Using aggregated, de-identified electronic health record data for multivariate pharmacosurveillance: A case study of azathioprine

- We use over 10 million electronic health records to study side effects of azathioprine.
- Liver damage was not increased relative to patients taking other drugs.
- Rounding from statistical de-identification did not significantly bias the calculated risks.
- Our approach successfully recapitulated classical knowledge of bone marrow suppression.