
Each center needed to create locally a patient file which comprised the following variables: unique patientID (linkable across the files), startdate (start of follow-up: may also be earlier than 2000 for those registered earlier), enddate (end of follow-up), date of birth (if no exact dates are available use midpoints (e.g. 15/07) and sex (M/F). This file was one of the Jerboa® input files (see appendix 1). Jerboa® forced one year of up to standard time and set the true start of follow-up based on the registration date, study period and the one-year of data requirement.

Outcomes:
The events of specific interest (ESI) have been defined by the EMA and comprise central neurological, peripheral neurological, autoimmune and other conditions:

1) central neurological disorders include: convulsion12, encephalitis (including encephalomyelitis and myelitis)14, demyelination (i.e. Multiple Sclerosis)14, Bells palsy18 19, transverse myelitis

2) peripheral neurological disorders include: (optic) neuritis, (systemic) vasculitis15 16, Guillain-Barré syndrome17

3) autoimmune disorders include: anaphylaxis13, autoimmune hepatitis, oculorespiratory syndrome20, thrombocytopenia21

4) other disorders: spontaneous abortion, and sudden unexplained death.
All of the ESI can be assessed in the available databases, except for spontaneous abortion and oculorespiratory syndrome, which were therefore excluded from the analysis. For all events, the start date was defined as the date of diagnosis.

Case definitions are according to the definitions developed by the Brighton Collaboration and literature in case of absence of a Brighton definition (Appendix 2). Corresponding disease codes were used to extract information on the outcomes from the databases. Owing to differences in disease coding schemes and the use of free text in most medical record databases, terminologies were mapped according to a common terminology system. Databases in PIV SANE use one of four nomenclature systems to describe the events: the International Classification of Diseases (ICD9-CM and ICD10); the International Classification of Primary Care (ICPC); and the READ Code (RCD) classification. These different terminologies were mapped using the Unified Medical Language System® (UMLS®) followed by extensive manual checking. The UMLS is a biomedical terminology integration system handling more than 150 terminologies including the four used in the PIV SANE project. Database query harmonization followed an iterative process with six stages: 1) event definition using clinical criteria established by the Brighton Collaboration; 2) identification of UMLS concepts corresponding to the event; 3) discussion among database owners regarding relevance and applicability of the UMLS concepts identified; 4) translation of the concepts into the terminology of each individual database; 5) extraction of data; and 6) creation of input files for Jerboa© and verification of output.

Each center created locally an event file which comprised the following variables: unique patientID (linkable across the files), date event (start of disease), event (acronym for disease, see Jerboa instructions). The date of the event was the date of diagnosis.
Case identification was entirely based on disease codes. There was almost no case validation in this first phase of the study (exception IPCI). In a subsequent phase, the positive predictive value of the disease codes will be calculated by conducting chart review on a selected number of cases.

Jerboa® calculated the incidence of disease while differentiating between chronic diseases and intermittent diseases (i.e. convulsions, anaphylaxis, thrombocytopenia and spontaneous abortion). For chronic diseases Jerboa® censored at the first occurrence of the disease. This means that all subjects with an event prior to the start of follow-up (prevalent) were excluded. The run-in period incorporated by Jerboa® to define incident events was varied between 0 years, 1 year and 3 years. Depending on the latency period of the disease a longer run-in may be chosen. For episodic diseases (i.e. convulsions, anaphylaxis, and thrombocytopenia) a run-in of 0 could be chosen. Jerboa® calculated rates of episodic diseases on repeated occurrence of the event, following a fixed episode duration of 4 weeks.