| Description |
Many public health entities maintain clinical data repositories useful for assessing the harm caused by adverse drug events (ADEs). In Utah, the state government has databases containing claims data for emergency department visits; claims data for inpatient visits; death certificates; medical examiner records; and controlled substance prescription records. These data are promising sources for surveillance of ADEs because of their ubiquity and standardized encoding, though they have limitations including restricted clinical content and lack of integration. To assess the value of these repositories for ADE surveillance in the public health context, the utility of three data source approaches was examined: a single data source; multiple, unlinked data sources; and multiple, linked data sources. Evaluation of a single data source ADE surveillance approach was conducted using inpatient claims data to monitor harm due to anesthesia drugs. The results demonstrate suboptimal performance; notably, very low sensitivity and tendency to identify ADEs not directly related to anesthesia agents. Multiple, unlinked data repositories were used to examine trends in methadone prescriptions for pain and associated ADEs. Emergency room visits, death certificates, and medical examiner cases showed a consistent trend of increasing morbidity and mortality as methadone prescription rates increased. The correlation of multiple independent data sources is important evidence, but the study was ecologic and additional work was necessary to further characterize methadone-related harm. A multiple, linked data source strategy was used to further investigate the ecologic results. Using probabilistic and deterministic record matching techniques, prescription records and death certificate data were linked into longitudinal patient histories. The linked data set was used to characterize deaths due to methadone, including quantification of prescription recipients, prescriber practice patterns, and decedent characteristics. Although linking several data repositories requires greater technical expertise, the results show that multiple, linked data sources offer a powerful means to conduct ADE surveillance. The lack of integration and restricted clinical detail of public health data is a concern for ADE surveillance. However, with appropriate use of informatics methodologies, surveillance for ADEs is satisfactory for public health assessment and intervention needs. |
