Methods for Retrospective Detection of Drug Safety Signals and Adverse Events in Electronic General Practice Records

Background: Examination of clinical data routinely recorded in general practice provides significant opportunities for identifying and quantifying medicine-related adverse events not captured by spontaneous adverse reaction reporting systems. Robust pharmacovigilance methods for detecting and monitoring adverse events due to treatment with new and existing medicines are required to estimate the true extent of adverse events experienced by primary care patients. Objectives: The aim of the study was to examine evidence of adverse events contained in general practice electronic records and to study observed events related to selective serotonin reuptake inhibitors (SSRIs) as an example of drug-specific pharmaceutical surveillance achievable with these data. Methods: Electronic clinical records for a cohort of 338 931 patients consulting from 2002 to 2007 were extracted from the patient management systems of 30 primary care clinics in New Zealand. Medical warnings files, prescription records and free text consultation notes were used to identify physician-recorded treatment cautions, including adverse events and medicines they were associated with. A structured chronological analysis of prescriptions, consultation notes and adverse events relating to patients prescribed the SSRI citalopram was undertaken, and included investigating reasons for switching treatment to another SSRI (fluoxetine or paroxetine) as a method for detecting evidence of drug safety signals. We compared the number of adverse events identified for patients at one practice with the number spontaneously reported to New Zealand’s Centre for Adverse Reactions Monitoring (CARM). Results: During the 6-year study period, 173 478 patients received 4811 561 prescriptions. There were 37 397 allergies, adverse events and other warnings recorded for 24994 patients (7.4%); adverse events relating to 65 different types of drug were reported. Medicines most frequently implicated in adverse event reports were antibacterials, analgesics, antihypertensive medicines, lipid-modifying agents and skin preparations. Citalopram was prescribed for 5612 patients, and 701 adverse events relating to citalopram were identified in the electronic health records of 473 (8.4%) patients. A total of 713 (12.7%) patients changed treatment from citalopram to another SSRI, and 164 reasons for the switch were identified: suspected adverse drug effects for 129 (78.7%), lack of effect for 29 (17.7%) and patient preference for 6 (3.7%). The most common adverse events preceding the switch were anxiety, nausea and headaches. Of the 725 adverse events and medical warnings recorded at one practice, 21 (2.9%) were spontaneously reported to the CARM. Conclusions: Routinely recorded general practice data provide a wealth of opportunities for monitoring drug safety signals and for other patient safety issues. Medical warning records and consultation notes contain a wealth of information on adverse events but structured search methodologies are often required to identify these.