A Pharmacoepidemiology Database System for Monitoring Risk Due to the Use of Medicines by New Zealand Primary Care Patients

Andrew M. Tomlin, David M. Reith, David J. Woods, Hywel S. Lloyd, Alesha Smith, John S. Fountain, Murray W. Tilyard

The use of large record-linked healthcare databases for drug safety research and surveillance is now accepted practice. New Zealand's standardized national healthcare datasets provide the potential to automate the conduct of pharmacoepidemiological studies to provide rapid validation of medicine safety signals. Our objectives were to describe the methodology undertaken by a semi-automated computer system developed to rapidly assess risk due to drug exposure in New Zealand's population of primary care patients and to compare results from three studies with previously published findings. Data from three national databases were linked at the patient level in the automated studies. A retrospective nested case-control design was used to evaluate risk for upper gastrointestinal bleeding (UGIB), acute kidney failure (AKF), and serious arrhythmia associated with individual medicines, therapeutic classes of medicines, and concurrent use of medicines from multiple therapeutic classes. The patient cohort available for each study included 5,194,256 patients registered between 2007 and 2014, with a total of 34,630,673 patient-years at risk. An increased risk for UGIB was associated with non-steroidal anti-inflammatory drugs (NSAIDs) (adjusted odds ratio [AOR] 4.16, 95% confidence interval [CI] 3.90-4.43, p < 0.001) and selective serotonin reuptake inhibitors (AOR 1.39, 95% CI 1.20-1.62, p < 0.001); an increased risk for AKF was associated with NSAIDs (AOR 1.78, 95% CI 1.73-1.83, p < 0.001) and proton pump inhibitors (AOR 1.78, 95% CI 1.72-1.83, p < 0.001); and an increased risk for serious arrhythmia was associated with fluoroquinolones (AOR 1.38, 95% CI 1.26-151, p < 0.001) and penicillins (AOR 1.69, 95% CI 1.61-1.77, p < 0.001). Automated case-control studies using New Zealand's healthcare datasets can replicate associations of risk with drug exposure consistent with previous findings. Their speed of conduct enables systematic monitoring of risk for adverse events associated with a wide range of medicines.